nVidia\'s GT300 specifications revealed - it\'s a cGPU!

苯苯小哥

4/22/2009 by: Theo Valich - Get more from this author
http://www.brightsideofnews.com/news/2009/4/22/nvidias-gt300-specifications-revealed---its-a-cgpu!.aspx

Hardware-Infos:
http://www.hardware-infos.com/news.php?news=2904

Overthe past six months, we heard different bits'n'pieces of informationwhen it comes to GT300, nVidia's next-gen part. We decided to staysilent until we have information confirmed from multiple sources, andnow we feel more confident to disclose what is cooking in Santa Clara,India, China and other nV sites around the world.

GT300 isn't the architecture that was envisioned by nVidia's Chief Architect, former Stanford professor Bill Dally, but this architecture will give you a pretty good idea why Bill told Intel to take a hike when the larger chip giant from Santa Clara offered him a job on the Larrabee project.

Thanks to Hardware-Infos, we managed to complete the puzzle what nVidia plans to bring to market in couple of months from now.

What is GT300?

Even though it shares the same first two letters with GT200architecture [GeForce Tesla], GT300 is the first truly new architecturesince SIMD [Single-Instruction Multiple Data] units first appeared in graphical processors.
GT300 architecture groups processing cores in sets of 32 - up from24 in GT200 architecture. But the difference between the two is thatGT300 parts ways with the SIMD architecture that dominate the GPUarchitecture of today. GT300 Cores rely on MIMD-similar functions [Multiple-Instruction Multiple Data]- all the units work in MPMD mode, executing simple and complex shaderand computing operations on-the-go. We're not exactly sure should wecontinue to use the word "shader processor" or "shader core" as theseunits are now almost on equal terms as FPUs inside latest AMD and IntelCPUs.

GT300 itself packs 16 groups with 32 cores - yes, we're talkingabout 512 cores for the high-end part. This number itself raises thecomputing power of GT300 by more than 2x when compared to the GT200core. Before the chip tapes-out, there is no way anybody can predictworking clocks, but if the clocks remain the same as on GT200, we wouldhave over double the amount of computing power.
If for instance, nVidia gets a 2 GHz clock for the 512 MIMD cores, weare talking about no less than 3TFLOPS with Single-Precision. Dualprecision is highly-dependant on how efficient the MIMD-like units willbe, but you can count on 6-15x improvement over GT200.

This is not the only change - cluster organization is no longerstatic. The Scratch Cache is much more granular and allows for largerinteractivity between the cores inside the cluster. GPGPU e.g. GPUComputing applications should really benefit from this architecturalchoice. When it comes to gaming, the question is obviously - how goodcan GT300 be? Please do bear in mind that this 32-core cluster will beused in next-generation Tegra, Tesla, GeForce and Quadro cards.
This architectural change should result in dramatic increase inDual-Precision performance, and if GT300 packs enough registers -performance of both Single-Precision and Dual-Precision data mightsurprise all the players in the industry. Given the timeline whennVidia begun work on GT300, it looks to us like GT200 architecture wasa test for real things coming in 2009.

Just like the CPU, GT300 gives direct hardware access [HAL] for CUDA3.0, DirectX 11, OpenGL 3.1 and OpenCL. You can also do directprogramming on the GPU, but we're not exactly sure would development ofsuch a solution that be financially feasible. But the point in questionis that now you can do it. It looks like Tim Sweeney's prophecy is slowly, but certainly - coming to life.
——————————————————————
日前，Hardware-infos得到了GT300的更多信息。
　　据透露，GT300将会采用全新架构，相比GT200每组包含的流处理器数量将会由24增加到32，这样流处理器的总数将史无前例地激增到512个；渲染集群数量将由10个增加到16个，每个渲染集群包含8个TMUs，所以GT300将具备128个TMUs，而且还会采用类似MIMD单元的MPMD单元。
　　由于产品还未Taped-Out，所以目前GT300的频率还不得而知，初步按照2GHz的shader频率来算，其单精度浮点运算能力至少会超过3TFLOPs，双精度运算能力则高度依靠MPMD单元，估计也会比GT200的有6到15倍的提升。
　　此外，GT300将采用动态集群组织，并增加更大的缓冲区，还将支持CUDA 3.0、DirectX 11、OpenGL 3.1以及OpenCL等技术标准。GT300的首款产品将会命名为GTX380，如无意外将会在今年年底发布。GT300必将成为NVIDIA GPU史上的又一个里程碑。

huangshidi

什么时候上市？

苯苯小哥

不知道多大个 弄个大机箱是对的

NewCastle

只是这个网站自己的猜测而已

leesd

功耗大户准备来了

nfsking2

呵呵，这篇文章的英文原文写得很有趣，特别是“GT300 isn't the architecture that was envisioned by nVidia's Chief Architect, former Stanford professor Bill Dally, but this architecture will give you a pretty good idea why Bill told Intel to take a hike when the larger chip giant from Santa Clara offered him a job on the Larrabee project”这段

Asuka

比较假

eDRAM

比较假
Asuka 发表于 2009-4-23 09:21

假在哪？你说说看

无敌二阶堂

学习。。。听高人评判下。。。

droganmaster

希望支持D11 否则就没意义了

Edison

苯苯小哥

We're not exactly sure should wecontinue to use the word "shader processor" or "shader core" as theseunits are now almost on equal terms as FPUs inside latest AMD and IntelCPUs.

koppie

这么多核心得多少晶体管啊

假在哪？你说说看
eDRAM 发表于 2009-4-23 10:44

ncxrt

夏卡长度可千万别超过GTX285啊.否则换卡就得换机箱了~

哈雷彗星

新技术在家用PC上的实际应用中大部分都用不到，玩家却还要为这些东西买单，象征意义大于实际意义，GT300能够真正普及恐怕也要明年年底见了

soga

考虑GT300时代，把我的G80 GTS 320M 换掉。
如果没啥显卡危机大作的话，还是可以继续等。

rickerlian

x87再次来袭，nv也就是做intel小弟的命

earcandy

先换电源再说

Asuka

假在哪？你说说看
eDRAM 发表于 2009-4-23 10:44

你看看各厂的半导体制程roadmap，就知道这篇文章假在哪里了

nvgf8800

新技术在家用PC上的实际应用中大部分都用不到，玩家却还要为这些东西买单，象征意义大于实际意义，GT300能够真正普及恐怕也要明年年底见了
哈雷彗星 发表于 2009-4-23 11:52

没错
卖的就是哪些花钱yy的