AMD: Zen chips headed to desktops, servers in 2016

The roadmap for upcoming AMD products has been pretty murky lately, but the company revealed some of its key plans for the next few years at its Financial Analyst Day today.

The zen of Zen

The biggest news has to do with Zen, the new x86-compatible CPU core meant to replace the current Bulldozer family. Today’s presentations didn’t go deep into the technical details of the new architecture, but what we learned was largely encouraging.  We know the general timeline and outlines of the first Zen-based products.

First of all, Zen is indeed a new high-performance core intended to compete against Intel’s best x86 processors.  AMD expects this core to deliver about 40% higher performance per clock cycle than today’s Bulldozer variants, as the slide above indicates. Zen looks to be more of a “brainiac” architecture like K8 and Broadwell than a “speed demon” like the Pentium 4 and members of the Bulldozer lineage.

Also, the first Zen-based cores are just the beginning.  Future “Zen+” variants of this architecture should improve per-clock performance even further.

The Zen core will feature simultaneous multithreading (SMT), or the ability to track and execute multiple threads per core. Although SMT can extend beyond this limit, Zen’s version of SMT will stop at two threads per core, like today’s big Intel cores. The inclusion of dual threads per core follows a proven template for success in big x86 CPUs, and it also should put AMD on more equal footing with Intel from a marketing standpoint.

The other reassuring technical detail is Zen’s addition of a “high-bandwidth, low-latency cache system.” Bulldozer’s caches have been considered something of a pain point since the first chips arrived, and they clearly don’t perform as well in directed subsystem tests as the caches on recent Intel CPUs.  Thus, this little nugget also seems like good news.

AMD expects to bring the first Zen-based silicon to market in 2016, and those chips will be based on a chip fabrication process that uses FinFETs, also known as 3D transistors.  The use of FinFETs is positive news in the sense that it should allow for faster switching speed and lower voltage operation than traditional planar transitors—and AMD’s major competitor will be producing chips on its second or third generation of 3D transistors by 2016. AMD didn’t reveal the specific foundry or process on which these chips will be made, but the obvious list of candidates is pretty small, including the 14- and 16-nm FinFET processes at Samsung, GlobalFoundries, and TSMC.

A many-core processor for high-end desktops and servers in 2016

What’s really intriguing about the first Zen-based silicon is the sort of systems it will target. My sense is that AMD will produce a single high-end CPU and aim it at two markets: high-end desktop PCs and servers.  The roadmaps for Opteron and FX processors describe a chip based on the Zen core with a “high core count with multi-threading,” ample memory bandwidth, and lots of built-in I/O.

On the desktop side of things, the Zen-based FX processors will be supported by a new AM4 platform that includes support for DDR4 memory. This same AM4 platform will also support smaller APU chips, unifying the company’s desktop offerings around a single socket.

Into 2017 and beyond

The downside of the news about a big, Zen-based chip in 2016 is that AMD apparently had to prioritize this one chip over other options. The company’s 2016 APU products for desktops and mobile systems will not yet incorporate the Zen core. Also, the schedule for the K12 core, the ARM-compatible sister to Zen, has been pushed back to 2017. CEO Lisu Su further revealed that Project Skybridge, the effort to make ARM- and x86-compatible CPUs share the same sockets and motherboards, has been nixed, reputedly due to lack of customer demand.

What we know of AMD’s plans for 2017 looks intriguing. One chip in particular, mentioned in the datacenter roadmap, appears to be something different from anything we’ve seen before: a “high-performance server APU.” This product may be the first from AMD to combine a best-of-breed x86 CPU with truly powerful graphics on the same chip. The major limitation preventing this sort of converged product in the past has been the bandwidth available in a CPU socket. Simply put, CPU sockets haven’t had the capacity to support a data-streaming processor like a GPU.

The CPU-socket bottleneck may be alleviated by the “transformational memory architecture” mentioned on the roadmap, which is almost certainly the high-bandwidth memory (HBM) subsystem that AMD has been developing for use with its GPUs. HBM situates the memory on the package next to the chip it serves, such as a GPU or an SoC, and promises substantially higher bandwidth than traditional DRAM in combination with lower power consumption. The total physical size of the chip-plus-memory solution can be much smaller with HBM, as well. AMD’s execs strongly hinted that HBM will make its way across the company’s products going forward, and this server-class APU looks to be one beneficiary.

That prospect is exciting because it could give AMD a foothold in markets like HPC and supercomputing, where it hasn’t really mounted a credible challenge to rivals like Intel and Nvidia in recent years. This big server-class APU could also begin to deliver on the promise of AMD’s HSA programming model, which has so far been full of promise but lacking truly compelling hardware implementations.

This big APU won’t just target servers, either. AMD hasn’t officially extended its desktop roadmap into 2017, but a source familiar with the firm’s plans has indicated to us that a beefy APU of this sort could make its way into client systems like laptops and desktops, as well. The vision here is for a single-package product with HBM to combine a high-performance CPU with a Radeon graphics solution that has real gaming chops—something well beyond the offerings we’ve seen from AMD’s own APU lineup and Intel’s Core i5/i7 processors to date. If it works out well, such a product could give AMD a product unlike anything else available: a decent gaming PC in a compact package, ready for gaming laptops, all-in-ones, and small-form-factor systems.

Comments closed
    • DarkMikaru
    • 4 years ago

    So, I’m just going to be blunt. AMD’s CPU’s are competitive for 90% of us out there. They just are. From Developers on down the line, not a single person at work (from any of my places of employment) have ever complained about any of my AMD builds performance. From FX4170’s down to A4s, people are able to get their work done just fine. And after upgrading a few of these machines to SSD’s (as these machines were built with WD Blue 500GB HDD’s 2+ yrs ago that started showing their age) people are seeing that storage speed, not CPU speed is what’s most important.

    It just seems to me that you all look at benchmark scores and rendering times only. And in reality, only a small percentage of us actually NEED an i5/i7 level of power for what we do. Horsepower means nothing to the average computer user…which is again..MOST OF US! I build 20 to 30 machines a year, between work, friends & family and my own personal clients. Of all those machines, less than 3 are Intel based builds. Why?

    1) What do you want to do with your machine now and in the future? Will you ever dabble in 3D graphic modeling, photo editing, video transcoding, gaming, etc.?

    2) What is your budget?

    Those two questions that intertwined with each other. Those clients who I’ve built i7’s for actually do video editing / 3D modeling for a living. That makes since as that is more of an investment. As they say “Time = Money”. However, no one needs an i7 for Facebook, Youtube or reading Techreport! 😉 We just don’t. Which is what most of us do even when we have a powerful system. Let’s face it, we aren’t working all the time.

    It just aggravates me when people say AMD isn’t competitive. Because in general everyday basic computing tasks they are more than qualified. It’s when you get into higher end computing / gaming scenarios where you’ll see the difference. But for most of us an AMD system is just fine. Give AMD the business on the lower end and recommend Intel for the high end for those who can afford it. But honestly, I’ve yet to have a single client come back and say.. “yeah, I think I want to go Intel on the upgrade next year.” They just don’t. You are either a power user or you aren’t. And AMD missed the boat on making that distinction and capitalizing on it, years ago.

      • Bensam123
      • 4 years ago

      Oh my, context! That’s where price/performance ratios come in and AMD is quite competitive there.

      • ronch
      • 4 years ago

      I agree. It’s the same way Mediatek makes SoCs for cheap phones. Sure, they are not Apple Cyclones or Qualcomm 8xx chips, but they work just fine. And it’s the same way a Toyota isn’t a BMW but hey, they work well enough too.

      You don’t see people bashing Mediatek or Toyota.

        • DarkMikaru
        • 4 years ago

        Exactly! In the grand scheme of things, in daily computing an AMD chip does the job just as well as any Intel chip. Period. Not competitive for that 10% of the market that actually could use that extra horsepower sure, but for the everyday computer user it’s more than enough.

        Recently had a friend who I built an A4-4000 APU based machine for send it back to me for help. She had picked up a nasty virus and couldn’t resolve it herself. So since she lived several states away I asked her if she wanted a Quad Core, needed more ram, bigger SSD etc. while I had it. Her response “No, performance has never been an issue.” She’s a professor teaching at University, so she spends much of her time on her computer. She’d tell me if the speed wasn’t up to par. Hell, my parents are still rocking an A4-3300 FM1 build from a few years back now and I never hear a word out of them with respect to it being slow or unresponsive.

        I’ve ranted long enough. But stopping to spread negativity about AMD’s lack of performance in that 10% of the market would probably do wonders to bring up consumer confidence.

      • UnfriendlyFire
      • 4 years ago

      AMD isn’t competitive in marketing and OEM boot-licking. That’s how Intel effectively dominated the laptop and desktop market, combined with their “Intel Inside” advertisements and OEMs stating “Intel Inside” on their products.

      I recall seeing a large Dell kiosk in a shopping mall that had a fair amount of Intel advertisements on them.

      There aren’t many laptops or desktops that have AMD’s APUs. And I know majority of gaming laptops have Nividia GPUs.

      The only area where AMD’s mobile GPUs have some sort of presence are in the premium business laptop (with dedicated GPU) area, and those laptops a minority of the overall business laptop market.

      The final kicker was that Intel secured the lucrative server market because performance-per-watt and idle power consumption was an major importance.

    • WaltC
    • 4 years ago

    Finally, just like Microsoft, AMD is also recovering from the “fruit of the poisonous tree” that Apple seems to inject into any major tech market it becomes involved in. With iPad sales continuing to tank and Cook recently stating he didn’t know when they’d rebound (“if ever,” between the lines), it’s refreshing to see that not even Apple is continuing to hawk the pernicious “post PC” nonsense that was never true/accurate from the first time Jobs coined the phrase (He was the inventor of the RDF, after all.)

    Anyway, it’s nice to see both Microsoft and AMD getting back into the saddle again with regard to x86 desktop & server tech. Win10 is coming along nicely–but still has a plethora of bugs to fix, but no show-stoppers at least for me with 10074–which I’m now using every day, actually, and booting 8.1 much less these days in comparison. AMD understands that it must put itself out front of Intel once again in the performance sector if it expects to survive as a company and Zen seems like just what the doctor ordered. HBM on its discrete gpus is also the proper direction.

    AMD is on the right track–my hope is that this time, should the company once again leapfrog Intel technically that it won’t sit on its laurels and wait for Intel to surpass it like it did the last time. Hopefully, the AMD upper management has learned that lesson. Intel is a tough competitor, just like Microsoft is a tough competitor–and any company hoping to compete with either in their primary market sectors needs to keep the R&D pedal to the metal constantly and without letup.

    Now that I know that AMD is putting the proper emphasis on the proper segments of the markets I have much confidence that the company will begin executing well again and force Intel to get off of its milking strategies. Every company that has gone toe-to-toe with Intel in x86 cpu markets has ceased to exist–every single one of them except AMD. That’s because AMD managed to leapfrog Intel technologically in the x86-compatible, Windows-compatible markets. AMD is going to have to do that again to regain it’s former ascendancy. They did it once (the only company to have done so) so they can do it again, at least theoretically…;)

    The thing about Intel that I admire is that the company realized the whole iPad/tablet thing was at best a short-lived fad (which I knew from day one as the dynamics driving the consumer x86 PC markets are unbeatable in terms of user-serviceability and user-configurability), and that the best bang-for-the-buck, the most consumer-friendly market there is in computer hardware is the desktop/laptop PC market–and Intel not once ever flinched from that position–as did both Microsoft and AMD to some degree, to their chagrin. But things are finally back on track for AMD and Microsoft, and the recent AMD announcements as to direction and strategy and product mix indicate that AMD’s “head” is once again on “straight”…;)

    Viva la’ competicion…! (Excuse my murdered Spanish, but some celbratory language seemed appropriate!…;))

      • K-L-Waster
      • 4 years ago

      Well, don’t get carried away – it’s a slide deck, not a blockbuster product launch.

    • RdVi
    • 4 years ago

    All the need to deliver for me to upgrade is close to haswell IPC, 8 cores, reasonable perf/watt and a price tag that is a little bit more enticing than Intel’s current high end offerings and I’ll jump on it. It would be my first AMD CPU since an Athlon 64 3500+. I’m at the point where I just don’t care how fast skylake is, I really can’t get past that I’d be buying yet another quad core. Intels 6 and 8 cores are just too expensive and are based on outdated architecture anyhow (not so much now, but usually).

    • ronch
    • 4 years ago

    Now I’m sure it CAN run Crysis!

    • TopHatKiller
    • 4 years ago

    According to Chips+Bits – a site that funnily enough does some actual investigative journalism [curse the idea!] the 40% ipc thingy is a deliberate understatement… they don’t say why, but the site isn’t usually right-off the mark on techy stuff. I would also posit that AMD’s default position is not to exaggerate on cpu benchs. [Please don’t laugh – that’s actually about true.]
    AMD was ever so cautious in providing no information that could possibly be misunderstood as actually comparative about anything to anyone. Why would they? They don’t compete until they provide the goods, why tip off the adversary before the fight?

    PS I’m in the UK and in election-hell tonight, a bottle of vodka and more smokes then I can handle doesn’t seem to be damping my terrifying dread of the result. My fellow citizens can’t be voting for those rightwing scumsucking filfth again, can they?! God please help. Earlier it was mentioned that an “AMD Defence Brigade” existed – if this is true and not a dirty tantalising lie, please… AMD Gods Everywhere launch an attack on us tonight.
    Sorry, know that’s off-topic.

      • chuckula
      • 4 years ago

      Here are a bunch of hammers to look at: [url<]https://www.google.com/search?q=hammers&hl=en-US&biw=360&bih=615&source=lnms&tbm=isch&sa=X&ei=uvtLVaPpB8OWNoyugdgC&ved=0CAYQ_AUoAQ[/url<]

        • TopHatKiller
        • 4 years ago

        Hammers will not do the job…not painlessly anyway. Thanks for the suggestion though…

        • anotherengineer
        • 4 years ago

        Nice selection, but I like the engineer’s hammer 😉

        [url<]http://www.hisltd.co.uk/Estwing-Engineers-Hammer.html[/url<]

      • TopHatKiller
      • 4 years ago

      Replying to myself. Mhmm.

      Rather then just downvoting, how about anyone says why? Don’t wont to face the horrible prospect of Zen being a killer architecture… Nor respond to Chips + Bits article. ? Or just shy?

      PS Either way, the election result for me is going to the bottle. I just cannot understand how my fellow countrymen can vote for tory rightwing scum. I’m at a loss,,,, looking at a future were all our poverty her in the UK is made much worse. And how can we cope? Sorry for being off-topic. I know. Sorry. I’m trying not to cry; but I’m struggling.

        • VincentHanna
        • 4 years ago

        If you want me to read an article, you have to link to it… And if you live in a country where [url=https://openmedia.org/blog/why-internet-users-everywhere-should-be-watching-eu-right-now<]Linking is illegal[/url<] then you need to get off the internets.

          • TopHatKiller
          • 4 years ago

          Far enough, except, i can’t be assed. i have to now put up with five fucking years of tory fifth assraping the poor and disabled in my country whilst the rich quaff champagne and laugh.

            • chuckula
            • 4 years ago

            From what I’ve seen of British politics the Tories are basically socialists, while Labour are basically fanatical socialists. Except that both sides pander to certain influential rich groups. Oh, and third-grade rants about the evil rich drinking champagne are just that, third grade rants.

            From the stories I’ve heard about people in England basically refusing to get jobs because it’s easier to live on welfare, I’m pretty sure that your problems have little to do with the “poor” not getting enough free stuff for not contributing anything to society.

            Example of how the “poor” are being “starved” to death in England: [url<]http://toprightnews.com/350-pound-woman-welfare-im-obese-dont-get-enough-government-money/[/url<]

            • TopHatKiller
            • 4 years ago

            No. Listen, mate.
            First off [1] political discussion is verging on trolling – but I started it – so I guess that’s on me.
            [2] Without being rude, your analysis on British politics is not…. wrong, but about as deranged as saying that the evil Martians rule the planet Earth.
            [2a] The tories go from centre to hard right. There are no socialists in that party.
            [2b] The Labour Party go from soft-right to leftist, with the occasional exception.
            [3] ‘Pandering to certain influential rich groups’: That would be different to what first-world political class in what country exactly?
            [4] ‘Quaffing champagne’ statement from me was I admit perhaps lazy, but it was an easy way
            to indicate the difference between the super-wealthy that we have in my nation and the rest of us. In Britain now we have more billionaires then anywhere else in the world [even then in the US] and an increasing problem with genuine poverty and suffering.
            If I was ranting, and frankly I don’t think I was, then it might be because I’m sickened by having to step over beggars in the street when not a few streets over their are un-taxed
            companies earning millions. In my own neighbourhood.
            [5] I have no idea ‘what stories you have heard’ about my country, but I do know – very very much – the repellent bias present in the corporate media: – even in my own country – never mind what horrid source you linked to. And I do know in real life – not fantasies made-up
            by political agitprop for the interest of the torries – but my own life experience where people
            struggling with their disabilities have been cruelly denied any help or support. Friends that struggle day-to-day to deal with what nature or accident has cursed them with, and the current [and now, new christ] government have denied them.
            Additionally: though I never met him: a gentlemen who came back from fighting in Iraq: he came home without legs and a complex injury to his back, this caused him horrendous pain and difficulty in his lower regions, and i guess because he was some kind of communist, he applied for disability benefits under the current governments law. Because he failed to ‘tick-every-box’ this government could wriggle out of providing him anything at all. It took over six months for his family and lawyers to actually secure him financial help.
            Does all this sound like the “poor” not getting enough for free?
            [6] Under the pervious government we, who live here, in Britain have suffered increases in homelessness, child poverty, absolute poverty… that’s what is actually happening here…
            [7] I’m rather happy about being British and living here, I think it’s quite a great place to be.
            Despite all out horrible problems, god knows, but I become angry with this sort of pig-ignorant shite vomited up from you.

            Now, anyway, the horrible government that has violated the promise made, after the second world war, to keep people safe – not just in blood, but in the victuals of common human life,
            has somehow come back in power. By a few percent. And there’s an enormous number of us
            who are gutted,depressed,perplexed by it.

            That’s My Speech Gone.
            ====== I apologise for everyone else ================
            ===I think Nvdia/AMD/Intel are crap? Discuss?? ===================

            • Suspenders
            • 4 years ago

            You keep using the word “socialist”. I do not think it means what you think it means.

            • VincentHanna
            • 4 years ago

            I read the Reuters article on the election… It made me think of you.

            Apparently Cameron barely lifted a finger to overthrow a Scottish separatist movement while winning [url=http://uk.reuters.com/article/2015/05/08/uk-britain-election-cameron-idUKKBN0NT21X20150508<]a landslide victory over[b<] fear.[/b<] [/url<] [quote<]Prime Minister David Cameron sealed a surprise election win by persuading Britons to choose the security of modestly rising living standards over an implausible pretender many feared could become the puppet of Scottish nationalists.[/quote<] Lol, and I thought Fox news was biased.

            • TopHatKiller
            • 4 years ago

            Thanks.

      • BaronMatrix
      • 4 years ago

      They should have said targeted for… With a real redesign using the best of everything since Hammer they should have a new chip for Cray to use…

        • Srsly_Bro
        • 4 years ago

        HE IS BACK! YES!

          • ronch
          • 4 years ago

          Bring Spigzone and abw back too! And Sschaem never deserved to be kicked out!

    • wingless
    • 4 years ago

    AMD needs to say “just kidding” and release Zen THIS YEAR! It can’t come soon enough.

    • ggaky00
    • 4 years ago

    I could never understand why all this hatred towards AMD.
    You don’t like their products? Great, there are alternatives.

    I keep seeing posts about “AMD will be late in execution as always”, “they have slow IPC so they have sh!t CPUs”, “their GPUS are awfully slow” “haha, Intel and nVidia will wipe the floor with AMD”. I never understood why the heck you care.

    Is it because it’s easy? Have we come to the point that we hate whatever we don’t like? So should I start hating Apple, HTC, Samsung, Philips, Creative etc. and start a flaming war towards companies that could NEVER care of what i think?

    You know, there’s a difference between hating and not liking.

    15 years ago people were “divided” but not fanatics.

    It’s becoming like soccer. I’m a Panathinaikos fan. Kill all Olympiakos fans. AAAAAAAAAAAAARG

    Oh well…

    Cheers !!!

      • VincentHanna
      • 4 years ago

      Frustration/pessimism and hate are different emotions. Hate implies malice. I want AMD to be firebombed by ISIS. That’s hate. I think the CEO of AMD should be hogtied and whipped. That’s hate.

      Clearly you do not know what hate is.

      Are we “at the point” where consistently disappointing performance leads to pessimistic outlooks? Absolutely. By definition, anything else is unrealistic.

        • ggaky00
        • 4 years ago

        So, wishing for a company to go out of business isn’t hate…. Is that what you mean?

        Cause, to my poor Greek brain, wishing for something bad to happen to someone has to do with hatred. Feeling hate isn’t always associated with physical consequences.

        And I don’t think that frustration or pessimism are associated with the flaming that AMD and other companies have had. Surely, they are emotions. But are totally different from all this that happen.

        Anyway, different opinions are always welcome as long as they are stated properly and in good manners. 🙂

        Cheers !!!

        P.S. (which i forgot to mention on the 1st post) Excuse my poor English, i’m only Greek 🙂

          • TopHatKiller
          • 4 years ago

          No point you have made is without decent merit. Critical comments you’ve received are frankly only supporting those same points.
          Personally, I suspect a significant element to unfounded attacks and criticism of AMD in this community is simply ‘post-purchase affirmation’ – those guys have spent their hard-earned [well, possibly hard-earned] on intel & nv, hence their going to attack competing products regardless of facts, sense or reason.
          Cheer! to You! Unpronounceable one.

          • K-L-Waster
          • 4 years ago

          No one is wishing them to go out of business – thing is, we’ve seen this movie before and it doesn’t usually have a happy ending.

            • VincentHanna
            • 4 years ago

            I’m kindof hoping they get bought out by someone like Samsung who has some cash to throw around, a reputation for cost management/solid design, their own chip foundry (with competitive lithographies), and a lineup of products that would directly benefit from AMDs patent portfolio…

            I’m not sure if that makes me optimistic, or pessimistic when it comes to the future of AMD.

            • ggaky00
            • 4 years ago

            AMD being bought from some other company be it Samsung/Apple/Sony/Mitsubishi or whatever, i think it makes things so complicated that it wouldn’t be allowed. There are so many patents at stake that lead to monopoly that FCC or whatever commission is involved wouldn’t allow such an act.

            Maybe the best thing for AMD is competitive products delivered on time along with Intel playing fair, or suddenly, a big money injection by a big investor.

            Cheers !!!

            • Suspenders
            • 4 years ago

            Im in the same boat. It would be nice if someone with big pockets were to buy them, because what htey really need to compete is some serious financial resources for R&D. I mean, Intel’s quarterly R&D budget alone is almost as much as AMD’s yearly revenue. How the hell are you supposed to compete with that?

      • chuckula
      • 4 years ago

      You’re confusing “We’ve heard this before and powerpoint slides aren’t a substitute for real silicon” pessimism for “hatred” when there really isn’t any.

      I’d say a solid 75% of TR readers would jump ship to AMD if they could just produce a desktop system that’s in the ballpark of what Intel offers. AMD wouldn’t even have to outright win, just be competitive to pick up quite a bit of business.

      Zen could be the chip that does the trick for AMD, but plenty of people here have heard AMD’s promises before.. with the Phenom and again with Bulldozer. The easiest way to change people’s minds is to actually get better products on the market. The rest takes care of itself.

      • BaronMatrix
      • 4 years ago

      Welcome to TechReport… Some of these folks may need therapy…

        • chuckula
        • 4 years ago

        Yeah!

        Here’s a list:
        1. BaronMatrix
        2. TopHatKiller
        3. Spigzone and his latest aliases

          • BaronMatrix
          • 4 years ago

          I’m sorry I don’t give a predatory monopoly convicted in every major market the benefit of the doubt for crapping on a non-profit…

          Wait, no I’m not… When AMD goes out of business, Intel may get some of my money…

          But we’ll see if AMD has enough CPU IP to drink some milkshake next year…

            • chuckula
            • 4 years ago

            Give it a rest with the cut & pasted drivel from 1998. Your precious AMD was run by a guy (Sanders) who bent over backwards to testify in favor of Microsoft at its antitrust trial, and his successor got kicked out for engaging in insider trading (he was lucky to avoid jail). Just this year, your precious AMD got sued by its own owners because the executives officially lied about the state of Llano production.

        • ggaky00
        • 4 years ago

        I’ve been reading TechReport for more than 12 years even though I never registered to write comments. I don’t think that any people need therapy. I just think that it’s easier to be sarcastic/blunt/cynical sitting in the comfort of your chair. Much like what happens in every website or online game.

        Damn, the whole human race is going down fast!!! :p

      • flip-mode
      • 4 years ago

      I don’t think you are reading the comments properly. Taken in context, the criticisms of AMD are not surprising, nor are they hateful. To do the work for you: people want AMD to offer a worthy alternative x86 product. AMD, for that matter, wants to do the same thing. AMD at one point had a superior chip. For several reasons AMD lost that edge. AMD has and continues to market its products as direct competitors to Intel’s. AMD has a history of setting expectations that their products can meet. So people criticize for two main reasons: first, AMD for how it markets itself, and second, for what is promised versus what is delivered. AMD sets the expectations. People criticize AMD when those expectations aren’t met. If AMD changed its message to “we are no longer competing in the mainstream x86 market and from now on our products will be targeting X, Y, and Z specific scenarios” and then delivers what is promised, that message would sink in after a while and you would stop hearing people criticizing AMD.

      • Bensam123
      • 4 years ago

      It’s easy… AMD is the dunce right now and possibly some over love for the other side. They also don’t have anything better to do.

    • wingless
    • 4 years ago

    They need to showoff real silicon this year. We need proof of this 40% bump to keep the company afloat.

    • Bensam123
    • 4 years ago

    Interesting, AMD has always had problems with adoption in the mobile area. Mainly due to efficiency, but people generally don’t like AMD as much as Intel (just like Nvidia vs AMD). If they had a choice between a low end IGP and a pretty high end one for a similar sum of money it’d be a no brainer. That’s something Intel can’t do without as close relationship to AMD or Nvidia as well.

      • K-L-Waster
      • 4 years ago

      In principle, APUs make perfect sense for laptops–provided they also fit in the power use and thermal envelope. Previous APUs were too power thirsty and ran too hot to be attractive to laptop makers. It remains to be seen how the upcoming chips will play out, but if AMD can deliver on what they have here they should be able to get some design wins.

        • brucethemoose
        • 4 years ago

        Llano was pretty dang good. The CPU was unlocked and clocked VERY conservatively, I got something like 40% more performance WITH an undervolt in my laptop.

        Unfortunately, llano was late, supply was short, and the GPU was a little small. Bulldozer basically killed any APUs after that.

          • Bensam123
          • 4 years ago

          Yeah… They have competitive power usage, just no one likes them. They may not have the same performance (CPU wise) as Intel stuff, but the video made up for it. No one bit though.

          Based on what we were just told it’s uping the ante even more on the GPU and the CPU may be quite a bit more power efficient. Win-win?

            • wimpishsundew
            • 4 years ago

            People didn’t bite because they didn’t even know it existed. The few laptops it was featured in were configured with abysmal specs and low quality construction. The problem was only enthusiasts know about the APU but it wasn’t targeted for enthusiasts. So that’s why half of AMD’s revenue is through China, where enthusiasts were fine with APUs because of their budget.

            • Bensam123
            • 4 years ago

            It wasn’t really targeted at anyone. Intel and Nvidia have a much larger marketing budget and bigger strings to pull to get their chips in everything under the sun.

            Hopefully a super beefy APU will change that. Sometimes things are so good you really can’t ignore them anymore.

    • ermo
    • 4 years ago

    So this basically confirms that the initial Zen core will have 3 instead of 2 ALUs per core? Just like Phenom II?

    On paper, this should improve IPC by 50%, but since you can’t expect to keep all pipelines fed all the time, +40% seems a decent number.

    ‘Progress’ indeed.

      • gruffi
      • 4 years ago

      IPC is far more than just the number of ALUs.

    • loophole
    • 4 years ago

    I just noticed that the socket next to the FX Vishera chips is listed as “AM3” rather than “AM3+”. For the FAD audience it’s probably not important, but I notice Kaveri is on “FM2+”, so they’ve been specific there.

    That just upsets one’s OCD…

      • ronch
      • 4 years ago

      AMD marketing is careless, as always.

    • anotherengineer
    • 4 years ago

    Funny nothing on future chipsets. Not that the A88X is bad or anything, just wondering if it will end up being an A95X or something and be used for both CPUs and APU’s?

    • jihadjoe
    • 4 years ago

    Interestly Toms Hardware did [url=http://www.tomshardware.com/reviews/processor-architecture-benchmark,2974-10.html<]an IPC comparison[/url<] of AMD and Intel chips 4 years ago. They ran each chip single core only at 3GHz, with turbo and power saving off. In most of those benchmarks it seemed like a 40% increase in IPC would have made Thuban competitive with Sandy Bridge, but Zambezi is starting off a bit further back than that.

      • maxxcool
      • 4 years ago

      I WOULD LOVE TR TO DO THIS NEXT YEAR!

    • ronch
    • 4 years ago

    This oughta shut up AMD haters who keep saying Zen is gonna be a small core. Also, it’s good how AMD has finally knocked some sense into themselves to know what people really want, instead of deluding themselves to thinking what people want. Welcome back to reality and sensibility, AMD. Oh, and I’m not sad to see SkyBridge go either. It is an interesting concept but I also kinda wondered how it would fit most of our lives, not just geeks who just wanna geek all over it.

      • maxxcool
      • 4 years ago

      I admit it. I was wrong! now, can they execute it ?

        • ronch
        • 4 years ago

        AMD is great at executing…. their own employees.

          • maxxcool
          • 4 years ago

          I will tell you what. I have HATED Amd’s plans and product since I bought my last thuban x6, and have been railing for years at the leadership, marketing and timelines.

          But if they can pull this off with the ”reduced staff” the have now, and prove my ACRES of pessimism wrong I will eat my shoes with a bottle taco bell hot sauce. gleefully.

            • DarkMikaru
            • 4 years ago

            What was wrong with your X6? I had a 1090T and loved it! Matter of fact, I hate that I ever sold that damn thing. I should of given it to my dad. He’d never need another upgrade again….EVER! LOL

            • maxxcool
            • 4 years ago

            Oh I liked the x6 itself , had 6 of them in total, 5 for a vmware lab for work that I built for myself and 1 that overclcocked the best for home gaming. But at the time the x6 was being put out the industry knew and the insiders knew that BD was not going to be a consumer solution, and it was not going to be anything close to intels mid range at the time. plus all the marketing was going to hell, it was spin-tastic…

      • Anonymous Coward
      • 4 years ago

      I’m still betting on something easy to design and easy to fab.

      [quote<]AMD has finally knocked some sense into themselves to know what people really want[/quote<] So you believe that AMD thought people wanted what, exactly?

        • ronch
        • 4 years ago

        APUs. Apart from space-constrained machines, who would seriously want them? Apparently, not many enough to make AMD lots of money.

          • Anonymous Coward
          • 4 years ago

          I see, so you think AMD should have packaged weaker graphics with their CPUs, surely a winning strategy.

            • maxxcool
            • 4 years ago

            devils advocate, having great graphics failed to deliver.

            • K-L-Waster
            • 4 years ago

            Having *ok* graphics failed to deliver. AFAIK even the best APUs were hard pressed to keep up with a discrete R9 250. Biggest blocker was memory bandwidth, since they were stuck on the DDR3 main system RAM.

            The new announcements may be better, if the APU compatible motherboards can also use DDR4.

            • Anonymous Coward
            • 4 years ago

            True, but it was a decent thing to try. They had (and still have) good graphics tech on hand. They needed to do something that Intel couldn’t, if they could justify any reasonable sale price. They tried to innovate with the CPU side of things, and it didn’t go very well.

            With the new Zen core, it seems they have given up being innovative, and will now simply be the cheap clone option. They are defeated.

            • maxxcool
            • 4 years ago

            as I told Ronch .. I LOTHE AMd of the last 5-6 years.. took every cheap shot on earth. *BUT* if they can get back to 90% of intels best for 200$ less. It will work. It won’t be ZOMG! .. but it will give them a third time under the recitation paddles. lets see what happens, because I do not want intel (as much as I like them) to be the ONLY game ..

            • ronch
            • 4 years ago

            No, it’s just that cramming CPU cores that will bottleneck any future graphics upgrades down the road and GPU cores that dont really interest gamers onto one piece of silicon is just… counter to sound logic. If you’re a gamer with money, just get a proper i5/i7+fast GPU. If you’re a gamer on a budget, just get an i3+R7 240 (or Pen’ium) and upgrade the CPU and GPU separately once you save up. If you don’t game and don’t need fast graphics, just get a Pentium/i3/i5/i7 and enjoy much better CPU performance.

            So really, I don’t see the value proposition of APUs. Even Scott seems to agree.

            • Anonymous Coward
            • 4 years ago

            I believe you are saying that AMD should have delivered Intel’s processors.

            Their all-CPU product had no identifiable advantage over Intel’s various offerings, and taking away their strong GPU part of the APU would have left them with absolutely nothing to talk about.

            • ronch
            • 4 years ago

            Honestly, I think Intel understands what people want and need better. If I were AMD, I would take out that GPU and give the CPU cores all the headroom to push farther. Come to think of it, that’s what those FX-4350 chips actually were.

            • Anonymous Coward
            • 4 years ago

            Lucky for you, AMD has been offering CPU’s without GPU’s without interruption. 4 threads, 6 threads, 8 threads.

            Let me quote you:

            [quote<]Also, it's good how AMD has finally knocked some sense into themselves to know what people really want, instead of deluding themselves to thinking what people want.[/quote<] Are you going to admit that this was a ridiculous statement?

            • maxxcool
            • 4 years ago

            As a power user, yes that’s what I would agree with. remove the gpu and add cores… or better add on die SRAM. 64mb or 128mb on die at speed sram would help the IPC of SR/EX by a good margin for the cache misses.

            but that would be another mask and sku and amd could not afford that then since they were knee deep into “NEW” zen.

            • K-L-Waster
            • 4 years ago

            For desktops this is true.

            For laptops and all-in-ones, APUs in theory make sense. However, those markets are also very power and thermal sensitive, so AMD needs to improve in those areas. Plus memory bandwidth to the IGP needs to improve, which could be handled by adding support for DDR4.

          • gruffi
          • 4 years ago

          The question should be, who doesn’t want an APU? Except high-end gamers I don’t see anyone. An APU is better in every way compared to an equally dimensioned CPU+dGPU combo. Most of AMD’s portfolio is based on APUs. The same applies to Intel. Every smartphone/tablet processor nowadays is an APU.

            • K-L-Waster
            • 4 years ago

            I can see the marketting pitch now: “Buy AMD, because you don’t really need all that CPU power the other guys are selling.”

    • jjj
    • 4 years ago

    Got to first note the delay for the ARM core, an extra year is a bit of a bummer.

    40% IPC doesn’t sound encouraging but final perf is not just IPC , we’ll see how all else works out and what clocks they can reach.
    The problem is that just matching Intel is not good enough in high end desktop. They need to beat them in single core perf and go 8 cores at 300$ so not only force Intel to stop behaving badly (4 cores+GPU at 300$) but cause a huge refresh cycle in this segment.
    Lots of people are just not in the mood to upgrade to 4 cores even if their system is rather old , give them a reason to get excited and they’ll upgrade but if you trigger that huge upgrade, you better have the best product so you can take advantage of that.

    Ofc there is also the option that AMD could downplay it’s perf gains for now. If you have an advantage, you don’t want your competitors to expect it this early.

    Was excited about y-day and watched the presentation in full but i’m less excited today. Was way boring. Not because it was a financial presentation, i’m used to that, it just lacked details , positive surprises or surprising strategies.
    The way they ferociously brag about the console wins when those wins were achieved before the current team and a few other points where they try to distort the truth, makes it very hard to have any faith in what they say about the future. You can’t just flat out lie about the now and the past and hope to have any credibility left when you talk about the future. Always liked the former WD CEO for being realistic as opposed to most others that are full of BS so maybe what bothered me the most y-day was the lack of realism.

    AMD also made it clear that they won’t go after phones and not sure i like that. Ofc they don’t have what is needed to go after phones at this point but if they had plans to do it in a few years, they wouldn’t have made it so clear that they are not interested in that at all.
    If you don’t go after phones and then glasses and w/e else comes, you might end up with no presence in personal computing long term. If AMD’s ( and Mvidia’s) long term strategy is to be server only, i got certain doubts about their ability to survive. They would have to compete with the guys that do the consumer side too and have more resources because of that.

    All in all, Zen is what matters now and we can still hope that it will be good so we’ll just have to wait and see.

      • _ppi
      • 4 years ago

      Re: Mobile chips

      Mobile chips are either about being super-low cost and use existing ARM designs (which is what MediaTek and Rockchip do), or do the same and leverage your process tech (Samsung), or have special features in the chips such as LTE modems (Qualcomm) or have better-than-stock-ARM cores in your chip (Qualcomm).

      AMD’s optimal way is the latest one – go ahead using custom ARM CPU and custom low-power GPU, that are better than stock ARM solutions. That would pit it right against Qualcomm, though, that has very good footprint in the industry (but at least, unlike Intel, it does not have process tech advange).

      But until AMD makes K12, and K12+Radeon SoC cannot be in 2W ballpark, with performance and power levels equal or preferably better than competition, there is no reason to pursue that market first, when AMD’s expertise is in high-performance chips and GPUs.

      To underline the above, look at some competition: nVidia has just withdrawn from mobile chips market, and Intel has to give away their chips for free to have any chance there.

        • jjj
        • 4 years ago

        I thought i made it clear in my post but guess not. So yes they don’t have what they need now (that’s less the core and more the modem and a bunch of other things) but they made it very clear that they won’t go after that market, suggesting that it won’t happen later either.. I wasn’t expecting them to do it now, because they can’t but they should do it soon.

        As for how the mobile market works, it’s more complicated than that.
        First everybody that sells to others ( so not Apple) and wants relevant share, ships mostly small cores. The volume for big chips that cost 40-50$ (with integrated modem) is very limited in phones, especially after you exclude Apple. There is no custom small core so Qualcomm does the same as Mediatek, Marvell or Spreadtrum (Rockchip is only now with the low end Intel trying to sell in phones, so they aren’t actually yet a player in phones). And the market is like any other competitive market you win or lose on perf and price and features and support and so on. The lack of a small custom core is actually a great opportunity for anyone that is not blinded by the irrational appeal of a big core. I could expand a lot on why and how a custom small core would be a great asset but this post is long as it is.
        On the LTE side ,that’s empty marketing. You don’t need cat 10 in a world that can’t sue it. Apple in it’s latest is only using cat4 and nobody seems to be complain about that. And a bunch of chips makers are shipping cat 4 and things are moving towards cat 6 and carrier aggregation. Nobody actually needs the most advanced modem, you just need ok perf and aim for small die area and low power.
        On the connectivity side a big factor is 5G and things might be very different. Nobody liked the Qualcomm early dominance and we might also see folks like ARM , PowerVR and other license 5G modems. There are at least some hints that things will be different.

        In the high end the focus on CPU is passe. A72 hits the good enough or even well above that.On the other hand the focus is shifting more and more on the GPU.
        Lets look at some chips and how they sue the die.
        Snapdragon 800 is 118mm2 , the 4 Krait cores 22mm2 the GPU 16mm2 , the modem 26mm2
        Kirin 920 is 125mm2, the quad A15 cores with cache 20mm2 , the quad A7 5mm2 , the GPU 21mm and the modem 26mm2
        Exynos 5433 is some 113mm2 , just under 20mm2 for the CPU and 30.9mm2 for the GPU
        Apple’s A8 is 89mm2, some 12.2mm2 for the CPU (plus some 4.5mm2 for a big L3 cache) and 19.1mm2 for the GPU
        By next year you could have some 40mm2 for the CPU cluster and modem together and 40mm2 or even more for the GPU, considering what’s needed for a 4k screen.
        So for someone like Nvidia or AMD having great perf per die GPU would be a big advantage.

        So bellow the high end it’s not a market one can ignore, it’s not unprofitable and there are smart ways to attack it.
        In the high end, being good at GPU would be a huge advantage and the GPU will only gain in importance.

        A key point are also the future form factors, if you are in phones, you are better prepared for glasses and what comes after that.

        As for other players. Some are failing like Intel or Nvidia, but others aren’t, like Spreadtrum and Mediatek. And some are fighting like Marvell or Samsung.
        For Intel it’s not surprising, they have this image that they are very good but that’s false. If you look closely, Intel messes up often in everything they do where they don’t have a monopoly. Winning a one-horse race is not much of a win and in everything else they stumble again and again.
        Nvidia gave up or maybe not, it’s stupid if they did but it’s not certain they won’t be back when 5G arrives. It might be that the Icera tech was just not cutting it and they don’t want to waste money on it when 5G might be easier to handle.
        In the end, this is a market like any other market but it’s important for anyone that wants to be in personal computing to have a presence.

        Edit: Just to add a small fact that might change some opinions about costs.
        Last year AMD’s spending on R&D was some 3x ARM’s R&D budget and Nvidia’s was some 4x . And ARM does a lot of things (http://www.arm.com/products/) with so little money.

    • VincentHanna
    • 4 years ago

    Is it bad that their roadmap apparently dead-ends at zen? That seems bad.

      • bfar
      • 4 years ago

      I see a Zen+ in that chart.

      • ronch
      • 4 years ago

      Why, were you expecting their next total redesign after Zen to be there?

        • VincentHanna
        • 4 years ago

        Aye… it takes 5+ years to redesign a core… The fact that they don’t have any irons in the fire seems not very promising.

    • A_Pickle
    • 4 years ago

    This needs to be AMD’s Conroe moment.

    • Mr Bill
    • 4 years ago

    I remember a decade ago when we were all so excited over Symmetric Multi Processing (SMP), and putting two or even 4 CPU’s on a single motherboard was all the rage. Its funny that Simultaneous Multi Threading (SMT) is now going to extend the paradigm on a thread basis.

      • just brew it!
      • 4 years ago

      Ummm… SMT isn’t new. It is essentially just another name for Hyper-threading, which has been around for more than a decade (2002-era Xeon and Pentium 4).

        • Krogoth
        • 4 years ago

        Actually, Hyper-threading is a marketing name for SMT. It was originally used in the P4 to address issues with its long pipeline. 😉

        SMT is almost as old as microprocessors.

          • just brew it!
          • 4 years ago

          Isn’t that essentially what I just said?

            • Kurotetsu
            • 4 years ago

            You’ll have to forgive Krogoth, in his zeal to remind everyone how old and worldly he is he sometimes forgets to read the posts he’s actually responding to.

        • Mr Bill
        • 4 years ago

        I guess I was just getting an impression of SMT as threads that were not just happening at the same time but were in some sort of (symmetric?) communication with eachother, actively coordinating the progress of the computation. I was envisioning thousands of threads running in a communication based symmetry on something like a graphics core.

          • just brew it!
          • 4 years ago

          Nope, nothing like that. It’s just a clever way of improving efficiency of a CPU core by giving the execution units something else to work on if a thread stalls because it is waiting for DRAM access. It’s basically a lightweight thread context switch that happens at the hardware (instead of OS) level. In order to backfill the execution units with work this way, the CPU needs to have an extra thread (per core) ready for execution. To facilitate this, a SMT (a.k.a. Hyper-threaded) CPU tricks the OS into scheduling additional threads by pretending it has more cores than there really are. That’s why a quad-core CPU with Hyper-threading shows up as an octo-core in Windows Task Manager.

      • A_Pickle
      • 4 years ago

      A decade ago was 2005, and by then we were all pretty sure that multi-core was how things were gonna happen…

      [url<]https://techreport.com/review/8616/amd-athlon-64-x2-3800-processor[/url<] I remember reading this review, and watching Intel get clobbered.

        • Mr Bill
        • 4 years ago

        OK I guess it was 15 years ago in 2001 when AMD came put with the palamino MP line of processors.

      • the
      • 4 years ago

      Are you referring to Quad FX (aka 4×4)?

      [url<]https://techreport.com/review/11353/amd-quad-fx-platform[/url<]

    • CuttinHobo
    • 4 years ago

    This is probably a dumb question, and I imagine the HBM stackup graphic is more for illustrative purposes… but with stacked dies grouped in the same package as a single-level die, how is the heat properly transferred to the heatsink?

    Wouldn’t there be a considerably thicker layer of thermal goop (between the dies and the heat spreader) over the single-level die than the stacked dies? Or is each layer so thin that multiple layers aren’t significantly taller than a single?

    • southrncomfortjm
    • 4 years ago

    Love how that graph shows way more than a 40% gain. Looks more like 400%.

      • chuckula
      • 4 years ago

      OMG! THE GRAPH MUST BE RIGHT! 400% IPC GAINS CONFIRMED!!

      Now I’m starting my stopwatch until this is a frontpage story on Fudzilla & Wccftech.

      • Mr Bill
      • 4 years ago

      You cannot infer that because there are no numbers on the vertical axis. However, all scales being uniform, Excacavator core was only, hmmm, maybe 3-5% gain over Bulldozer.

    • HisDivineOrder
    • 4 years ago

    In regards to the IPC improvements, I still remember the promises they made about Bulldozer’s performance in comparison to the architecture it replaced. I do not tend to trust AMD’s statements after the comments about that (and Phenom II and Phenom).

    They have this tendency to exaggerate and I don’t believe they even think they’re lying when they do it. They believe their own BS and that means they’re convincing to those who either want to be lied to or don’t know any better.

      • Sam125
      • 4 years ago

      Meanwhile, back in 2015. Some news happened that seems pretty positive and at the very least is news that shows AMD is on track to not be such a huge loser anymore. 😉

        • HisDivineOrder
        • 4 years ago

        Meanwhile, back in 2015, you need to consider what happened in the past to help you decide whether or not to trust them when they give you that “news that shows AMD is on track to not be such a huge loser anymore.”

        Because I don’t think I do. I hope they pull it off. I really do. I just don’t think they’re trustworthy at this point and have a LOT to prove to anyone. More even than words alone will do.

          • wimpishsundew
          • 4 years ago

          Kind of hard to say whether you can believe or not since the company is ran by completely different people.

          At least now, AMD has people that seems to know what they’re saying. Last that AMD has competent management, they took Intel by storm.

          But as always, their marketing is abysmal.

            • K-L-Waster
            • 4 years ago

            Fair point, as far as it goes. However, keep in mind that at the same time, Intel had made a massive wrong turn in the form of the Netburst architecture. Even if you don’t like Intel as a company, it’s pretty clear that their architecture isn’t a disaster this time around.

        • ronch
        • 4 years ago

        Designing some of the most sophisticated pieces of technology known to mankind is hardly the work of huge losers. Designing a CPU core is very, very tough.

      • ronch
      • 4 years ago

      I think AMD’s new management is keen on fixing the company’s honesty reputation. They could still try and deploy their own version of RDF but they’d do well to be honest. It’s just the right way to do PR.

      • gruffi
      • 4 years ago

      Every IPC number I’ve seen on an AMD slide was true so far. Being it Shanghai/Deneb, Llano, Kaveri and so on. So, what performance promises were not true? You shouldn’t get confused with other companies. For example, Intel is much better in exaggerating performance improvements. Remember anyone stuff like this?

      [url<]http://pics.computerbase.de/1/7/7/7/4/1-1080.434309994.jpg[/url<] Promised more than 40% improvement in gaming for Penryn compared to Core2. It was only ~5% on average on the client side and even less in gaming. Big LOL. To tell you the truth, AMD is much more honest with their performance projections. Maybe not always precise. But who is? You just should know what are official slides and what are fake slides.

    • LoneWolf15
    • 4 years ago

    I’d love to see it. But marketing departments need to talk big to give investors confidence, and the market as well.

    I’ll be a happy camper if Zen is released and it lives up to what they say. But for me, I’m going to call this a “show me the money” moment.

    • fredsnotdead
    • 4 years ago

    “Just wait ’til next year!” -AMD every year.

      • HisDivineOrder
      • 4 years ago

      Well, to be fair, I think a couple years ago they started saying, “Just wait till 2016!” because of the magic of Jim Kellar and Zen.

      Before that, yes, it was “next year” every year.

    • TheSeekingOne
    • 4 years ago

    Strange…ExtremeTach claims that AMD’s 7th generation APUs will have Zen inside:
    “It’s assumed that Zen will also power the 7th-generation of desktop and mobile APUs that are shown here, though AMD only explicitly stated that Zen was coming to the desktop. ”

    In short, AMD didn’t explicitly state that Only the upcoming FX series chips are gooing to have Zen cores.

      • K-L-Waster
      • 4 years ago

      There’s a world of difference between ExtremeTech assuming something and it being reality. Not to say it is impossible, but really, no news is no news.

      • chuckula
      • 4 years ago

      ExtremeTech obviously misinterpreted the slides from AMD that go out of their way to say that the “7th generation” APUs do NOT include Zen.

      I can’t blame them too much because you could get a college degree in AMD slide interpretation, but here’s how you know it’s true.

      Look at this slide: [url<]https://techreport.com/r.x/amd-fad-2015/desktop-mobile-roadmap.jpg[/url<] (also in the story). Look very carefully at 2016. NOT at the fun-filled top graph that says "Zen", but at the parts below it that just say "7th Gen A-series". Note what's missing from those parts of the graph? The word Zen -- because those APUs don't include Zen cores. Believe me, it's not because AMD is trying to artificially lower expectations.

    • Chrispy_
    • 4 years ago

    Man, I really want AMD to succeed (the industry pretty much [i<]needs[/i<] it at this point) - but 40% IPC performance over Excavator just doesn't cut it. AMD's best IPC at the moment is Steamroller, AKA the A10-7850K, a 3.7GHz part. Intel's 2500K [i<]from 2011[/i<] also runs at 3.7GHz (on boost). The bad news is that the 2500K from 2011 is [i<]at least 40% faster[/i<] than the A10; In [url=http://www.anandtech.com/bench/product/288?vs=1200<]a lot of these tests[/url<] the 2500K is closer to 60% faster than the A10, and it's probably not always even running at 3.7GHz all the time, meaning that it has even more than 60% of an IPC advantage. Basically, Intel's Skylake is expected to be the biggest IPC gain in years and it will be here before this year is out. AMD are promising a chip that will be along another year after Skylake and that isn't quite as good as Sandy Bridge, from 2011, but "at least it's better than those awful <construction equipment name here> products we've crapped out for the last six years". AMD: 40% gain is good, but it's not the 100% you'll probably need by 2016.

      • Sam125
      • 4 years ago

      That’s an interesting comparison tool. Kaveri vs Haswell shows a similar [url=http://www.anandtech.com/bench/product/1200?vs=837<]~60% lead for single threaded benchmarks[/url<] which is also on a finer process node with FinFET already incorporated. Multithreaded however, is where Intel shines which only goes to show how much of a disaster CMT has been for the BD line of cores. AMD has at least been able to keep up with each iteration of IPC improvements. A +40% increase plus a higher rate of improvement rate each iteration is a good thing, not bad as you're trying to make it out to be.

        • gruffi
        • 4 years ago

        That comparison is complete nonsense. You compare a full CPU vs a castrated CPU. You should compare 2M/4T Bulldozer vs 2C/4T Core or 4M/8T Bulldozer vs 4C/8T Core. Everything else just doesn’t make any sense. Cinebench also isn’t the best benchmark to compare AMD and Intel due to its strong Intel bias. There is also something very wrong with these numbers. The Kaveri POV-Ray score can’t be just 17.5% of the i5. Kaveri usually scores better than an i3 in POV-Ray. Which means the i5 should be only ~50% faster.

        At the moment Intel has ~40% more single thread IPC than AMD on average (Haswell/Broadwell vs Steamroller/Excavator). 40% more for Zen would close the gap. The question is, how much IPC improvements will Skylake have.

      • chuckula
      • 4 years ago

      Yeah, 40% IPC ….. plus clockspeeds in at least the high 3GHz range and better in the low 4GHz range …. plus a sufficient number of cores at these high speeds… plus power/thermals that aren’t insane…. keeps AMD in the game.

      It does *not* destroy Intel or make Skylake obsolete. Oh, and Skylake will be about 1 year old by then if Zen launches ontime.

      Zen isn’t the death of Intel, it’s (hopefully) a life preserver for AMD.

        • Ryhadar
        • 4 years ago

        I planned to throw AMD a bone with Bulldozer if they could have matched Intel in games. When that didn’t happen, I figured that as long as they had a decent mATX option I could still justify it (they had good encoding figures anyway).

        Hoping they get it right this time.

          • just brew it!
          • 4 years ago

          Yeah, they don’t quite seem to “get it” on the platform front. For anyone who thought AMD dropped the ball on the 9xx series chipsets, take a look at the selection of mATX AM3+ boards sometime.

          Sure, it is the OEMs that design and build the motherboards; but something is definitely out of whack on the mATX front. I get the feeling that AMD encouraged the motherboard vendors to NOT provide decent mATX options for Socket AM3+, in order to make the APUs look like a more attractive option in comparison. Quite a few people would actually LIKE a beefier CPU in a smaller form factor, and this segment is currently not being served AT ALL by the AMD ecosystem.

          To this day, most of the mATX AM3+ boards on the market are based on a 760G+SB710 configuration. Radeon 3000 DX10 IGP? Check. Buggy, half-decade old southbridge with no 6Gb/sec SATA support? Check. That’s a travesty! I hate to be cliche, but 2009 called, and they want their chipset back!

          The only (tiny) bright spot for these boards is that mobo vendors really do seem to be recycling their old Socket AM2-era designs with just enough modifications to support Socket AM3+. How is this a bright spot, you ask? Well, it means you can get a VIA-based onboard audio codec instead of Realtek.

          WTF.

          At least it looks like they are unifying their desktop sockets with AM4. If they manage to pull Zen (and an accompanying chipset) off, there should be less stupidity on the platform front.

        • BaronMatrix
        • 4 years ago

        The funny thing is some sites use 2600K in their benchmarks and IvyBridge doesn’t always win…

        Workloads can be affected by so many things you can’t say there’s an overall winner… If AMD goes for brute force like the original Opteron\FX how can anyone say they won’t succeed…? Especially with them hiring the Opteron guy back to do it…

        I actually hope it’s ARM and phones all over again… Intel deserves a comeuppance…

      • A_Pickle
      • 4 years ago

      That isn’t true. Intel’s most recent cores haven’t been performance improvements, they’ve been power consumption improvements. Sandy Bridge performs pretty close to Ivy Bridge which performs pretty close to Haswell which performs pretty close to Devil’s Canyon. From Sandy Bridge to Devil’s Canyon there’s probably a nice bump in performance, but… even still, Sandy Bridge is a respectable chip even today. Paired with a solid GPU, it can play the most modern games.

      People bought Phenom II’s despite the Nehalem i7’s being better. They weren’t quite as fast, and they weren’t quite as power efficient… but they were still pretty daggum fast, and ridiculously cheap. If AMD can do that again… they’ll have some winners on their plate.

        • just brew it!
        • 4 years ago

        Bingo. They don’t need to beat Intel in all aspects, they just need to provide reasonably competitive performance at an attractive price point. Hopefully the architectural improvements and the move to a modern fab process (being stuck at 28nm puts them at a serious cost disadvantage) will get them back in the game.

        • Krogoth
        • 4 years ago

        Intel CPU performance ranking (assuming clockspeed is equal) : Sandy Bridge = Ivy Bridge < Haswell = Broadwell < Skylake.

        Power efficiency ranking: Sandy Bridge < Ivy Bridge < Haswell = or < Skylake < Broadwell

        Phenom IIs were actually the fastest x86 chips you could get on the market (They were slightly faster than Yorkfield) for about a period of six months before Bloomfield made their depute.

          • sreams
          • 4 years ago

          Might want to re-post those rankings. They are confusing as written. It looks like you are saying Sandy Bridge is greater than Ivy Bridge, which is greater than Haswell (for your power efficiency rankings). Shouldn’t it read:

          Sandy Bridge < Ivy Bridge < Haswell

          ?

        • w76
        • 4 years ago

        Agreed, I think fanboys have their expectations too high (shocking, I know). AMD can’t hope to compete with Intel for a the higher-end, i7 K-series sort of market. That’s not even the point or probably their honest objective.

        40%, if they hit it, means they can compete in the mid-range* and the low range, whereas right now they compete largely on the basis of pity for most use cases. But that would be okay! Most people probably buy mid-range chips. It’d be up to their incredible (!!!) marketing department and sales team to overcome not even being in the running for the performance crown with halo products. It shouldn’t be too hard; that they don’t segment features as heavily as Intel (like ECC RAM) brings a lot of value to some people, as long as performance doesn’t suck.

        * If Intel allows them to. Intel could destroy AMD tomorrow with a product range performance bump or cost reduction, or more slowly by going full-throttle on their product pipeline, but they’re afraid of monopoly status even if they earn it. So Intel will probably allow AMD to sit at the mid-range table and eat some scraps, at least if they’re smart.

        • Chrispy_
        • 4 years ago

        [quote<]Intel's most recent cores haven't been performance improvements[/quote<] Not sure I can agree with that statement, just because there are tons of benchmarks that prove it wrong. In [i<]some[/i<] scenarios, the performance gains from generation to generation are small - for example, gaming sees benefits of 3-5% at best. But in the vast majority of tests and measurements, Haswell is (on average) about 25% quicker than a similarly-clocked Sandy. In some cases it is almost 100% quicker. Just use the tool I linked earlier to compare a 3.4GHz 2600K to a 3.5GHz 4770K. The median improvement across a battery of 33 tests is something like 25%, even though the clockspeed difference is less than 3%. That's IPC, damnit. Intel have improved IPC by 25% in a tick-tock cycle without people even noticing. Suddently AMD's 40% improvement after 6 years doesn't sound so great :\

      • anotherengineer
      • 4 years ago

      If that 40% IPC is true it should put it slightly above Sandy. And if they sell a true 8 core, for the right price it will be a good seller just like Thuban was. If software continues to improve it’s multi-threading capability, a true 6-core zen might even be on par against a quad i5 skylake in multi-threaded programs.

      Here’s to hoping so at least.

      • anotherengineer
      • 4 years ago

      If I recall, the i5-2500K is a true quad and the A10 is 2 modules or 2cores with AMD hyper-threading so to speak??

        • Chrispy_
        • 4 years ago

        Correct, but that doesn’t affect the results of the single-threaded tests, and AMD spent a lot of effort trying to convince people that a module was closer to two seperate cores than a single core with Hyperthreading.

        You’re right, it’s not apples-to-apples, but it’s the closest we can get for comparison, and looking at single-threaded tests puts an Intel “core” against an AMD “module” anyway, eliminating the discrepancy between the way the cores are counted.

      • ronch
      • 4 years ago

      Using the Bench webpage over at Anandtech, I think AMD needs twice the IPC of Bulldozer to attain parity with Haswell. The 4790K totally zooms past the FX-8350 in both single- and highly-threaded stuff. In particular, the 4790K rapes the FX-8350 in single-threaded performance.

      • TopHatKiller
      • 4 years ago

      OFG’s sake. Performance = ipc x clock x corecount
      x or / uncore scaling-distribution-latency
      & cache hierarchial performance.
      therefore ?

      We have no info about any of the above. Assume nothing dumbasses, for Christ’s sake. Zen could be tonnes faster then Haswell/Skylark etc or tonnes slower, no one knows except AMD.
      And AMD ain’t telling NOBODY till next year sometime.

      …the sun is brighter then my lamp, my lamp is brighter then my penlight, my penlight is brighter then the night…. how bright is the sun? in candela?.

    • Vergil
    • 4 years ago

    [quote<]The downside of the news about a big, Zen-based chip in 2016 is that AMD apparently had to prioritize this one chip over other options. The company's 2016 APU products for desktops and mobile systems will not yet incorporate the Zen core.[/quote<] What about Bristol Ridge, a quad-core Zen APU with GCN 2.0 and full HSA 1.0?! [quote<] Also, the schedule for the K12 core, the ARM-compatible sister to Zen, has been pushed back to 2017. CEO Lisu Su further revealed that Project Skybridge, the effort to make ARM- and x86-compatible CPUs share the same sockets and motherboards, has been nixed, reputedly due to lack of customer demand.[/quote<] What about the 2W low-power dual-core Styx APU with GCN 2.0 and HSA 1.0? lack of demand? Not sure about that, could be the lack of ARM support in Windows 10 on Desktop... As for Socket compatibility, who really cares for that? Every Intel series has its own socket, most i7 CPUs have different socket than i5s and nobody seems to complain about that.

      • Deanjo
      • 4 years ago

      [quote<]As for Socket compatibility, who really cares for that? Every Intel series has its own socket, most i7 CPUs have different socket than i5s and nobody seems to complain about that.[/quote<] A lot of AMD's minuscule cpu sales over the last few years were because a lot of owners of older chips were still able to upgrade because of the socket compatibility. Had many of those users been forced into a new socket, it is likely that they would have completely jumped ship to intels offerings since they could not reuse their existing motherboard.

      • the
      • 4 years ago

      [quote<]As for Socket compatibility, who really cares for that? Every Intel series has its own socket, most i7 CPUs have different socket than i5s and nobody seems to complain about that. [/quote<] I for one complain about that a lot. We've had four different sockets form Intel that support the same basic dual channel DDR3 + integrated graphics schema over the past 6 years. With a little for thought, I'm pretty sure they could have cut at least one of those from being released, if not two. While Intel has been slowly progressing in terms of chipset support, what would the average user truly miss if they were able to plug a socket 1151 Skybridge chip into a Z68 motherboard? Native USB 3.0 is nice but most of those motherboard had it by virtue of a 3rd party controller anyway. NVMe support on Z97 is nice too but that feature will likely come of age with Z170 and SkyLake.

        • flip-mode
        • 4 years ago

        At this point a new socket is an easy concession to make for a competitive chip from AMD. Fine, sure, great, gimme the new socket. That would be the easiest shortcoming to deal with. If AMD needs to put out a new socket EVERY YEAR to keep up with Intel: no complaints from me. Because the alternative is that instead of buying new AMD motherboards every few years, I just have to buy new Intel motherboards instead.

      • chuckula
      • 4 years ago

      [quote<]What about Bristol Ridge, a quad-core Zen APU with GCN 2.0 and full HSA 1.0?![/quote<] I dunno, what about the fact that AMD's slides explicitly say that the APU you just described isn't going to exist until 2017 at a minimum? (Hint: AMD flat-out said that its APUs in 2016 DO NOT INCLUDE ZEN period.) [quote<]What about the 2W low-power dual-core Styx APU with GCN 2.0 and HSA 1.0? [/quote<] Yeah, where was that in the presentation from yesterday? No, I'm not talking about the faked slides, I mean the real presentation that took place on this planet called "earth". AMD made zero mention of this magical Styx part. I can see why you think AMD's products are so amazing. I bet they wish those products were actually on their own roadmaps.

    • Deanjo
    • 4 years ago

    I leave you with this on past AMD cpu performance speculation…..

    [url<]http://www.nordichardware.com/CPU-Chipset/amd-bulldozer-performing-on-par-with-hexa-core-core-i7-processors.html[/url<] [quote<]source says that performance is really close to that of Intel's Gulftown architecture with hexa-core Core i7-980X leading them.[/quote<] Keep in mind that the 980x was already a year old chip at the time of these rumours and even through the refinement of Piledriver the FX8350 was still getting beaten by the then three year old i7-980X.

      • chuckula
      • 4 years ago

      If TR still has a Gulftown rig in the corner somewhere… I’d *really* like to see how it does next year when the time comes. There will definitely be some cases where it’s almost impossible for Zen to lose… like AVX benchmarks or AES-NI workloads… but I’d be *awful* curious to see if an 8 core Zen can finally beat a 6 core 2010-era Intel part in general purpose workloads.

        • auxy
        • 4 years ago

        To be fair, Gulftown parts bottomed out at like $600 and I don’t really think any of these Zen parts will reach that high…

          • Deanjo
          • 4 years ago

          Ya and I remember AMD slapping a thousand dollar price tag on the initial batches of the FX-9590…..

          • gruffi
          • 4 years ago

          And it was a 12-thread processor. Not really fair to compare it against an 8-thread processor. The funny thing is, despite its implementation failures Bulldozer could beat or at least match Gulftown in several multithreaded apps. 4 CMT modules beat or match 6 SMT cores. Not that bad at all.

            • chuckula
            • 4 years ago

            Uh.. haven’t you been drinking your koolaid properly registered this morning AMD fanboy?

            Zen also has SMT and those 8-core Zen parts we’ve all been promised* have 16-threads. So you have the typical conundrum: We call hyperthreading stupid and illegal when Intel does it, but what happens when AMD copies Intel? AMD invented the technology and did it right! That’s what we say!

            * By faked slides that is. AMD has never said that consumer-grade hardware actually gets an 8 core CPU, that’s just optimistic conjecture.

            Oh, and the fact that Bulldozer… which AMD produced with parts having $900 pricetags and is still selling as the top-end part in 2015… manages to sorta kinda win a few benchmarks over vastly more affordable, energy efficient, and more advanced Intel parts from 2010 is not a sign of AMD’s strength, rather it’s a sign of AMD’s incompetence, and no amount of Jim Keller worship will overcome the laws of physics.

      • Vaughn
      • 4 years ago

      And in 2015 gulftown is still faster than current fx chips.

        • gruffi
        • 4 years ago

        Don’t forget that the “current FX” chips are from 2012. Gulftown is not much older. 😉

    • maroon1
    • 4 years ago

    Another AMD slides with big numbers. 40% in AMD slides could mean 10 or 20% in real world.

    Remember how AMD show a slide showed that mantle being 3x times faster than DX11
    [url<]http://cdn2.wccftech.com/wp-content/uploads/2014/01/Mantle-vs-DirectX-635x357.jpg[/url<] AMD often use a cherrypicked benchmark that no one cares about to show how great they are. Better to wait for reviews

      • the
      • 4 years ago

      Well actually that did turn out to be true:

      [url<]http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm/3[/url<] Depending on the card, gains over 3x were observed. The catch is that using DX12 also nets a similar increase in performance over DX11.

        • shank15217
        • 4 years ago

        Thats because DX12 used many of the principles that were part of mantle. I’m sure you’re not so naive thinking that DX12 would have been what it is without mantle are you?

        • maroon1
        • 4 years ago

        I don’t know why you are twisting my words. WHen did I deny that it was true ?

        I said it is cherrypicked benchmark. Please read my comment properly. AMD will only show the best case scenario, and in real world you won’t get similar or even close in most games. There are some cases where mantle will give zero improvement.

        AMD could also use cherrpicked benchmarks to show 40% improvement for zen. The average improvement could be way less than that.

        • VincentHanna
        • 4 years ago

        Cherrypicked benchmark that nobody cares about…
        Thanks for proving his point.

      • NovusBogus
      • 4 years ago

      Fair point, but the fact that they’re claiming meaningful IPC improvements at all instead of just “yo dawg I herd you like cheap APUs…” is still pretty encouraging. If they can close the gap back to the point where $150 gets you a really nice chip, it’ll be good for AMD and good for the industry.

      • CBHvi7t
      • 4 years ago

      [quote<]40% in AMD slides could mean 10 or 20% in real world[/quote<] 10% at best, very much depending on the case. This is assuming an additional thread per core. Those 40% come mainly from two sources: 1. lower clock rate 2. better utilization due to SMT. Per thread performance will go down once you raise utilization.

      • gruffi
      • 4 years ago

      You can see that this is a slide about Star Swarm on Kaveri? IPC is a completely different story.

    • tipoo
    • 4 years ago

    40% IPC over Excavator, so 1.68x Piledriver in total? That’s Haswell territory, if not above it…Hm. Granted Skylake will be out then too, but they’ll be closer than they have been in a very long time.

    • UnfriendlyFire
    • 4 years ago

    “a decent gaming PC in a compact package, ready for gaming laptops”

    Um…

    *Checks the amount of laptops with only Intel CPU*

    Yeah, good luck breaking into that market share. I don’t know any decent Kaveri laptops other than HP’s overpriced Elitebooks, and they started selling Broadwell Elitebooks about a month ago.

      • brucethemoose
      • 4 years ago

      With Llano, AMD burned OEMs so bad that they SUED. Trinity was also late and barely better than llano, Kaveri was super late… So ya, OEMs don’t really trust AMD APUs right now.

        • UnfriendlyFire
        • 4 years ago

        I thought it was the investors that sued? I know that GF delivered the APUs months too late.

          • chuckula
          • 4 years ago

          The OEMS might have done something worse than lawsuits: After they got burned by AMD, they refused to buy the chips once they finally showed up leading AMD to go from a supply shortage (due to GloFo) to a supply glut (due to lack of purchases).

          That massive inventory bubble then led to the investor lawsuit.

          • nanoflower
          • 4 years ago

          Yes it was the investors that sued over the promises AMD made and failed to deliver with Llano.

            • ronch
            • 4 years ago

            It was the investors that sued AMD, but not because of the products themselves, IIRC, but because AMD overstated demand and told them Llano was selling like pancakes when there were crickets all over it.

    • Meadows
    • 4 years ago

    A Powerpoint slide isn’t worth too much credit but if the quoted numbers have any truth behind them, then this is going to be at least significant.

    • UnfriendlyFire
    • 4 years ago

    And then the fab plants report production issues. And then AMD cuts more staff, including engineering, and telling the remaining marketing team to “make something happen”.

    • TheFinalNode
    • 4 years ago

    That presentation was pretty good, and the Q&A in particular gave me more faith in the company’s direction. I presumed the analysts’ questions would be all about margins and whatnot like the bean-counters ask on the earnings conference calls but there were some tough and more-technical-than-expected questions asked, which the team handled fairly well.

    Really excited for their FinFET GPUs! It seems that 2015 will have a new architecture/memory system (HBM) with a focus on performance and then 2016 will transition to a new node with a focus on power efficiency, as well as applying HBM to more of the product stack.

    • Tristan
    • 4 years ago

    Only 40% IPC improvement ? Weak. They need 100% to match Haswell

      • gruffi
      • 4 years ago

      Nonsense.

    • Mikael33
    • 4 years ago

    So that will put them back in line with the Phenom II IPC eh?
    (I kid)

    • loophole
    • 4 years ago

    For anyone interested in seeing the whole roadmap slide deck, it’s here:
    [url<]http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-presentations[/url<] -> Q1 2015 Investor Presentation So I guess all those "leaked slides" from last week were fakes. Interesting! They looked pretty darn good, and managed to cause a lot of buzz ahead of the actual analyst day. Now that I've connected the dots, it's time to get my tinfoil hat out 😛

    • shank15217
    • 4 years ago

    AMD is definitely going back to roots. A good initial architecture means improvements will also bring the pain.

      • chuckula
      • 4 years ago

      [quote<]AMD is definitely going back to roots.[/quote<] If you mean making clones of Intel chips, then I can buy that since some of those slides looked like they were dusted off from the Sandy Bridge launch in 2011. [quote<]A good initial architecture means improvements will also bring the pain.[/quote<] AMD definitely has a high pain tolerance!

        • nanoflower
        • 4 years ago

        I’m thinking the people in pain are AMD investors. Maybe they will do well in 2016 if Zen actually gets released then but that’s a long ways away.

    • Sam125
    • 4 years ago

    Thanks for the analysis and summary. It sounds like everything AMD has endorsed heavily for the past four years is going to materialize with the Zen line of APU/CPUs: HSA, HBM, compelling APUs and possibly a competitive server part for the first time in a long time.

    Combine that with an advanced 14nm FinFET process and it becomes pretty clear that as long as AMD can execute then things are looking on the up-and-up.

      • chuckula
      • 4 years ago

      Aside from a few vague references to the obvious fact that they intend to make Zen APUs in the future, there is basically zero crossover between Zen and anything else in AMD’s roadmap for 2016.

      Those “7th generation” APUs are *not* using Zen cores and there was zero mention of any “Zen” product that uses HSA. In fact, while HSA & heterogenous computing in general got some attention, I think that AMD noticeably dialed back their emphasis in that area, with HSA not being center stage in particular.

        • Sam125
        • 4 years ago

        The Zen based APU is slated for 2017. That was plainly stated in the article.

        [quote<]Those "7th generation" APUs are *not* using Zen cores and there was zero mention of any "Zen" product that uses HSA.[/quote<] Starting from Kaveri on, all of AMD's APU are HSA 1.0 compatible. Meaning when the first software packages supporting HSA are available, everything from Trinity and on will be supported. Although Trinity and Richland will likely need some bridging software to bypass features that weren't available for those two APUs. Look at the bigger picture here. A [b<]lot[/b<] of things that AMD bet on when they weren't doing well are poised to bear fruit within the next few years. AMD's qualitative graph also shows that not only does Zen offer +40% IPC improvements but the following iterations will also increase IPC at a higher rate than the BD based cores. I would say it's a perfect storm 'a brewing, but that's just too cliche. 😉

        • ronch
        • 4 years ago

        HSA will go the way of 3DNow!, FMA4, and XOP. Without support, and especially with the renewed efforts to advance per-core performance, it’s done for except for the few devs here and there who are adventurous enough to play with it.

      • Deanjo
      • 4 years ago

      [quote<]Combine that with an advanced 14nm FinFET process and it becomes pretty clear that as long as AMD can execute then things are looking on the up-and-up.[/quote<] Only if you believe that intel has been sitting around doing nothing for the last few years and have no plans to further develop their own product. If AMD gets lucky they may be able to match what intel has been offering for a few years now.

        • Sam125
        • 4 years ago

        Yeah, I guess we’ll have to see how anti-competitive Intel becomes if AMD can start to compete again as that’s a reliable metric for how well AMD is doing. 😉

          • Deanjo
          • 4 years ago

          Already pulling out the “but intel cheats” card?

            • Sam125
            • 4 years ago

            Hahah nah, I was kidding. There’s already legal precedent that makes it easier for AMD to go after Intel if they resort to monopolistic anti-competitive practices again.

            Realistically, AMD is playing catchup not only in terms of CPU architecture but also the process node on which their chips are made. How well that’ll realistically translate to real world performance, I wouldn’t know. Maybe Dean Kanter can answer that question this Thursday. 😉

            • Deanjo
            • 4 years ago

            [quote<]There's already legal precedent that makes it easier for AMD to go after Intel if they resort to monopolistic anti-competitive practices again.[/quote<] I'm not sure that a prior settlement can be used as evidence or precedent for an anti-competitive practice lawsuit.

            • Sam125
            • 4 years ago

            Due to the terms of the settlement, I believe you’re correct. What I meant is that AMD might be more willing to litigate if they believe the competition is being extremely unfair since they have kind of legally won against Intel in the past.

            However, the real story, IMO, is whether AMD’s products can at least be competitive with Intel again from a technical and performance standpoint. That’s much more interesting.

            • faramir
            • 4 years ago

            David Kanter.

            • ronch
            • 4 years ago

            Dean Kanter? Who he?

    • tipoo
    • 4 years ago

    Is the pipeline shorter than Bulldozer and family? Would that 40% IPC increase then be offset by its viable factory clocks being lower?

    It would still be a gain in efficiency, but I wonder how that will play out.

      • the
      • 4 years ago

      It may not matter: Bulldozer never really hit its targeted clock speeds which were supposed to be north of 5 Ghz. Well they sort of got there if you don’t mind a 220W CPU.

        • tipoo
        • 4 years ago

        That wouldn’t change a reduction in clocks relative to shipping Bulldozer architectures.

          • the
          • 4 years ago

          Indeed. Short pipeline and the same clocks that Bulldozer shipped with wouldn’t be a bad thing.

        • Mat3
        • 4 years ago

        We heard a lot about how BD didn’t reach expected clock speeds, but I don’t think AMD was aiming for 5Ghz+ on non-overclocked CPUs, even high end ones. It would have been pure folly for them to think that, and we don’t need hindsight to say that.

        • Anonymous Coward
        • 4 years ago

        One can imagine that a shorter pipeline will reduce the clock speeds, regardless of how it went with Bulldozer. AMD has no time or money to fine tune the design, except in the years following launch.

        • loophole
        • 4 years ago

        I remember hearing AMD’s target frequency for 1st generation Bulldozer was 30% higher than Thuban. Although a little vague, if we take it at face value, they were hoping for a top-end SKU of Zambezi to have a base clock speed of 4.3GHz with a turbo speed of 4.8GHz.

        It would be interesting to know what the Vishera’s projected clock speeds were expected to be if things had gone as AMD had hoped with the design. I wonder how far it is from the 4GHz base / 4.3GHz turbo that the 125W TDP SKUs ended at.

        This whole train of thought assumes that the 30% number we got told came from data from AMD’s engineers and was at least once within the realm of possibility, rather than being a number that the marketing folks came up with without input.

        • gruffi
        • 4 years ago

        Bulldozer hit its targeted clock. Clocks were supposed to increase by 20-25%. Phenom II started at 3 GHz base clock, Zambezi started at 3.6 GHz base clock. That’s exactly a 20% increase. There were never any plans to start at 5 GHz or even higher. That’s complete nonsense.

        The problem with Bulldozer was its IPC. Targeted IPC was about the same as “Stars” with some percent loss due to architectural changes (0-5%). That was a failure in two ways. First, the execution. They didn’t lose just 5% maximum. It was more like 15-20%. Something went wrong during the design or implementation stage. Second, the concept. Targeted IPC should have been at least 10-15% above “Stars” to match at least Nehalem.

          • robliz2Q
          • 4 years ago

          The subsequent BD based cores had some cache improvements, where BD had severre bottle neck due to the cache hierarchy. Actually Kanter talks about the problem around 20min in on the ‘cast.

      • TopHatKiller
      • 4 years ago

      Look, AMD made a huge boobie in the original Zambezie/Bulldozer cpu. They ain’t anytime soon admitting what the multiple errors were;
      [ for a start there was a massive ‘bloat’ problem with the chip itself that resulted in a punishing low transistor density – no admission or explanation from them]
      [ they missed their power target and clock as well – no admission again]
      [ they expected software to be recompiled for their feature set – nope, intel prevented that]

      the original ideas/floated were a high ipc, but low transistor-count & power design, then it was a mediocre ipc and high clock, but still low-transistor design…and of course they were to going to continually iterate and improve the design with new chips every year… but soooooooo long ago AMD decided to say ‘fcuk it’ quit, not tell nobody, but move to another design. We can go on about what went right/wrong about 15h, but we’ll never know until the engineers tell us – and that can’t happen till about 2017, when I’d guess 15h products are all history. [If then.]

      There isn’t anything wrong – in theory – with 15h’s cluster-mt design, or their odd cache arrangements, but in truth all that got from their work was a design that could just about compete with sandybridge and be okay… but left them exposed so terribly in the server market which is after all either amd or intel care about [growing market $50bn-way plus]

      Zen is totally different and will perform as it does when it come out. IPC is only one element to a chip, and without more reliable info we can assume nothing and nothing at all. Anyone and Everyone who assumes are just hopeful/bigoted/ignorant/delusional frigging idiots.
      Wake me up when someone coughs up real datapoints, that I might be able to get my teeth into. Uhm, or socket…. Teeth in socket, tried that.. hurt & resulted in embarrassment.

    • K-L-Waster
    • 4 years ago

    As much as there are some good technology goodies in here, to me the most encouraging thing about it is the fact they are dropping Skybridge. Really, this never made much sense to me. Making x86 motherboards makes sense, and I can see the business logic behind making ARM motherboards, but I cannot for the life of me see why you would want to try to design one motherboard that is pin compatible with both architectures and whose peripherals work with either chip family interchangeably. Sounds like a lot of development expense to deliver a product that, as far as I can tell, no one was asking for.

    It always seemed like a product line that only makes sense to the Marketing department. The fact it has been dropped is a good sign that there are responsible adults running the show for a change.

      • dodozoid
      • 4 years ago

      *there…

        • K-L-Waster
        • 4 years ago

        Fixed. (Mee rites gud mee dus….)

    • DPete27
    • 4 years ago

    Meh, I remember all sorts of marketing hype from AMD about how awesome Bulldozer was going to be also. I’ll believe these performance claims when someone actually runs the chip through some REAL benchmarks.

    • anotherengineer
    • 4 years ago

    I really hope Samsung’s FinFET is on par with Intel’s and that GF can produce at volume with it, because it would be nice to see some real competition again!!

    The HBM is kinda funny that they pawn it off at up to 50% more energy efficient than GDDR5, however it will probably only be on big GPU cards, where the GPU power absolutely dwarfs the vram power, well unless there is 16GB of it maybe 😉

    • Platedslicer
    • 4 years ago

    [quote<]The downside of the news about a big, Zen-based chip in 2016 is that AMD apparently had to prioritize this one chip over other options. The company's 2016 APU products for desktops and mobile systems will not yet incorporate the Zen core.[/quote<] What? CPUs > APUs? What ever happened to "the future is fusion"? Besides being run over by the tanker truck of reality, I mean.

      • nanoflower
      • 4 years ago

      The future is fusion.. The future is not now..

        • willmore
        • 4 years ago

        “The future is already here — it’s just not very evenly distributed.” — William Gibson

        • chuckula
        • 4 years ago

        The future was yesterday!
        The present is fusion.
        The future is Zen.

      • Meadows
      • 4 years ago

      They’ve dropped that slogan some time ago.

    • achaycock
    • 4 years ago

    Sadly, even if this proves to be an amazing CPU, it will be too late for me. After 15 years of using AMD in my main rig, I’ll be using Skylake when it comes out.

    I do hope AMD manage to do well with this though, we need the competition.

      • just brew it!
      • 4 years ago

      I should be OK with my FX-8350 until then. If Zen looks halfway decent I won’t need to jump ship for Intel.

        • ronch
        • 4 years ago

        Same here. We’ve reached the point when a Core i5 Ivy or Haswell or.any 8-core FX should suffice until it breaks.

      • anubis44
      • 4 years ago

      Don’t do it. Skylake leaked benchmarks show no appreciable improvement over the previous 3 generations of Intel processor. You’ll regret it, and feel dirty betraying AMD on the very eve of their victory to boot. 🙂

    • Growler
    • 4 years ago

    [quote<]Zen looks to be more of a "brainiac" architecture like K8 and Broadwell than a "speed demon" like the Pentium 4 and members of the Bulldozer lineage.[/quote<] Translation: Prepare for disappointing clock speeds, and more "Megahertz Myth"-type justifications! I do hope this turns out to be a good product, though. Strong competition is welcome.

    • brucethemoose
    • 4 years ago

    [quote<] The roadmaps for Opteron and FX processors describe a chip based on the [b<]Xeon [/b<]core [/quote<] AMD's new marketing plan: Utterly confuse potential customers. It's brilliant! EDIT: Ninja'd

    • cygnus1
    • 4 years ago

    Damage,

    [quote<] The roadmaps for Opteron and FX processors describe a chip based on the Xeon core with a "high core count with multi-threading," ample memory bandwidth, and lots of built-in I/O. [/quote<] Xeon cores on the Opteron and FX roadmaps?

      • chuckula
      • 4 years ago

      Oops! We thought we had changed all the reference from those totally-not-copied Sandy Bridge design docs there.

      Zendy Bridge will be amazing!

      — AMD

    • cygnus1
    • 4 years ago

    I’d love to see where K10 would fit on that graph where they show the 40% improvement in IPC between the dozer and zen cores…

      • Waco
      • 4 years ago

      Ditto. If I recall correctly the K10 (in the Phenom II X6 version) is faster per clock than even the newest Steamroller cores by a measurable amount in many things.

        • Geonerd
        • 4 years ago

        Absolutely.
        Per clock, my Flintstone-era Thuban is tens of percent faster than ‘Roller.

          • brucethemoose
          • 4 years ago

          Energy usage is pretty comparable too.

        • ronch
        • 4 years ago

        AFAICT Piledriver delivers about 92% the IPC of K10, depending on the workload and instruction mix, so Steamroller should at least equal K10’s IPC. That’s not saying much, tho.

      • Meadows
      • 4 years ago

      Probably the lower end. A Steamroller or Excavator core should be roughly equivalent to one old-style K10 core, give or take.

      I recall when Bulldozer originally launched it had something like a 10-20% IPC disadvantage compared to the previous tech (depending on workload), but later improvements over the years have made up for that, more or less. So IPC is back where they set out from six years ago, but clocks are significantly higher and there are more cores for a given price, too.

      The only problem is that the competition has advanced twice as much.

        • Chrispy_
        • 4 years ago

        You appear to think clockspeed and IPC are related; They are not.

        I would agree that Steamroller/Excavator probably makes up the 10-20% IPC disadvantage over K10 that Bulldozer had. Likewise, Intel are expected to move the IPC goalpost forwards about 15% with Skylake, which will likely have launched, sold-out and been superseded before Zen is available for real. Skylake is damn near confirmed for 2015, compare to AMD who cite “2016” which usually means “a paper launch in Q4 with supply shortages running on for months”

          • Meadows
          • 4 years ago

          No I don’t think they’re related.

          This is just R&D differences in action. AMD wanted to make use of as much of the sunk cost as possible over the years, while intel could iterate on their existing designs and start work on the next big thing simultaneously without any of the managers going into a shock.

          • just brew it!
          • 4 years ago

          I wonder if Intel might drag their feet on the Skylake launch given AMD’s acknowledgement that their next architecture isn’t coming until 2016.

            • Krogoth
            • 4 years ago

            Nah, Skylake is certainly on course for a late 2015 into early 2016 launch.

            • just brew it!
            • 4 years ago

            I didn’t say it wasn’t. I was merely speculating that Intel might delay it on purpose if there isn’t going to be any credible competition from AMD until later in 2016.

            • _ppi
            • 4 years ago

            AMD claims Zen is coming in 2016 for quite some. So no surprises for Intel here.

            Intel needs Skylake rather to get people update from their own old CPUs, and boost sales, as those on Sandy Bridge+ are still happy campers.

            • Chrispy_
            • 4 years ago

            Yeah, anyone with Sandy on Desktop is quite content to ignore everything since.

            I’m just playing with Core-M at the moment and it proves that Intel are [i<]miles[/i<] ahead in the mobile space, but they don't really offer anything to tempt people running old S1155 setups at all.

      • eofpi
      • 4 years ago

      I’d like to see whether that graph is for integer, FP, or vector computation. And I’d really like to see if Zen does away with the silly shared FPU architecture.

      • gruffi
      • 4 years ago

      K10 has similar IPC as Steamroller. So, it would be very close to Excavator.

    • NeelyCam
    • 4 years ago

    Jim Keller is bringing the pain

      • chuckula
      • 4 years ago

      Ask not for whom Jim Keller brings the pain.
      For he brings the pain for thee.

      • Geonerd
      • 4 years ago

      ‘Tis a shame they didn’t install him and his team two years earlier… :/

    • chuckula
    • 4 years ago

    Some of you have already caught it but there’s a forum thread here: [url<]https://techreport.com/forums/viewtopic.php?f=29&t=115444[/url<]

    • divide_by_zero
    • 4 years ago

    Wow, really surprised that they’re openly claiming a 40% boost in IPC. Very exciting if they’re able to pull it off, especially if they can do it on the schedule mentioned in the article.

    And for those making the “Will ship Dec 2016 in limited quantities” jokes, honestly I’d be fine with that. Considering the scope of this undertaking – new architecture, process shrink, FinFET, SMT – I’ll be ecstatic if they’re able to actually be shipping final units in that time-frame as opposed to their customary long delays on new designs.

    • flip-mode
    • 4 years ago

    Availability in 2016 = shipping to OEMs in late December 2016 in limited quantities, limited SKUs.

    • Kretschmer
    • 4 years ago

    From user Phaleron’s live writeup:

    [quote<]2015 Focus: Stabilize the PC Business, invest in enterprise, embedded, and semi-custom, [b<]continue to reduce Operating expenses[/b<], manage cash and liquidity, lay foundation for growth and profitability FUTURE OPERATING MODEL Revenue: [b<]PC: Flat to down GPU: single digit % growth[/b<] FirePro, server, semi-custom, embeded: double digit % growth in each segment Operating Margins: [b<]PC & GPU: Mid-single digit % growth[/b<] FirePro, server, semi-custom, embeded: double digit % growth[/quote<] This suggests more staff cuts and weak competitive positioning of CPUs, APUs, and Gaming GPUs. Not good news for an enthusiast! Edit: Delineated between 2015 and 2016+.

      • chuckula
      • 4 years ago

      AMD tries to put a brave face on it, but every interesting product launch* on AMD’s calendar is in 2016 and they are really treading water for at least the next year (hopefully everything is ready to go by mid-2016 for AMD’s sake).

      * Exception made for R9-390X, but it won’t generate ginormous profits even if successful simply because it’s an uber high-end GPU that doesn’t sell in massive quantities.

        • the
        • 4 years ago

        High end GPUs can occasionally sell in bulk. AMD was caught flat footed with the popularity of their GPUs for crypto-coin mining. For awhile there, I could have sold my Radeon 7970 for more than I paid for shortly after it launched.

        The crypto-coin mess is also one of the reasons why we haven’t seen a Radeon 285X as AMD ordered more chips to meet demand but by the time they had hit retail, the demand had died down with an excess of inventory to sell through.

        I also think AMD is seeing VR as the next crypto-coin GPU rush and preparing for it.

        • nanoflower
        • 4 years ago

        Yes, AMD is really going to have problems since they don’t have anything to sell today that will interest customers in any market. If they are cutting operational budgets that means they can’t hire new engineers and so they are dependent on the current staff and designs to win back market share. Should anything go wrong (problems with production, the design not working as well as they hoped, Intel coming out with something better than expected) that’s going to lead to even more budget cuts leading to a steady slow decline as they just won’t have the budget to compete.

        They really need Zen to put them back on the map in the price-performance arena as well as keeping the power usage down so that the product can compete head to head with Intel. Without that it’s hard to see how they convince people to buy AMD over the next year. Now maybe the HBM designs can start to win them more contracts but will an AMD APU with HBM be price and performance competitive?

      • Vergil
      • 4 years ago

      Who is this Phaleron, and what does his unrelated “totally scientific write-up” have to do with this article and AMD’s 2016 roadmap?
      Haven’t seen you here before, not trying to be accusative, but you should know that stealth marketing and bashing on specific companies is against the rules and regulations.

      V the Vigil

        • nanoflower
        • 4 years ago

        Except that his writeup came from the AMD information today. He was doing a play by play whereas Scott did a writeup after the end of the conference call. No stealth marketing involved especially since much of the news from AMD is bad for this year.

    • Kretschmer
    • 4 years ago

    Interesting concepts; we will have to see how the execution is handled.

      • Mr Bill
      • 4 years ago

      Errr…. 40% faster!

    • jokinin
    • 4 years ago

    So… maybe in the end, AMD will have time to catch up with intel.
    That would be very interesting indeed!

      • just brew it!
      • 4 years ago

      Intel will not be standing still in the meantime.

        • shank15217
        • 4 years ago

        They aren’t exactly moving like a speed demon these days either. Intel’s strongest suite is their manufacturing, I wonder how well AMD’s products would do if they were made in Intel’s fabs. Skylake is supposed to bring around 10-15% performance gains per comparable SKUs.

          • flip-mode
          • 4 years ago

          First of all and as doesn’t really need to be said, I’ll believe AMD’s 40% when I see it. And that 40% might be in very limited scenarios, very specific workloads or instructions.

          Secondly, 40% just kinda puts AMD up in Intel’s ballpark for IPC, it doesn’t give them any kind of lead.

          Thirdly, clock speeds matter. If AMD could have made Bulldozer and sons hit 5-6 GHz and at low power envelopes, there wouldn’t be a problem with Bulldozer and sons. We’ll have to see what kind of clock speeds AMD can reach with the new architecture.

          Fourthly, Intel has been exceedingly conservative with the clock speeds. Intel could raise the clocks on its whole line of chips by 40% right now today if it wanted to.

          Other than the Prescott blunder, Intel has not had any trouble staying out in from of AMD. AMD isn’t going to change that with 40%. AMD will need to do several things to bring a chip that legitimately challenges Haswell. AMD will need to execute. AMD will need to deliver on time and with real production volume. AMD will need to realize 40% additional IPC on average rather than best case. AMD will need to maintain clockspeeds in the 4 GHz range. AMD will need to bring power consumption way down. Meanwhile, Intel will need to pretty much put its whole staff on vacation for 12 months.

          I’m not at all optimistic. When it comes to AMD, it is alway, always, always believe it when you see it. And then it is always, always, always something less than what was promised, later to market than promised, hotter than promised, slower, less available, and so on.

            • _ppi
            • 4 years ago

            “Other than the Prescott blunder, Intel has not had any trouble staying out in from of AMD.”

            And other those couple years between K7 and Prescott were what? I would not exactly call that period time when Intel had lead, at all.

            AMD’s biggest issue has always been manufacturing (at least compared to Intel, since that is what is the real Intel’s advantage), and I am afraid is what can damage Zen’s launch.

            • flip-mode
            • 4 years ago

            I hesitate to say that there is any one thing that is AMD’s biggest issue. There are so many issues: revenue/income (translation: simply a lack of financial resources to do the R&D necessary to compete with Intel’s R&D efforts), manufacturing, market penetration (fighting Intel’s entrenchment), chipset/motherboard inadequacies, ineffectual leadership, and since Bulldozer the processor architecture itself has been an actual issue. I don’t think any one of those trumps the rest, but the come together to form a monumental headwind to success.

            Bulldozer is/was so damn bad it is hard to overestimate. It is amazing AMD has survived. It’s amazing AMD managed to polish Bulldozer enough to make it somewhat marketable at some points along the way.

            • jihadjoe
            • 4 years ago

            FX series isn’t exactly starved for clocks though. I imagine it being manufactured at Intel’s would have helped a lot with thermals and power consumption, but at the end of the day the IPC just isn’t up to snuff.

            Also, the P6 core was rightfully awesome. Prior to K8 and the Athlon getting DEC’s EV6 bus and DDR RAM support the P6-IIIs were faster than K7 clock-for-clock, and still managed to hang on to K8 afterward.

            In fact, Intel’s Conroe and Kentsfield recovered from Netburst and went to creaming AMD by doing exactly what they should have done to begin with: A P6-core with a faster FSB and support for DDR RAM.

            • Flatland_Spider
            • 4 years ago

            [quote<] Thirdly, clock speeds matter. If AMD could have made Bulldozer and sons hit 5-6 GHz and at low power envelopes, there wouldn't be a problem with Bulldozer and sons. We'll have to see what kind of clock speeds AMD can reach with the new architecture. [/quote<] That was pretty much the situation. The 4GHz procs were starting to come on plane, but the fab process was running out of steam at that point. I'm guessing 2GHz - 3GHz will be the normal range with a few in the 3GHz - 4GHz range. As long as AMD is in the ballpark performance and heat wise, has a competitive chipset feature set, and is cheaper I'm okay with that.

        • chuckula
        • 4 years ago

        Good news for AMD: In honor of the 10th anniversary of Prescott going off the deep end, Intel has decided to reduce the IPC of Skylake by 80% and give you two cores on two separate hunks of silicon!

        We love us some aughties retro!

        • jokinin
        • 4 years ago

        Yes, or course, that’s why I said “maybe”.
        I really hope AMD can make some competitive and high performing CPUs again.

        • Mr Bill
        • 4 years ago

        Actually, this will be good news for Intel fans because
        (1) Intel will roll out some faster tech that they would otherwise have sat on indefinitely
        (2) and it will go for nearly the same price to steal AMD’s thunder
        (3) and that price will be far below what it would have been if not for AMD being in the market.

      • Liron
      • 4 years ago

      I want to belieeeeve!

        • ronch
        • 4 years ago

        So lame….

Pin It on Pinterest

Share This