By: David Kanter (dkanter.delete@this.realworldtech.com), July 25, 2012 11:25 am
Room: Moderated Discussions
none (none.delete@this.none.com) on July 25, 2012 3:18 am wrote:
> Very interesting article, thanks!
>
> I have a little comment: in the text, you
> take into account the area added by IVB GPU; but GPUs also have graphics
> dedicated units that are completely useless for GPGPU. Is it difficult to put
> them apart?
I'm not sure I parsed your question correctly, but let me try and answer.
1. I can't tell the graphics-specific portions of a GPU from a die photo. Texture units are hard to see, and in some cases they cannot really be removed. ROPs contain the L2 cache which is necessary. In contrast, I have a high resolution die photo that clearly shows the GPU area of IVB.
2. As I mentioned above, in some cases you cannot remove the graphics functionality without rendering the GPU useless.
David
> Very interesting article, thanks!
>
> I have a little comment: in the text, you
> take into account the area added by IVB GPU; but GPUs also have graphics
> dedicated units that are completely useless for GPGPU. Is it difficult to put
> them apart?
I'm not sure I parsed your question correctly, but let me try and answer.
1. I can't tell the graphics-specific portions of a GPU from a die photo. Texture units are hard to see, and in some cases they cannot really be removed. ROPs contain the L2 cache which is necessary. In contrast, I have a high resolution die photo that clearly shows the GPU area of IVB.
2. As I mentioned above, in some cases you cannot remove the graphics functionality without rendering the GPU useless.
David
