Message boards :
Number crunching :
Where can I find CPU and GPU comparasons
Message board moderation
Author | Message |
---|---|
Send message Joined: 13 Apr 15 Posts: 19 Credit: 22,543,893 RAC: 0 |
Is there a spreadsheet or graph showing relative performance of various CPU and GPU configurations? I have a bunch of AMD GPU's from the olden-days of Bitcoin and am wondering if it's worth the effort to restore some of these GPU's for use with this project. Right now, I have several CPU servers running on this project, and am collecting some credits, but I don't know how the GPUs would work in comparison to CPUs. The only GPU rig I have pretty much ready to go is a dual 7950 rig, but I do have dozens of Raedon GPU cards just sitting on a shelf... Just don't know if it's worth it or not to build up rigs again... PS, im a total NOOB here, just discovered Boinc SW last week... |
Send message Joined: 24 Jan 11 Posts: 708 Credit: 543,263,280 RAC: 140,951 |
Well first requirement is that the GPU cards have to have double precision math capability. And yes, it is very worthwhile to compute MW tasks on the GPU since they are much more powerful compared to CPUs. I just saw a nice chart comparing the various generations of GPUs on their GPGPU capabilities on another forum. http://www.videocardbenchmark.net/directCompute.html Find your old cards there and see if it is worthwhile putting them to use in another rig. Cheers, Keith |
Send message Joined: 13 Apr 15 Posts: 19 Credit: 22,543,893 RAC: 0 |
Thanks. That link took me to passmark, which is a great site. Im specifically looking for details for comparason as it relates to Boinc and/or MilkyWay@Home specifically. The GPU's I have include R9-2x0X as well as 79xx series. Im wondering how much work you can accomplish on these cards when compared to Intel CPU's. I haven't yet figured out exactly to tell how my CPU's are going, since I only started a week ago. I just recently got a pair of Dell machines with dual X5680 CPU's (24 cores per machine), and a HP DL980G6 box with 128x 2Ghz cores running and combined they seem to push out about 100k credits per day. Wondering if the GPU's can meet or exceed that value, and by how much. Ill probably see if I can fire up my box with Dual 7950's and see how that compares. |
Send message Joined: 24 Jan 11 Posts: 708 Credit: 543,263,280 RAC: 140,951 |
Even those Dell machines would get overmatched by even modest GPUs. Even though I don't run the CPU MT application, I do know that application is a CPU hog requiring all cores be used. Have you looked at the statistics pages for the project yet? They show the computing power for various GPUs and CPUs. You might look at the GFLOPS ratings on the CPU page: http://milkyway.cs.rpi.edu/milkyway/stats.php http://milkyway.cs.rpi.edu/milkyway/cpu_list.php The top computers page can often show you what similar hardware platforms are capable of generating with regard to RAC. http://milkyway.cs.rpi.edu/milkyway/top_hosts.php See if you can match your CPU and GPU to someone else's computer for comparison. Cheers, Keith |
Send message Joined: 8 May 09 Posts: 3319 Credit: 520,327,271 RAC: 21,209 |
Thanks. That link took me to passmark, which is a great site. Im specifically looking for details for comparason as it relates to Boinc and/or MilkyWay@Home specifically. Each gpu can do roughly 10 times the work of a single cpu. This of course varies on how old the gpu and cpu is, the newer ones generally do better in both cases. Go here http://milkyway.cs.rpi.edu/milkyway/stats.php and here http://milkyway.cs.rpi.edu/milkyway/top_hosts.php and see the top cpu's and gpu's here at MW. The one thing you can't tell is how many units at one time a gpu is running, yes the better gpu's can run multiple units at one time increasing that 10 times number even further. Yes your 7950's can crunch here and do very well here and probably wipe the floor compared to your cpu's. Your cpu's only saving grace could be the sheer number of them. This guy is using a 7970, only a bit more powerful than your 7950, and is doing gpu units like this: 38.50 1.79 106.88 MilkyWay@Home v1.02 (opencl_amd_ati) This guys top cpu unit is: 6,122.01 15,360.47 255.71 MilkyWay@Home N-Body Simulation v1.48 (mt) You can see the gpu is doing units a TON faster and getting fully half the credits of the cpu unit that takes multiple cpu cores to complete. |
Send message Joined: 22 Jan 11 Posts: 375 Credit: 64,657,871 RAC: 0 |
Errm, guys, how about my benchmark thread for WU time comparisons?? http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=3551 Updated table is near the end of the thread (or go to the AnandTech thread linked there). I'm pretty sure my HD 5850 could do nr 100k/day o/c to 850 MHz! ;) Those high end GPUs, especially the 7900s & R280Xs, will destroy pretty much any CPU in MW output. Team AnandTech - SETI@H, DPAD, F@H, MW@H, A@H, LHC, POGS, R@H, Einstein@H, DHEP, WCG Main rig - Ryzen 5 3600, MSI B450 G.Pro C. AC, RTX 3060Ti 8GB, 32GB DDR4 3200, Win 10 64bit 2nd rig - i7 4930k @4.1 GHz, HD 7870 XT 3GB(DS), 16GB DDR3 1866, Win7 |
©2024 Astroinformatics Group