Message boards :
Number crunching :
How many gpus can you run on an AMD AM4 socket motherboard? (Ryzen 7 1900, 2700 etc)
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 · Next
Author | Message |
---|---|
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
You could have probably bought one of those 'working' Splitters with the money you've spent on cables. I'm still using the same cables I was a year ago on my mining machine. I did buy a few newer cables, didn't make any difference on the slots that report as pcie 2. I do have a spare board, one day I might swap it out and see how many of those slots report pcie2 instead of 3. Right now I'm kinda busy with other things. Simple test. This slot reports as pcie2, that slot says 3. Shut off machine, swap cables. Reboot, does the 2 slot now say 3 or not? If not, then it's Not the cable. |
Ian&Steve C. Send message Joined: 28 Sep 99 Posts: 4267 Credit: 1,282,604,591 RAC: 6,640 |
I have 2 of them. And have owned them for over a year. Mine work fine with good cables. *shrug*. Seti@Home classic workunits: 29,492 CPU time: 134,419 hours |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
You could have probably bought one of those 'working' Splitters with the money you've spent on cables. I'm still using the same cables I was a year ago on my mining machine. I did buy a few newer cables, didn't make any difference on the slots that report as pcie 2. I do have a spare board, one day I might swap it out and see how many of those slots report pcie2 instead of 3. Right now I'm kinda busy with other things. My simple test is figure out which gpu is "failing" and then put a new/known good cable in place. And 99.44% of the time it stops failing. Since the AMD MB I am using basically only has 2 slots that run gen 3. How would your technique help? Sigh, lets face it. We are all cranky old men (anyone over 40 mind you). Tom A proud member of the OFA (Old Farts Association). |
Mr. Kevvy Send message Joined: 15 May 99 Posts: 3776 Credit: 1,114,826,392 RAC: 3,319 |
Some cleanup was required and posts have gone away. Please keep it civil as per the policies on the left displayed always when writing a posting. This would be a good time for a reminder of the Filtered Users setting in your profile's Community Preferences, which can be used to put someone on ignore, and the tiny (and too often easily-missed) red X button at the bottom of every post between the post ID and Reply button. If a post is found to be offensive, it's always better to use that to report it rather than respond in kind and throw more fuel on the flames. Thank you. :^) |
Ian&Steve C. Send message Joined: 28 Sep 99 Posts: 4267 Credit: 1,282,604,591 RAC: 6,640 |
back on topic, Tom, how are those new cables working out so far? Seti@Home classic workunits: 29,492 CPU time: 134,419 hours |
Bernie Vine Send message Joined: 26 May 99 Posts: 9954 Credit: 103,452,613 RAC: 328 |
As the argument started again straight away can I suggest that you both put each other on ignore. Meanwhile another lock till things cool down |
ML1 Send message Joined: 25 Nov 01 Posts: 20283 Credit: 7,508,002 RAC: 20 |
Dear All, There are many ways to do things, and many approaches. Please keep to the facts? Please note an apt quote: “Before you speak ask yourself: Is it kind, is it necessary, is it true, does it improve upon the silence?†Regards, Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
I am currently running 9 gpus on a Biostar TB350-BTC motherboard with a Ryzen 7 2700. I have re-deployed /added 1 gtx 1070ti and 2 gtx 1070's to the system, removing the equivalent gtx 1060's. To re-deploy 2 more gtx 1070ti's I would need to add a couple of 8 pin power leads. This system now has a fairly rapidly increasing RAC and looks like it should hit or exceed a 300,000 RAC. Since I am running a 8c/16t cpu at 90% of the available cpu cores, this means I a pretty small number of cpu threads left to run cpu tasks on. Yes I am running a cpu to gpu ratio in the app_config.xml file. And -nobs on the commandline for the gpus at this time. May try some other combos to see if I can maintain production while running more shared cpu threads. I suppose the other interesting thing to note is the Intel box that I borrowed 2 gpus from and replaced those with gtx 1060 3GB's is still holding steady near 300,000 RAC. Even though I am now crunching 10 cpu threads on a non-Seti project. Tom A proud member of the OFA (Old Farts Association). |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
I am currently running 9 gpus on a Biostar TB350-BTC motherboard with a Ryzen 7 2700. Today I have wondered into the room where this box is running and it has been off (powered down) twice. So I have reset the cpu speed to "auto" which should drop it down to around 3GHz and I have moved a small fan that was blowing across the system back up onto the table. Lets see if it runs all night. Tom A proud member of the OFA (Old Farts Association). |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
back on topic, I have not had ANY lost gpu experiences since I switched to the UGREEN 1' 6" cables. The description calls them "data transfer" cables for USB drives. All I can say for sure is they sure are humming along. Tom A proud member of the OFA (Old Farts Association). |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
I am currently running 9 gpus on a Biostar TB350-BTC motherboard with a Ryzen 7 2700. Nothing has quit since I dropped the cpu back to "auto". Today is Sunday afternoon. I guess I will try my previous lowest turbo speed 3.7GHz and see if it runs at that. ----edit--- apparently I already have reset it to 3.7GHz. I don't remember. I thought I had set it down to "auto". I have been trying to run it faster than 3.7GHz but apparently I have an institutional limit. Nothing I have will run faster than 3.7Ghz no matter what model... :) ---edit--- Tom A proud member of the OFA (Old Farts Association). |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
The Biostar TB350-BTC has 3 gen 3 pcie slots and 3 gen 2 pcie slots. A couple of days ago I was re-reading my motherboard manual and noticed this. The Nvidia X server utility/app under Linux was report 2 cards with Gen 3 but all the rest with Gen 2. So I just got done swaping my 1 to 4 extender card into a Gen2 slot since it was only running at Gen 2 anyway. And now I have a gtx 1070Ti and two gtx 1070's running at Gen 3 speeds. Tom A proud member of the OFA (Old Farts Association). |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
As part of my "never ending quest" for the "optimum" combination of cpu and gpu threads to generate the maximum total production I have switched off the -nobs command line (because any experiment with it on causes my cpu's to peg to 100%). And I am now running 0.49 cpus per gpu. Prior to this with the local Boinc Manager set to 90% cores I had 9 gpu/cpu threads and 5 cpu threads. I now have 9 gpu/part cpu threads and 10 cpu threads. So far it looks like about half my cpu task threads are allocating 1% of the available 6% to gpu tasks. About half of my cpu threads are running at a steady 6%. And because I don't have -nobs enabled, the gpus are running/stopping cycling between 0% and 1%. And yes the gpu tasks seem to be running slower. I will be watching my production graph to see if it suddenly stops going up and sorta flattens out. Tom A proud member of the OFA (Old Farts Association). |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
You must have turbo enabled. The base clock for the 2700 is 3.2Ghz. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
You must have turbo enabled. The base clock for the 2700 is 3.2Ghz. I do have the CPB enabled. But my last experience with the multiplier set on "auto" showed it running at just below 3GHz even though the CPB was enabled. The bios may not support other turbo features. At least not under something I have recognized. Since it appears it will run reliably at 3.7GHz I think I will leave it there. Yes, I have boosted the cpu voltage offset to the recommended 1.35 volts and still had the system stop as previously reported. No this bios doesn't appear to have LCC (sp). Oh, well. Unless I try above 3.7GHz it seems to be perfectly stable. Tom A proud member of the OFA (Old Farts Association). |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
You would never use a core voltage offset of 1.35V. Maybe 0.135V. You can set core voltage to a manual level of 1.35V for a fixed multiplier of 37X. Your motherboard must be very strange compared to my ASUS C7H boards. I just leave core voltage on Auto with the fixed multiplier of 40X or 40.25X and the BIOS adjusts the core voltage accordingly to compensate for loading. I have CPB Enabled and Performance Boost Override set to Auto. Under BOINC loading my core voltage ends up around 1.33-1.34V as automatically set by the BIOS. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
That is the final voltage. Not the offset. Your right though. I think this bios/mb is essentially a version 1 because I don't recognize any Precision Boost toggles at all. And presumably a Ryzen 7 2700 has that. I am going to drop the voltage back a bit lower though. The voltage offset has "green, yellow, and red". It takes a red level offset to get the cpu voltage up to where you and another mentioned said you were running. It is still running stable and none of the gpus has decided it doesn't want to play anymore. And so far I am getting a 3 lanes gen 3, 3 lanes gen 2. Tom A proud member of the OFA (Old Farts Association). |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
That is the final voltage. Not the offset. I googled around and ran into a review that had a Ryzen 7 2700 running 4GHz using 1.431 volts. I have managed 3.9GHz@1.410 volts. 3.95GHz crashed. So far 3.9GHz has run "overnight". I also checked the on the date of the latest bios update and it is still March 5. https://www.biostar.com.tw/app/en/mb/introduction.php?S_ID=882#download Recap of current settings: So I am currently running a cpu at 3.9Ghz, without -nobs which is allowing me to run 9 gpu threads and 10 cpu threads. The local Boinc Manager is set to 90% of available cores (8c/16t) so in theory I am limited to 14 cpu threads total. But I am running the ratio of cpus to gpus at 0.49 cpus to 1 gpu. My RAC is still climbing at about the same slope as before so it is possible I haven't slowed anything down. And may have not increased my final production either. Tom A proud member of the OFA (Old Farts Association). |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
Boy that didn't last long. I am back to the 3.7GHz because after trying to push it faster, it would not stay stable at 3.9GHz. And the graph was looking iffy so I switch back to the gpu maximum setup ( 1 to 1 ratio with -nobs in the app_info.xml commandline). When I hit my RAC "plateau" I still want to see what happens with more cpu threads and the gpus running slightly slower. Tom A proud member of the OFA (Old Farts Association). |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
I doubt your cheap Biostar motherboard has the necessary robust VRM components to allow much more than stock. Probably only a 1 + 4 VRM design like much of the B350 motherboards. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
©2024 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.