AMD Unveils World's First 7nm Datacenter GPUs with PCIe 4.02 Interconnect

Discussion in 'HardForum Tech News' started by cageymaru, Nov 6, 2018.

  1. cageymaru

    cageymaru [H]ard as it Gets

    Messages:
    19,803
    Joined:
    Apr 10, 2003
    AMD unveiled the world's first lineup of 7nm GPUs for the datacenter that will utilize an all new version of the ROCM open software platform for accelerated computing. "The AMD Radeon Instinct MI60 and MI50 accelerators feature flexible mixed-precision capabilities, powered by high-performance compute units that expand the types of workloads these accelerators can address, including a range of HPC and deep learning applications." They are specifically designed to tackle datacenter workloads such as rapidly training complex neural networks, delivering higher levels of floating-point performance, while exhibiting greater efficiencies.

    The new "Vega 7nm" GPUs are also the world's first GPUs to support the PCIe 4.02 interconnect which is twice as fast as other x86 CPU-to-GPU interconnect technologies and features AMD Infinity Fabric Link GPU interconnect technology that enables GPU-to-GPU communication that is six times faster than PCIe Gen 3. The AMD Radeon Instinct MI60 Accelerator is also the world's fastest double precision PCIe accelerator with 7.4 TFLOPs of peak double precision (FP64) performance.

    "Google believes that open source is good for everyone," said Rajat Monga, engineering director, TensorFlow, Google. "We've seen how helpful it can be to open source machine learning technology, and we're glad to see AMD embracing it. With the ROCm open software platform, TensorFlow users will benefit from GPU acceleration and a more robust open source machine learning ecosystem." ROCm software version 2.0 provides updated math libraries for the new DLOPS; support for 64-bit Linux operating systems including CentOS, RHEL and Ubuntu; optimizations of existing components; and support for the latest versions of the most popular deep learning frameworks, including TensorFlow 1.11, PyTorch (Caffe2) and others.
     
  2. virtualheretic

    virtualheretic Limp Gawd

    Messages:
    237
    Joined:
    Sep 6, 2005
    But how many Plex transcodes can it handle?
     
    Sufu and Esso like this.
  3. 128 threads? You tell me.
     
  4. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    Welcome to the age of trickle-down innovation.
    AMD, Intel, and NVidia are focusing on products for the 0.00001% (Google, Facebook, Amazon), who don't care what the chips cost.
    Maybe some day, we peons will benefit from these super-mega-ultra-chips, but don't hold your breath.
     
    Icon_Charlie likes this.
  5. Riccochet

    Riccochet Off Topic Award

    Messages:
    21,702
    Joined:
    Apr 11, 2007
    99.9999% of people have no use for this product. Or any other datacenter specific product.
     
  6. You do understand that we do right? And it has ALWAYS been like this. Lots of consumer CPUs etc have been cut down server designs, and even back in the day people were BIOS modding consumer GPUs into their workstation GPUs, as everything else was just about the same. How you think we have not benefited from the tech and designs from the enterprise level, I have no idea. As consumer level stuff has little demand on it, server however has a lot of factors that need to be improved. The people on the [H] are NOT normal consumers, and make up a super small fraction of the consumer base.
     
    Pocatello, Tsumi, Darth Kyrie and 5 others like this.
  7. criccio

    criccio Fully Equipped

    Messages:
    12,294
    Joined:
    Mar 26, 2008
    Additionally, don't we see the trickle down affect in literally ever consumer GPU we have right now?
     
  8. That_Sound_Guy

    That_Sound_Guy 2[H]4U

    Messages:
    2,317
    Joined:
    Apr 29, 2002
    You just mentioned 3 companies that almost everyone on the planet with an internet connection uses. Google for schools that converted to Chrome books and google docs could consume entire data centers of capacity and its only growing. Facebook is the junker in the list but there's no arguing how much impact it has today and how much traffic is rolls. Amazon has AWS, small business lifeline of the current and future. AMD and Intel know who will buy their cutting edge gear by the crate, it's not johnny best buy/tablet computer family.
     
  9. Stoly

    Stoly [H]ardness Supreme

    Messages:
    6,278
    Joined:
    Jul 26, 2005
    45636461_10157971181794942_2780231707654094848_o.jpg?_nc_cat=102&_nc_ht=scontent.fntr6-2.jpg

    Man, that thing looks bigger than a Raspberry Pi
     
  10. pcgeekesq

    pcgeekesq [H]ard|Gawd

    Messages:
    1,403
    Joined:
    Apr 23, 2012
    I don't use any of them, except for having an Android phone, and that's not using Google's servers.

    The point is, don't expect any of what was announced today to produce any improvement in gamer's rigs any time soon.
     
  11. That's it everyone, stop ALL R&D work, its not going to affect gamers anytime soon, you heard it here first people!
     
  12. Sharps97

    Sharps97 Limp Gawd

    Messages:
    414
    Joined:
    May 12, 2000
    False. 7 nm Ryzen will be out on the desktop in the first half of next year with both IPC and clock speed boosts.
     
  13. Spidey329

    Spidey329 [H]ardForum Junkie

    Messages:
    8,677
    Joined:
    Dec 15, 2003
    I'd be OK if this makes it into the consumer side GPU's (like the Nvidia tech did).

    There's a lot of data-scientists having to use Nvidia equipment because AMD is lagging in the ML/AI compute field.
     
  14. Nobu

    Nobu 2[H]4U

    Messages:
    3,138
    Joined:
    Jun 7, 2007
    That 14nm chip is bigger than I expected, but I guess it needs to be in order to handle all the data that's thrown at it by those cores.
     
  15. Nolan7689

    Nolan7689 [H]ard|Gawd

    Messages:
    1,348
    Joined:
    Jun 5, 2015
    Yeah that ring is pretty big.
     
    OnceOver, Anijuz, motqalden and 6 others like this.
  16. notarat

    notarat [H]ard|Gawd

    Messages:
    1,697
    Joined:
    Mar 28, 2010
    It should...They just gave Intel the Raspberry

    Queue Dark Helmet Quote....

    dark-helmet.jpg
     
    boocake, Ranulfo, Riccochet and 3 others like this.
  17. N4CR

    N4CR 2[H]4U

    Messages:
    3,802
    Joined:
    Oct 17, 2011
    That's no moon!
    Weird though, not the active interposer design I expected from them.
     
  18. Criticalhitkoala

    Criticalhitkoala [H]ard|Gawd

    Messages:
    1,670
    Joined:
    Apr 17, 2015
    I feel entitlement in your statement, we've benefitted a shit ton for a long time now from their innovations. Or are you typing on [H] with a Cyrix chip right now?
     
  19. cdabc123

    cdabc123 2[H]4U

    Messages:
    2,257
    Joined:
    Jun 21, 2016
    i love new hardware because in a few years i normally get to play with it. and of course for anyone who didnt know you can by a ibm power 9 workstation that supports pcie 4
     
    spintroniX likes this.
  20. ChadD

    ChadD 2[H]4U

    Messages:
    4,069
    Joined:
    Feb 8, 2016
    Well we will see consumer cards based on this soon enough.

    Having said that. Who do you think will be using all the cloud services powered by this hardware ? Do you not use Google ? I'm sure you are one of the few not on facebook.

    It isn't trickle down... its powering the services real people use.
     
    Vercinaigh likes this.
  21. cdabc123

    cdabc123 2[H]4U

    Messages:
    2,257
    Joined:
    Jun 21, 2016
    you may be suprised how much cheaper this hardware can be obtained for sometimes. for example in a few years when amd releases another gen they often liquidate all stock they have for DIRT cheap if you know the people who get first dibs you can get parts for pretty dang cheap. computers move very fast it may be possible to obtain one of these for a reasonable price withen the year if its a popular product.
     
  22. Cloud computing for games and visuals.

    While this is not big news for most of us, it's where most of the profits are. And AMD being a more profitable & competitive company does benefit us in the end one way or another.
     
    Last edited by a moderator: Nov 6, 2018
  23. idiomatic

    idiomatic [H]Lite

    Messages:
    70
    Joined:
    Jan 12, 2018
    Not like relational databases were built to keep track of the parts of Saturn, or that thread handing and a tonne of other shit, was first invented for the LEM guidance computer, all stellar examples of everyday consumer hardware.

    Its actually a change from the recent exception of consumer gaming hardware pushing basically everything. nVidia crushed it parent SGI, now the circle completes.
     
    Shadowed likes this.
  24. DejaWiz

    DejaWiz Oracle of Unfortunate Truths

    Messages:
    18,850
    Joined:
    Apr 15, 2005
    PC tech has ALWAYS been trickle-down in this manner because the Enterprise/Business segments are the bulk of sales volume and revenue for manufacturers.

    The consumer segment pales in comparison.

    ...the enthusiast portion (most of us here) of that consumer segment is but a tiny sliver.
     
  25. zainali

    zainali Gawd

    Messages:
    763
    Joined:
    Aug 8, 2006
    where is this picture from? i cant in any of the press releases.
     
  26. Red Falcon

    Red Falcon [H]ardForum Junkie

    Messages:
    9,891
    Joined:
    May 7, 2007
    Ha! Give it five years or less and we will be seeing the benefits.
    As the others have stated, it has always been this way, even since the 1990s.

    I remember in the late 1990s, only the military had access to 1000Base-T NICs and gear, and by 2001, it was fairly mainstream in mid-range desktops.
    Even CPUs - VT-x and VT-d were both only offered on high-end Xeon CPUs from Intel, and within two years it was then available on Core 2 CPUs and a bit later even Atom CPUs.

    So, by you saying "any time soon", that really means within a few years, at least going by history of computing, software development, and feature sets within the last 25 years.
     
  27. N4CR

    N4CR 2[H]4U

    Messages:
    3,802
    Joined:
    Oct 17, 2011
    Maybe he doesn't realize the Vega chip in workstation/datacentre currently is the same as the Vega in the desktop gaming version...
     
    Darth Kyrie and Red Falcon like this.
  28. Teenyman45

    Teenyman45 2[H]4U

    Messages:
    2,294
    Joined:
    Nov 29, 2010
    Then the gamer equivalent consumer variant of this gpu and memory stack (like Vega Frontier to Vega 64) would be, what, a 20-30% improvement over Vega 64 and on par with a 2080?
     
  29. Stoly

    Stoly [H]ardness Supreme

    Messages:
    6,278
    Joined:
    Jul 26, 2005
    Tweaktown's facebook page
     
    zainali likes this.
  30. N4CR

    N4CR 2[H]4U

    Messages:
    3,802
    Joined:
    Oct 17, 2011
    This is about what I estimated too.
    They reckoned they wouldn't' release it though.. I think they may at the price the 2080 is..
     
    Deleted member 93354 likes this.
  31. they need to price it like the 2070 if it has any hope. It's missing the RTX extensions.
     
  32. freeloader1969

    freeloader1969 2[H]4U

    Messages:
    2,241
    Joined:
    Dec 13, 2003
    10 years from now you'll get one on Ebay for $100.
     
    Icon_Charlie likes this.
  33. ChadD

    ChadD 2[H]4U

    Messages:
    4,069
    Joined:
    Feb 8, 2016
    Gimick marketing.

    Nvidia has had "RTX" tech since their volta chips. They can call there tensor units "RTX ray tracing McGimmick" all they like it doesn't change anything. Its just 16 bit tensor hardware. Nvidia wanted to find a use for it as it is baked into volta and beyond and finding a consumer use for it is logical. (They are not ever designing a real ground up game chip ever again, on one is NV, AMD, Intel. Money is in the server market.) No one wants to leave 20-30% of a die doing nothing when they use those same chips in consumer parts, or worse fuse it off so people can't use consumer cards over their big expensive server tensor enabled parts. So you have to hand it to NV Marketing they got out there and spun a great story. (I have ZERO doubt the game developers that have been talking about RTX... have also had their games running on beta AMD hardware.)

    From AMDs MI60/50 release;
    "ROCm has been updated to support the TensorFlow framework API v1.11.....
    These low-level instructions implement compute operations all the way from single bit precision to 64-bit floating point. The most beneficial instruction for the acceleration of deep learning training is a float 16 dot product which accumulates into a 32-bit result, maintaining the accuracy of the operation."

    AMD can call their machine learning (tensor flow bits) super compute units or whatever they want. Bottom line its the same type of hardware you find on Nvidias volta and beyond server parts as well. The bottom line is Tensorflow is a google thing... they opensourced it and it became a major standard for use in machine learning. Both NV and AMD have built hardware to accelerate it. (don't let marking fool you... NV isn't doing anything new or unique they didn't invent tensor flow... they are simply building to Googles API like everyone else is / will be, Intel is on the way as well and no doubt there cards will also be aimed at the server market and the tensorflow api)

    When AMD releases their next consumer card. Don't worry all the partial real time ray tracing developers are talking about will work just fine... and most likely via Vulkan for either vendor.
     
    jeffj7, {NG}Fidel, N4CR and 2 others like this.
  34. Tsumi

    Tsumi [H]ardForum Junkie

    Messages:
    13,007
    Joined:
    Mar 18, 2010
    What does soon mean for you, 5 days?

    I'm sure we'll see it on consumer level (probably limited to prosumer level, but consumer level nonetheless) within 5 months of its actual launch.
     
  35. sirmonkey1985

    sirmonkey1985 [H]ard|DCer of the Month - July 2010

    Messages:
    21,361
    Joined:
    Sep 13, 2008
    she showed it on stage and everyone went ape shit trying to take a picture of it.. i'd be surprised of the only ones that actually got a good picture of it were the ones in the front row and the only 2 people there that were using real camera's, lol.
     
  36. Meaker

    Meaker Official representative of powernotebooks.com!

    Messages:
    924
    Joined:
    Jan 10, 2004
    That thing will have NOTHING on the Intel chip in terms of size though, that package is going to be HUGE with 1400mm^2 silicon on it.
     
  37. Hisshadow

    Hisshadow n00b

    Messages:
    55
    Joined:
    Mar 11, 2017
    Right off the bat, if this read as a co-cpu giving you all that it talked about... without video at all.. i'd buy one and put in my pc.


    but if this is a video card, then it either runs on part with a rtx 2080 ti or.. i'm sticking with nvidia.

    so if anyone out there at amd is listening, start making ai coprocessing cards and ill buy one of those.
     
  38. Trimlock

    Trimlock [H]ardForum Junkie

    Messages:
    15,141
    Joined:
    Sep 23, 2005
    These ARE GPU's but not video cards, they are used as accelerators like you wanted. If all you cared about is speed for AI processing then NVidia also offers accelerators for that too (already available).
     
  39. UnrealCpu

    UnrealCpu 2[H]4U

    Messages:
    2,736
    Joined:
    Jun 20, 2003
    funny how AMD is all in the news when the stock dropped more than 40% in two weeks.

    Feel sorry who ever owned this stock
     
  40. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,425
    Joined:
    Nov 19, 2008
    Depends where you purchased the stock. Although with that said its over $20 again.