Make Home Page
Add to Favorites
1.
Just another concept thats gonna be paper launched... with a load or crap performance...
Just hot air blowing out of their a$$
Just hot air blowing out of their a$$
[Posted by: Anti-AMD Fanboy! | Date: 12/05/07 11:01:07 PM]
Discussion
Discussion on Article:
ATI May Be Considering “Multi-Die GPUs” for R700 Family – Rumours.
Started by: Anti-AMD Fanboy! | Date 12/05/07 11:01:07 PM
Comments: 14 | Last Comment: 12/10/07 09:13:35 AM
Expand all threads | Collapse all threads
[1-7]
2.
I think the R700 will beat nvidia8800GTX performance ; )
[Posted by: Marign | Date: 12/06/07 06:13:42 AM]
3.
This article has some bad grammar. Specifically, the right phrases are “multi-die” or “two-die”. Multi-dice does not work. Just like Multi-core is OK, but multi-cores is wrong."Two physical dice" is correct.
Also, when referring to computar patterns instead of small cubes, the plural form is usually "dies". Still, "multi-dies" would be bad as well.
Also, when referring to computar patterns instead of small cubes, the plural form is usually "dies". Still, "multi-dies" would be bad as well.
[Posted by: Cowdancer | Date: 12/06/07 06:54:37 AM]
4.
The simplist solution for AMD to increase performance would be to just use more TMUs and ROPs!!
That being said I do believe that suck a solution could be possible now only WITH AN EFFICIENT HIGH SPEED INTERCONNECT. I say this because of the way the ATI GPUs have their shaders. If the schedulers can communicate properly it would be effectively like having one GPU with 640 sharers. You would have one scheduler making all the decisions and passing information to the second GPU. Even though that would increase shader power I could see Communications dealing with the texture units to work together as if they were on the same GPU and form one image QUICKLY as the main bottleneck. Oh well I guess we'll see.
That being said I do believe that suck a solution could be possible now only WITH AN EFFICIENT HIGH SPEED INTERCONNECT. I say this because of the way the ATI GPUs have their shaders. If the schedulers can communicate properly it would be effectively like having one GPU with 640 sharers. You would have one scheduler making all the decisions and passing information to the second GPU. Even though that would increase shader power I could see Communications dealing with the texture units to work together as if they were on the same GPU and form one image QUICKLY as the main bottleneck. Oh well I guess we'll see.
[Posted by: Megamanx00 | Date: 12/06/07 10:54:52 AM]
5.
I think you are on to something there Megaman
[Posted by: Bill Gates | Date: 12/06/07 05:03:03 PM]
6.
AMD is like Midas king , but whatever they touch turns into worthless crap. "multi die gpu's " from ATI are exciting like Cyrix news
[Posted by: muzzle8410 | Date: 12/09/07 01:30:17 PM]
7.
Could be interesting. A multi-GPU on a single die.
1> Perhaps they could do a bunch of simple GPUs on a single die - say 16 or 32... And then depending on what the system load is GPUs could be toggled on or off for power savings, or performance.
2> They might be able to create an effective Tile-based GPU architecture. So while ATI and nVidia may have not had the most amazing successes in terms of X-Fire/SLI due to cost to consumer, and complexity of drivers... A low level approach may yield better results.
Hopefully AMD/ATI can rally. Competition is good. AMD's push to 64 bit x86, and multi-core pushed Intel ahead. And we all benefited from faster, more efficient chips.
1> Perhaps they could do a bunch of simple GPUs on a single die - say 16 or 32... And then depending on what the system load is GPUs could be toggled on or off for power savings, or performance.
2> They might be able to create an effective Tile-based GPU architecture. So while ATI and nVidia may have not had the most amazing successes in terms of X-Fire/SLI due to cost to consumer, and complexity of drivers... A low level approach may yield better results.
Hopefully AMD/ATI can rally. Competition is good. AMD's push to 64 bit x86, and multi-core pushed Intel ahead. And we all benefited from faster, more efficient chips.
[Posted by: Woodie | Date: 12/10/07 09:13:35 AM]
[1-7]
You must log in to add comments.
Unfortunately, the old registrations do not work anymore. Please register again. We apologize for the inconvenience.
Our Sponsors
