AMD lays out its Ryzen and Radeon plans for 2018 and beyond at CES

Few companies in the world of computing were as busy as AMD was last year. The company refreshed its mainstream desktop processor lineup from top to bottom with Ryzen 3, Ryzen 5, and Ryzen 7 CPUs, bringing more competitive cores and threads to more affordable price points than ever. In high-end desktops, Ryzen Threadripper CPUs and the X399 platform offered competitive performance and better platform features than the Intel competition on the X299 platform. In the data center, AMD took the wraps off a full range of Epyc server CPUs that challenged Intel’s Xeon Scalable Processor family on bang-for-the-buck. In notebooks, Ryzen Mobile APUs joined the battle with Intel’s Kaby Lake Refresh chips. And in graphics, Radeon RX Vega graphics cards proved competitive with Nvidia’s GTX 1070 and GTX 1080 at the high end of the market, even if they remain difficult to get ahold of today.

Even if AMD didn’t win in every benchmark on every platform, the company can inarguably claim the most competitive product lineup it’s enjoyed in years. The challenge for AMD now is to keep that momentum going as we enter 2018. Ahead of CES, the company laid out a case for how it intends to keep the pedal to the metal.

The most natural place to chart a path through AMD’s 2018 starts with the desktop CPU, and enthusiasts will be happy to learn that AMD is bringing Raven Ridge APUs to the AM4 platform next month. That launch will be followed by second-generation Ryzen CPUs featuring the Zen+ architecture and built on GlobalFoundries’ 12-nm LPP process technology. Those chips will launch in April alongside a new X470 chipset. In the second half of the year, AMD will release the second generation of Ryzen Threadripper CPUs and update its Ryzen Pro platform with the new chips.

Ever since the launch of Ryzen CPUs last year, desktop users have been clamoring for a desktop APU that integrates Zen CPU cores with Radeon graphics processors. Those folks can rejoice soon, because Raven Ridge is coming to the AM4 platform. AMD will offer two such APUs next month: the Ryzen 5 2400G and the Ryzen 3 2200G.

  Cores/

threads

Base

clock

(GHz)

Boost

clock

(GHz)

GPU

compute

units

GPU

peak

clock

L3

cache

TDP
Ryzen 5 2400G 4/8 3.6 3.9 11 1250 4 MB 65 W
Ryzen 3 2200G 4/4 3.5 3.7 8 1100

The Ryzen 5 2400G marks AMD’s first fully-enabled Raven Ridge product. This chip will offer four Zen CPU cores alongside 11 Radeon Vega compute units. Those CPU cores will run at up to 3.9 GHz boost speeds with a 3.6 GHz base clock, and the complete package will sell for $169. Meanwhile, the Ryzen 3 2200G will drop back to four cores and four threads running at 3.7 GHz boost speeds and 3.5 GHz base speeds, and it’ll offer eight Vega compute units on board. That part will go for an especially spicy $99 suggested price. Both chips fit into a 65 W TDP, they use soldered heat spreaders, and they’ll be fully unlocked on both the CPU and graphics side of the die for overclockers to work their magic. AMD will launch them February 12.

Although we’d expect no less, AMD teased the performance of the Ryzen 5 2400G by putting it up against the integrated graphics processor of Intel’s Core i5-8400. Unsurprisingly, AMD’s testing reveals that the fully-enabled Vega 11 IGP walks all over the UHD Graphics 620 on board the i5-8400. More surprising, perhaps, is that AMD believes it’d take a GeForce GT 1030 to bring the i5-8400’s gaming performance on par with that of the Ryzen 5 2400G. To be clear, I don’t believe that it’d be necessary to pair a $200-ish processor like the Core i5-8400 with the GT 1030 to get comparable performance. One could easily get a Core i3-8100 for $130 and pair it with that same $70 or $80 graphics card and get competitive numbers to the effect of AMD’s claims for the i5-8400 setup.

Even with that in mind, AMD’s real advantage is that the platform cost of the Ryzen 5 2400G and a B350 motherboard would likely land around $250, and one could easily overclock the 2400G thereafter. The locked Core i3-8100 requires a relatively expensive Intel Z370 motherboard to run, as Intel’s partners haven’t released inexpensive H- and B-series motherboards in the 300-series generation yet. With such a motherboard, one might charitably need to spend $300 or more between the graphics card, CPU, and motherboard for an entry-level Core i3 gaming system. The bottom line is that the Ryzen 5 2400G looks like a sweet value for entry-level gamers and small-form-factor builders without room for a full- or even half-height graphics card.

Even for folks who don’t game, the Ryzen 5 2400G and Ryzen 3 2200G open up a wide swath of systems that were difficult for AMD to claim a foothold in with first-generation Ryzen CPUs. Since those chips lacked integrated graphics, folks with basic productivity needs had to scrape the bottom of the barrel for a basic discrete graphics card to light up their displays. That added expense ate into AMD’s platform-cost advantage a bit, and the Ryzen G-series parts neatly iron out that wrinkle. For basic computing tasks, a Ryzen 3 2200G seems like an extremely competitive part for $100. We should have an opportunity to find out soon.

 

Second-gen Ryzen CPUs promise positive evolutions

AMD’s Raven Ridge Ryzen APUs offer the first hints of what’s coming in AMD’s second-generation Ryzen desktop CPUs, as well. Although AMD didn’t offer many details of what to expect of these parts, folks hoping for some kind of foundation-shaking rework of Zen to fix its deficiencies relative to the Intel competition will be waiting for a while yet. Zen+ will likely offer meaningful improvements in performance, but from what AMD told us, I would expect a nice evolutionary improvement rather than a huge leap.

First off, Zen+ will be fabricated on GlobalFoundries’ 12LP process that we learned of in September of last year. AMD said the move from 14-nm LPP to 12LP will result in lower power usage and higher frequencies, two improvements that will be nothing but upside if they play out. The company also alluded to improvements for cache speed, memory speed, and latency, another bit of polish that has the potential to address some of the complaints enthusiasts have had about Ryzen chips. Finally, second-generation Ryzen CPUs will use the Precision Boost 2 boost algorithm that the company introduced with Ryzen Mobile APUs. Instead of relying on a one-core to two-core and then an all-core boost speed, Precision Boost 2 allows the processor to boost each core as it’s able depending on thermal conditions and workload severity. That more granular approach could offer better performance in workloads that previously straddled two-core and all-core boost conditions.

To mark the introduction of second-generation Ryzen CPUs, AMD will also introduce a new high-end motherboard platform called X470. We don’t know a ton about this new platform yet, other than that AMD claims it’s “optimized for second-generation Ryzen CPUs” and that it could offer lower power consumption than its predecessor. Despite any optimizations X470 might offer, second-generation Ryzen CPUs will be backwards-compatible with today’s X370 motherboards, and that combo shouldn’t require anything more than a firmware update to function properly.

Today’s Ryzens get a storage boost and price cuts

Old SEPs on the left, new on the right

For the moment, AMD is using the lag time between the maturity of the first generation of Ryzen processors and the arrival of the second generation to potentially put more competitive pressure on Intel. We’ve been seeing deep discounts on a lot of Ryzen CPUs from retailers of late, but AMD is making those lower prices official with a change to its suggested price list. The deepest cuts come at the top of the stack, where the Ryzen 7 1800X plummets from $499 to $349, the Ryzen 7 1700X falls from $399 to $309, and the Ryzen 7 1700 drops $30 to end up at $299. Those prices aren’t far off the deep discounts we’ve seen of late, but it’s nice to know that builders can reliably enjoy those stickers rather than waiting for the discount winds to blow the right way again.

In the middle of the market, AMD’s Ryzen 5 1600X drops from $249 to $219, the Ryzen 5 1600 drops from $219 to $189, and the Ryzen 5 1500X drops from $189 to $174. Of these drops, the Ryzen 5 1600’s is probably the most important, as the value proposition of that CPU was among the strongest of any chip released last year even at its suggested price. For around the same price as a locked Core i5-8400 with six cores, especially thread-hungry builders can get an overclockable CPU with a respectable stock cooler and 12 threads, and that continues to be an enormously attractive proposition for AMD. The Ryzen 5 1500X drops to $174 from $189, but that gap probably isn’t enough to make the four-core, eight-thread part appealing versus the only slightly-more-expensive 1600.

At the lower end of the Ryzen stack, the Ryzen 5 2400G slots into the spot that the Ryzen 5 1400 occupied previously—and just below the 1500X. AMD says the Ryzen 5 1400 will continue to be produced, but unless someone really needs a wide-and-slow chip, it’s hard to imagine choosing the 1400 over the higher-boosting and Radeon-equipped 2400G. A similar fate would seem to await the Ryzen 3 1200, whose relatively low clocks and lack of an IGP look hard to justify next to the Ryzen 3 2200G with its higher clocks and powerful onboard graphics.

A couple more platform perks await Ryzen builders in 2018. First up, AMD is attempting to blunt Intel’s vendor-locked Optane Memory storage accelerator with a software partnership from Enmotus. The companies have joined forces to offer Enmotus’ FuzeDrive storage-tiering utility to Ryzen owners for just $20. FuzeDrive can join hard drives, SSDs, and even RAM together in a tiered caching approach that stores frequently-used files on faster tiers of storage. AMD cheekily suggests that builders could even use Optane modules as intermediary storage tiers if they wanted, and we see no reason to doubt it since Optane formats just like any other NVMe device when it’s not paired with its companion software. The downside of FuzeDrive is that since it’s a block-level caching solution, losing a storage device from the tiered pool or experiencing a power failure while the system is operating will likely mean data loss. AMD admits as much, but falls back on the justified advice that users should be backing up their files anyway.

Finally, AMD is updating its popular Wraith Max cooler to more fully capitalize on the RGB LED craze. Where the first Wraith Max only had a single-color LED ring encircling its fan, the Wraith Prism will offer controllable RGB LEDs around its fan’s perimeter to allow for a rainbow effect. The cooler’s fan hub will also be illuminated this time around, and the fan’s rotor is now translucent instead of opaque to allow the RGB LEDs in the center to shine through. The current Wraith Max is now available for $45 at retail for those who want to cool their Ryzen chip without going overboard on the blinkenlights, too.

 

Ryzen Mobile branches out

While AMD’s Ryzen 5 2500U and Ryzen 7 2700U have shown promise as AMD re-enters the mobile space, those high-end and high-performance parts don’t give potential AMD partners all the flexibility they might want in designing notebooks for every market segment. That’s where the Ryzen 3 2300U and Ryzen 3 2200U come in. This pair of mobile processors will fill out the Ryzen Mobile lineup for entry-level and lower-midrange systems that don’t need the absolute highest gaming performance.

  Cores/

threads

Base

clock

(GHz)

Boost

clock

(GHz)

GPU

compute

units

GPU

peak

clock

L3

cache

TDP
Ryzen 7 2700U 4/8 2.2 3.8 10 1300 MHz 4 MB 15 W
Ryzen 5 2500U 2.0 3.6 8 1100 MHz
Ryzen 3 2300U 4/4 2.0 3.4 6
Ryzen 3 2200U 2/4 2.5 3.4 3 1000 MHz

AMD will also be launching a full lineup of Ryzen Pro mobile APUs in 2018, and it expects to beat Intel’s eighth-generation vPro CPUs to market with these business-ready systems. I wrote in-depth about the capabilities of Ryzen Pro desktop systems last year, and those same security and manageability features are coming to Ryzen Pro notebooks this spring.

  Cores/

threads

Base

clock

(GHz)

Boost

clock

(GHz)

GPU

compute

units

GPU

peak

clock

L3

cache

TDP
Ryzen 7 Pro 4/8 2.2 3.6 10 1300 MHz 4 MB 15 W
Ryzen 5 Pro 2.0 3.6 8 1100 MHz
Ryzen 3 Pro 4/4 2.0 3.4 6

As a brief refresher, Ryzen Pro processors support transparent system memory encryption, remote management using the DASH open consortium’s standards, 18-month stable system images, and more. AMD will offer a Ryzen Pro Mobile chip in every one of its product tiers.

RX Vega hits the road

Just a couple hours ago, Radeon RX Vega graphics made a splashy midrange debut on Intel’s eighth-generation G-series CPUs for mobile devices. While we don’t know any details of the chip yet (or whether it’s related to the chip that Intel is using at all), AMD has its own plans to bring RX Vega graphics into mobile PCs.

The company teased a smaller Vega chip with a single stack of HBM2 RAM on board during the conference, but it didn’t share any further details regarding the graphics-processing resources on board.

AMD did tell us to expect the RX Vega Mobile chip to power thin-and-light gaming notebooks, but details are still frustratingly thin. We expect to hear more about the RX Vega Mobile GPU later this year.

7-nm makes its debut with a Radeon Instinct chip

Perhaps the biggest news from AMD’s GPU roadmap is that the first 7-nm GPU it will build is a Vega chip, and that it won’t be a consumer graphics chip.

Instead, this Vega GPU will be a Radeon Instinct product. AMD says it’ll have dedicated instructions for accelerating deep-learning training and inference, new I/O capabilities for high-speed communication among nodes in general-purpose HPC clusters, and support for the company’s hardware-managed multiuser GPU . This chip will be sampling to customers in 2018, though AMD didn’t offer further details about its availability.

Along with this new chip, AMD announced that its complete deep-learning software stack has achieved production-ready status, and it touts that stack’s open-source accessibility in contrast to the closed environment of CUDA. The ease of implementation and widespread popularity of CUDA will be difficult to overcome, to put it mildly, but AMD is laying the foundation for customers interested in its hardware.

AMD has laid out an exciting series of potential products for 2018. Stay tuned as we get our hands on some of this hardware and put it to the test.

Comments closed
    • Unknown-Error
    • 2 years ago

    R5 2400G – 3.6/3.9 GHz
    R5 1500X – 3.5/3.9 (XFR) GHz
    R5 1400 – 3.2/3.4 GHz

    R3 2200G – 3.5/3.7 GHz
    R3 1300X – 3.5/3.9 (XFR) GHz
    R3 1200 – 3.1/3.4 GHz

    There is a bumb in clock speed (in relation to the previous gen SKU) but could a hypothetical R7 2800G/2900G hit 4.4 GHz or 4.5 GHz or will it max out at 4.2 GHz?

    • Shouefref
    • 2 years ago

    Shouldn’t they redesign their Ryzen’s to tackle Spectre and Meltdown?

      • chuckula
      • 2 years ago

      Why would they do that?!?!

      RyZen is immune to anything that can hack an Intel chip. And by “anything” I mean “everything”.

      • tipoo
      • 2 years ago

      AMD CPUs don’t allow less-privileged code to perform even speculative memory accesses to more privileged memory locations when those accesses would result in a page fault, so they’re immune to the worse bug, Meltdown. They do have Specter though, but I would assume the patch would be correspondingly smaller in impact.

      Future hardware will probably respond to that too, but they don’t have to scramble as hard as Intel and ARM here (what they should scramble on is capitalizing on Intels performance losses!)

      • just brew it!
      • 2 years ago

      Canceling the launch of all CPUs currently in the pipeline, this late in the game, could very well be fatal to AMD. Too much already invested.

      A real fix for Spectre is going to require both hardware and software support. We can’t just ditch speculative execution entirely; that would be a big step back in performance and, perhaps more importantly, performance/watt. What we [i<]can[/i<] do is provide robust hardware support for disabling speculative execution in situations where it could potentially leak information. Then compilers (and especially JavaScript JIT compilers, since web browser security is the most glaring hole exposed by Spectre) need to take advantage of that.

    • Paine
    • 2 years ago

    Need Threadripper gen2. Maybe then I can afford gen1.

    • Unknown-Error
    • 2 years ago

    By the way…..7nm? These process naming schemes are getting ridiculous. Intel is the closest with their 14nm gate length being around 20nm (minimum). But remember that is the gate length. The contacted gate pitch is 70nm in the 14nm process. So what is GF going call it process when or if their gate length actually reaches 7nm? The [b<]-14nm[/b<] process? Quantum Mechanics is gonna be pissed.

      • DancinJack
      • 2 years ago

      who cares what the names are? As long as they increase performance and reduce power at some point, I’m good with whatever they wanna call it.

    • chuckula
    • 2 years ago

    Sure AMD, take your foot off the gas on CPUs when the competition is down.

    • smilingcrow
    • 2 years ago

    Does Raven Ridge use Zen+ cores on 12LP?

    • ronch
    • 2 years ago

    Excited to see how Ryzen 2 (or Ryzen+) will improve over the first iteration. And of course, every succeeding iteration over the next few years.

    • WhatMeWorry
    • 2 years ago

    “One could easily get a Core i3-8100 for $130 and pair it with that same $70 or $80 graphics card and get competitive numbers to the effect of AMD’s claims for the i5-8400 setup”

    Ok, so $130 i3 + $80 graphics card = $210. Whereas 2400G is $169. Plus you only have one chip/vendor to deal with. Isn’t that a clear win for AMD here?

      • Beahmont
      • 2 years ago

      No, because the Intel chip is going to have better IPC than the AMD chip. Core for Core Intel still wins pretty handily.

      Second, you have to take AMD’s numbers that their APU’s GPU is competing with a 1030 with a large grain of salt. Especially because the metric they used is an artificial benchmark and not real world numbers with Frame times and 99th percentile figures to judge actual playability and not just marketing FPS numbers.

      It’s also worth pointing out that in the Intel setup another $20-$50 dollars gets you a 1050 or 1050ti that will stomp that AMD APU into the ground. Another $20-$50 dollars in the AMD build does not deliver the same performance increase on the GPU side.

        • xeridea
        • 2 years ago

        AMD IPC is pretty close to Intel, lower clocks though.

        So comparing $170 to $260 now? That’s fair. You could also pair the APU with a 1050TI and get same performance for cheaper.

          • mczak
          • 2 years ago

          Since the i3-8100 lacks turbo, the cpu clocks are actually slightly lower than the cpu clocks of the 2400G (2400G is 3.6 – 3.9 Ghz, whereas the i3-8100 is a fixed 3.6 Ghz, albeit I’m not sure if the 2400G can sustain its cpu base clock together with the gpu max clock if both are used). So I’d guess indeed the cpu part of the 2400G won’t be all that much slower overall than the i3-8100 for single thread performance if at all (maybe 5% or so) – but OTOH it will very easily beat that i3-8100 on multi thread performance (as the latter lacks HT).
          Sounds like the 2400G would be a very attractive option indeed, even though I remain sceptical that it can match a GT 1030 for graphics outside some 3dmark benchmarks.
          Of course, if you do pair the i3-8100 with a GTX 1050, graphics will be much faster than the 2400G. As much as I like the GT 1030 as a lowend chip (nvidia didn’t have a desktop low-end gpu last generation, since the gm108 maxwell chip was strictly mobile only, missing display outputs for instance), its price/performance ratio is indeed pretty bad compared to GTX 1050. But the 2400G is quite a bit cheaper than i3-8100 + GT 1030 already, boards are cheaper too, so arguably such a solution is really in a different price bracket.

          • Beahmont
          • 2 years ago

          No. I’m comparing a $270 CPU and GPU with a $310 APU and GPU.

          A 1050 is the low end of mainstream gaming. These new APUs do not change the fact that even a good MOBA box needs at least a 1050 at a normal gaming resolution for 2018. Unless you are gaming at less than 60fps on crappy 720p that $30ish extra on the APU buys you nothing.

          Yes, the APU gets you into really crappy gaming for $30ish more. But no one stops there for gaming, they go on to buy a dGPU. At that point your $30 bucks is better spent on more dGPU than an APU. Ryzen and Vega have still not changed the fundamental problem for APUs in the gaming space.

          AMD needed APUs this gen to compete in the workplace no dGPU desktop box space. These might still do that. But not because the iGPU is that good, but merely because it finally exists and like the Intel iGPU it is good enough.

        • WhatMeWorry
        • 2 years ago

        Thanks for the information. Then do you feel more kindly towards the Intel/AMD 8809G?

          • Beahmont
          • 2 years ago

          I’m not really sure. EMIB is a massive game changer in how we think of laptop performance. At least presuming Intel can get the right yeilds for the right material costs for EMIB manufacturing.

          Even if Intel’s numbers on the direct performance benefits are rosey, the other gains in space saving and thermal solutions should be indirect performance gains because of less throttling and more battery in the same space.

          It is also a different problem because Raja chips are supposed to go in places that you can’t upgrade and add a dGPU anyway.

          The Raja chip also has the advantage of actually doing what APUs can’t at this point and that is actually game at reasonable resolutions for 2018.*

          That said, I am still cautious. On paper it looks very good. In practice, who knows?

          *Presuming Intel did not screw the pooch and screw up the frametimes.

    • jihadjoe
    • 2 years ago

    1800X looks mighty tempting at $350. I’d like to wait for Zen+ if it can get just 500MHz more though.

      • shank15217
      • 2 years ago

      Microcenter has been selling the 1800X for months, now everybody else can join in!

        • Beelzebubba9
        • 2 years ago

        Yeah I went in planning on buying a 1600X and saw the 1800X for $349 and snapped one up.

        It’s nice!

    • DragonDaddyBear
    • 2 years ago

    Maybe I missed something but when can we expect a replacement for the RX580 and lower? I’m guessing we can expect a rebadge as the RX680 line this year with even more stupid-high clocks?

      • watzupken
      • 2 years ago

      May be you won’t see any new models at all. I doubt they can push Polaris chips any higher in clockspeed without a huge bump in power requirement. Already their chips are no where near Pascal efficient, so pushing the power envelope further is not going to help them with the diminishing returns.

        • DragonDaddyBear
        • 2 years ago

        You’re probably right. It’s frustrating, though. The mid-range isn’t really mid-range right now. There is a huge gap thanks to the stupid mining craze that’s right where I like to buy my cards.

        Maybe they’re secretly planning a shrink and production volumes that will actually keep the cards affordable this time. Or I could come back to reality and roll the green team because I can actually buy their cards. >_<

    • mark625
    • 2 years ago

    So what I take away from this is that Intel’s new Kaby-G with Radeon graphics will stomp all over the new Ryzen APU pieces.

    I can see why AMD would want to make a quick buck by selling parts to Intel. But why would they allow Intel to put out a processor that will run rings around AMD’s own equivalent parts?

      • willmore
      • 2 years ago

      Compare the costs of the chips. AMD probably makes as much from the GPU they sell to Intel as they do from these chips. Then Intel has to add HBM2, a CPU, and do the packaging.

      The Kaby-G chips will outperform the announced AMD APU chips, but they’ll cost 2x to 3x as much. So, that’s not really the same market segment.

      AMD probably said “we’re not going to have a chip in the 20-24 (with HBM2) CU range, so why not sell chips to Intel?”

      • derFunkenstein
      • 2 years ago

      As it should, considering it’s going to be far more expensive.

      • mark625
      • 2 years ago

      To answer my own question, I have read elsewhere (sorry TR!) that this is more of a shot in Nvidia’s direction, to take a big bite out of their mobile GPU sales. AMD is willing to let Intel have the Vega chip in order to put a damper on Nvidia’s mobile market, while also getting a chunk of revenue from Intel.

      If that is their logic, then I guess I can buy into that. Hopefully it will not hurt them in the long run.

        • Shobai
        • 2 years ago

        Fair enough, but you could have read that here, too – it’s in the article

      • NTMBK
      • 2 years ago

      Raven Ridge has both a significantly lower price and a significantly lower TDP. You’re comparing a 15W part to a 65W part!

        • Anonymous Coward
        • 2 years ago

        Notably, both the 2400G and 2200G are 65W parts, which apparently is good enough for a lot of laptops. But price… obviously there is going to be a big difference in price.

      • watzupken
      • 2 years ago

      I doubt they are competing for the same market/ price range. People who need an integrated graphic solution may not need a super powerful one.

      Also, it is beneficial to AMD since Intel chips tend to sell better, and AMD may end up selling more Vegas chips.

    • Kretschmer
    • 2 years ago

    Has AMD marketing ever released a comparison graph that was even ballpark close to real world performance?

      • chuckula
      • 2 years ago

      Who cares.
      These comparison graphs are clearly 437% more accurate than their old comparison graphs so everything’s fine.

        • maxxcool
        • 2 years ago

        bahahahaha

        • Redocbew
        • 2 years ago

        The graphs are also 25% thinner than the graphs of the previous generation.

      • Puiucs
      • 2 years ago

      their CPU comparisons when Ryzen launched vs the Intel were fairly accurate. (like the r7 1800x vs i7 6900k slides)

    • blastdoor
    • 2 years ago

    [url<]https://www.semiwiki.com/forum/content/6713-14nm-16nm-10nm-7nm-what-we-know-now.html[/url<] The table comparing Intel 10nm to foundry 7nm sure is interesting — it makes it look like foundries will take the lead. Does anyone have better info than this?

      • tipoo
      • 2 years ago

      Will be interesting seeing a world where Intel can’t lean on a fab advantage.

        • blastdoor
        • 2 years ago

        Understatement of the decade!

        • psuedonymous
        • 2 years ago

        Depends on how long “EUV is one year away!” continues and who manages to roll out first. Which is mostly down to who has the cash reserves to gut and rebuild their fabs. Both GloFo and TSMC have mentioned moving to partial-EUV processes as part of 7nm rather than an outright switch, so they’re clearly hewing towards keeping their current equipment running for as long as possible.

          • freebird
          • 2 years ago

          I’m hoping for a late 2019, but probably 2020 for the High Volume Manufacturing with EUV. Samsung may be the 1st since they plan on using it in DRAM production which should be simpler to implement than SoCs/CPUs.

          EUV has many things that are holding it back… it doesn’t really involve “Which is mostly down to who has the cash reserves to gut and rebuild their fabs” because they all already have some EUV on-site and have been playing with them for quite some time. The cross over point becomes when will EUV be able to handle High Volume Manufacturing at an acceptable Uptime availability versus doing Triple or Quad Optical patterning at 7nm. There are also additional issues that will need solved in the move to 5nm.

          Here is a relatively recent/good article that explains many of the issues involved:
          [url<]https://semiengineering.com/issues-and-tradeoffs-for-euv/[/url<]

    • Welch
    • 2 years ago

    Damnit… Just built 5 systems because I couldn’t wait for those APUs to drop. That 2400g would be perfect and the price sure is right. That 2200g though… An office machine wet dream on price. You can have a B350 board and chip for probably around $150-170…. The RAM is where it gets pricey. $110 for the cheapest 3200 ATM for 2x4GB.

    I wonder why on Earth AMD didn’t make a push for those APUs to release much sooner, like mid last year. It feels like they missed a good opportunity for office refreshes this last year. Hoping it’s a strong year for AMD and prices come dropping down due to competition.

      • shank15217
      • 2 years ago

      You mean raven ridge? I think it’s still going to be a great buy this year. The desktop APUs should be exciting with their much higher TDPs. What would be neat is a Thread Ripper socket APU.

        • Welch
        • 2 years ago

        Yeah, the Raven Ridge stuff for desktop sockets…

        Sad, I don’t think we will see a threadripper in all of it’s glory in an APU format. Maybe if there is some sort of use for an IGP in the business world that can take advantage of something Vega has to offer.

        Between these chips and the Intel IGP Vega… I think AMD is going to have a record year on it’s hands. Calling that now.

    • willmore
    • 2 years ago

    Looks like the Ryzen Pro 7 base clock is wrong. 3.8 is a bit hard to believe. How about 2.2?

    • Spotpuff
    • 2 years ago

    Finally the elusive Threadriprer has a release date.

      • nico1982
      • 2 years ago

      +1 to you for noticing.
      -1 to AMD for the effort.

    • strangerguy
    • 2 years ago

    With the 2 new desktop APUs on the table what’s the point of the 1300X/1500X anymore? AMD should just retire them.

      • NTMBK
      • 2 years ago

      Somewhere to dump failed dies?

      • Zizy
      • 2 years ago

      1500X still has more L3 cache, it isn’t completely redundant. It has been pointless because of 1600 since the beginning though.

      1300X was likely retired as there is no new price in the second column. But AMD hasn’t released the replacement chip yet.

      • shank15217
      • 2 years ago

      Some of us like our discrete GPUs TYVM

    • ET3D
    • 2 years ago

    The difference in price between the 2200G and 1300X suggests a difference in performance. They have the same clock speed, so perhaps the smaller L3 cache has a significant effect.

    In any case, the 2200G sounds like great value, and would likely find a place in a lot of entry level PC’s. (Now if only RAM prices dropped back to a reasonable level.)

    • NTMBK
    • 2 years ago

    7nm Vega port in 2019, Zen2 on 7nm in 2019… everything is lining up for a 2020 next generation console.

      • tipoo
      • 2 years ago

      The CPU is where they could really differentiate from the 8th gen, since the .5 consoles already used a die shrink on GPU gains. Hopefully it would indeed be Zen 2 and not some successor to the cats.

      • ronch
      • 2 years ago

      That’s assuming AMD bags design wins from Microsoft and Sony again. I sure hope they will. And going from Jaguar to Zen++ is gonna be a giant leap.

        • tipoo
        • 2 years ago

        Still the only game in town for a high GPU performance APU, probably neither the RajaChip or Intels new GPU will change Intels interest in the low margin console space. I’d give AMD an 85% chance unless Sony/MS want to move off of APUs entirely.

        • Puiucs
        • 2 years ago

        There is no longer a question of whether or not AMD will win the next generation consoles. It’s pretty much a given. Both Intel and Nvidia just don’t have anything close to AMD’s APU technology and they also can’t really compete in the perf/$ side.
        Couple that with compatibility issues that might come from moving away from AMD and… you get the picture.

          • Zizy
          • 2 years ago

          Nah, there are many other options. I don’t think none are as likely as a pure AMD choice, but they aren’t impossible either:
          Sorted by probability, I would say probability split is like 60 for AMD / 40 for not-AMD. Out of that 40%, I split those 5 cases in 40/30/20/10/0. Maybe 3 goes way higher in ranking in case MS and Sony want to chase after Switch, but I think they won’t as they know it won’t work.

          0) AMD only. Good enough, cheap enough, and above all, easy enough solution.
          1) Intel + NV combo. KL-G like chip, just with NV instead of AMD. These two are the best at the moment and both MS and Sony are planning X2 and PS5 now, so they might get a nod.
          2) AMD+NV combo. Intel doesn’t want to offer as good CPU price, so a KL-G like chip is made with nearly as good AMD’s CPU and NV’s GPU. Would be a possible choice of MS and Sony if they deem AMD is too far behind in GPU efficiency to compete in the next gen.
          3) NV only. ARM CPU, NV’s GPU. Something like 4x A76 or whatever will be the name of the next gen cores. Barely any CPU advantage over Jaguar but great efficiency. Enables PS5 and X2 to be portable as Switch, yet as performant as X1X when docked and compatible with existing games. Continuation of the current console generation.
          4) Intel only. Promise of a great CPU and a good enough GPU gets the deal from MS and Sony.
          5) Intel + AMD combo. Backup of Intel only, as Intel’s GPUs still suck and NV is harder to work with.

            • NTMBK
            • 2 years ago

            There’s always the possibility of a dark horse like Qualcomm coming in- if Sony or MS decides to make a Switch competitor, then tablet SoCs are back on the table.

            • Anonymous Coward
            • 2 years ago

            Nobody is going to touch AMD so long as consoles continue with a similar form factor. The Zen core is solid, its affordable, its efficient, the GPUs are fine if clocked a little conservative.

            It would be insane for anyone to try to break into such a low margin space against a competitor with a toolbox like AMD has.

            • ronch
            • 2 years ago

            I think the biggest reason to buy CPU and GPU from AMD is that if there’s a problem with how those two parts work together, AMD and only AMD is held accountable. Getting Intel and Nvidia will entail lots of finger pointing.

          • renz496
          • 2 years ago

          lol if that’s the case why nintendo go for nvidia solution after using AMD/ATI GPU for whole three generation? nvidia did have their own “APU” called Tegra. in the past tegra main focus is mobile market. but that’s no longer the case.

            • Anonymous Coward
            • 2 years ago

            Nintendo targeted a form factor that AMD could not reasonably work in, so your point holds no water.

            • renz496
            • 2 years ago

            Let me tell you something: AMD also have their own SoC type of chip even if the GPU portion of the chip probably not as fast as nvidia. Remember project Discovery? It is AMD attemp to make SoC for tablet like device. So your saying of Nintendo targeted a form factor AMD could not reasonably work in is not accurate. Even if the home console have much bigger power budget it is not limitless. Right now nvidia can be significantly faster for the same die size and power budget (GP102 vs Vega 10). Sure Vega can can be very fast if you take advantage of it’s feature like RPM but even game develeoper themselve said the usage of FP16 is quite limited on modern graphically complex game (and there is still optimization issue) Compared to GP102 which developer can pretty much tap into it’s raw performance without supporting exclusive feature.

            • Anonymous Coward
            • 2 years ago

            Funny, I don’t remember Project Discovery. From Google, I see they had actual silicon running a demo in 2014. So AMD did have silicon, which is more than I expected, but their situation looks dire next to the steady releases of Tegra by nVidia. I wonder if they had already abandoned it by the time Nintendo was ready to go in 2016, with an nVidia SoC first launched in 2015. So I’ll agree they theoretically could have powered the Switch, but its not surprising they didn’t.

            The only problem with nVidia’s GPUs in the larger consoles is what CPU to strap them to. In AMD’s corner, you have existing integration between Vega and quad-core Zen, works great, they just need more GPU and some HDR memory. From nVidia, you’d need serious work on the CPU side to even match current consoles, or using an Intel-nVidia glued hybrid, you’d be at a big price disadvantage. No other player is worth discussing. Not seeing how AMD can be beaten.

    • Unknown-Error
    • 2 years ago

    Not bad. Price cuts always helps. Future doesn’t look awesome but it certainly looks steady. Only thing is, whenever we get our hope high about AMD things just come crashing down. If things continue like this then at least from a CPU stand-point AMD will remain relevant. Not sure how the GPU side will go though.

      • enixenigma
      • 2 years ago

      With no new consumer desktop GPUs announced (custom Vegas don’t count in my book), it’s not looking great.

    • the
    • 2 years ago

    This road map, while not aggressive, is currently better than Intel’s which is currently in chaos. AMD just need to continues executing while Intel stumbles to take the lead.

    The real variable will be Intel’s roadmap in response. The CEO of Intel will have an opportunity to discuss this tomorrow at CES.

      • willmore
      • 2 years ago

      While he dodges questions about Meltdown. Should be exciting.

    • RtFusion
    • 2 years ago

    Very surprising to see Vega on 7NM despite this last official slide saying that the next iteration of Vega would be on 14NM+:

    [url<]https://www.anandtech.com/show/11404/amd-updates-gpu-architecture-roadmap-after-navi-comes-next-gen[/url<] But with being on 7NM, it matches closely with this leaked Instinct Roadmap several months ago: [url<]https://videocardz.com/65521/amd-vega-10-and-vega-20-slides-revealed[/url<] If AMD can execute, it will be very interesting times in the datacentre for them. EDIT: Ryzen lineup does also differ with the one that was also leaked a while back: [url<]https://i.redd.it/c7fix5jkqy201.jpg[/url<] No Ryzen 7/5 Mobile Gaming chips.

      • Zizy
      • 2 years ago

      There are, AMD just decided to brand them as R5/R3 instead, indicating CPU performance – R5 2400G and R3 2200G.

Pin It on Pinterest

Share This