Message boards :
Number crunching :
Setting up Linux to crunch CUDA90 and above for Windows users
Message board moderation
Previous · 1 . . . 98 · 99 · 100 · 101 · 102 · 103 · 104 . . . 162 · Next
Author | Message |
---|---|
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13736 Credit: 208,696,464 RAC: 304 |
Decided to take the plunge and moved 2 boxes to Linux. Should give the output a good lift. Grant Darwin NT |
Lionel Send message Joined: 25 Mar 00 Posts: 680 Credit: 563,640,304 RAC: 597 |
Decided to take the plunge and moved 2 boxes to Linux. Possibly. I have removed a gpu from syphilis (now down to 2 gpus from 3). Smallpox (2 gpus) hasn't been running for a few months so output has been zero on that front. I do take your point though, it will be interesting to see where they land. Syphilis was averaging around 42k prior to the GTX 780 TI coming out so will be interesting to see if the throughput keeps things constant or not. |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
Decided to take the plunge and moved 2 boxes to Linux. . . They should achieve 6 digit RACs. Stephen :) |
Lionel Send message Joined: 25 Mar 00 Posts: 680 Credit: 563,640,304 RAC: 597 |
Decided to take the plunge and moved 2 boxes to Linux. How's the Ryzen project going? |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
How's the Ryzen project going? . . It's still running Windows and SoG. I have tried a few times to get Linux up and running SETI but while I can get it going, after a short while it loses the eternet port and goes permanently offline. But when I boot back into Windows the port is just fine again. There is something funny about the Linux drivers and this Ryzen mobo. So far it is above my pay grade. Stephen <shrug> |
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
I have only had that type of a problem on 1 computer - happens that is a 1700x/Gigabit board. In my case it is when a PCI extender is flaky and drops offline. That takes down the Ethernet, be it wirer or wireless adaptor. All I can say is leave NVidia-smi open to see if they go down together. |
JohnDK Send message Joined: 28 May 00 Posts: 1222 Credit: 451,243,443 RAC: 1,127 |
Some weeks ago there was talk about a new version (99b1?) of the special app been tested by petri, as I recall it sounded as it was close to be ready for release. Any news about that? |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Still in development I guess. Haven't heard a peep of any progress report. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
I have only had that type of a problem on 1 computer - happens that is a 1700x/Gigabit board. . . No extenders in this rig, it is running just the two 1060s. A shame because it should be my best producer. Stephen :( |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
How's the Ryzen project going? What Ryzen motherboard is it? Have you updated the Linux firmware so that it has the latest motherboard drivers? Do you have a lot of SATA and M.2 drives installed? The Ryzen cpu and PCH chipset can only support certain configurations with certain devices. The host runs out of PCIe resources. I just found this really neat image of the new X570 motherboard PCI layout with how it handles resources. This demonstrates why for example you can't plug a gpu into the X4 slot and also install a M.2 drive that is served out of the PCH. One or the other, but never both can be installed. Notice how the Wi-Fi and Ethernet port are served out of the PCH. If you have a gpu installed into the X4 slot you have to configure the BIOS to separate the PCI resources coming out of the PCH. For example this is part of my BIOS configuration. M.2_2 PCIe Bandwidth Configuration [Auto] Auto-detects the M.2_2 slot. If an M.2 device is detected, PCIEX16_1 and PCIEX8_2 will run at x8/x4 mode, respectively. [Disabled(X8 mode)] Disable M.2_2 for PCIEX16_1 and PCIEX8_2 running on high performance. (x16 or x8/x8) [Disabled(X4/X4 mode)] Disable M.2_2 with PCIEX8_2 running on x4/x4 for PCIE Raid card. PCIEX16/X8_1 Mode This item allows you to set the link speed for PCIEX16/X8_1 slot. Configuration options: [Auto] [GEN 1] [GEN 2] [GEN 3] PCIEX8/X4_2 Mode This item allows you to set the link speed for PCIEX8/X4_2 slot. Configuration options: [Auto] [GEN 1] [GEN 2] [GEN 3] PCIEX4_3 Mode This item allows you to set the link speed for PCIEX4_3 slot. Configuration options: [Auto] [GEN 1] [GEN 2] Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
. . OK, this is the hardware list... ASUS PRIME X370-Pro mobo 2 x Corsair Vengeance DDR4 2666MHz 8GB ram EVGA SuperNova G2 750W Gold PSU Samsung 850 EVO 250GB SSD on 0-0-0 (Windows drive) Sandisk 120GB SSD on 1-0-0 (Linux drive) Sandisk Ultra 16GB USB 3.0 flashdrive on Port 4 Hub 1(SETI data drive for Windows) Seagate 1TB ST1000 HDD on 4-0-0 CD Drive ASUS DRW 24D5MT on 5-0-0 GTX1060-6GB on PCIe x 16 slot 1 GTX1060-6GB on PCIe x 16 slot 3 (which runs in x4 mode) . . I believe that should be conflict free for on board mobo resources. Stephen ? ? |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
That is a lot of storage drives attached to your computer. They all need resources. I would refer you to section 1-6 in your manual for the IRQ interrupt table which shows which slots, devices and controllers share IRQ resources of which one of them is the LAN chip served out of the PCH chip. I never had more than 3 storage devices connected and have never lost the LAN port except for the motherboard I killed when I scraped a SMD device off the backside of the board near the cpu socket. That was self-inflicted. [Edit]Not sure about your designations on your drives. Maybe port#'s? The 4 and 5 ports are served out of the Asmedia controller and there is a disclaimer in the manual not to use those ports if you intend to use a graphics card in one of the PCIe slots are the speeds will get pulled down to X4 or something. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Lionel Send message Joined: 25 Mar 00 Posts: 680 Credit: 563,640,304 RAC: 597 |
Might sound like a silly question, but is there a thread around where you can see/post comments on the, I think it is, x41p_V0.98 optimised app with Linux? |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
I would say either this thread or the High Performance Linux clients at SETI Seti@Home classic workunits:20,676 CPU time:74,226 hours 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 |
Missed this news yesterday. It looks like the .ISO image for Ubuntu 19.10 is going to come with the proprietary Nvidia 390 and 418 drivers in it. So you will be able to install them after the initial default Nouveau driver installation without having any network connection. https://www.phoronix.com/scan.php?page=news_item&px=Ubuntu-19.10-NVIDIA-On-ISO Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Stephen "Heretic" Send message Joined: 20 Sep 12 Posts: 5557 Credit: 192,787,363 RAC: 628 |
Missed this news yesterday. It looks like the .ISO image for Ubuntu 19.10 is going to come with the proprietary Nvidia 390 and 418 drivers in it. So you will be able to install them after the initial default Nouveau driver installation without having any network connection. . . That's a good move ... Stephen :) |
Lionel Send message Joined: 25 Mar 00 Posts: 680 Credit: 563,640,304 RAC: 597 |
|
Wiggo Send message Joined: 24 Jan 00 Posts: 34748 Credit: 261,360,520 RAC: 489 |
My hat's off to Petri for the Cuda code and to TBar for the packaging the deal and supplying a fixed version of BOINC (now if that stupid Notices tab could be moved all the way to the far right....), but if Seti's project url was also included in the readme that would just make it just so much easier for us dummies to just copy'n'paste). :-) As to an update on what's going on with my rejuvenated 2500K w/ dual GTX 1060's rig, here's a run down so far over the last 20hrs. The CPU app is about on par with the Lunatics AVX app, but I'm now doing 3 tasks in the same time as running 2 under Windows (so there's a 50% plus to start with). I didn't really pick a good time to do the change over for with all the noise bombs about, but those good blc25's that took around 11mins 30secs to do under the Windows SoG app went down to around 3mins 11secs and the blc26's that take 9mins under Windows are being done in around 2mins 30secs. The 1060's in the 2500K performance under Windows were achieving an APR of 264.99 GFLOPS and now with the special sauce under Linux that has now gone up to a whopping APR of 1,118.20 GFLOPS (and still rising). My total RAC stayed flat for yesterday, even with my 3750K rig's RAC still dropping and the 2500K rig being offline for around 10hrs (for cleaning, parts additions and replacements), just 1 real drawback I found out to having well hidden cables and wires was tracing them all back out again, but today it's already heading ballistic. :-O Cheers. |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
+1 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 |
Are you running with the "-nobs" in the commandline for the app_info.xml? If you decide you want to wring the very last production out, that will. Needs a core per gpu. Tom 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.