Tagrineth said:
Anttaru, I'm just going to say one thing.
Folding@home for GPUs relies on having strong programmable pixel shaders. Ever heard of the GPGPU movement? It stands for General Purpose GPU, referring to programmable shaders that are so complex and open that the GPU can actually potentially run CPU-like general purpose code (of course, still having severe limitations compared to an actual CPU but still out of the normal boundaries of graphics). A well known 3D graphics site had a contest about a year ago to make GPGPU style programs for DirectX9, one of the top 10 was a version of Frogger that ran purely on the GPU - physics, sound, collision were all done using pixel shaders.
This is how folding@home GPU edition works.
The PS3's RSX, while being capable of an astounding 1.8 TFLOPs in graphics calculations, probably can't run folding@home GPU edition.
I'm not 100% positive on this, as for some reason my usual sources say nothing about how powerful RSX's shaders are. =c
But any PC with a GeForce 7k or Radeon 10k model would be able to do the same thing, and probably just as fast. =x And an SLI or CrossFire PC could easily outperform a PS3's folding@home...
Edit again: I read your Folding@home on GPUs link.
After reading it, I guarantee the PS3 Folding@home will NOT run on RSX.
I cannot guarantee the RSX will render for Folding@home. Folding@home previously ran off of Nvidia's hardware in the past according to Stanford.
Yes they are more interested in ATI, cause ATI can mash Nvidia. In the end it's all a matter of what kind of driver is running off of the hardware. Software can fix such issues.
Tagrineth said:
Since the RSX was developed based on an existing nVidia core foundation (much like the NV2A being based on the GeForce3 architecture) - I believe in this case it is similar to the GeForce 7900 line - I very highly doubt folding@home will be able to tap into its 1.8TFLOPs in any way whatsoever.
Folding@home started in 2000. Before the NV2A's time. Stanford did start off with Nvidia according to the official information posted in their website. There is no reaosn why they can't do it again. They just have their heart set on ATI, because ATI smokes Nvidia.
Tagrineth said:
This puts folding@home's effectiveness on the PS3 into a very different light, I'd say. They've probably improved their code since that early client's demonstration, but one wonders how much. They're also limited to an absolute ceiling of 218 GFLOPs, while on a current PC client -
Which means, a high end PC with just one ATi X19x0 card would outperform the PS3 client by about double, and if the system had two in CrossFire, that number would soar.
PS3's contribution to... what, now?
I'll say again. The Cell will donate around 100 of it's Giga floating point operations. The rest will be riding on the GPU most likely.
Official Standford University information: Folding@home PS3@cure project
Now in 2006, we are looking forward to another major advance in capabilities. This advance utilizes the new Cell processor in Sony’s PLAYSTATION 3 (PS3) to achieve performance previously only possible on supercomputers. With this new technology (as well as new advances with GPUs), we will likely be able to attain performance on the 100 gigaflop scale per computer. With about 10,000 such machines, we would be able to achieve performance on the petaflop scale. With software from Sony, the PlayStation 3 will now be able to contribute to the Folding@Home project, pushing Folding@Home a major step forward.
It's not a matter of who can solo this faster. And it is obvious there will be more PS3's out there than R600's and 1900XT as everyone has acceptable graphic cards already. If you want to compare numbers, we'll probably have a chance when they post results. 10,000 PS3's will do petaflop scale digits, that is more than what they can hope for.
If the 360 were to do this and one console did better than the other in progress reports, you don't just finger the system that got less work done. You welcome it with open arms cause their presance makes the point of the operation better. This is how it will be for PC and PS3. And if Microsoft does this with the 360, then good, let them do it without being harassed.
Tagrineth said:
Edit: Okay, I lied. I'm gonna add a little to this.
I never said I don't believe GPUs are capable of more. I'm well aware of the GPGPU concept and have been for years, actually. I'm just doubting the PS3's ability to run GPGPU code. Also, keep in mind that GPGPU operation is usually very wasteful because it takes very complicated pixel shaders, much moreso than it would on a CPU. It's still very fast, though, because GPUs are very highly parallel by nature.
It's funny how you're trying to teach me about GPU processing other tasks after i told you about it, when in you're last post you said:
Tagrineth said:
The 1.8TFLOPs is on the RSX graphics processor which will BE INOPERABLE WHILE RUNNING FUCKING @HOME because it can't execute general code, only graphics.
And
Tagrineth said:
The FLOP rating of the graphics chip is misleading because graphics processors are by necessity designed to do many parallel operations at once... but they can't just process anything - they can only process graphics.
Will you repeat the above today ?
cause now you saying stuff like
Tagrineth said:
But it takes a very advanced GPU to be capable of GPGPU operation,
Advanced GPU is faster and is now modern, but these methods has been happening for a long time now. I've stressed out a 64 meg graphics Rage series. It wasn't too bad but it's freaking slow compared to my 9550. The rage got fried eventually.
Tagrineth said:
and historically, saying "PS3 is a 2TFLOP system because the GPU offers 1.8TFLOPs!" is a very misleading statement (note, also, that FUD stands for Fear, Uncertainty, Doubt - which refers to marketing tactics that involve befuddling the consumer with misleading statistics, pretty much).
It's not like i said it was a 2TFLOP, i said it was near that. At my first estimate 6 seconds. 8 seconds would have been closer and 10 seconds would have been there.
The CPU does 218GFLOP and the GPU 1.8TFLOP that is 8GFLOP above 2TFLOPs of calculation occuring within the system. Hmm. There is no sense seperating the colaberation happening in a single package.
Tagrineth said:
Also.
PS3's Cell processor runs at 3.2GHz. Period.
It isn't variable.
It can't go higher.
~218 GFLOPs is the *maximum* of PS3's 7-PE Cell. It isn't a "variable" "current speed", it's the maximum. The ceiling. It won't go faster than that.
I never said the Processor inside the PS3 would magically get faster.
Tagrineth said:
It also won't go as fast as that without *FLAWLESS* programming and absolutely no hiccups like cache misses, branch misprediction, memory latency issues, multiprocessor ring bus holds, et al. There are a lot of reasons why it's more difficult to get full performance out of a multiprocessor system, and the more processors you add, the more difficult it becomes. Cell has a lot of features which try to alleviate those problems and make more processors more efficient, but there are limits. Especially once Cells get into 16-32+ elements, I'd hazard that, say, a 16 Cell core would likely only be able to run up to the theoretical peak of at most 14-15 of those processors, with the wasted power due to memory, bus, and code inefficiency being at least equivalent to an entire PE. Maybe Cell is so clever that it might take more PEs total to get to the equivalent point of one PE being wasted, I'm not sure, but look at it this way:
A dual processor PC will generally only be about 50% faster than a single processor of the same make and model.
Remember 1 of 8 SPEs is reserved for redundancy. That can be the hardware solution to poor programming. I'm sure the programers musn't ignore that when developing software for the PS3 version of the cell.
Overkill is often the solution to dealing with lag issues. Like for graphic cards .... Why do we have cards that are doing hundreads of fps when we cannot see past 30fps ? The answer is when it does lag we do not notice it.
If software will lag out the machine, fix it. If it cannot be done, then save it for the next advancement in hardware.
In closing i would like to add. I haven't much more to say. I've had to repeat myself a few times, and highlight alot of what i said. I don't see the reason for battle here.
Assuming this is based on myself and my beliefs i'll have to say.
- I don't support console war to the extent of destruction, i don't give a shit if 1 system has "the best of any", cause i will have them all.
- I do favour Sony, Wii and 360 never touched PS2 for me yet I have no choice but to look foward to PS3 more than Nintendo and Microsoft. However if PS fails to be number 1 again, That's because we are all playing number 1. So what ?
- I believe PS3@cure is a major contribution. If it opens people hearts, it will also set a many, lovely mood, for so many strangers, friends, and family. I cannot understand negativity towards it. Sure there are 2 sides to everything, one would dislike the fact Sony may be doing this to sell more PS3. Now everyone hates cold business, but we all love our videogames. So who cares, it's all good.
- I have always been scolded for being a firm ATI man. So this is just plain irony i find.
- Everytime i wrote about the Cell processor outdoing itself, i wrote about what they will be doing with it in other machines. Like Toshiba's claimes of the cell processor recording 49 mpeg2 videos at the same time, or IBM making a 16TFLOP machine. I never said the one in the PS3 is of that caliber.
- I threaten no one. I'm proud to be able to keep my cool when being cursed for unacceptable reasons. I have no desire to be better than anyone. I also know i am not the best.
- I do not dissmiss everything my opponet has shown me. Nor do i feel i ever debated against the materals usied against me it to begin with.
- I do not think the PS3 "alone" is a gift from the highest.
- I don't assume my grammar is perfect when presenting argument. I can use Word if it's ever an issue. Which it is not.
As true as they are. I hope my points end this foolishness Tagrineth. I've had enough of this for now. At this point It's taking alot of time to edit and write these posts. No argument is worthy of so much time.