Nvidia announces Tesla P100 data-center GPU

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Apr 5, 2016.

  1. Hilbert Hagedoorn

    Hilbert Hagedoorn Don Vito Corleone Staff Member

    Messages:
    36,850
    Likes Received:
    5,937
    GPU:
    AMD | NVIDIA
    Nvidia just announced a HPC data-center module based on Pascal GPU architecture. The Tesla P100 is a daughter add-in board that holds 150B transistors, the GPU itself however will hold 15B transistors...

    Nvidia announces Tesla P100 data-center GPU
     
  2. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,885
    Likes Received:
    2,253
    GPU:
    5700XT+AW@240Hz
    Yep, and he also announced that Tesla P100 has 600mm^2.
    That chances them making it work was "0",... , nada, no way of making it. (to make it look more amazing, that they built something what is impossible to make)

    He is so inconsistent and irrational as ever. Especially his insulting way of talking to audience like to idiots.
    Saying that 384 bit Maxwell has only 384 traces for memory against 4000 ones going to Pascal's HBM2.

    Considering pinout of one GDDR5, he is very wrong, but he makes such statement to make next thing look bigger.
     
    Last edited: Apr 5, 2016
  3. schmidtbag

    schmidtbag Ancient Guru

    Messages:
    4,654
    Likes Received:
    1,481
    GPU:
    HIS R9 290
    Seems there's a typo for the M40 "FP64 CUDA Cores / SM" and "FP64 CUDA Cores / GPU" rows.

    Anyway assuming this is stable, this will be some pretty impressive hardware. But considering the sheer amount of technical hurdles this accomplished, I wouldn't feel too comfortable using this for any mission-critical datacenters. There is too much risk for failure for something this different.


    Anyway it's nice to see nvidia supporting HBM2, but it does get me to wonder how AMD will fare against that. HBM was a pretty big win for AMD but with nvidia having access to it too, Polaris has some catching up to do.
     
  4. cowie

    cowie Ancient Guru

    Messages:
    13,202
    Likes Received:
    291
    GPU:
    GTX
    pretty sure they are in the wild but I am so hardened to these carnivals at this point.

    a lot of cool stuff they got going thuo no matter the size or working order of the die.
    there new mid range gaming cards based on these should plow thou dx12 asyc or not just going by the numbers compared to Maxwell

    well I would think so anyway

    not for nothing but did they really race unmanned race cars or was just "future foward looking" that would be pretty cool like a giant matchbox race track car :)
     

  5. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,885
    Likes Received:
    2,253
    GPU:
    5700XT+AW@240Hz
    I am sure they have those cards. Huang was running around reciting commercials trying to sell every single one and any they will make in future.

    I would not call it racing, but there is video from last year with guy using his AI driven car and he got speeding ticket because AI overlooked something (sign). Or maybe that sign was not even there as there was road construction. Either way, AI just used internal data from map for speed limit in that area.

    Those things will be popular targets for mayhem lovers. Either by laser confusing cameras or by drawing street arts making it look like there is different road somewhere. Or just finding way to disturb device through RF (not even hack necessarily, just something forcing it to reboot).
     
  6. BLEH!

    BLEH! Ancient Guru

    Messages:
    5,918
    Likes Received:
    91
    GPU:
    Sapphire Fury
    This the NV-link thing?
     
  7. cowie

    cowie Ancient Guru

    Messages:
    13,202
    Likes Received:
    291
    GPU:
    GTX
    no I seen that one I was talking about where "the fonz" was showing a nv race car and talked about a ten car race on a track.

    i don't get to hyped up over anything till Hilbert get gets one in his hands, i can see through this campaign trail np.
     
  8. Noisiv

    Noisiv Ancient Guru

    Messages:
    6,664
    Likes Received:
    504
    GPU:
    2070 Super
    "Parallel to the P100 announcement Nvidia is announcing the DGX-1, a deep learning super computer.
    It holds two Xeon processors and a lovely eight Tesla P100 units each holding 16GB of HBM2 memory.
    Priced at only $129,000, but it is considered to be a super-computer."


    170 TFLOPS almost gets you a spot in the latest TOP 500 LIST

    so even technically it should be considered a supercomputer. Within a mere 3U rack.

    [​IMG]
     
  9. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,885
    Likes Received:
    2,253
    GPU:
    5700XT+AW@240Hz
    ^ It surely is impressive. I do not mean performance, that's OK, but Density is impressive. I wonder what is performance per watt there.

    Price is kind of not nice. 170TFLOPS for $129k. As I look at Radeon Pro Duo having 15TFLOPS for $1,5k. If only AMD had it with more vram... Well, next generation.

    Anyway, 20 TFLOPS at 15B transistors... It's good as Fiji has only 8.6 TFLOPS with 8.9B Transistors. Only question is how that translates to Consumer type of Pascal and at what clock P100 ticks.
     
  10. cowie

    cowie Ancient Guru

    Messages:
    13,202
    Likes Received:
    291
    GPU:
    GTX
    that's good and all maybe finding a cure for cancer but the thing cant even do 3dvantage at least not the 3d part !@#$ it cant even do crisis

    all kidding aside 10 get you into the top 25 ish thats just nuts
     
    Last edited: Apr 6, 2016

  11. Ieldra

    Ieldra Banned

    Messages:
    3,490
    Likes Received:
    0
    GPU:
    GTX 980Ti G1 1500/8000
    Dude are you drunk ?!

    170TFLOPS of FP16, 21 Tflops of FP16 per chip

    If you're gonna compare with Fiji compare FP32 no ? That's 8.6Tflops vs 10.6Tflops stock
     
  12. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,885
    Likes Received:
    2,253
    GPU:
    5700XT+AW@240Hz
    Yes, but those ASICs are not that much better than last generation to prevent having same place with 60~100 GPUs. In other words there is reason why those TOP super computers use CPUs.

    Nuts would be if P100 could outperform Xeon based processors in those operations which last generation apparently sucked at. And doing it at lower price and power consumption.

    Because those more modern SCs around position 25 use between 40000 and 80000 CPU cores, and suck around 1~2MW.
    And having 10 times DGX-1 would suck only 0.032MW.
    I'll just blame it on Hilbert for writing it up into one very long sentence. Cheers. Any more hate? And thanks for making me less impressed.
     
    Last edited: Apr 6, 2016
  13. Noisiv

    Noisiv Ancient Guru

    Messages:
    6,664
    Likes Received:
    504
    GPU:
    2070 Super
    it doesn't

    90% this is HPC only due to existance of NVLink and huge investment into DP.
    x2 more transistors than GM200 for mere 16% more Cuda Cores says this chip does not care about SP - at all.

    As far as I know AMD is not even a player in deep learning,
    and as for Intel - looks like their future Knights Landing is gonna get thrashed in compute raw specs:
    3 TFLOPS FP64 vs 5,10,20 TFLOPS (FP64,FP32,FP16)

    300W might look worrying, but then again why not:
    http://ark.intel.com/products/series/75809/Intel-Xeon-Phi-Coprocessor-7100-Series#@Server
     
  14. Shadowdane

    Shadowdane Maha Guru

    Messages:
    1,306
    Likes Received:
    12
    GPU:
    MSI RTX2080Ti
    Yes this is NVLink... that is only going in super computers and server boards.


    Consumer cards will still be PCI Express, I don't see that changing anytime soon.
     
  15. Noisiv

    Noisiv Ancient Guru

    Messages:
    6,664
    Likes Received:
    504
    GPU:
    2070 Super
    not quite. PCI express is still used, for example in DGX1

    "While NVLink primarily focuses on connecting multiple NVIDIA Pascal GP100 GPUs together
    it can also connect Pascal GP100 GPUs with IBM Power CPUs with NVLink support.
    "

    [​IMG]
     

  16. Ieldra

    Ieldra Banned

    Messages:
    3,490
    Likes Received:
    0
    GPU:
    GTX 980Ti G1 1500/8000
    Haha there was no hate there at all, genuinely thought you were drunk :p awesome hardware
     
  17. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,005
    Likes Received:
    139
    GPU:
    Sapphire 7970 Quadrobake
    Trusting the NVIDIA CEO 100% is akin to a Darwin Award. It's like believing AMD about Crossfire support.
     
  18. Ieldra

    Ieldra Banned

    Messages:
    3,490
    Likes Received:
    0
    GPU:
    GTX 980Ti G1 1500/8000
    Trust issues aside, I love watching his talks, he's so awkward

    ONE HUNDRED AND FIFTY BILLION TRANSISTORS
     
  19. Musouka

    Musouka Master Guru

    Messages:
    326
    Likes Received:
    0
    GPU:
    ZOTAC GTX970
    Haha, did he really say that? :banana:

    I guess it's excusable since Chinese (and the languages it influenced like Japanese and Taiwanese) have a 10,000-base for big numbers instead of the 1,000-base used in many other languages including English.

    So in his mind he might have been thinking 150 (100 million unit) but only half-converted it to its equivalent in English :infinity:
     
    Last edited: Apr 6, 2016
  20. Ieldra

    Ieldra Banned

    Messages:
    3,490
    Likes Received:
    0
    GPU:
    GTX 980Ti G1 1500/8000
    He was talking about the transistors on the whole board, not just the gpu die :p

    It was very confusing when he said it though, I did a double take... Thought 'either this guy is coked up or I heard wrong' lol
     

Share This Page