NVIDIA Tesla K80 “GK210-DUO” Graphics Card Heading Out To The Professional Market – Features Over 2 TFlops of Double Precision Compute
NVIDIA is gearing up for a big launch in its Tesla line of professional graphics cards with the upcoming Tesla K80. In a leak, Videocardz managed to spot references of the Tesla K80 graphics card along with a few GPUs based on the Maxwell architecture however launch dates for the new cards are not known at the moment so we will get on with the specifications.
NVIDIA Tesla K80 "GK210-DUO" Graphics Card Heading Out To The Professional Market
Since the Tesla K80 is a dual-chip graphics card, it will feature two GK210 chips which are codenamed as the "GK210-DUO". The Tesla K80 will essentially be two Tesla K40s on a single board which will result in improved performance and efficency as compared to using two separate Tesla K40 cards. The specifications will include two GK210 chips featuring 5760 CUDA cores (2880 CUDA Cores each), 448 TMUs and 96 ROPs. Judging from the Tesla K40 board, we can expect a memory size as high as 24 GB on the Tesla K80 board since the Tesla K40 featured 12 GB of GDDR5 memory. The memory will be featured along a 384-bit bus interface and will feature a TDP of around 375W, similar to the GeForce GTX Titan Z.
The core and memory clock speeds are currently unknown but NVIDIA's previous Tesla K10 which was based on two GK104 cores featured lower clock speeds compared to its consumer counterpart, the GeForce GTX 690. For cooling, the card can adopt a passive heatsink design if the thermals are under control. The other variants that have been leaked include several upcoming Maxwell based graphics cards for the Quadro and Tesla series that include the recent GM107 based cards including the GM204 based series which will be branded as the Tesla M and Quadro M series.
A slide posted in NVIDIA's HPC oriented Heterogeneous Computing Solutions presentation shows off the Tesla K80 "GK210-DUO" graphics card in a performance slide which provides a peak bandwidth of 500 GB/s and total double precision performance rated at over 2 TFlops. The single precision performance for the card is rated at over 7 TFlops which is impressive from a HPC perspective. The card will retail for a high price of around $7000 US or even more but for users who want to stick something close to the Tesla K80 can just grab the GeForce GTX Titan Z for $3000 US. Even the GeForce GTX Titan Z is getting major price cuts through system builders with a few OEM builders offering a $1000 US price cut on the card dropping it down to $2000 US which still is a huge price for a graphics card but a lot cheaper than its original MSRP. No launch date has so far been confirmed for the Tesla K80 but we expect to see it in the next couple of months.
NVIDIA Tesla K80 Specifications:
|NVIDIA Tesla K80||NVIDIA Tesla K40||NVIDIA Tesla K20X||NVIDIA Tesla K20||NVIDIA Tesla K10|
|CUDA Cores||2880 x 2||2880||2688||2496||1536 x 2|
|Boost Clock||TBC||810MHz, 875MHz||N/A||N/A||N/A|
|Memory Clock||6GHz GDDR5||6GHz GDDR5||5.2GHz GDDR5||5.2GHz GDDR5||5GHz GDDR5|
|Single Precision||~7 TFLOPS||4.29 TFLOPS||3.95 TFLOPS||3.52 TFLOPS||4.58 TFLOPS|
|Double Precision||~2 TFLOPs||1.43 TFLOPS (1/3)||1.31 TFLOPS (1/3)||1.17 TFLOPS (1/3)||0.19 TFLOPS|
|Transistor Count||14.2 Billion||7.1 Billion||7.1 Billion||7.1 Billion||7 Billion|
|Manufacturing Process||TSMC 28nm||TSMC 28nm||TSMC 28nm||TSMC 28nm||TSMC 28nm|
|Architecture||Kepler GK210||Kepler GK110||Kepler GK110||Kepler GK110||Kepler GK104|