![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
![[Spectre]](/data/avatars/m/89/89728.jpg?1455675927){kind=avatar}
pxc said:
Yeah, ATI shows a 30% improvement in branching performance on one of their slides: 14 cycle "branch A" executes in 14 cycles on X1K vs 20 cycles on "green" (vs G70 on other slides). GROMACS on GPU sticks out if you look at relative performance on GPUBench, HMM and the BLAS benchmarks (R520 vs G70, both are in the same ballpark). So yes, it is unfortunate that Stanford worked so closely with ATI on GROMACS on GPU because performance is not in the same range as other non-ATI assisted GPU applications.ElMoIsEviL said:
Again that makes little sense from the samples on gpgpu.com and nvidia has been working with Havok from the beginning, so I doubt this is Havok FX performance. IMO that 14x difference is using ATI's own routines (optimized for X1K in ways that cripple it for other architectures) and not Havok FX and others (Cry Engine 2, etc). There is absolutely no configurations (software, drivers or even cards used) given, so ATI's marketing materials on the subject are very suspect.ideally, the gpu is awesome for folding, since it has such an absurd amount of power.. so yes, for that very reasonAtomicMoose said:
If your machine is on all the time anyway, you lose nothing....Heat is a moot point. Proper cooling and you're fine.Demon_of_The_Fall said:
Sorry about that. I just wanted to bring up that it's likely only going to be an option for X1000 series cards and why. I'm sure people with GPUs of any flavor will want to try. Unfortunately decisions were made that will exclude over 1/2 the potential field.AtomicMoose said:
I see your point, and it is valid. Feel free to start another thread to discuss what you brought up....maybe enough discussion will prompt Stanford to have one foot on each side of the fence, as it were.pxc said:
It may even be more then 14x. You don't actually need benchmarks to tell you it will be significantly faster. Think of it this way, you have two cards.. both are fast.. one has non slip rain tires while the other has summer tires. Under normal circumtances both would be somewhat equal. But in this case they both want to race in the rain. Who do you think will end up making it to the finish line first?pxc said:Yeah, ATI shows a 30% improvement in branching performance on one of their slides: 14 cycle "branch A" executes in 14 cycles on X1K vs 20 cycles on "green" (vs G70 on other slides). GROMACS on GPU sticks out if you look at relative performance on GPUBench, HMM and the BLAS benchmarks (R520 vs G70, both are in the same ballpark). So yes, it is unfortunate that Stanford worked so closely with ATI on GROMACS on GPU because performance is not in the same range as other non-ATI assisted GPU applications.
And do you have a non-ATI source for physics benchmarks? ATI is claiming a 14x difference between the X1900 and 7800. The 7900 is even "slower" than the 7800 according to ATI's slides.
No decision was made to exclude anyone. The 7x00 series is a DX9C level accelerator. It was not mean't for Ultrathreading purposes or anything more then Graphics acceleration. The x1K series were built as a step towards ATi's goal of a full Unified shader multipurpose accelerator for more then Graphics acceleration. ATi's is following Microsoft's lead. nVIDIA believes in being cautious and as such have not been working on a unified architecture. As a result they may end up slightly behind the curve (they are right now but out of sheer brute force, reminicent of the x-3Dfx years, they're able to remain competitive).
And I'm sure ATi doesn't plan on placing more resources towards fixing there Linux drivers until Linux stops sucking and starts workingunhappy_mage said:No. Because I don't have a high-end graphics card, and don't plan on buying ATI until their Linux drivers stop sucking and start working.
ElMoIsEviL said:And I'm sure ATi doesn't plan on placing more resources towards fixing there Linux drivers until Linux stops sucking and starts working
Oh and starts gaining marketshare, starts getting more then a handful of badly programmed games.. and ummm.. starts actually being useable by the mainstream population.
MrNasty said:Wow - that's the most arrogant statement I've ever heard. Linux's share of the computer pie is bigger than ATi's in terms of installed user base.
MooseEdit: Please, no name calling. edit kk - I hapen to like linux is all, and you've got to admit it was a funny comment
) and how to fix that.I would if I could...BriSleep said:
AtomicMoose said:
Im gonna disagree. My machine is like an atomic heater, and its worse when im gaming. No friggin way im gonna let heat build up in my room like that - I cant afford to keep the AC where i like it even when its not running...AtomicMoose said:
AtomicMoose said:
lithium726 said:btw, i think you were looking for "mute" point, moot means its arguable