Hmm, this should be interesting. Hope it ends up being useful as a benchmarking tool for gpu's.

so they arent making it for all gpu's just nvidia cuda gpu's ? laim

Rebo&Zootyso they arent making it for all gpu's just nvidia cuda gpu's ? laim

fugger @XS says they are keeping him out of the loop on that one. There is a link in the source to XS' thread about it!

I am very interested to see how fast my 8600gts calculates 1m of PI...

It would be interesting to see how my 2900pro compares to a 8800gt at pi.

Oh Wait! ATI doesn't support CUDA! :shadedshu

flashstarIt would be interesting to see how my 2900pro compares to a 8800gt at pi.

Oh Wait! ATI doesn't support CUDA! :shadedshu

Well look at FAH GPU...that was ATI ONLY for quite a while, and only shortly in the future is a version being released for NV support...I suppose if it's quicker for them to release or easier until someone finds an easier way...those that have NV can give it a shot. Hard saying if it will truly be justifiable for anything gaming related, but for benchers out there I suppose there's some interest in it.

I'm sure an ATI version will be released...if ATI gpu's can fold, I'm sure they can SuperPI! :toast:

It's pretty much pointless to have a benchmark that only runs on a single manufacturer's hardware. There isn't any way to compare scores between architectures, except if you overclock your 8800gt so that it's faster than a 8800gts, etc.

SuperGPi anyone?

superpi does have a multithreaded flavour around, its just not popular because the dual core crowd hate losing to quads :P

If this works on ATI and Nvidia, i'm all for it!

funny CUDA*
that proves that SuperPI is bad :wtf:

CUDA this CUDA that . . . New motto ?

I think it wont be long before ATi respond with something simular. Im sure it will be better too. Then we move on. Im all for benchmarking tools, but like flashstar said, It's pretty much pointless to have a benchmark that only runs on a single manufacturer's hardware.

And it's a CUDA app :ohwell:

PauliegHmm, this should be interesting. Hope it ends up being useful as a benchmarking tool for gpu's.

If what others here suggest is true, ie only for NV cards, it's worthless. If you can only compare a few cards from the same series and the generations after that but not the real competition then it's pointless.

btarunrAnd it's a CUDA app :ohwell:

in the thread linked to from the link here on TPU, there is a quote from someone saying that ATI can indeed run this program.

That said its a 3rd party quote with no reference so it may not be true.

Musselsin the thread linked to from the link here on TPU, there is a quote from someone saying that ATI can indeed run this program.

That said its a 3rd party quote with no reference so it may not be true.

Interesting, they didn't say it's based on CUDA but that the new code was helped by CUDA. If it means that the app uses CUDA based code to run Pi calculations on a NVidia GPU, and that there's another part of the app that communicates with ATI GPU's differently, that could be a possibility. After all, the object of the app is to test the arithmetic ability of the GPU, and it does it by making shaders do Pi calculations, they're 'fully-programmable' shaders afterall.

Sort of like those "Can you run it? system requirements lab" tests are run on IE browser using ActiveX, on Mozilla (and derivatives) using Java applets.

That would suck if its only for Nvidia.

KursahWell look at FAH GPU...that was ATI ONLY for quite a while, and only shortly in the future is a version being released for NV support...I suppose if it's quicker for them to release or easier until someone finds an easier way...those that have NV can give it a shot. Hard saying if it will truly be justifiable for anything gaming related, but for benchers out there I suppose there's some interest in it.

I'm sure an ATI version will be released...if ATI gpu's can fold, I'm sure they can SuperPI! :toast:

the reasion F@H didnt have nvidia support was first the 7 seirse couldnt produce decent performance foldin, then later because the 8 seirse drivers had issues that would give inaccurage results, pcper forums we had a long talk about this, alot of g80 owners said it was bias, but a few who acctualy understood it explained the problems, and they where all nvidia driver based.....basickly nvidia has a bad habbit of not fixing driver buggs that dont relate dirrectly to getting higher scores in 3dmark or more fps in the top bench games of the month.