The internet is ablaze with rumors of a leaked lineup of AMD's hotly-anticipated Ryzen CPUs. The leak in question first originated on Chinese forum Coolaler. The post has since then been removed, but Google has a cached version here. The list shows a total of 17 Ryzen models, ranging from a so-called R3 1100 with four cores, all the way up to the R7 Pro 1800X with the eight-core, 16-thread design that AMD has touted since the first Zen announcements. Perhaps the most interesting nugget in the leak is the presence of four six-core, 12-thread models. Drop the wimpy little salt shaker and grab the 40-pound bag of ice-melting magnesium chloride from the garage as we dig into this.
Source: Coolaler
The full Ryzen CPU list includes four R3 models with four-core, four-thread designs, four R5 chips with four cores and SMT, four R5 models with six-core, 12-thread setups, and five R7 models packing eight cores and 16 threads. The rumors suggest a rather narrow range of clockspeeds among all Ryzen chips, since the eight-core versions have a range of 3.0 to 3.6 GHz. All lesser models have an even smaller range between the lowest and highest clock speeds. It looks like we're looking at base clocks, and there's no word on how high turbo clocks might be. The meaning of the "X" and "Pro" suffixes are similarly shrouded in mystery.
On the surface, the leak suggests that AMD will adopt a similar naming scheme to the one the it used in the era of Radeon 200- and 300-series graphics cards, then seemingly dropped after releasing the Polaris-based Radeon RX cards last year. Thinking along the same lines, the application of the "R7" terminology to the range-topping processors seems strange. Another possible scenario is that the processors are named in order to ease comparison to Intel's Core models, which have used i3, i5, and i7 prefixes to stand in for good, better, and best performance, for seven generations over the course of most of a decade.
Given that Zen is made of CPU complexes (CCXes) with four cores each, the existence of six-core Ryzen CPUs has been the subject of hardware forum debate for some time. A six-core model could possibly be manufactured by mixing and matching fully- and partially-enabled CCXes. WCCFTech speculates that six-core products may be manufactured by pairing two CCXes with one disabled core each. Alternatively, one could theoretically obtain six cores by joining a fully-enabled CCX with a half-disabled unit.
The rumor of models with partially-disabled four-core CCXes likewise raises the possibility that we'll see Zen-based replacements for current Kabini APUs. Sub-$100 processors practically require integrated graphics, but once AMD completes work on the rumored Raven Ridge Zen-based APUs, bargain-basement models with less than four cores may be a reality.
The rumor goes on to say that AMD will begin selling eight-core Ryzen models on March 3, and that quad- and six-core models will soon follow. Pricing remains an open question, though no matter what happens, some gerbils will not be impressed.
Intel is already planning to release the 7740K and 7640k, the first i5 with HT enabled. All those 7700K & 7600K buyers last month will surely appreciate it.
If Intel wanted to compete all they would have to do was make the 10-core parts affordable and they could simply win by default with the extra cores. But Intel loves its margins too much to do that.
Dammit you kids settle down or you’re both grounded!
[quote=”jts888″<]Where did you hear that cache is where they made up the density?[/quote<] L3 is 3.1 mm^2 smaller for the same capacity. L2 is 0.6 mm^2 larger per core for double the capacity. [url<]https://techreport.com/news/31402/amd-touts-zen-die-size-advantage-at-isscc[/url<] Given a that the table is comparing 4 core / 8 thread processors, the Zen processor L2/L3 caches look be 0.7mm^2 smaller. That still leaves 4.3mm^2 or 86% if the space savings to account for. Given that the reported feature sizes at ISSCC are all larger than Intel's, the logical explanation is missing features and/or more space efficient layout (at what cost?). [quote="jts888"<]That said, beware overly dense cache structures, since additional read/write porting has direct density costs. Intel has had a 64B L2<->L1 path and a 2*32B one from L1 (and 1*32B to) since Haswell, where Zen's internal per-clock bandwidth is half that, the same as Sandy/Ivy Bridge.[/quote<] Excellent considerations given that Zen's 6T SRAM structure is 37% larger than (?)Skylake's. I'll be very interested to see how the bandwidth and latency of these caches stack up.
Jim Keller, our geeky lord and savior, Chris Lattner, if not the creator of Swift then at least it wouldn’t exist without him, Matt Casebolt, designer of the Macbook Air…Some really really top talent guys leaving Apple for Tesla right now. And money clearly isn’t the problem, I think all these guys miss that feeling of being on the verge of something big, especially after Project Titan fizzled out.
I say you guys settle this like real men and have a slap fight.
Clearly this can’t be true, a 600Mhz difference between the slowest 8C/16T Zen and the fastest and this results in Zens changing from being competitive with the i7 6900K to being competitive with the i7 7700 non-K? Really?!?
And a similar situation exists with the 6C/12T parts?!?
Is there anyone here that believes this?
This is why I get so angry every time there’s a discussion on architects like Keller and someone has to butt in with “nah this is down to the hundreds of other guys who worked on it”.
Guess what? Those hundreds of other guys did jack shit without Keller and are utterly replaceable.
Also: Why not just name it based on the number of cores? R4 = 4 core, R6 = 6 core, R8 = 8 core
Psst, Voldenuit and Duct Tape Dude
The catch there is that if Zen’s SMT is more effective than we’re used to (which looks likely), that leaves single-thread performance correspondingly below expectations.
Muhahahaha.
I don’t care if RyZen puts Intel out of business or fails to boot properly, I’ve already won!
Forever emblazoned.
[url<]http://i.imgur.com/AeBc6J0.png[/url<]
Deflection at its finest with details you never stated in your comments. How about you man up to your previous little tirade. Show everyone how you can actually act like a grown up. Or get ready to eat some hat on Ryzen day.
I’m sorry that my comments in which I stated that AMD’s word of “well before the end of Q1” is untrustworthy and there was no chance of them releasing 4C8T CPUs (the latter of which is refuted by the contents of the news post) were so rage-inducing for you.
PROJECT BUFFALO IS A SUCCESS!
How am I being aggressive when you said you would do exactly what I’m holding you to? And you have the gall to state how MY behavior in that other thread was allowed. Priceless LOL! It’s all him, it’s all him, it’s all him, I’m innocent. Go ahead and get me ban’d, that seems to fit a passive aggressive style. Like the payoff of the bet would ever happen anyway…
In fairness to Intel, none of the systems thus far with SGX support NVDIMMs. The convergence was scheduled for Sky Lake-EP but the first wave of Sky Lake-EP won’t support Xpoint/Optane. Other forms of NVDIMM support on Sky Lake-EP is an open question at this juncture. I would expect some additional forms of memory encryption but right now it is wait and see.
Not to impugn Mr. Kanter’s statements, obviously, but it is clear there are workloads in which, without a clockspeed or thread advantage over a competing Intel produce, AMD is able to offer competitive performance, and it doesn’t appear that AMD had to modify the benchmark in any untoward way in order to achieve those results. The fact that you have some vendetta against AMD doesn’t change that.
I want the 1800 Pro, because it sounds professional. They should price it at $350.
[quote<]AMD has already shown a performance demo[/quote<] Great. Where are the real reviews by people who aren't AMD? Incidentally, how does Zen get Broadwell level IPC while having half the vector execution resources? [url<]http://www.linleygroup.com/mpr/article.php?id=11666[/url<]
AMD has already shown a performance demo (and provided the specific demo that was run to the public) that demonstrates Ryzen’s single-threaded performance is roughly comparable to that of a Broadwell-E processor, so I think all of the trepidation about AMD’s performance there is probably not needed. [b<]IF[/b<] this leak is real, it could be that AMD is targeting those Intel products by price and not necessarily performance. There's really not enough information to tell either way.
Calling out a “jerk” is okay. But you kinda shoot yourself into the foot and destroy your own argument if you are being a jerk yourself in the process.
You want to appear superior? Either ignore jerks or argue with style, tranquility, cold hearted facts and intelligence.
Insulting people in order to “get a point across” just makes you look like an immature childish person and close to never gets the actual point to the recipient.
I wasn’t being passive-aggressive. I just never believed AMD would release a half-disabled CPU. Turns out (apparently) I’m wrong, and now bitterman is…doing stuff in this comment section. I might be biased (since I’m the target), but seriously, between his behavior in that comment thread and his behavior in this one, how is this allowed?
I don’t normally and this truly is a one-off but this durf hit a nerve…not going to lie. Look away if you must cuz it ain’t going to be pretty.
Get ready for a feast you will never forget.
Excellent.
I don’t know what happens next. Perhaps the power of the Wasson will trigger a TR exclusive Ryzen review sample. Or perhaps chuckula just gets another “achievement unlocked” popup.
Jeff’s tweeted about the Buffalo. It has even more coverage than before, so uh. Yeah!
correct sir..
“對應Intel型號” literally mean “corresponding Intel model”
and as a average user, I don’t care about single-thread performance, as long as it’s much cheaper that Intel counterpart
I don’t read Chinese but the Intel column is interesting. I’m guessing it lists Intel processors that are considered the equivalent of Ryzen.
So, to compete with a i5-7600 requires a 6C12T Ryzen “R5”?
To compete with a i7-7700 requires an 8C16T Ryzen “R7”?
This could portend very bad news for single-threaded performance if it takes 1.5-2 times as many Ryzen cores to equal an Intel part. At least, I’m expecting in multi-threaded workloads performance will be comparable unlike the utter fiasco of Bulldozer.
Hahah! Just read those. It looks like Bittermann has a reason to be bitter. 😛 But still, as ronch said, we should at least keep it ‘civilized’.
But it does look like derFunkenstein is gonna eat his hat.
Where did you hear that cache is where they made up the density? I would have guessed the less beefy AVX was where any gains would have come from.
That said, beware overly dense cache structures, since additional read/write porting has direct density costs. Intel has had a 64B L2<->L1 path and a 2*32B one from L1 (and 1*32B to) since Haswell, where Zen’s internal per-clock bandwidth is half that, the same as Sandy/Ivy Bridge.
AMD just gave a talk at ISSCC that puts a Ryzen CCX at 44 mm^2 with the comparable structure from Kaby Lake at 49 mm^2
That in and of itself isn’t really that surprising; the thing that gets me is WHERE AMD made up that space advantage, and that’s in the cache, where they seem to be beating Intel at density. That’s been a traditional advantage for Intel, so it will be interesting to see how they managed to do it.
[url=https://en.wikichip.org/w/images/6/6f/skylake_%28quad-core%29.png<]4C[/url<] [url=https://en.wikichip.org/w/images/e/e9/skylake_%28quad-core%29_%28annotated%29.png<]Skylake[/url<] [url=http://www.anandtech.com/show/9505/skylake-cpu-package-analysis<]is[/url<] 122 mm[super<]2[/super<], and the cores/cache take about 1/3rd of that, making 44 mm[super<]2[/super<] for a CCX sound pretty reasonable.
44mm2 is crazy. Just compare that to a 22nm 4C Ivy that’s something like 160-180mm2. Granted it had graphics but it only took 1/3 the entire die. Either Zen is amazingly compact and/or it’s built on an amazing process node.
My branch prediction unit is down today.
If we ignore all the PRO items on the assumption they’re just the model of the same number plus a couple extra business-oriented features (as seen on some current models IIRC), it makes a bit of sense. That results in a low-clock and high-clock variant of each core/thread config, plus one more SKU for the halo (such as it is).
It definitely should be R3/R5/R7/R9 = 4C4T/4C8T/6C12T/8C16T or something similar, though. They’re so close to making that first number actually mean something! Also that would (mostly) eliminate the need to stick an X on the end of the high-clock names.
What if they changed from R# to RX on the graphics side so that R# would be available to distinguish CPUs? (Not that it worked well or anything.) GPUs have a much more clear performance progression than CPUs, and CPUs could use the extra digit to clarify as per my previous paragraph (if the marketing dept didn’t fumble it).
These names make me think of the R500-based ATI cards (X1000 series), which probably isn’t the association they’d most appreciate.
My work here is done.
Hmmm, only four down votes.. You guys can do better than that… How about if I add that my numbers were correct since PRO skus are OEM only…?
That means 3 octos, 2 hexas and a bunch of quads…
Just so you know, Jim’s now working for Tesla. Holy cow this guy is DA BOSS!!!
[url<]https://electrek.co/2016/01/28/tesla-jim-keller-apple-processor-architect-2/[/url<]
Sure they can. All it’ll take is AMD HQ calling JIM over the phone, screaming, “HELP!!!!”
OK, now I gotcha. I’ve updated my comment.
You gotta take a launch as it is. The good, the bad, and the ugly.
I tend to just let jerks be all the jerk they want. I’m not gonna play their game. I used to want to put them in their place but you know what? They already ARE in THEIR place!!
Are you joking? Well in case you are not. SCONE is a new kernel module. That paper is literally about a new software version of Linux Containers which now includes SGX support. Sheesh
The applications which run inside of a container don’t need to be recompiled because Linux Containers is a type of virtualization technology. The applications aren’t requesting SGX access. The kernel module is. This paper is about each container having it’s own enclave in SGX using the kernel module.
Who wants to speculate how many of those SKUs will be available in 2017?
My guess = 4. I’m feeling generous today.
Chuckula started it right before Skylake launched, and at the time we were all very confused about how it worked. The podcast after that had a segment where some very confused people tried to parse it.
How is it that I just now saw this? As before, I’m honored [url=https://techreport.com/news/31388/rumor-amd-ryzen-cpu-lineup-leaks-with-six-core-models-in-tow?post=1020158<]just for being chosen[/url<].
Interesting, the announcement that the 4 core 14nm Zen Die comes in at 10% less size and power than the 14nm 4 core Intel die, and that’s with twice the size of L2 cache; 512KB Vs 256KB.
[url<]http://www.eetimes.com/document.asp?_mc=RSS_EET_EDT&doc_id=1331317&page_number=2[/url<]
Edit your post that appears in the top comment to be Jeff saying Buffalo.
Buffalo is a TR NDA thing from the days of Scott being unable to talk under NDA about the product at hand but able to discuss hypothetical situations to which the answer is or isn’t “buffalo”.
I’ve always thought having a good product is what makes a good launch. Rumors of it being awesome prior to launch can only go downhill in case it’s not awesome. If there’s rumors of it being awful, and then it’s awful, oh well no big loss.
I’m also a little confused about what I should be doing right now. I’m just honored to have been chosen.
So, OEM and maybe standalone will have Polaris 12, or whatever you wish to call it, as a “video output” solution.
Approx 30-60% of a GTX 1050 based on very limited benchmarks found.
Care to refute this instead of downvote?
As a 14nm CU is about 4.25mm^2 as Fritzchens fritz on Flickr shows… 123 4.25 8 * – is 89. 89 6 – is 83. 6 is minimum die size for 64b MC.
Purportedly I don’t know what to do but I’m glad to be included.
Maybe I’ll just treat myself to buffalo wings.
If Intel strikes a massive return punch, great for consumer 😀
I dislike if this happens as I far favor AMD over Intel. But consumer is the winner in an Intel counter pinch that crushed AMD again.
Than ARM in ~5, maybe 10, years beats this.
I know everyone is excited about Ryzen, but we still have yet to see how Intel will react. It would be a shame if AMD’s victory is short lived. The next couple of years will be great for consumers, but long term we will have to wait and see if they can consistently bring that pressure to Intel
I tend to think you’re right. It’s just too damn confusing and I count myself as an enthusiast, so how is a casual supposed to figure it out.
The question of the day is will it overclock the way Intel chips have not since Sandy Bridge? If so, it not only has the potential to be competitive, but also to win the hearts and minds of the overclocking and enthusiast crowd.
I think this scheme is baed on a proposal out of which the company will choose 4 or 5 possibilities. Some junk found in a garbage can which probably just escaped the shredder.
I have to admit I’m really excited. The CPU space hasn’t been this interesting since Sandy Bridge in 2011.
> over the yokels at AMD
Do you ever look at the time and emotional energy you spend on this vitriol and wonder where else it could be spent in your life? It’s not healthy, man.
Yeah, AMD. The same company that pretended a 6700K can’t play dawn of the ancients in order to make RyZen look good.
How about a real academic paper from a real third party: [url<]https://www.usenix.org/system/files/conference/osdi16/osdi16-arnautov.pdf[/url<] Relevant quote: [quote<]SCONE increases the confidentiality and integrity of containerized services using Intel SGX. The secure containers of SCONE feature a TCB of only 0.6✓–2✓ the application code size and are compatible with Docker. Using asynchronous system calls and a kernel module, SGX-imposed enclave transition overheads are reduced effectively. For all evaluated services, we achieve at least 60% of the native throughput; for Memcached, the throughput with SCONE is even higher than with native execution. [i<]At the same time, SCONE does not require changes to applications or the Linux kernel besides static recompilation of the application and the loading of a kernel module.[/i<][/quote<] And that's just one implementation. Looks pretty good to me, and frankly I trust real software developers over the yokels at AMD any day of the week.
Man, you really have some thin skin. It’s the internet, please don’t take things personally.
I’ve said my peace. My intent was not to lock horns with anybody that day and never was or is. Sometimes you have to call out jerks for just being jerks and that one stuck in my craw for a while. I’ll be back when Ryzen is released.
SME isn’t a very technically complicated feature, but it does offer protection from cool-boot and other physical attacks that SGX doesn’t.
It honestly seems absolutely insane to me that Intel didn’t add comparable functionality before introducing XPoint/Optane NVDIMMs.
Gross.
Oh yeah I remember. I even commented on it.
I’ve never locked horns with derFunk but yeah, I suppose he was being aggressive there. As I’ve said, I’m not taking sides. I’m saying we should keep it nice and friendly around here. That goes for everyone, not just these two.
Oh, you missed the fun?
[url<]https://techreport.com/news/31266/in-the-lab-intel-core-i3-7350k-cpu?post=1016926#1016926[/url<]
Naw I don’t think so….sunshine. [url=http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2013/12/AMD_Memory_Encryption_Whitepaper_v7-Public.pdf<]How about AMD's whitepaper?[/url<] Let's see what they say... [quote<] SME is a general purpose mechanism that is flexible, integrated into the CPU architecture, scalable from embedded to high-end server workloads, [b<]and requires no application software modifications.[/b<][/quote<] Let me guess AMD is lying too? I would say "Thanks for playing" but it looks like you are just playing with yourself at this point. Maybe instead of trolling you could oh I don't know look up the information first before talking. It may be new to you but hopefully you can figure it out.....for the sake of all of us.
[quote<]Do you not know the difference between something that requires software to be compiled for something vs a hardware based solution that does not?[/quote<] A lot better than you do, trust me sunshine. [quote<]All you have to do is look it up. Here you go.[/quote<] Oh look, you linked to Wccftech when trying to pretend that you actually have a clue about anything. You lose. Thanks for playing.
Um no. Saying that something has to be compiled to support a feature does not mean it is compatible to all CPUs. Software has to be compiled to support SGX. SME aren’t instructions. Now SEV will probably be something you have to compile for but SME no.
Do you not know the difference between something that requires software to be compiled for something vs a hardware based solution that does not?
All you have to do is look it up. [url=http://wccftech.com/amd-zen-encryption-sme-sev-hw-based-sha/<]Here you go.[/url<]
[quote<]SGX is implemented in software. [/quote<] Great then it should work fine on... I dunno... Bulldozer. Saying that SGX is implemented in software is like claiming that a 386 has full AVX-512 support because in order to actually use AVX-512 the software has to emit the actual instruction that tells the chip to perform an operation. You are confusing the fact that SGX is [b<]used by[/b<] software with the blatantly incorrect statement that it is [b<]implemented in[/b<] software. That's flat wrong. Additionally, I guarantee you that you are going to have to have software that actually uses SME in a RyZen chip, it's not magical feature that floats down from the sky.
SME is NOT the same thing as SGX. SGX is implemented in software. What’s in Ryzen encrypts all of memory space regardless of what data is being accessed. They operate on completely different levels.
While you may not give a crap about all of us because, well, you don’t know us, and we don’t really give a crap what kind of person you are either because we don’t know you also, you might as well play nice here with none of your bad dude antics otherwise you’ll only give people here a bad time. Now I don’t know what derFunk said to piss you off and I’m sure as heck not taking sides here, but I suggest you cool off. It’s not cool holding on to grudges here. While that doesn’t really affect us, it will affect you and does say something about what kind of guy you are. In the end, it’s yourself you’re killing, not anyone here. As we’ve established, we do not know each other. We are far far away. And if you keep this up you will be footballed outta here in no time.
Don’t give a crap what you all think. You don’t know me and how I normally act. He took what I thought was an innocent remark up to 11 on the ahole scale right away and continued to push it. Didn’t see anyone calling him out for that crap cuz I guess he is a regular. I had nothing against him before that. I’ll be back when it is released.
AMD explicitly said that 3.4GHz was the base Zen speed – let’s assume that’s 8C Zen to allow some downward movement for 4C/6C SKUs.
Therefore the entire table reads like someone’s guesses. Maybe they heard the names, but the clock speeds look like guesses, and likely the product matching is guessed too.
AMD’s plan with Zen is to make 8C the new mainstream and enthusiast, and 4C the new budget, with a 6C bridge between the two.
Btw AMD said today that a 4C Zen CCX with 8MB L3 is 44mm^2.
Looking forward to seeing these on a price/performance chart against Intel’s lineup. Until then I find it hard to speculate. There’ll probably be a big early adopter tax on these, but if they’re in any way competitive, we can look forward to the price of enthusiast grade CPUs falling over the next couple of years.
Rx 460?
But a Polaris 12, probably 8 CU, chip is coming. Given architectural improvements might be as fast as a 7750.
But it might be OEM/Intel only.
Intel has been in the same boat for years. It is much worse for their Xeon brand. You need a cereal-ring decoder to make any sense of it.
Killjoy! Don’t you realize rumors have always been an integral part of every imminent product launch?? It makes the launch complete, lock, stock and barrel!!
On another but related note, it seems Ryzen will also include support for Windows 7. That’s good news for Win7 holdouts like me. Say what you will about Windows 10 but I’m not touching it with a 10-ft. cattle prod.
Le obligatoire link:
[url<]http://www.pcmag.com/news/351565/amd-ryzen-processors-will-support-windows-7[/url<]
You just KNOW AMD marketers are bound to Bork something. What terrific people!!
Yeah. Very not cool. There’s a term for guys like him though but I’ll let someone else say it.
Let’s keep it friendly, guys. No crap like that.
Maybe because you are the only one that got Butthurt over the Ryzen? People need to see it succeed to finally have a fair game from Intel.
You ever thought about that? No! You only think about yourself.
That’s rude, man. It does not make you look ‘cool’. And frankly my dear, very immature!
Are you… er.. bitter?
Given how it’s obviously prudent to take all this with a bag of salt, I think it’s also highly probable that these are pretty much what will make it to final products, if history has told us anything about pre-launch AMD rumors (BD performance numbers, BD die shots, K10 core die shots, etc.) Unless, of course, it’s a deliberate ruse to throw Intel off.
I can go with the relatively low clock speeds for the R7 and 6C/12T R5 models because they’re pretty much in line with what we’ve been used to from Intel, but the 4C parts seem to be a little slow especially when there are just 4 cores. What do you think?
R3, R5 and R7 should just be 2C/4T, 4C/4T, and 4C/8T, respectively, just like Intel. R9 should be 6C/12T, and (wait for it…) R[u<]11[/u<] should be 8C/16T. AMD goes up to 11!!! Intel is toast, dude!!!
Why would it be odd? The FX series has been getting along without an integrated GPU since Day 1.
I think people mainly downvote you because you’re kind of a dick.
APU’s are around the corner, probably closer than most of us would like. Hasn’t the cry since FX came out been too many APU’s, not any CPU’s?
[quote<]Another possible scenario is that the processors are named in order to ease comparison to Intel's Core models, which have used i3, i5, and i7 prefixes[/quote<] You really think this is the [i<]other possible[/i<] scenario?
“Mr. CEO, we cannot allow a bad product naming gap!”
Maybe AMD’s CPU division fancies themselves as pirates: RRRrrrrrrr.
Press Question: “When was the last time you were net-profitable?”
AMD response: “RRRRrrrrrrr.”
thanks but i’d rather trust TR’s review over some snipped spreadsheet.
is it a nacho hat? i love those!
[url<]https://www.youtube.com/watch?v=wOsWfzzvlkU[/url<]
[url<]http://i.imgur.com/ixRjRYE.jpg[/url<]
Is this where I say ‘username checks out’? 🙂
I’d love to be in that hat’s place. Hot.
Oops, that was supposed to be a PM.
Has AMD stated that Ryzen chips won’t have APUs on them? The high-end X370 boards have DVI, HDMI and DP ports on them. Are we returning to an era where motherboard makers are putting iGPUs on their motherboards? Or is this some sort of backwards compatibility thing?
EDIT: Ok looks like there are going to be Zen APUs in the future and we’re just getting straight processors to start with, which I’m actually really happy with.
Not really.
The CPUs in question are meant go up against Socket 2011 chips a.k.a HEDT market. They also lack integrated GPUs.
Somehow the goofiness of the names makes the rumor more believable to me. Nobody outside of AMD marketing could come up with something so ridiculous. Well, nobody except intel marketing.
Fixed, thanks for the heads-up.
Raven Ridge APUs (4 Zen cores + Vega-based GPU) are supposed to arrive in the 2nd half of the year.
Not particularly new features, just different names for SGX.
[Edit: For all the ignorant downthumbers, care to post a single shred of evidence that you’re actually right? Or did you just do the usual pavlov’s dog drooling at any buzzword that AMD put in a powerpoint and act like they invented the Internet or something?
Here’s a real page with real software that not only uses SGX but extends it to VMs using hardware that you can buy today: [url<]https://github.com/01org/kvm-sgx/wiki[/url<] ] And if those are considered "big" features then transactional memory sure counts too.
I think the AES-128 encryption on the memory controllers and the hardened virtualization zoning is considered a fairly big deal for cloud providers and their customers.
Polaris 12, 8(?) CU, <82Mm^2 die size (123 – 41 or more), 1GB VRAM? <20W if 900mhz. 20-30 dollars to OEM?
Besides OEM no matter anyways.
Polaris 12: [url<]https://videocardz.com/64787/amd-polaris-12-spotted-in-linux-patches[/url<]
I can see them saying “they are all available unlocked, just not with ECC enabled. If you want an unlocked 1200, the 1200X is available.”
Edit: Another theory is that they’re higher-voltage/TDP/clock models, either with no stock cooler or a Wraith cooler.
Naw, he kept pushing it…
Chill.
Looks like local tough guy derFunkenstein will have to eat my hat when one of these Ryzen cpu’s come in sub $200 at release (re; i3-7350K comments). Get ready as I’ll be wiping my arse with it for the next couple of weeks so you can get a real strong flavor and enjoy every bite.
That would contradict the statement from AMD that all Ryzen CPUs will have unlocked multipliers, so it could be something else (or, you know, this “leak” could be entirely fake.)
You could buy a R7 370X to go with your R7 and X370 too!
I suspect a lot of these ‘rumors’ are leaked outdated info. Maybe it’s intentional.
There’s no profit in ultra basic low-end parts. That’s probably why AMD’s lowest part this generation is the RX 460. If you don’t need anything more than that AMD is still clearing out old inventory of Oland (re-labelled, re-spun 7730) and Bonaire (re-laballed, re-spun 7790) stock that doesn’t sell because it’s as useless as an IGP.
I promise I wasn’t trying to be difficult. 🙂 I’m just trying to make sure I understand the perspective and scale of the market for it. And yes, it’s a shame AMD hasn’t got it ready to go… Maybe next time.
Given the X models are being compared to Intel K models, I would assume it just means unlocked. I fear Pro may mean ECC support, but maybe it just means “not X”.
TSX is enabled on the 6700K, 6600K, 6700, and 6600 along with their Kaby Lake counterparts, and that covers the majority of SKUs that people on this site care about.
That’s not to mention the fact that it’s very widespread in Xeons even going all the way down to the Xeon-D.
As for being a bit buggy in desktop haswell systems, it’s a highly complex feature that AMD still isn’t even attempting to emulate 4 years later.
I’m really just referring to the desktop space. Other than installing a few Xeons in blades at my last job, I have no experience in the server space. Given that, I believe my comment makes sense: nobody really cares what the underlying feature set is or what the heritage of the processor is as long as it performs well for the intended workload and for the given price.
> It seems odd that AMD (or nVidia) haven’t released an RX 410/GTX 1010 for half the price with all the cutting edge technology but having removed the unnecessary GPU power that makes up a large proportion of the chip and cost.
The trick is that most systems tend to revolve around inexpensive commodity parts with IGPs good enough for straightforward tasks, which makes the need for a low-end GPU irrelevant. We are a very long way away from the days of cheap onboard GPUs with shaky drivers from the likes of SiS. Anyone buying an IGP-less performance part will probably pair it with a capable discrete GPU, or grab the cheapest thing that will function in a basic way . The low end of the market doesn’t reward much in the way of R&D expenses: Fermi and GCN 1.0 (and even VLIW, for the 5450 and 6450 bargain bin selections) are still good enough to render a desktop, play video, and connect to VGA, DVI, and HDMI. Even a late bloomer like the Geforce GT 720 seems more like a repurposed laptop GPU; splitting out a subset of a bigger, more competent architecture takes research, testing, and money that won’t be made back for years. They’ll put it off for as long as they have to.
true, this way Ryzen cannot be comptetitive for most users with celeron, pentium and i3. They must buy also cheap graphics card, that costs some 50$ and increase power draw
I had a few GPUs as hand me downs before, but the X1650Pro was the first GPU I bought with my own money, good times. 512MB for no reasonable reason for that chips performance. But when I saw muzzle flash light up the dark scene in Half Life 2s Ravenholm I remember thinking it looked uncannily good.
Forgive me, wasn’t TSX found to be substantially glitchy in Haswell chips and disabled via microcode updates? It’s an important feature looking forward, but isn’t ubiquitous on Skylake parts either; maybe AMD just assumed its market penetration wasn’t great enough to merit delaying Ryzen even longer to incorporate support. If the company survives long enough to realize Ryzen 2, maybe it’ll show up then.
I’d be awfully curious to see a legitimate RyZen pre-order listing with full model number and specification information.
For example, from last September we have this for Kaby Lake: [url<]https://techreport.com/news/30681/report-estonian-retailer-accepting-kaby-lake-cpu-preorders[/url<] --> Fascinating how AMD fanboys who claim to be excited about RyZen don't actually want any hard information about what the real products will be like. Something tells me that they don't want to leave the neighborhood of make-believe that AMD put on last year for the harshness of the real world.
AMD isn’t leaving us much of anything but rumors at this point…weird.
If RyZen is actually supposed to be a competent server chip then plenty of people will care. And the pricing of a server CPU is a rounding error compared to the other costs involved in running a high-end server.
Unless of course, RyZen really isn’t targeting high end servers.
Rebuttal: Nobody cares as long as the performance is there and commensurate with pricing.
Yeah, except that there are no features in RyZen that are actually new.
If anything, it still lacks some 4 year old features like transactional memory that were the subject of rants and gnashing of teeth when they weren’t included in the 4770K back in 2013.
For example: [quote<]Virtualization and TSX features are very important, as is AES. It is very disgusting decision not to include them on i3, thus screwing budget buyers.[/quote<] [url<]https://techreport.com/discussion/24879/intel-core-i7-4770k-and-4950hq-haswell-processors-reviewed?post=735461[/url<] So if Haswell --including core i3s -- without transactional memory in 2013 was a disaster, what is RyZen without transactional memory in 2017?
[quote<] ...five R7 models packing headling...[/quote<] Possibly an extra word there that slipped past editing.
Yeah, this is too many. AMD doesn’t need confusion any more than it needs to give away CPUs for free.
I’m not talking about the missing iGPU.
I’m talking about the potential market created by that missing iGPU. Anyone who doesn’t need a powerful 3D rendering GPU still needs to buy an RX 460/1050 or above to get features like high resolution monitor outputs.
It seems odd that AMD (or nVidia) haven’t released an RX 410/GTX 1010 for half the price with all the cutting edge technology but having removed the unnecessary GPU power that makes up a large proportion of the chip and cost.
The low end cards on the market are currently based on Kepler and GCN 1.0 from back in 2012. The demand driven by mid-range iGPU-less Zen chips seems like the ideal time to release updated versions.
This IS a rumor…
Several things wrong with the specifics…
AMD said all Ryzen are 3.4GHz+…
Even if they disable 4MB of L3 on 1800 series, they still only have from 3.4GHz to around 4.2GHz (with boost or without) so the chips don’t have much to differentiate on clock…
And they also said they wouldn’t go above 6900X prices… I think that’s a mistake especially if they can really nudge above 4GHz (might be 6950X level)…
Anyway I predict if they have 4/4 they will start at $150, 4/8 will start at $250, 6/12 will start at $400 and the 8/16 will start at $550…
But maybe not…
Kind of like the A64, eh?
Yeah, it looks like they are trying to line their products against the competing Intel i series, but there isn’t a good equivalent for the six core chips.
The arbitrary R# isn’t very useful in its current form. They should do away with the arbitrary # scheme and make them represent something like the core count. (I.E. R4, R6, R8) If they really feel the need to add suffixes, make them useful as well. (I.E. R8s for SMT) Alternately, they could use the proceeding odd number (assuming no odd numbers of cores) to represent the SMT models:
R4 – 4C4T
<R4s or R5> – 4C8T
<R6s or R7> – 6C12T
<R8s or R9> – 8C12T
When you consider how long APU’s have actually been around, it’s more normal to have a CPU paired with an OEM gpu. Not like AMD’s current apu’s are any good anyway. Much better to slap in a 460 derivative.
“Buffalo”
– Jeff Kampman, The Tech Report
Odd? I thought that’s what enthusiasts spent the last few years screaming at Intel for, processors decoupled from the GPU if you were using a more powerful dedicated one anyways.
The laptop based ones will of course be APUs with GPUs.
Repeat after me: RYZEN IS A SERVER CHIP (that we crammed into a desktop form factor).
That comes with all the purported benefits and drawbacks of what being a server chip entails.
It seems a bit odd for AMD to be launching high end processors with no integrated graphics, without having an up to date basic GPU available that can drive multiple 4K screens and provide HEVC decoding.
I need to get on Ebay so I can buy an X1600 to go with a 1600X!
No kidding. We’re supposed to believe they’ll have FIVE 8C16T models at launch? I can understand 12 SKUs. That gets us an unlocked/enthusiast chip, a workstation chip, and a locked/consumer chip in each of the configurations. Beyond that we’re getting kind of silly for launch.
Incidentally, for the talk about 17 individual SKUs launching at once, go back in time to Bulldozer’s launch and you’ll see there were a grand total of three SKUs at launch.
[url<]https://techreport.com/review_full/21813/amd-fx-8150-bulldozer-processor[/url<] I'm expecting more than 3 RyZen SKUs at launch, but 17 sounds outright dumb.
“Buffalo”
[i<]- Jeff Kampman, The Tech Report[/i<] (as per the request from chuckula in [url=https://techreport.com/news/31372/poll-how-are-kaby-lake-and-ryzen-shaping-your-pc-building-plans?post=1019432<]this post[/url<]) Paging [b<]Ben Funk[/b<] and [b<]tipoo[/b<].... Operation Buffalo above this line -------------------------------------------- Original post below this line Man, that naming scheme is so bad I don't even know where to start. Confusion with Radeon models? R5 being used for both quad and hexa models? Why not R3, R5, R7, R9?! Mixing in both X suffix and Pro models but never together? Unnecessary prefix/suffix syndrome! [quote<][b<]I'll have an R7 1700X + MSI X-Power X370 + XFX R7 370 XXX please.[/b<][/quote<] FFS if I facepalm myself any harder I'll need forehead surgery to repair the damage....
If I had to guess at the “X” versus non-X CPUs, I’d bet only one of them has the funky temperature based auto-overclocking enabled.
Yeah, I’m very skeptical of anything non-official at this point.
Until my trusted reviewers have samples in hand and official AMD retail prices, I’m just going to pretend everyone is blowing smoke.
Interesting about the CPU complexes, repeat of the Athlon X3s possible where you could re-enable defunct dies, even if at a lower clock than the other cores?
Aside from all the other red flags that are being raised (ryzed?) with this “leak” it directly contradicts the [i<]other[/i<] rumors that the high-end launch chips have a 3.7GHz base frequency. Contradictory RyZen rumors: CONFIRMED!
FIRST!
#TheCNNEffect
The issue I have with this rumor is that the official reports and clear remarks from people who know better have said that all of the CPU SKUs (4 and 8 core) would be released at the same time with full availability at retailers the day of release. AMD has played the paper launch game before and knows that won’t be good for them this time.