AMD does not create custom hardware for developers. You can contact AMD about having them sponsor your academic project.
Academic hardware request:
Thanks for your answer !
The following questions are to estimate time schedules and to evaluate
how many GPU-version tests to be expected:
Q1. What is the latest GPU model for OpenCL developers ?
Q2. Is it the same model as the one AMD's developers use ?
Q3. What is the schedule for the final AMD OpenCL GPU ?
Thanks in advance!
Q1) We recently release the HD6990 I believe, which is a dual-gpu cayman.
Q2) We must develop on all the cards that we support, so everything from a RV710 to a dual-gpu Cayman, there isn't a particular card that we use.
Q3) ?? I don't understand, I don't every expect their to be a 'final' GPU as long as AMD makes graphic cards.
it's pretty obvious to me that the OP is talking about an AMD solution for what NVIDIA has as Tesla card line. I know you have the FireStream I just wonder why you guys don't talk about it (up to 4GB of VRAM with one single GPU)! Plus, it seems that you have been long working to make it integrate very well with opteron processors as co-processor units, specially for float point calculations (Torrenza). I've even seen blades with 2 of these 9370 + 4 opterons. As much as the APUs seems to be a very good choice on this aspect, it will never pair the power of a Opteron+Firesteam couple.
I understand that Radeon line card is perfectly suitable to OpenCL calculation (by the way, the GPU is the same for Radeon and FireStream), but not even evoking FireStream cards is really intriguing. Well, it is not of my business, but still. May I ask why FireStream is not brought to the discussion when one talks about professional solutions? By the way, is there any FireStream based on the HD6900 series GPU?
Definition for Q3:
"final" = all OpenCL 1.1. features implemented and tested by AMD - with one multi-core CPU and at least two(2) GPUs serving OpenCL apps simultaneously
~ full implementation of 2D and 3D image
~ double-precision fully implemented
~ from developer's point of view:
a) AMD's developer version = consumer versions of CPU/GPUs ready for OpenCL 1.1 apps system tests
b) AMD's professional CPU/GPUs available with warranties that all OpenCL 1.1 features work in pilot installations in customer premises
c) AMD's professional CPU/GPUs available with warranties that all OpenCL 1.1. features work in industrial installations
ref: Does AMD support image for AMD Phenom(tm) II X4 955 Processor ?
AMD's answer: AMD support image for AMD Phenom(tm) II X4 955 Processor is only in beta <-- schedule for final version ?
Is CLINFO for HD6990 available in web and where ?
If not, could AMD provide CLINFO about AMD GPU's ?
Someone having that card can post his clInfo. But I think you can get all important information about 6990 here:
The main reason is you are communicating with developers, and not marketing or sales, on this forum and as such we use consumer devices. As you mentioned, the chips are the exact same, AMD doesn't gimp the consumer chips as one of our competitors does. So from OpenCL's perspective, a FireStream device looks almost identical to a Radeon device based on the same chip. The main differences between FireStream and the Radeon line is that AMD builds and tests the FireStream cards, provides direct support for them, length of warranty and amount of VRAM. While important for professional setups, except for VRAM amount, these items really not related to developer support.
More information can be found here: http://www.amd.com/us/products...stream-9370-9350.aspx
Originally posted by: MicahVillmow Laobaruca, The main reason is you are communicating with developers, and not marketing or sales
figure that I thought about it actually xD
We don't give out schedules on our products publicly, so I can't answer these questions. You will need to contact an AMD representative about getting this information privately. Most likely it will require some sort of business relationship and an NDA.