Featured Articles

TSMC: Volume production of 16nm FinFET in 2H 2015

TSMC: Volume production of 16nm FinFET in 2H 2015

TSMC has announced that it will begin volume production of 16nm FinFET products in the second half of 2015, in late…

More...
AMD misses earnings targets, announces layoffs

AMD misses earnings targets, announces layoffs

AMD has missed earnings targets and is planning a substantial job cuts. The company reported quarterly earnings yesterday and the street is…

More...
Did Google botch the Nexus 6 and Nexus 9?

Did Google botch the Nexus 6 and Nexus 9?

As expected, Google has finally released the eagerly awaited Nexus 6 phablet and its first 64-bit device, the Nexus 9 tablet.

More...
Gainward GTX 970 Phantom previewed

Gainward GTX 970 Phantom previewed

Nvidia has released two new graphics cards based on its latest Maxwell GPU architecture. The Geforce GTX 970 and Geforce GTX…

More...
EVGA GTX 970 SC ACX 2.0 reviewed

EVGA GTX 970 SC ACX 2.0 reviewed

Nvidia has released two new graphics cards based on its latest Maxwell GPU architecture. The Geforce GTX 970 and Geforce GTX…

More...
Frontpage Slideshow | Copyright © 2006-2010 orks, a business unit of Nuevvo Webware Ltd.
Thursday, 17 May 2012 10:58

Kepler GK110 detailed by Nvidia

Written by Slobodan Simic

nvidia tesla logo

Total of 2880 cores in 15 SMX modules


Nvidia has officially published the Kepler GK110 whitepaper detailing all the specs that that had us wondering for months. The Kepler based GK110 will first show up as a Tesla K20 graphics card is built for intense computing applications that include data analytics, weather modeling,, computational chemistry and pshysics and aimed at both server and workstation systems.

As far as the specs are concerned, the GK110 features 7.1 billion transistors and promises up to three times the performance per watt when compared to previous Fermi architecture. It packs a total of 2880 cores organized in 15 SMX modules. Each SMX feature 192 single-precision CUDA cores, 64 double-precision units, 32 special function units (SFU) and 32 load/store units. The GK110 GPU features six 64-bit memory controllers, which adds up to a 384-bit memory interface.

nvidia gk110 1

The memory subsystem inside the Kepler based GK110 includes 64KB of on-chip memory for each SMX that can be allocated with a bit more flexibility when compared to the previous Fermi architecture, enabling 32/32KB split between shared memory and L1 cache. In addition to L1 cache, the GK110 Kepler also has 48KB of Read-Only Data cache. In case you lost the number, this adds up to 960KB of shared memory.

As far as the L2 cache is concerned, the Kepler GK110 packs 1536KB of L2 cache, double the amount of L2 cache found in the Fermi architecture and offers up to twice the bandwidth per clock. As expected the GK110 also has ECC memory protection support for all register files, shared memories, L1 and L2 cache and DRAM memory.

In addition to these specs, Nvidia also included a couple of Kepler features that will show up in the GK110 and that include Dynamic Parallelism, Hyper-Q, Grid Management and Nvidia GPUDirect. We already wrote about these features and if you are looking for more details, you can always check out the Nvidia GK110 whitepaper located here.

All in all, the GK110 looks like a pretty impressive GPU, but we honestly doubt that we'll see a Geforce graphics card based on the GK110 anytime soon.

 nvidia gk110 die

Last modified on Thursday, 17 May 2012 11:06
blog comments powered by Disqus

 

Facebook activity

Latest Commented Articles

Recent Comments