| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Chat Room
Thread: PS3 / GPU vs PC / CPU |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 9
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hey
I know this Topic is old and probably already written about at thousand times. But there is so much info when I search, I cannot sort threw it and I cannot find my questions answered, which is: Whats really the difference between the PS3 work and PC work? Is it really like Apples and Oranges? Would the PS3, if it would do the same work as a PC at WCG, be about the same speed as a new PC? Or would it still be faster, and if yes how much? If its a lot faster, why is almost nobody developing more GPU / PS3 applications? I am also asking because, I love that my computer is doing work for the whole of humanity, and I leave my PC on at night to do more crunching, and try to convince friends to install Boinc etc. but if a PS3 / GPU is so much better, I am really asking myself what for? I mean, I am looking at these PS3 numbers http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats and thinking even if I deinstall Boinc now, and buy a PS3 in 3 years, and then let the PS3 go to work, it will make up the time I have not used Boinc for the last years, within days. So it seems to me its not really worth having Boinc installed and getting excited about. So a proper explanation, best scientific, the difference between CPU and GPU crunching would be nice, so I can sleep better at, night, or well get a PS3 sooner than later. Thanks a lot in advance for you "scientific" answer |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Not better: different.
Yes, we've had this conversation a thousand times. GPUs can do a few things very fast, by working in parallel. Nvidia tried to show off their fancy GPU technology with the help of the Mythbusters folks. They showed a robot firing paintballs one by one to draw a smiley face. They called that a CPU. For the GPU, they had a massive contraption that fired hundreds of paintballs at once, and painted a crude Mona Lisa in a split second. The problem for Nvidia is that the analogy fits TOO well. The "CPU" can fire its paintballs anywhere, with total control. The "GPU" takes forever to set up, and once you have done so, you can only paint another Mona Lisa. So, is the GPU really faster? No. It only gets more done if you have a very specific task, and are prepared to accept the limitations of the GPU, such as less precise mathematics. Will this always be the case? Yes and no. No, because GPUs are already solving the problems of imprecise mathematics and unsuitability for general purpose computing. But, on the other hand, CPUs are expected to take advantage of the same massively parallel pipelines used by GPUs. In short, the boundary between CPU and GPU is getting fainter. Manufacturers that say "forget this general purpose nonsense" (ATI) do manage to create chips that are faster for graphics. But manufacturers that try for a general purpose chip end up with something more and more like a CPU. PS: Don't waste your time looking at the Stanford numbers. They are mostly science fiction, since each number measures a completely different thing. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello Thorsten171,
The PS3 is based on the IBM Cell processor with (I think I recall) 7 SPE processors and 1 PPC processor. So, 8 processors in all. None of the processors allow OOP (Out of Order Processing - a technique that speeds up x86 computers) and the SPE processors have very limited memory addressing, which makes keeping their caches filled the key bottleneck - a programming problem here will cut speed drastically. The Cell processor sounded almost like magic when it was first discussed, back in the day of single core CPUs. Going up against quad-cores today, it is much less overwhelming. In the not too distant future, it will be faced with Nehalem-derived octo-cores. And the quad-cores and octo-cores will be easily programmed by mature, well-supported compilers. And they will have larger memories than the PS3, though this is not very important. A lot can be fit into the PS3 size of memory. It just takes tight programming. And clock rates continue to go up - the PS3 is beginning to show its age. So, it is nice to be able to use the PS3 but the modern CPUs are beginning to pass it in the race for speed. Lawrence |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Observing 744 users reported for GPUgrid this moment after how many months of go-live? The client is still heavily tested just for them, hence a client that carries the 6.3 moniker.
----------------------------------------cheers http://boincstats.com/index.php?pr=bo
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks a lot for your answers lawrencehardin and Didactylos I really appreciate it :-)
|
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
PS: Don't waste your time looking at the Stanford numbers. They are mostly science fiction, since each number measures a completely different thing. OK re: the PS3, if Stanford #s are BS, then are PS3Grid's as well? http://www.ps3grid.net/forum_thread.php?id=219 How do we assign credits? On standard PC, BOINC assign credits based on the average between the floating-point and integer performance of the machine according to a set of benchmarks performed by the client, regardless of the real performance of the application on the machine. These benchmarks on the Cell processor are of course wrong because they do not use the SPEs. The way we assign credits takes into account these facts. We have bench-marked Cell MD on a reference machine based on a Opteron 2Ghz (read this paper). This machine returns the following benchmark by the BOINC client: CPU type AuthenticAMD AMD Opteron(tm) Processor 146 [Family 15 Model 5 Stepping 10][fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx mmxext lm 3dnowext 3dnow up] Number of CPUs 1 Measured floating point speed 1707.25 million ops/sec Measured integer speed 2969.64 million ops/sec The average is therefore 2338.4 MIPS (million instruction per second) or equivalently 9.7 Cobblestone/hour (each unit of BOINC credit, the Cobblestone, is 864,000 MIPS). Cell MD is currently 16 time faster than this reference machine on the PS3, therefore we will credit 155.9 Cobblestone/hour on the PS3 for Cell MD. This is in line with the SUSTAINED floating-point performance measured by us which for Cell MD is between 25 and 30 GFLOPS. In fact, the ratio between average number of operations and floating-point operations is 1.27 which gives between 132 and 158 Cobblestone/hour. [Edit 1 times, last edit by Former Member at Nov 13, 2008 6:46:59 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
esoteric17, BOINC credits don't compare equally with other BOINC projects, let alone different architectures. It's sweet that they try to come up with a fair conversion, but don't kid yourself that it means anything.
It's the Big Lie perpetrated by BOINC - the idea that projects grant equal credit for equal work. They simply don't. Some give comparable credit, but over the entire set of BOINC projects, there is huge variation. |
||
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges: 
|
OK re: the PS3, if Stanford #s are BS, then are PS3Grid's as well? http://www.ps3grid.net/forum_thread.php?id=219 I don't know about the PS3 but I've processed ~165 wu's on GPU's at GPUGrid. Every WU requests and receives 3232.06 credits even though run times range between 4 and 20 hours. Personally, I like the shorter ones.    |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks for the explanation.
|
||
|
|
|