cuorematto

GPU backend and CPU backendtogheter  for a second node

Discussion created by cuorematto on May 12, 2008
Latest reply on May 23, 2008 by michael.chu
Hi Michael like i see i can choose if to enable CPU backend or GPU backend
my question is :
1) Is possible in a new revision of CAL SDK to enable CPU backend and GPU backend togheter ?

( i will build a second node reserved only for stream computing calculation; instead the first node i will run like a Master Node 1 for compilation and other applications + lower stream computing calculation , node 2 the Slave used for only streaming computing and massive calculation CPU and GPU used at 100% for stream computing.
I have configured successfully the MPI program for build my GPU and CPU clusters.
Now im waiting a new revision of CAL SDK wich support MicrosoftVS 2008 and waiting for my new R700 GPU )

2) Can i have a customized Graphic card with 45 Nm process and 800 stream processor for each GPU and 64 TMU for each gpu ?
(Note: the dual gpu version must have 1600 SP and 128 TMU)

3) i have found on internet a website called GPU review (look this link)

http://www.gpureview.com/Radeon-HD-4870-card-564.html

There is the future 4870 card and you can setting up the features ; but i think this specifications are not the real specifications.
Like i see if i increase the GPU core Clock from the 4870 the Calculation in Teraflop will be 1008 GFLOPS=1000000008 Floating Point operations per second with a clock core from 850 Mhz. If i increasing the clock core to 950 Mhz the Teraflop calculation will become 552000 GFLOPS=552000000000 Floating point Operation
per second.
Is this increment showed at this website true ?
Can you show this to the engineers and tell me more about what is true and what is fake ?
Thanks Michael i love stream computing.

Outcomes