Message boards :
Number crunching :
Really dinky or old cpu models driving "huge" gpus....
Message board moderation
Author | Message |
---|---|
![]() Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 ![]() |
Here is a thread to talk about your story. I saw a Pentium 2 core apparently driving one or two gtx 1080Ti's as an example. The ones I am looking at are in the top 100 producers so they are getting VERY respectable levels of production. So talk about what you did and why you chose to... Tom A proud member of the OFA (Old Farts Association). |
![]() ![]() Send message Joined: 5 Sep 99 Posts: 1652 Credit: 1,065,191,981 RAC: 2,537 ![]() ![]() |
I think this computer fits that category. A dual core pentium cpu driving 1080ti and a 970. Its cpu starved and im doing alot of magic in cron to increase thread priority to setiathome executable. If i would exchange that motherboard, cpu etc to a new one the RAC wouldnt increase that much after all. https://setiathome.berkeley.edu/show_host_detail.php?hostid=8676036 ![]() _________________________________________________________________________ Addicted to SETI crunching! Founder of GPU Users Group |
Ian&Steve C. ![]() Send message Joined: 28 Sep 99 Posts: 4267 Credit: 1,282,604,591 RAC: 6,640 ![]() ![]() |
Not for SETI, but my Plex transcode server has a GTX 1660 attached to a Intel C2758 Atom processor lol. Only PCIe 2.0 8x though. https://imgur.com/a/3cpk9AQ Seti@Home classic workunits: 29,492 CPU time: 134,419 hours ![]() ![]() |
![]() Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 ![]() |
Not for SETI, but my Plex transcode server has a GTX 1660 attached to a Intel C2758 Atom processor lol. Only PCIe 2.0 8x though. Once upon a time I think I tried to run an Atom-powered computer on Seti@home. It probably took a VERY long time to process a cpu task.... I wonder if I should amend the title to include "really old model cpus" like here: https://setiathome.berkeley.edu/show_host_detail.php?hostid=8611128 I know I appear to be picking on Viper but he is just a "target of opportunity". :) Tom A proud member of the OFA (Old Farts Association). |
![]() Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 ![]() |
I haven't owned one of these in "several" years: https://setiathome.berkeley.edu/show_host_detail.php?hostid=8785790 And here is another "elderly" cpu model: https://setiathome.berkeley.edu/show_host_detail.php?hostid=8785790 Wasn't this one of the top producers of its generation? https://setiathome.berkeley.edu/show_host_detail.php?hostid=7284511 I skipped over several other Viper systems but he does seem to be specializing in driving pretty respectable sized gpus with tiny cpus. I also skipped over Anonymous" who is doing similar things to Viper. Here is another one: https://setiathome.berkeley.edu/show_host_detail.php?hostid=8738726 Tom A proud member of the OFA (Old Farts Association). |
![]() ![]() ![]() Send message Joined: 1 Apr 13 Posts: 1858 Credit: 268,616,081 RAC: 1,349 ![]() ![]() |
Here is a thread to talk about your story. https://setiathome.berkeley.edu/show_host_detail.php?hostid=8729943 Is a Xeon E5450 (basically a 3ghz Core2Quad) w/ 4xGT980s. Currently pulling a pretty respectable 180k+ RAC doing GPU-only crunching, which puts it #37 in the top 100, two behind the Z600. Definitely no head-room to do CPU tasks, but with 1 core per GPU it seems to do OK. Doubt I'll tempt fate and try to add more GPUs with a switched riser. Thing is, that old Socket 775 mobo just kicks butt, and it seems to outperform the HP boxes when it comes to servicing the PCIE slots. It also seems to have faster memory access. I think it's just how Gigabyte built the i/o architecture. The PC itself was a freebie from a neighbor, so even with some better memory, a better CPU and cooler and a decent power supply I got into it for pretty cheap. Might add an SSD at some point, though I'm not sure how much speed that would add. The HP Z600 https://setiathome.berkeley.edu/show_host_detail.php?hostid=8807028 has dual X5675s, and seems to be a bit faster than the Gigabyte. I just scored a 4-slot switched riser, so I'm running 1x980 internal and 6 GPUs external (3x980 and 3x750ti) external. Also seem to have the headroom to do all that and still crunch away on 9xCPU tasks while keeping the utilization reasonable. No clue what the RAC will end up at on that one, as I just added the riser, but it looks to be passing through 200k as I write. Will have to wait to tweak until it levels out a bit. Why, because dual CPU boxes are cool and I couldn't bear to have all those threads going to waste. Besides, it's cool here now and I need the heat. I just rearranged the equipment layout to place it where all that nice heat rises up the basement stairs.) Or perhaps I'm just really, really bored right now ... I did spend the afternoon sorting nuts,. bolts and screws :) Bottom line, I have no plans to add more boxes at this point, so it's a game of optimizing what I have to increase performance while maintaining or reducing power consumption. 2000 kwh / mo is a bit nuts ... 980s are power hogs, but they were cheap and I have them, so ... Later, Jim .... ![]() ![]() |
Ianab Send message Joined: 11 Jun 08 Posts: 732 Credit: 20,635,586 RAC: 5 ![]() |
Thing about that machine is the G3258 is basically 1/2 of an i5-4xxx CPU. So while it's only got 2 cores, it's single thread performance is pretty good. Only problem is it's only got 2 threads. But if you only need to feed 2 GPUS, and flag away the CPU processing completely, it'a going to drive the GPU just as well as a multi-core multi-threading beast. Possibly better because the CPU wont be doing anything else, so it gets to make best use of it's cache and memory access, without having to fight with 10 other tasks that are running. Could swapping to an i7 could potentially drop the RAC? Anyway, if you look at an older PC like that, and say "What's the best way to make this crunch?". Then throwing in a big PSU and a couple of decent GPUs is going to do a LOT more than upgrading the CPU. |
![]() Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 ![]() |
Just found this new "dinky" cpu..... driving 4 gpus.... in the top 20.... https://setiathome.berkeley.edu/show_host_detail.php?hostid=8707336 Tom A proud member of the OFA (Old Farts Association). |
Ian&Steve C. ![]() Send message Joined: 28 Sep 99 Posts: 4267 Credit: 1,282,604,591 RAC: 6,640 ![]() ![]() |
Just found this new "dinky" cpu..... driving 4 gpus.... in the top 20.... maybe you posted the wrong link. a 7700k isnt that "dinky" and that system appears to be driving 7 GPUs, not 4. Seti@Home classic workunits: 29,492 CPU time: 134,419 hours ![]() ![]() |
![]() Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 ![]() |
Just found this new "dinky" cpu..... driving 4 gpus.... in the top 20.... Odd. That url is displaying a GenuineIntel Intel(R) Pentium(R) Gold G5400 CPU @ 3.70GHz [Family 6 Model 158 Stepping 10] (4 processors) [4] NVIDIA GeForce GTX 1060 6GB (4095MB) driver: 440.26 OpenCL: 1.2 On the leaderboard page. But it is clearly linking to that 7700k in the subsidiary page.... Odd. A proud member of the OFA (Old Farts Association). |
juan BFP ![]() ![]() ![]() ![]() Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 ![]() ![]() |
Just found this new "dinky" cpu..... driving 4 gpus.... in the top 20.... Maybe you need to refresh your browser cache because here it said: Tipo de CPU GenuineIntel Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz [Family 6 Model 158 Stepping 9] Número de processadores 8 Coprocessadores [7] NVIDIA GeForce GTX 1060 6GB (4095MB) driver: 440.26 OpenCL: 1.2 Exactly as posted by Ian. ![]() |
![]() ![]() ![]() Send message Joined: 6 Nov 99 Posts: 716 Credit: 8,032,827 RAC: 62 ![]() ![]() |
and the stderr.txt show 8 gpu ... setiathome_CUDA: Found 8 CUDA device(s): |
juan BFP ![]() ![]() ![]() ![]() Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 ![]() ![]() |
Seems like he has one of that "complicated to keep working" hosts look what shows now and no WU in process. <core_client_version>7.14.2</core_client_version> <![CDATA[ <stderr_txt> ome_CUDA: CUDA Device 7 specified, checking... Device cannot be used Cuda initialisation FAILED, Initiating Boinc temporary exit (180 secs) setiathome_CUDA: Found 6 CUDA device(s): Device 1: GeForce GTX 1060 6GB, 6075 MiB, regsPerBlock 65536 computeCap 6.1, multiProcs 10 pciBusID = 1, pciSlotID = 0 Device 2: GeForce GTX 1060 6GB, 6078 MiB, regsPerBlock 65536 computeCap 6.1, multiProcs 10 pciBusID = 2, pciSlotID = 0 Device 3: GeForce GTX 1060 6GB, 6078 MiB, regsPerBlock 65536 computeCap 6.1, multiProcs 10 pciBusID = 3, pciSlotID = 0 Device 4: GeForce GTX 1060 6GB, 6078 MiB, regsPerBlock 65536 computeCap 6.1, multiProcs 10 pciBusID = 4, pciSlotID = 0 Device 5: GeForce GTX 1060 6GB, 6078 MiB, regsPerBlock 65536 computeCap 6.1, multiProcs 10 pciBusID = 9, pciSlotID = 0 Device 6: GeForce GTX 1060 6GB, 6078 MiB, regsPerBlock 65536 computeCap 6.1, multiProcs 10 pciBusID = 12, pciSlotID = 0 In cudaAcc_initializeDevice(): Boinc passed DevPref 7 setiathome_CUDA: CUDA Device 7 specified, checking... Device cannot be used Cuda device initialisation retry 1 of 6, waiting 5 secs... setiathome_CUDA: Found 6 CUDA device(s): A bad connector riser cable? Who knows? ![]() |
![]() Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 ![]() |
A page refresh didn't seem to clear it. But a browser cache clear helped the next day. I can "see clearly now...." (sung in an off tune voice). Tom ps. I was wondering around the WWW and found someone trying to sell an (old) Amd Ryzen 1700 for more than the price of a new 2700.... A proud member of the OFA (Old Farts Association). |
rob smith ![]() ![]() ![]() Send message Joined: 7 Mar 03 Posts: 22652 Credit: 416,307,556 RAC: 380 ![]() ![]() |
As ever one has to be very careful about prices on some websites. I recently upgraded one set of subs, many places were offering them at less than 1k each, but one load of cheeky wotzitz were asking over 3k each for exactly the same thing (used - and using shots from the manufacturer's website of the serial number plates to "show they were genuine"). The web is well named - there are lots of unscrupulous spiders out there to trap the unwary. Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
©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.