Sign in with
Sign up | Sign in

Nvidia Debuts GK110-based 7.1 Billion Transistor Super GPU

By - Source: Tom's Hardware US | B 72 comments

Nvidia's taken the covers off of what it calls the "most complex" commercial chip on the planet.

We know all about Nvidia's GK104 chip, which has most recently been flying through our labs in a dual configuration in the sexy GeForce GTX 690. While that card is the king of gaming (for now), the big daddy of Nvidia Kepler-based GPUs isn't even here yet.

This week at the Nvidia GPU Technology Conference in San Jose, the graphics company took the wraps off of the Kepler-based GK110 GPU that will power the Tesla K20 – a professional-level graphics card for serious business.

Nvidia Tesla K20Nvidia Tesla K20The big reveal at this conference from a hardware standpoint definitely is the GK110, which packs an astonishing 7.1 billion transistors on a 28nm process. It also promises to have all the compute features that some were feeling missing from the GK104. Nvidia CEO Jen-Hsun Huang said at a post-keynote Q&A that the GK110 is "the most complex IC commercially available on planet."

7.1 billion transistors in the GK1107.1 billion transistors in the GK110

In comparison, next in complexity and transistor count is a chip from Xilinx called the Virtex-7 2000T FPGA, which integrates 2 million logic cells and 6.8 billion transistors. To help put that in better perspective, Intel's 10-core Xeon Westmere-EX has 2.6 billion transistors.

The GK110 features 15 SMX units with 192 CUDA cores per unit, which gives a grand total of 2,880 CUDA cores. Nvidia hasn't yet revealed full specifications on the Tesla K20 products yet, but indicated that not all boards will have all 15 SMX units running. Regardless, people can safely expect the use of around at least 2,496 CUDA cores from most Tesla K20 implementations.

The memory bus has been upgraded to 384-bit with six 64-bit controllers in parallel. As for memory capacity itself, Nvidia did not specify. When pushed for an answer, Huang said simply, "Not enough."

To clarify, he added, "As much fast memory as possible behind 384 bits," but no matter what, it will "likely not be enough, because the problems [the K20 is] trying to solve are so huge."

Unfortunately, the GK110 isn't quite finished yet, so we won't be seeing this one until Q4 2012. When it does become available the GK110 GPU is expected to be incorporated into the new Titan supercomputer at the Oak Ridge National Laboratory in Tennessee and the Blue Waters system at the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign.

For those who want a Kepler-based Tesla product today, Nvidia also announced was the GK104-based Tesla K10, which is available immediately. This accelerator board features two GK104 Kepler GPUs that deliver an aggregate performance of 4.58 teraflops of peak single-precision floating point and 320 GB per second memory bandwidth.

The Tesla K10 has already found use in the oil and gas industries, as well as signal and image processing.

  Nvidia Tesla K10Nvidia Tesla K10

"Fermi was a major step forward in computing," said Bill Dally, chief scientist and senior vice president of research at Nvidia. "It established GPU-accelerated computing in the top tier of high performance computing and attracted hundreds of thousands of developers to the GPU computing platform. Kepler will be equally disruptive, establishing GPUs broadly into technical computing, due to their ease of use, broad applicability and efficiency."

As Nvidia CEO Jen-Hsun Huang detailed at his keynote, the Kepler-based Tesla cards feature three new innovations that help add to the edge over Fermi. They are:

  • SMX Streaming Multiprocessor -- The basic building block of every GPU, the SMX streaming multiprocessor was redesigned from the ground up for high performance and energy efficiency. It delivers up to three times more performance per watt than the Fermi streaming multiprocessor, making it possible to build a supercomputer that delivers one petaflop of computing performance in just 10 server racks. SMX's energy efficiency was achieved by increasing its number of CUDA architecture cores by four times, while reducing the clock speed of each core, power-gating parts of the GPU when idle and maximizing the GPU area devoted to parallel-processing cores instead of control logic.
  • Dynamic Parallelism -- This capability enables GPU threads to dynamically spawn new threads, allowing the GPU to adapt dynamically to the data. It greatly simplifies parallel programming, enabling GPU acceleration of a broader set of popular algorithms, such as adaptive mesh refinement, fast multipole methods and multigrid methods.
  • Hyper-Q -- This enables multiple CPU cores to simultaneously use the CUDA architecture cores on a single Kepler GPU. This dramatically increases GPU utilization, slashing CPU idle times and advancing programmability. Hyper-Q is ideal for cluster applications that use MPI.
     

Read more at our liveblog of the Nvidia GTC keynote, and find out what applications Nvidia has planned for gaming in the cloud with GeForce Grid.

Read more from @MarcusYam on Twitter.

Display 72 Comments.
This thread is closed for comments
Top Comments
  • 26 Hide
    tipoo , May 17, 2012 10:18 PM
    fb39ca4BTW where is the fan for that thing? Is the reference board water cooled?


    These things would be sitting in clusters with powerful push/pull fans in the casing. It's not meant to be a standalone card in a PC.
  • 23 Hide
    bardacuda , May 17, 2012 10:17 PM
    2880 cores? Booya! Last rumored specs I saw were 2304 cores. Hope to see the gaming version of this card...even if it's just to drool over and never buy.
  • 22 Hide
    Hellbound , May 17, 2012 11:17 PM
    fb39ca4If it will help oil companies, its a bad thing.


    Oil is everywhere in our lives.. From fuel, to paints, to cosmetics, clothing, skin treatments, lubrication, waxes, asphalt, sulfur.. Not to mention the many practical purposes for petrochemicals..

    Our very livelihood depends on oil...
Other Comments
  • 1 Hide
    fb39ca4 , May 17, 2012 10:13 PM
    BTW where is the fan for that thing? Is the reference board water cooled?
  • 23 Hide
    bardacuda , May 17, 2012 10:17 PM
    2880 cores? Booya! Last rumored specs I saw were 2304 cores. Hope to see the gaming version of this card...even if it's just to drool over and never buy.
  • 26 Hide
    tipoo , May 17, 2012 10:18 PM
    fb39ca4BTW where is the fan for that thing? Is the reference board water cooled?


    These things would be sitting in clusters with powerful push/pull fans in the casing. It's not meant to be a standalone card in a PC.
  • 18 Hide
    tipoo , May 17, 2012 10:23 PM
    tipooThese things would be sitting in clusters with powerful push/pull fans in the casing. It's not meant to be a standalone card in a PC.


    Here's how they look
    http://www-sop.inria.fr/dream/pmwiki/uploads/Public/Public/nvidia-tesla.jpg
  • 15 Hide
    tipoo , May 17, 2012 10:32 PM
    Combat WombatBill Dally is a pretty smart cookie... I think they will put fans on it heh heh heh.This will make a prime Xmas Gift... being released in Q4.. Anyone feel free to buy me one hahahah.



    You guys are aware this $2000+ card would not play any of your games at all, right? Nor would it even work in a PC. I'd buy you one if you gave me 4000 dollars though :p 
  • 18 Hide
    upgrade_1977 , May 17, 2012 10:34 PM
    And yet still no gtx680 - 690 boards anywhere to be seen.
  • 7 Hide
    hajila , May 17, 2012 10:45 PM
    But can it play Crysis?
  • 16 Hide
    Combat Wombat , May 17, 2012 10:47 PM
    hajilaBut can it play Crysis?

    Hahahah, I knew someone would have to say it.
  • 9 Hide
    matt_b , May 17, 2012 11:07 PM
    Very impressive numbers.
    So if the 690 runs at $1000, what on earth will the pricetag for this be (and yield)?
  • 3 Hide
    bardacuda , May 17, 2012 11:09 PM
    matt_bVery impressive numbers.So if the 690 runs at $1000, what on earth will the pricetag for this be (and yield)?


    The Tesla card will no doubt be several thousand dollars. So far there is no word if there will be a more affordable gaming version yet. If there is I would expect it to be almost $1000.
  • 22 Hide
    Hellbound , May 17, 2012 11:17 PM
    fb39ca4If it will help oil companies, its a bad thing.


    Oil is everywhere in our lives.. From fuel, to paints, to cosmetics, clothing, skin treatments, lubrication, waxes, asphalt, sulfur.. Not to mention the many practical purposes for petrochemicals..

    Our very livelihood depends on oil...
  • -1 Hide
    vestibule , May 17, 2012 11:35 PM
    Combat WombatIt's more the bragging rights more than anything.


    Even for that reason, it wouldn't make any sense. It has 2880 cores, the GTX 690 has 3072 cores...

    There's a reason people don't get workstation cards meant for professional use to use as a gaming card lol, this is not any different. I wouldn't be surprised to see the GTX 690 outdo it in gaming, even if just by a little.
  • 13 Hide
    tomfreak , May 17, 2012 11:39 PM
    It is likely to have a Desktop Gaming card with the compute unit disabled. I dont see a reason for Nvidia NOT to launch a Gaming GK110 card. We will see GK110 coming out @ GTX770/780 series, while the GTX680 will likely to have change of shorter reference design like GTX670 and rebrand as GTX760 Ti. Typical Nvidia style.
  • 6 Hide
    sykozis , May 17, 2012 11:50 PM
    TomfreakIt is likely to have a Desktop Gaming card with the compute unit disabled. I dont see a reason for Nvidia NOT to launch a Gaming GK110 card. We will see GK110 coming out @ GTX770/780 series, while the GTX680 will likely to have change of shorter reference design like GTX670 and rebrand as GTX760 Ti. Typical Nvidia style.

    If the rumors are true, it's TDP is 300watts.....which would go against nVidia's "performance per watt" claims..... It would in fact be the single most power hungry consumer graphics card ever produced. Due to the complexity of the processor, it's likely to be more expensive to produce a graphics card than what it cost to produce GTX690....

    vestibuleEven for that reason, it wouldn't make any sense. It has 2880 cores, the GTX 690 has 3072 cores...There's a reason people don't get workstation cards meant for professional use to use as a gaming card lol, this is not any different. I wouldn't be surprised to see the GTX 690 outdo it in gaming, even if just by a little.

    Tesla isn't a workstation card... It's a compute card. It doesn't do graphics.
  • 5 Hide
    razor512 , May 18, 2012 12:10 AM
    Hopefully tomshardware can benchmark that card.

    I want to see how well it can work as a physx card, or provide hardware acceleration for CUDA capable video editors such as adobe premiere pro


    PS tomshardware, if you get the chance, it would be interesting if you can benchmark the GTX 690 with quake 1 :) 
  • 13 Hide
    Anonymous , May 18, 2012 12:47 AM
    You notice the lack of fan but completely miss the lack of video outputs!!!

    We all want to get excited but go look at previous tesla cards to know this ain't going in your desktop pc
Display more comments