Performance Tab still broken
Message boards :
Server and website :
Performance Tab still broken
Message board moderation
| Author | Message |
|---|---|
|
Send message Joined: 4 Aug 14 Posts: 266 Credit: 2,219,935,054 RAC: 0 Level  Scientific publications   
|
The Performance tab was always a great source of information on gauging how each GPU was performing in comparison to other GPUs. I have now resorted to scanning the Volunteers \ Hosts list for this info. Below is a quick summary taken 2nd November 2020 2:00am UTC. In this listing the Recent Average Credit (RAC) has been used for the comparison due to the variability of the work unit runtimes. Toni has stated that the credit formula has been uniform for several years. This list is not definitive, just an indicator at best. There are many factors that could affect this listing. NOTE: Hosts with multiple GPUs have been exluded. Recent Average Credit has been rounded The best performing GPU from each type has been listed Rank GPU RAC 23 RTX 2080 Ti 1032000 61 RTX 2080 Super 760000 65 RTX 2070 Super 747000 74 GTX 1080 Ti 712000 85 RTX 2080 658000 94 RTX 2070 625000 116 RTX 2060 Super 585000 138 GTX 1080 528000 155 GTX 1660 Ti 511000 156 RTX 2060 510000 166 GTX 1070 Ti 502000 194 GTX 1070 468000 216 GTX 1660 Super 436000 276 GTX 1660 396000 335 GTX 1650 Super 353000 408 GTX 1060 6Gb 310000 459 GTX 1060 3Gb 288000 490 GTX 1650 274000 809 GTX 1050 Ti 193000 960 GTX 1050 160000 Below is a list of GPU efficiency (based on the list above) Again, list is not definitive and should not be taken too seriously. There are many factors that could change this listing. NOTE Watts are estimated for each GPU type. GPU Watts RAC/Watt Rank GTX 1660 Ti 130 3931 1 GTX 1650 Super 100 3530 2 GTX 1660 Super 125 3488 3 RTX 2070 Super 215 3474 4 RTX 2080 Ti 300 3440 5 GTX 2070 185 3378 6 GTX 2060 Super 175 3343 7 GTX 1660 120 3300 8 GTX 1650 85 3224 9 RTX 2060 160 3188 10 GTX 1070 150 3120 11 GTX 2080 215 3060 12 RTX 2080 Super 250 3040 13 GTX 1080 180 2933 14 GTX 1080 Ti 250 2848 15 GTX 1070 Ti 180 2789 16 GTX 1060 6Gb 120 2583 17 GTX 1050 Ti 75 2573 18 GTX 1060 3Gb 120 2400 19 GTX 1050 75 2133 20 |
|
Send message Joined: 21 Feb 20 Posts: 1112 Credit: 40,764,483,595 RAC: 7,379,314 Level  Scientific publications 
|
the 2080ti TDP number is way high. most are 250w, some a little more. but 300 would be atypical. personally i run my single one at 225W, and my 5x system at 215W each. 
|
 ServicEnginIC Send message Joined: 24 Sep 10 Posts: 591 Credit: 11,739,836,510 RAC: 9,833,974 Level Scientific publications 
|
I have now resorted to scanning the Volunteers \ Hosts list for this info. Great (and laborious) job! Very interesting, thank you for this effort. |
|
Send message Joined: 4 Aug 14 Posts: 266 Credit: 2,219,935,054 RAC: 0 Level Scientific publications
|
the 2080ti TDP number is way high. most are 250w, some a little more. but 300 would be atypical. Setting the Watts was a difficult choice. For the RTX 2080 Ti, highend cards are 300W, reference TDP is 250W. If I was to pick the best performing GPU, I was making an assumption, it was a highend card. And then there are user who do modify the power limits. (Me included) The table is definitely not perfect. Hence the caveat on the post. Nice to know you run your GTX 2080 Ti at 225W, as it was your GPU in the list. I may post this list occassionally. I will update your GPU TDP if it appears in the next list. |
|
Send message Joined: 4 Aug 14 Posts: 266 Credit: 2,219,935,054 RAC: 0 Level Scientific publications
|
I have now resorted to scanning the Volunteers \ Hosts list for this info. This was a manual task, but I have almost finished a script to automate the process. This should provide more GPUs in the list and hopefully include median RAC figures for GPU types. I will also revert to reference TDP for efficiency calculations on the median RAC. Time permitting, will attempt to publish a list every month. |
|
Send message Joined: 4 Aug 14 Posts: 266 Credit: 2,219,935,054 RAC: 0 Level Scientific publications
|
Finished the script and gathered some info on the first 3000 Hosts (Volunteers Tab) which I have listed below. Hosts with Multiple GPUs: 320 (which are excluded from list) List compiled: 5th November 2020 0:00UTC. If you have a keen eye, you will see a RTX3070 and 2 RTX3090 GPUs. (Any RAC they have, would be from a compatible card that was removed) GPU Model Count GTX 1070 197 GTX 1060 6GB 194 GTX 1080 175 GTX 1080 Ti 158 GTX 1050 Ti 142 GTX 1060 3GB 114 GTX 970 114 RTX 2070 SUPER 111 RTX 2060 101 RTX 2080 Ti 96 GTX 1660 Ti 94 RTX 2070 89 GTX 1650 74 RTX 2080 SUPER 70 GTX 1070 Ti 67 RTX 2080 65 GTX 1050 64 GTX 960 63 GTX 1660 SUPER 60 RTX 2060 SUPER 53 GTX 750 Ti 50 GT 1030 39 GTX 1650 SUPER 37 GTX 1660 37 GTX 980 35 GTX 980 Ti 24 GTX 1060 18 GTX 750 16 GTX 950 15 GT 730 11 GTX 760 11 GTX 770 11 Quadro P1000 11 Quadro K620 10 Quadro P2000 10 GTX 650 8 GTX 660 8 GTX 780 7 Quadro P4000 7 GTX 650 Ti 6 GTX 960M 6 Quadro K2200 6 TITAN V 6 TITAN X Pascal 6 GTX 1060 with MaxQ Design 5 GTX 980M 5 GTX TITAN X 5 MX150 5 Quadro RTX 4000 5 GTX 1650 with MaxQ Design 4 GTX 680 4 P106090 4 Quadro K4200 4 Quadro M4000 4 Quadro P5000 4 Quadro P600 4 RTX 2070 with MaxQ Design 4 Tesla M60 4 940MX 3 GT 640 3 GTX 745 3 GTX 750 980MB 3 GTX 780 Ti 3 GTX 950M 3 GTX TITAN Black 3 Quadro P2200 3 Quadro P620 3 Quadro T1000 3 RTX 2080 with MaxQ Design 3 TITAN Xp COLLECTORS EDITION 3 840M 2 GT 740 2 GTX 1650 Ti 2 GTX 1660 Ti with MaxQ Design 2 GTX 650 Ti BOOST 2 GTX 660 Ti 2 GTX 670 2 GTX 765M 2 GTX TITAN 2 MX130 2 MX250 2 P104100 2 Quadro K4000 2 Quadro K5000 2 Quadro M2000 2 Quadro T2000 2 RTX 3090 2 Tesla K20Xm 2 Tesla T4 2 Tesla V100PCIE16GB 2 GT 650M 1 GT 740M 1 GTX 1050 with MaxQ Design 1 GTX 1070 with MaxQ Design 1 GTX 1080 with MaxQ Design 1 GTX 645 1 GTX 770M 1 GTX 780M 1 GTX 880M 1 GTX 970M 1 Quadro K6000 1 Quadro M1000M 1 Quadro M1200 1 Quadro M3000M 1 Quadro M6000 1 Quadro P3000 1 Quadro P3200 1 Quadro P400 1 Quadro P4200 1 Quadro RTX 3000 1 Quadro RTX 8000 1 Quadro T2000 with MaxQ Design 1 RTX 2070 Super with MaxQ Design 1 RTX 2080 Super with MaxQ Design 1 RTX 3070 1 Tesla K20c 1 Tesla K40c 1 Tesla K80 1 Tesla P100PCIE12GB 1 TITAN RTX 1 |
 Send message Joined: 8 Aug 19 Posts: 252 Credit: 458,054,251 RAC: 0 Level  Scientific publications 
|
Very much appreciate your effort, Rod4x4. You've provided me with a better idea of how they stack up running MDAD tasks than I previously had. Thanks. |
|
Send message Joined: 8 Aug 19 Posts: 252 Credit: 458,054,251 RAC: 0 Level Scientific publications
|
One thing that throws a wrench into rod4x4's comparison is the ADRIA-Bandit WU. Every time one of my machines gets one, that host's RAC takes a beating. I've also found that if restarted they throw an error, even on a single GPU host. That showed up during my last driver update. Don't know if anybody else had this happen. |
 Send message Joined: 13 Dec 17 Posts: 1404 Credit: 8,898,646,190 RAC: 7,548,451 Level  Scientific publications
|
I think Ian has mentioned the same. |
|
Send message Joined: 21 Feb 20 Posts: 1112 Credit: 40,764,483,595 RAC: 7,379,314 Level Scientific publications
|
I know I've seen that before, but don't remember specifically which ones. I think I saw that behavior on the PABLO tasks.
|
|
Send message Joined: 8 Aug 19 Posts: 252 Credit: 458,054,251 RAC: 0 Level Scientific publications
|
I remember Ian mentioning it about ADRIAs when they first came out and comparing them to PABLO WUs, as they are similar in size but ADRIA WUs give around half the points that PABLOs did. (I searched the forum for "ADRIA" and "Adria" but couldn't find it.) Speaking of broken tabs, the donation page and the profile creation page haven't worked for me since I started last year. There seems to be a problem with the "I am not a robot" verification picture is missing. |
|
Send message Joined: 4 Aug 14 Posts: 266 Credit: 2,219,935,054 RAC: 0 Level Scientific publications
|
One thing that throws a wrench into rod4x4's comparison is the ADRIA-Bandit WU. Statistically, all volunteers should experience similar issue. Outliers will still occur, but will only have a small and temporary effect on the statistics. |
|
Send message Joined: 8 Aug 19 Posts: 252 Credit: 458,054,251 RAC: 0 Level Scientific publications
|
https://www.gpugrid.net/forum_thread.php?id=5125&nowrap=true#55153 Ian C wrote: The VillinAdaptive WUs also pay a lot less credit reward as compared to the pablo tasks, per time invested. Found it! Now, these are labeled ADRIA_NTL9Bandit100ns. I might be comparing apples to oranges, but both seem to credit about half the normal cobblestones. |
|
ServicEnginIC Send message Joined: 24 Sep 10 Posts: 591 Credit: 11,739,836,510 RAC: 9,833,974 Level Scientific publications
|
Ian C wrote: It's been happening since long time ago. I found this older "Immensive credits difference: ADRIA vs. PABLO tasks" thread even expresely mentioning it. And regarding this rod4x4 thread "Performance Tab still broken" topic, I agree that Adria work units may be a good tool for a performance comparative, due to their consistence in number of calculations per task. |
|
Send message Joined: 8 Aug 19 Posts: 252 Credit: 458,054,251 RAC: 0 Level Scientific publications
|
Thanks ServicEnginIC, since nothing's changed I guess "Ours is not to question why, ours is just to crunch, not cry!"
I'm all for that, as long as times are used instead of points per day for rating the cards. That would negate the effects that errors and WU granted credit variances have on a host's RAC. I see credit acquisition rates as a function of the entire host machine's output, rather than just the GPU. (Capt. Obvious, I am) Time for me to display more newbie naivete...🤔 Is it possible to write a script to glean only the data from ADRIA tasks? (Edit) Thanks again for your work on time comparisons of GPUs running the PABLO tasks, ServicEnginIC! |
|
Send message Joined: 4 Aug 14 Posts: 266 Credit: 2,219,935,054 RAC: 0 Level Scientific publications
|
I'm all for that, as long as times are used instead of points per day for rating the cards. That would negate the effects that errors and WU granted credit variances have on a host's RAC. I see credit acquisition rates as a function of the entire host machine's output, rather than just the GPU. I cannot use task runtimes for several reasons: 1. I don't have access to the backend data The frontend data on the volunteer tab is the only easily accessible data. That is why I always quote this page as the source. Scanning 3000 hosts takes less than 270 seconds. I have calculated that scanning 100 tasks for 20 hosts on each of the 150 volunteer pages scanned could take several days. As a comparison, a SQL query on the backup data would only take a few minutes to run (or less) 2. variable length of runtime of the MDAD task makes for a complex calculation. The variable runtime of the tasks would need a defined measure of the work unit calculation to allow for a meaningful comparison. It would not be a full comparison if we only concentrate on one kind of work unit. It should also be considered that there are runtime variabilities in each generation of ADRIA work units. The variance on the RAC is uniform across the board, so does not distract on the overall performance results.(Each user will see the same variance) It is correct to say credit is a function of the work unit output as completed on the GPU and host, ...as is the runtime... In my original post, I also pointed out that Toni has stated that credit calculation has been consistent over the years. It is the only constant we have for making a comparison at the frontend. I do admit the comparison is definitely not perfect, but useful enough to make a generalized comparison. It would be good and welcomed if better methods can be highlighted. More importantly, we would not have this dilemma if the Performance tab was working. Instead, I simply started this thread as an alternate way for performance comparison to be made, help GPUgrid users, take some workload off GPUgrid admins (fixing Performance tab would distract them from more important project work) and stimulate discussion. I hope this post is a vehicle for sharing ideas, provide support for GPUgrid and engender open discussion. |
|
Send message Joined: 8 Aug 19 Posts: 252 Credit: 458,054,251 RAC: 0 Level Scientific publications
|
I get it now, mate. Users can't access the level of data that Grosso can, and the existing server webpage interface is no longer providing pertinent data so it is essentially useless. I hope somebody on the team actually reads your post and corrects this problem. Thanks for your kindly responses and thanks double for your contribution to benchmarking GPU performance while Processing ACEMD platform based tasks. |
|
Send message Joined: 4 Aug 14 Posts: 266 Credit: 2,219,935,054 RAC: 0 Level Scientific publications
|
I have calculated that scanning 100 tasks for 20 hosts on each of the 150 volunteer pages scanned could take several days. As a comparison, a SQL query on the backend data would only take a few minutes to run (or less) I must have missed my morning coffee when I made that calculation. Having a further think about it, the script should take just under 30 minutes to grab the tasks for each host, not several days (what was I thinking). That then makes it viable to grab ADRIA task runtimes for each participating host. Might have this done by the end of the week. |
|
ServicEnginIC Send message Joined: 24 Sep 10 Posts: 591 Credit: 11,739,836,510 RAC: 9,833,974 Level Scientific publications
|
I've catched on past week in my systems some ADRIA work units, few, but enough for me to confirm some suppositions. * Graphics card ASUS TUF-GTX1650S-4G-GAMING, based on GTX1650 SUPER GPU, running at PCIE gen 3.0 x16 slot, got three. Execution times (seconds): WU#1: 18628,53 ; WU#2: 18660,36 ; WU#3: 18627,83 * Graphics card ASUS DUAL-GTX1660TI-O6G , based on GTX1660 Ti GPU, running at PCIE gen 2.0 x16 slot, got two. Execution times (seconds): WU#1: 16941,52 ; WU#2: 16998,69 -1) So far, execution times for ADRIA tasks are relatively consistent when executed at the same card model and setup. * Graphics card ASUS ROG-STRIX-GTX1650-O4G-GAMING, based on a factory overclocked GTX1650 GPU, got three. Execution times (seconds): WU#1: 23941,33 ; WU#2: 23937,18 ; WU#3: 32545,90 -2) Ooops!!! What's happened here? Execution time for WU#3 is very different to WU#1 and WU#2, being executed at the same graphics card model. The explanation: WU#1 and WU#2 were executed on a card installed at a PCIE gen 3.0 x16 slot, while WU#3 was executed on a card installed at a PCIE gen 3.0 x4 slot. Both are installed at the same multiGPU system, but due to mainboard limitations, only PCIE slot 0 is running at x16, while two more slots 1 and 2 are runing at x4, thus limiting performance for ADRIA tasks. Conclusion: Execution times are not only showing a particular graphics card performance itself, but also its particular working conditions at the system where it is installed. rod4x4, in my opinion, your decision to discard hosts with multiple GPUs makes complete sense due to this (and other) reason(s). It's a pity that ADRIA work units availability is unpredictable, and usually very transient... -3) rod4x4, thank you very much again for your labour. More importantly, we would not have this dilemma if the Performance tab was working. +1 |
|
Send message Joined: 13 Dec 17 Posts: 1404 Credit: 8,898,646,190 RAC: 7,548,451 Level Scientific publications
|
That GPUGrid tasks are heavily dependent on PCIE bus speeds has been posted and commented multiple times by Ian. I too see a noticeable slowdown on all GPUG tasks when comparing the same RTX 2080 cards (x3) from the two cards at X8 speeds compared to the card running at X4 speed. Not as big a difference as your test comparing X16 to X4 though. Einstein sees the same kind of differences though not as extreme compared to GPUG. |
©2025 Universitat Pompeu Fabra