Folding Forum

Mod note: This and related posts are off-topic to the discussion of the blog post and have been placed in their own topic.

ATI doesn't really exist in folding, so I can understand Vijay not mentioning the GPU manufacturer regarding new project. If Vijay wants more GPUs folding, then PG needs to get their hands out of their arses and kick AMDs butt so them faulks would help out with their cards.
Currently only nvidia has viable GPU folding 'scheme' so to say, and it is significant contribution, but market leader AMD's 'scheme' is complete failure and everyone is in the dark what is going on with their cards and optimisations. Last update on fahcore_16 and new WU was announced like 8 months ago and was a step backwards (even though it was said that performance increased, which it did not) and we (beta testers) are yet to hear(after 8 months) an answer what happened.
Also, where abouts is that newly released WU? I cannot find any threads related to it in beta section.

Jesse_V wrote:...
ATI wasn't specifically mentioned, so I think that can be assumed AFAIK.

NV wasn't spcifically mentioned either!

7im wrote:

Jesse_V wrote:...
ATI wasn't specifically mentioned, so I think that can be assumed AFAIK.

NV wasn't spcifically mentioned either!

but we can with great certainty assume that he meant nvidia only as project number is similar to what nvidia gpus are currently receiving and that AMD gpu folding isn't really a factor

We who? How certain can you be with a completely new type of project and new benchmarking test?

AMD is ALWAYS a factor.

Don't assume anything, unless you like to be wrong a lot.

Unfortunately AMD in CPU market not really a factor anymore with what is going on inside the company. And AMD GPUs as folders are definitely a non factor. Whoever thinks that AMD GPU folding in current state is a factor is completely delusional. I will gladly admit that I am wrong when someone from PG will tell me for sure that I am wrong, but I think it is safe to say that won't really happen, will it?

Besides why not release one project for just one manufacturer who actually has a base of folders who would actually give any info or suggestions related to the points. Then when beta testers are satisfied with formulas and point amount and the calculating procedure, then PG can release new projects(renames) for every manufacturer

well Muziqaz,

I will agree with you that the support given to AMD GPU is not very good.
One need only look at the Gromacs posting for hire calling for Cuda Programers, and no mention of open CL.

However, looking at the STATS tells a different story.

2131 FLOPS is currently the largest.

Sure NVDA generates 2116 with less than half the cards, but AMD is still the largest total.

10x the total of all WiNDOWS computers is nothing to sneeze at.

I don't know how you can say 2/5 of the total is a non factor.

I share your frustration that optimized cores are delayed for years...but this move only validates the extreme importance of the GPU and its future to the project.

Something you and I have argued against the point "reducers" for many years.

First post on the new forum for me, calling for points to reflect power of the GPU.

viewtopic.php?f=10&t=447#p3794

Five years later....better late than never.

I don't really care about the stats. The fact is - AMD folding development is stagnant at best. How many of that total are new AMD cards of 5,6 and 7 series? If my electricity wasn't free, I would tell Stanford to take those ATI GPU WUs and shove them where the sun don't shine. It is the same a receiving Core_11 WUs on 5 or 6 series GPUs, just absolute waste of time and electricity.
Too bad about Gromacs, opencl has huge potential, but I guess vendor locking and instruction disables on competing products is more fun than open standard

I was not expecting this discussion on this thread. Just thought I'd note that the number of active GPUs (red) has remained roughly the same for years now. I'm curious to see if they'll be a rise due to this. Points matter.

How many of that total are new AMD cards of 5,6 and 7 series?

well, 6 series actually is the one working best. Open cl core actually works on 6850, and 6870.

It is the 7 series where the wheels came off again. (7950 and 7970)

But I was looking at Steam the other day, and AMD 7950 and 7970 make up something like .06 percent of the installed base.
There are a huge number of 6850, 6870, 7850,7870, and 7770 cards out there that can FOLD.

Like I said, I share your frustration, and won't FOLD with my 7970 until it does some efficient work.....but we do need to keep things in perspective.

Points matter.

My second mantra after my sig.

Nathan_P wrote:I'm assuming that this will not just be restricted to nvidia GPU's?

Why are some of you who responded to Nathan_P assuming that the announcement is restricted to either NV or to AMD(ATI) when neither was mentioned? Consider the possibility that it might be an OpenCL project which is released to both. Just because (AFAIK) they've never done that before doesn't tell us anything. FAH has never used QRB on GPUs before either.

Real opencl project for both manufacturers? No that would be a shocker. For one, nvidia is too tied up with CUDA, and for them nothing else matters, and second, if there was anything remotely close to true open cl project taking advantage of AMD gpus to fullest - we would get separate announcement from PG.

mdk777, losing 2.5k ppd with last (8month old) update can be called 'works best'? I would call it - it works, and that's all.

mdk777, losing 2.5k ppd with last (8month old) update can be called 'works best'? I would call it - it works, and that's all.

Well my 6850 went up.(on the initial open cl)

Haven't run it recently, so I guess it could have changed.

And I suspect that it will be CUDA based (the announcement)

But as I mentioned in another post : NVIDIA has sold 350,000,000 CUDA capable cards and only 6000 of them are currently FOLDING.

6 thousand out of 350 million is a statistical error. Glad to see they are taking steps to capture some of this potential even if it isn't with the most cost effective cards on the market.

As I understand it, the CUDA language is proprietary and quite similar to OpenCL in many respects, but with just enough extra features to keep the customers using it.

As far as the HD8 series, it's in the same boat as NV's Kepler. There has been no progress at optimizing either one since they came out. I simply see nothing in the announcement implying any answer to the question of which will be supported, hence my suggestion that it might be OpenCL only. NVidia's CUDA devices also support OpenCL, and if the Fahcore supporting Opencl works with AMD, it probably will work equally well with NV. Why should FAH need two different Fahcores for GPUs?