Message boards :
Number crunching :
GPU FLOPS: Theory vs Reality
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 . . . 17 · Next
Author | Message |
---|---|
Al ![]() ![]() ![]() ![]() Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 ![]() ![]() |
I had actually acquired one of EVGA's hybrid coolers with the intention of water cooling my 1080, but I really don't see the need at this point, from all indications it looks like the ACX 3.0 cooler they have on it is doing a pretty darn good job, running right now clocked at 2050 mhz it is running at 35C with the fan still at 100%, and is running at stock voltages. I think I will instead put it on the 980Ti, that is running in the mid 50's to high 60's when it's crunching hard, and could use it a lot more than this one could. Just need to find a stock blower type cooler on Fleabay, and then break out the scalpels, Dr. Al has entered the room! lol ![]() ![]() |
![]() ![]() Send message Joined: 29 Nov 99 Posts: 358 Credit: 5,909,255 RAC: 0 ![]() |
Hey Shaggie76! I'm trying to collect data to make the best computation/power-usage choices possible for upgrading my modest farm.From what I have heard, the GTX 750 Ti is in the sweet zone wrt best computation/power consumption and price, plus it only requires power from the PCIe bus! My 2 primary rigs are identical HP Z400 Xeon W3550 with 1 GTX 750 Ti (except for RAM & Windows version) ( 7996377 & 8010413 ) but I've been doing much tweaking on them lately including swapping tasks between CPU & GPU in order to maximize throughput (based on time). [edit1] Right now, 7996377 is processing 2 WU/GPU with Cuda50, and the other one: 1 WU/GPU with SoG_r3472.exe [/edit] The reason the RAC is so different is because I had the 2 GPUs in 1 rig until a week ago. I looked in my import of host.gz from the end of May and there are ~2000 PCs with at least 1 GTX 750 Ti in them. Here are the host_IDs of those with a RAC >10k with only 1 GTX 750 Ti: id RAC p_model 5501972 19,710.27 Intel(R) Core(TM) i7-4770K CPU @ 3.50GHz [Family 6 Model 60 Stepping 3] 7436798 18,246.78 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 6309460 16,394.15 Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz [Family 6 Model 60 Stepping 3] 6748333 16,149.52 Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz [Family 6 Model 58 Stepping 9] 7923604 15,738.89 Intel(R) Xeon(R) CPU E5-2687W v3 @ 3.10GHz [Family 6 Model 63 Stepping 2] 5974961 15,330.76 Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz [Family 6 Model 60 Stepping 3] 7953544 15,269.04 Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz [Family 6 Model 62 Stepping 4] 7028448 15,026.19 Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz [Family 6 Model 42 Stepping 7] 7978786 14,831.74 Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz [Family 6 Model 94 Stepping 3] 7379595 14,313.28 Intel(R) Xeon(R) CPU X5690 @ 3.47GHz [Family 6 Model 44 Stepping 2] 7908428 14,209.08 Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz [Family 6 Model 60 Stepping 3] 6825981 13,925.73 Intel(R) Core(TM) i5-2500 CPU @ 3.30GHz [Family 6 Model 42 Stepping 7] 7416916 13,700.43 Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz [Family 6 Model 63 Stepping 2] 7612658 13,566.00 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7258237 13,548.33 Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz [Family 6 Model 42 Stepping 7] 7319278 13,416.58 Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz [Family 6 Model 60 Stepping 3] 7913336 13,364.99 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7454279 13,231.54 Intel(R) Core(TM) i5-4460S CPU @ 2.90GHz [Family 6 Model 60 Stepping 3] 7284006 13,203.66 AMD FX(tm)-8350 Eight-Core Processor [Family 21 Model 2 Stepping 0] 6709467 12,961.88 Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz [Family 6 Model 63 Stepping 2] 7901799 12,955.97 Intel(R) Core(TM) i7-5960X CPU @ 3.00GHz [Family 6 Model 63 Stepping 2] 7851657 12,868.96 Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz [Family 6 Model 60 Stepping 3] 7904960 12,791.95 Intel(R) Core(TM) i5-2320 CPU @ 3.00GHz [Family 6 Model 42 Stepping 7] 7986324 12,749.98 Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz [Family 6 Model 45 Stepping 7] 7262806 12,730.06 Intel(R) Core(TM) i7-4770K CPU @ 3.50GHz [Family 6 Model 60 Stepping 3] 7310710 12,680.98 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7377680 12,611.13 Intel(R) Core(TM) i7-4790S CPU @ 3.20GHz [Family 6 Model 60 Stepping 3] 7227246 12,519.56 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7573103 12,408.44 Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz [Family 6 Model 58 Stepping 9] 7865587 12,353.76 Intel(R) Xeon(R) CPU E5620 @ 2.40GHz [Family 6 Model 44 Stepping 2] 7354663 12,344.53 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 5760259 12,331.10 AMD Phenom(tm) II X6 1055T Processor [Family 16 Model 10 Stepping 0] 7909699 12,231.40 AMD FX(tm)-8300 Eight-Core Processor [Family 21 Model 2 Stepping 0] 6637790 12,220.64 AMD Phenom(tm) II X6 1090T Processor [Family 16 Model 10 Stepping 0] 7538403 12,087.77 Intel(R) Core(TM) i7-3930K CPU @ 3.20GHz [Family 6 Model 45 Stepping 7] 7645887 12,057.22 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7359113 11,969.71 AMD FX(tm)-8350 Eight-Core Processor [Family 21 Model 2 Stepping 0] 7476673 11,909.16 Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz [Family 6 Model 94 Stepping 3] 6340434 11,902.15 Intel(R) Core(TM) i7 CPU X 990 @ 3.47GHz [Family 6 Model 44 Stepping 2] 7414744 11,646.36 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7942637 11,606.73 Intel(R) Core(TM) i7-6700K CPU @ 4.00GHz [Family 6 Model 94 Stepping 3] 7833959 11,551.70 Intel(R) Core(TM) i5-4440 CPU @ 3.10GHz [Family 6 Model 60 Stepping 3] 7599512 11,474.69 Intel(R) Core(TM) i5-4460 CPU @ 3.20GHz [Family 6 Model 60 Stepping 3] 7770856 11,324.23 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7167010 11,122.04 Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz [Family 6 Model 58 Stepping 9] 7928301 11,114.64 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7919916 11,083.28 Intel(R) Core(TM) i5-6500 CPU @ 3.20GHz [Family 6 Model 94 Stepping 3] 6792171 10,978.81 Intel(R) Xeon(R) CPU E5-1620 0 @ 3.60GHz [Family 6 Model 45 Stepping 7] 7854550 10,900.42 Intel(R) Core(TM) i5-4670K CPU @ 3.40GHz [Family 6 Model 60 Stepping 3] 7907798 10,698.89 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7600313 10,599.64 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7917835 10,597.81 Intel(R) Core(TM) i5-6600 CPU @ 3.30GHz [Family 6 Model 94 Stepping 3] 7859134 10,551.84 Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz [Family 6 Model 94 Stepping 3] 7336487 10,539.69 Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz [Family 6 Model 42 Stepping 7] 7971571 10,395.83 Intel(R) Core(TM) i5-6500 CPU @ 3.20GHz [Family 6 Model 94 Stepping 3] 7337570 10,387.94 Intel(R) Core(TM) i5-3570K CPU @ 3.40GHz [Family 6 Model 58 Stepping 9] 3678155 10,376.33 Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [Family 6 Model 15 Stepping 11] 7517543 10,371.19 Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz [Family 6 Model 60 Stepping 3] 7323951 10,328.15 Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [Family 6 Model 15 Stepping 11] 5503553 10,288.52 Intel(R) Core(TM) i7 CPU 860 @ 2.80GHz [Family 6 Model 30 Stepping 5] 7855653 10,198.51 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 7147692 10,154.19 Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz [Family 6 Model 58 Stepping 9] 7842573 10,134.67 Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz [Family 6 Model 60 Stepping 3] 7447166 10,125.54 Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz [Family 6 Model 60 Stepping 3] 6732079 10,122.73 Intel(R) Core(TM) i7-3770S CPU @ 3.10GHz [Family 6 Model 58 Stepping 9] 7924712 10,120.94 AMD FX(tm)-8350 Eight-Core Processor [Family 21 Model 2 Stepping 0] 7956967 10,043.47 Intel(R) Core(TM) i7-3820 CPU @ 3.60GHz [Family 6 Model 45 Stepping 7] 5541574 10,038.80 Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz [Family 6 Model 23 Stepping 10] 5403642 10,036.27 Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz [Family 6 Model 23 Stepping 10] 7997014 10,022.78 Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz [Family 6 Model 94 Stepping 3] 8006285 10,011.96 AMD FX(tm)-8320 Eight-Core Processor [Family 21 Model 2 Stepping 0] [edit2] I'm guessing the GTX 750 Ti does a max of ~7.5K/day[/edit] Let me know if you'd like similar stats for other card, Cheers, Rob :-} |
![]() ![]() Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 ![]() ![]() |
My single EVGA GTX 750 Ti FTW had a RAC in 9-10k with two tasks running at once. Power consumption was in the 40-45W range. At the moment I'm planning on sticking it in my Celeron J1900 system to see how well that CPU can run the GPU. SETI@home classic workunits: 93,865 CPU time: 863,447 hours ![]() |
W-K 666 ![]() Send message Joined: 18 May 99 Posts: 19655 Credit: 40,757,560 RAC: 67 ![]() ![]() |
Shaggie, just for complete accuracy, in your chart, you may want to add FTW to the 1080, it isn't a massive difference, but it should probably be noted, as it is a factory overclocked version, which I have then bumped up even higher. I am still amazed at how cool it is running compared to my 980Ti, night and day. I wonder if I have a bit more headroom on it to OC it some more? I thought that heat was the limiting factor when overclocking, and if so, it appears there is more room to run. But, I don't feel the need to go nuts with it, I would like it to do long term, reliable crunching service for me. Now that these are starting to appear out in the real world, I'll have to do a little looking around to see what others experiences are and compare notes. I think if you look at some of the review sites, they say the 1080 the restriction for overclocking is a lack of power on the standard model with 6pin connector. Photo's of the cards showed that one of the power regulator positions were not populated. They were hoping that the OEMs might change the connector to an 8pin and populate this position. ![]() Shows bottom, out of 7 positions, power regulator components missing. |
![]() ![]() Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 ![]() ![]() |
I scanned a few dozen from that list of 750Ti hosts Stubbles69 posted and pulled out the top half-dozen: the average is just over 300 CR/h which is pretty impressive for a 60W part! |
![]() ![]() Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 ![]() ![]() |
I just got confirmation from Kent Harald Rasmussen that his RX 480 is running 2 GPU tasks at once which seems to suggest about 420CR/hr for it. This is surprisingly low -- under half of what the R9 Nano was doing so I'd really like to see more RX 480 data! |
Al ![]() ![]() ![]() ![]() Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 ![]() ![]() |
I think if you look at some of the review sites, they say the 1080 the restriction for overclocking is a lack of power on the standard model with 6pin connector. Photo's of the cards showed that one of the power regulator positions were not populated. They were hoping that the OEMs might change the connector to an 8pin and populate this position. I'll do some research soon, prob tomorrow afternoon when it's raining and I can't work outside. As for mine and power, looks like I am more than good to go, as it has two 8 pin connectors, so it has power coming out of its ying yang! lol ![]() ![]() |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13928 Credit: 208,696,464 RAC: 304 ![]() ![]() |
I scanned a few dozen from that list of 750Ti hosts Stubbles69 posted and pulled out the top half-dozen: the average is just over 300 CR/h which is pretty impressive for a 60W part! And keep in mind that's generally with the GPU not fully loaded, the power load is often less than 75% (less than 45W) if the maximum 60W. Grant Darwin NT |
![]() ![]() Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 ![]() ![]() |
I scanned the host export and found about a dozen hosts with Ellesmere cards (the RX 480 is the only released Ellesmere card I think). I can't tell if they're doing unusual things like running multiple tasks but the numbers aren't that far apart so I doubt it. Host Id: Credits/Hr 7492259 420.6506388 7431180 647.7853343 8034949 367.8899388 8037810 498.1638136 Average 483.6224314 I also scanned for Fiji parts; there's quite a few in the db and I can't tell what version they are (Nano, Fury, Fury X, etc) Host Id: Credits/Hr 8001648 341.4652324 8001994 557.7983477 8003231 661.4473237 8003833 413.5674201 8013353 551.747657 8014347 385.3597423 8029489 336.9615636 Average 464.0496124 Note that RueiKe's R9 Nano's do a lot better than these do on average -- my guess is his water-cooling setup helps a lot because the Nanos are reputed to throttle when they get hot as I'm sure they do when left crunching for hours. |
![]() ![]() ![]() Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 ![]() ![]() |
Note that RueiKe's R9 Nano's do a lot better than these do on average -- my guess is his water-cooling setup helps a lot because the Nanos are reputed to throttle when they get hot as I'm sure they do when left crunching for hours. Yes, watercooling makes a big difference on the Nanos, as they will definitely thermally throttle running BOINC. I did a video on it. Also, I have optimized SBS and period iterations based on the DOE on published in another thread. I am also overclocking them, achieving same GPU clock as the Fury X. GitHub: Ricks-Lab Instagram: ricks_labs ![]() |
![]() ![]() Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 ![]() ![]() |
I scanned the host export and found about a dozen hosts with Ellesmere cards (the RX 480 is the only released Ellesmere card I think). I can't tell if they're doing unusual things like running multiple tasks but the numbers aren't that far apart so I doubt it. Personally I've been disappointed in the performance of the Fury cards compared to my R9 390X. With less than 70% of the shaders of a Fury Nano or Fury X it still manages to churn through MB tasks in ~6 minutes. With the fans set to auto it does run up to 68ºC but they are still silent at ~40%. THe only config settings I use are -hp -cpu_lock & I think that -cpu_lock might be depreciated in the current version app. So it might not be doing anything. SETI@home classic workunits: 93,865 CPU time: 863,447 hours ![]() |
![]() ![]() Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 ![]() ![]() |
It's been a crazy few weeks of work for me but I've resumed my puttering. I grabbed a recent host export from SETI and scanned it for hosts that have updated in the last few weeks and have submitted enough credit; I grouped them by installed GPU and started harassing the task PHP interface to pull down tasks stats to identify down hosts that had completed enough tasks with their GPU to provide a good sample. After getting enough sample hosts I then sorted to take the top 10 for each GPU to try to rule out people running more than one task concurrently on the same GPU. The results weren't too crazy: ![]() Error-bars are standard-deviation of the data. Source for my hacks are on GitHub Next I need to some tedious data-entry for the final leg of my study: I want to get TDP and theoretical FLOPS out of Wikipedia so that I can augment my results further. |
![]() ![]() Send message Joined: 29 Nov 99 Posts: 358 Credit: 5,909,255 RAC: 0 ![]() |
All I have to say for now is: WOW! Great work :-D |
![]() ![]() Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 ![]() ![]() |
Ok one more picture for tonight: ![]() Remarkably this confirms what was probably common knowledge: the GeForce 750 series is exceptionally good performance per watt! I just need more data for some of the newer cards. |
![]() ![]() ![]() Send message Joined: 29 Jun 99 Posts: 11451 Credit: 29,581,041 RAC: 66 ![]() ![]() |
Shaggie76, this is real good stuff you are trying to do.The only real question I have ATM is the validity of your data sample as I am sure you do. |
![]() ![]() ![]() Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 ![]() ![]() |
Remarkably this confirms what was probably common knowledge: the GeForce 750 series is exceptionally good performance per watt! It would be interesting to see where Fury/Nano falls within this distribution. HBM is supposed to give a power advantage... GitHub: Ricks-Lab Instagram: ricks_labs ![]() |
![]() ![]() Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 ![]() ![]() |
Shaggie76, this is real good stuff you are trying to do.The only real question I have ATM is the validity of your data sample as I am sure you do. The only thing that's keeping me from casting a wider net to build confidence in my results is my respect for the health of the SETI PHP server. If there was a 'task/workunit' database dump I'd download that and be even more thorough. If you're curious you can see the thresholds I used in the scripts -- I think it was at least 25 valid work units per host and at least 10 hosts per GPU. Sadly this sets the bar too high for some of the newer cards that I was hoping to get data for. |
![]() ![]() Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 ![]() ![]() |
Remarkably this confirms what was probably common knowledge: the GeForce 750 series is exceptionally good performance per watt! I'll definitely get those stats -- I'm curious too! I'll probably have to do it by hand because there aren't enough in circulation out there to qualify with thresholds I set. |
Al ![]() ![]() ![]() ![]() Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 ![]() ![]() |
Say Shaggie, if you're interested, I have a _ton_ more data that has been logged since I installed emfers program, just let me know and I'll send it to you to review. ![]() ![]() |
![]() ![]() Send message Joined: 29 Nov 99 Posts: 358 Credit: 5,909,255 RAC: 0 ![]() |
Hey Shaggie, How much knowledge would be needed to adapt your script to find out how many of the top 10,000 top computers/hosts have an "anonymous platform" on the PC's "apps" page (at the bottom)? Cheers, Rob |
©2025 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.