AMD as a platform?

L

LaChupacabra

Limp Gawd
Joined
Aug 29, 2007
Messages
251
Hi, I wasn't sure where to post this (it's about AMD as a platform) but this seemed as good a place as any.

Those leaked benchmarks from the Enquirer (reputation of source aside) that said a phenom with crossfire 2900's scored 30k in 3DMark06 got me thinking. Would it be possible for AMD to off-load some of their CPU workload onto the GPU or the other way around?

This (http://www.nordichardware.com/news,6849.html) article posted on NordicHardware makes it look like AMD is focused on making their AMD products work well together and it would seem to be a logical first step towards their on-die GPU. Whose to say that GPU needs to be limited to processing graphics? Nvidia certianly doesn't think so. Is a modern GPU capable of picking up some slack for the CPU?
 
LaChupacabra said:
Hi, I wasn't sure where to post this (it's about AMD as a platform) but this seemed as good a place as any.

Those leaked benchmarks from the Enquirer (reputation of source aside) that said a phenom with crossfire 2900's scored 30k in 3DMark06 got me thinking. Would it be possible for AMD to off-load some of their CPU workload onto the GPU or the other way around?

This (http://www.nordichardware.com/news,6849.html) article posted on NordicHardware makes it look like AMD is focused on making their AMD products work well together and it would seem to be a logical first step towards their on-die GPU. Whose to say that GPU needs to be limited to processing graphics? Nvidia certianly doesn't think so. Is a modern GPU capable of picking up some slack for the CPU?
Click to expand...

http://ati.amd.com/products/streamprocessor/index.html
 
http://ati.amd.com/products/streamprocessor/index.html

(don't know how to do that fancy "quote" thing)


Yea, like the folding@home gpu client. What I was trying to say was if AMD has built some sort of new extensions into their processors chipsets and gpus that automatically off-load some of the work. So, isntead of running a F@H client on your CPU and a client for your GPU, if it would be possible for hardware level off-loading to the GPU with certain instructions without having to code it.
 
LaChupacabra said:
Whose to say that GPU needs to be limited to processing graphics? Nvidia certianly doesn't think so. Is a modern GPU capable of picking up some slack for the CPU?
Click to expand...
We have already seen the beginnings of non-traditional applications for GPUs when Stanford released a F@H client for ATI video cards. More non-video applications are sure to come in the future.

Edit: Hadn't seen your post above before I replied.
 
LaChupacabra said:
http://ati.amd.com/products/streamprocessor/index.html

(don't know how to do that fancy "quote" thing)


Yea, like the folding@home gpu client. What I was trying to say was if AMD has built some sort of new extensions into their processors chipsets and gpus that automatically off-load some of the work. So, isntead of running a F@H client on your CPU and a client for your GPU, if it would be possible for hardware level off-loading to the GPU with certain instructions without having to code it.
Click to expand...

The answer to that is stage3 fusion. When stage3 fusion is released it will be a common CPU and GPU architectire. It will be both in one design. The architecture will be pretty amazing, when it gets done.
 
duby229 said:
The answer to that is stage3 fusion. When stage3 fusion is released it will be a common CPU and GPU architectire. It will be both in one design. The architecture will be pretty amazing, when it gets done.
Click to expand...


where is the info on this stage 3 fusion =p I've never heard of stages for it
 
Digital Viper-X- said:
oh si i see, he was referring to the step 3 of plan ^^ i thought stage 3 fusion was another product haha
Click to expand...

Actually, all three are separate products. The first will mainly show economy improvements (both $$$ and power) and will be released first, then the second will introduce some integration in terms of performance, and the third will show the most efficient integration with the most exploitation of GPU performance.

So in other words, the "stage 3" is their end goal with fusion, but we will see products along the way showcasing the other two stages of integration.
 
You must log in or register to reply here.
Back
Top