Alleged NVIDIA GK104 Specifications

[Lonbjerg]

Stumbled over this:

Old roadmap - PEAK performance.

New roadmap - SUSTAINED performance:

And "Kepler is even higher in sustained than the previous peak number...and so is "Fermi" lol.

I wonder if Intel's "Knights Corner" has something to do with that...since Intel's gave it's performance in sustained, not peak, too...

 

[Madcatatlas]

roadmaps are relative pal


Given the acrimonious nature of this conversation, and the history of inflammatory tit-for-tat that both you and Lonbjerg have, referring to one as "pal" is no less baiting/inflammatory as referring to one as "chump" in this context.

You both know what you were doing...and judging by the volume of reported posts your incendiary interactions precipitated here I can frankly say the community is tiring of it.

None of us care what you two think of each other, but if the baiting and inflammatory posting style does not immediately cease then the community is going to compel its moderators to step in and manually force it to cease ala vacations, etc.

That should be sufficiently clear.

Administrator Idontcare

 

[Lonbjerg]

roadmaps are relative pal

But peak or sustained isn't chump...


Given the acrimonious nature of this conversation, and the history of inflammatory tit-for-tat that both you and Madcatatlas have, referring to one as "chump " is no less baiting/inflammatory as referring to one as "pal" in this context.

You both know what you were doing...and judging by the volume of reported posts your incendiary interactions precipitated here I can frankly say the community is tiring of it.

None of us care what you two think of each other, but if the baiting and inflammatory posting style does not immediately cease then the community is going to compel its moderators to step in and manually force it to cease ala vacations, etc.

That should be sufficiently clear.

Administrator Idontcare

 

[Madcatatlas]

i think you ment to say and instead of or pal

 

[Madcatatlas]

and "are not" ofcourse, dont want to give you that for free

 

[Lonbjerg]

i think you ment to say and instead of or pal

No.
Peak is a different metric than sustained chump.
Both NVIDIA and AMD (Hell even consoles are using peak number for PR) used to use the peak perfornance...because peak was a higher number than sustained.
But you never get peak numbers in the real world.
And that Intel broke that little PR "rule"...it's fun to watch NVI*DA follow suit...now we only need AMD to follow suit.
So stop twistning my words...chump.

I meant what I wrote cobber...


See your mod comment above, enough with the "chump" talk.

Admin Idontcare

 

[Lonbjerg]

This on is for you, find someone else to get your hard on with:
http://vimeo.com/15412319

That was was free from me...get the point...or should I "spell" it our for you? :thumbsdown:

 

[Olikan]

Since the days of NVIDIA's very first DirectX 10 GPUs, NVIDIA has been using different clock domains for the shaders and the rest of the GPU (geometry domain). Over the past few generations, the shader clock has been set 2x the geometry domain (the rest of the GPU). 3DCenter.org has learned that with the next-generation "Kepler" family of GPUs, NVIDIA will do away with this "Hotclock" principle. The heavy number-crunching parts of the GPU, the CUDA cores, will run at the same clock-speed as the rest of the GPU.

It is also learned that NVIDIA will have higher core speeds overall. The clock speed of the GK104, for example, is expected to be set "well above 1 GHz", yielding compute power "clearly over 2 TFLOPs" (3DCenter's words). It looks like NVIDIA too will have some significant architectural changes up its sleeve with Kepler.

http://www.techpowerup.com/157039/NVIDIA-Kepler-To-Do-Away-with-Hotclocks.html

 

[tviceman]

Since the days of NVIDIA's very first DirectX 10 GPUs, NVIDIA has been using different clock domains for the shaders and the rest of the GPU (geometry domain). Over the past few generations, the shader clock has been set 2x the geometry domain (the rest of the GPU). 3DCenter.org has learned that with the next-generation "Kepler" family of GPUs, NVIDIA will do away with this "Hotclock" principle. The heavy number-crunching parts of the GPU, the CUDA cores, will run at the same clock-speed as the rest of the GPU.

It is also learned that NVIDIA will have higher core speeds overall. The clock speed of the GK104, for example, is expected to be set "well above 1 GHz", yielding compute power "clearly over 2 TFLOPs" (3DCenter's words). It looks like NVIDIA too will have some significant architectural changes up its sleeve with Kepler.

http://www.techpowerup.com/157039/NVIDIA-Kepler-To-Do-Away-with-Hotclocks.html

I wonder how true this is, as shader clocks were 2x that of core clocks. Making them the same thing sounds like a huge compromise on the surface, but obviously there is more than that going on under the hood. So it seems that Kepler will have gone under quite an overhaul.

 

[Olikan]

more speculation:

GK100 features 1500 CUDA Cores, GDDR 5 memory with memory interface of 512-bit, the performance is close to the dual-core GeForce GTX 590, it’s scheduled to debut in Q2 2012. While GK104 features 1000 CUDA Cores, GDDR 5 memory with memory interface of 256-bit/384-bit as well as core/Shader frequency of over 1GHz, the performance is likely to be slightly higher than that of GeForce GTX 580, it’s expected to arrive in January or February 2012.

http://en.expreview.com/2011/12/21/nvidia-kepler-gk100-and-gk104-specifications-spotted/19764.html