cs4 is there for r600 and cs4.1 is there for rv770. If you notice, you'll find that the cs 4.x specs are the intersection of the capabilities of r600 and g80 and rv770 and g80 respectively. cs5 is there to standardize on dx11 hw and to amd a chance to get their gpu hw up to scratch in computing.
How will NVidia counter the release of HD5xxx?
- Thread starter Miksu
- Start date
- Status
cs4 is there for r600 and cs4.1 is there for rv770. If you notice, you'll find that the cs 4.x specs are the intersection of the capabilities of r600 and g80 and rv770 and g80 respectively. cs5 is there to standardize on dx11 hw and to amd a chance to get their gpu hw up to scratch in computing.
NVIDIA's could be a strict superset of 4.1 too, it just removes addressing restrictions.
I don't think that's fair, ATI's problem was a lack of high level compiler (Brook was just a little too abstract). The hardware would have been plenty sufficient otherwise. The private writes are nasty, but not an insurmountable issue given the just plain superior FLOPs. Regular algorithms like Matrix multiplication and FFTs are not exactly unimportant ... and for those the lack of random writes is a non issue.
read global write private is a bad idea, atleast for shared memory. Some times, you *need* a faster scatter gather than what global memory can provide.
As far as their compiler goes, I wont say brook was more abstract, but I'd say brook as distributed by amd was (worse than?) useless. Overall, it has been a combination that has kept amd gpgpu efforts from taking off, but I am hoping for a working drivers (if not high quality) for opencl to be launched along with rv8xx.
As far as their compiler goes, I wont say brook was more abstract, but I'd say brook as distributed by amd was (worse than?) useless. Overall, it has been a combination that has kept amd gpgpu efforts from taking off, but I am hoping for a working drivers (if not high quality) for opencl to be launched along with rv8xx.
I don't know what compute-specific stuff they don't support, if any, but to be CS 5.0 compliant you need to have all shader model 5.0 features, not just specifically compute shader related. I think it's a safe bet to say they don't support the GatherCmpAlpha() call for instance.
Same thing with CS4.0 and CS4.1. Nvidia can't do CS4.1 without supporting Gather(), which they currently don't.
PhilTaylor
Newcomer
lack of support for CS 5 by 300 series? if so, interesting.
I wonder how nV could have architected a part that didnt include full support.
Reminiscent of the D3D9 issue, where they refused to sign the DX9 Input Agreement after the fiasco over D3D8 and Shaders. And didnt "see" the 24-bit precision parts of the spec coming, that advantaged ATI in that generation.
But nV did sign the IA after the spanking the 9700 put on the 5900 leaf-blower...so I dont understand how the 300 series could not have full support since nV knows what is in the spec now.
Given Jen-Hsun's comments about 10.1, its a bit more understandable how they decided to not give gate count for what turned into CS 4.x.
Wow, they're going with the "PhysX >>> All!" stragety?Well here's your answer. Pathetic.
That I did not expect.
spacemonkey
Newcomer
"The gaming world has moved to dynamic realism, which depicts actual physical movement more realistically than ever before."
Yep, I always knew visual realism was overrated. Next from NVIDIA - olfactory realism
Yep, I always knew visual realism was overrated. Next from NVIDIA - olfactory realism
Well that was low. In my opinion the best thing for them to do is... nothing Every half-sensible person would know that companies will naturally try to promote their own product and somewhat belittle the competition. So in this day and age where most of the people who will give few hundred bucks for a video card are just a google search away from a ton of reviews. And when they see that NVidia's $100 product isn't actually faster but a lot slower than ATi's $400 product then NV will garner a lot of negative attention toward themselves. On the other hand thanks to their TWIMBP program there are lots of not so tech-savvy users to whom NVidia's brand is way more familiar than ATi and will most likely opt for their cards.
I have to admit this was a lot more structured in my head than it came out in form but the bottom line is that IMO here applies the old saying "there is no bad publicity" especially when you are trying to promote Batman's moving cape against a card which supports a whole new level of DX plus is supposedly faster in some situations than even your dual chip card. So there you have it - for me the best policy for them is to keep it low, business as usual instead of focusing attention towards ATi who are generally not as PR aggressive as NV and by themselves probably won't generate as much interest in their new generation as NV's anti-marketing would at least in the more tech illiterate crowd because the others know where (and for what) to look anyway
Every half-sensible person would know that companies will naturally try to promote their own product and somewhat belittle the competition. So in this day and age where most of the people who will give few hundred bucks for a video card are just a google search away from a ton of reviews. And when they see that NVidia's $100 product isn't actually faster but a lot slower than ATi's $400 product then NV will garner a lot of negative attention toward themselves. On the other hand thanks to their TWIMBP program there are lots of not so tech-savvy users to whom NVidia's brand is way more familiar than ATi and will most likely opt for their cards. I have to admit this was a lot more structured in my head than it came out in form but the bottom line is that IMO here applies the old saying "there is no bad publicity" especially when you are trying to promote Batman's moving cape against a card which supports a whole new level of DX plus is supposedly faster in some situations than even your dual chip card. So there you have it - for me the best policy for them is to keep it low, business as usual instead of focusing attention towards ATi who are generally not as PR aggressive as NV and by themselves probably won't generate as much interest in their new generation as NV's anti-marketing would at least in the more tech illiterate crowd because the others know where (and for what) to look anyway
- Status
Similar threads
