Message boards :
Number crunching :
Ryzen and Threadripper
Message board moderation
Previous · 1 . . . 11 · 12 · 13 · 14 · 15 · 16 · 17 . . . 69 · Next
Author | Message |
---|---|
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Now downloaded. I re-ran benchmarks and re-read prefs. Is there anything else I need to do to make BOINC aware? I still don't see any enumeration of the OpenCL drivers in your hosts' Details page. You should see OpenCL 1.2 right after the CUDA driver version for the GTX 1060 on the Coprocessor line in Details. To be sure look at the beginning of the Event Log and you should two entries for the 1060. One for the CUDA drivers followed by one for the OpenCL drivers. You do have to reboot the system to use any newly installed graphics driver. A quick and easy way to see if you have both CUDA and OpenCL components is to run clinfo from the Terminal. It prints out the CUDA driver version, the OpenCL versions and how many detected graphics platforms on the system. sudo apt install clinfo 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 |
I see you have OpenCL installed now and I also see a opencl_sah task in the cache. No completed OpenCL tasks so far. The schedulers will send the host the three different varieties of the OpenCL apps with one of them the SoG type. Eventually as more tasks are returned, the schedulers will settle on the SoG application as the fastest and then that type of task will be all that you get for the gpu. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
jsm Send message Joined: 1 Oct 16 Posts: 124 Credit: 51,135,572 RAC: 298 |
Yes indeed thanks to you for the guidance on these drivers. I thought the readers might like to see the 'helpful' guidance given by ASUS suoort after I passed the problem of loss of performance every 19 seconds. I quote:- This is the another recommended resolution of our Product Engr. Troubleshoot kindly: Go to BIOS and check if there is temperature spikes every 19 seconds (5+ or 10+ degrees higher) For Ubuntu, Keep Ubuntu updated. Check for bent pins on the socket Use another Ubuntu distribution version- 18.10 or Ubuntu 19.04 and see if issue resolved or not Last and not lest, final solution is that the processor not working properly; please RMA the processor. Interesting? jsm |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
IOW, they don't have a clue about anything to do with Linux. The comment about watching for temps is the best deductive reasoning, If the temp is spiking past Tmax, the processor WILL reduce core clocks to reduce temps. Yes, the cpu freq_governor has been updated in later versions of Linux to be more compatible with AMD processors. The version in Ubuntu 18.04 is still biased more toward Intel processors with rudimentary accommodation for modern AMD processors. But I don't see the kind of behavior you show in your graphs in any of my AMD hosts. I understand how the processors work and make allowances for their weaknesses so that they work at their best. I have very good cooling on the cpu so that I always stay below the Tmax of 68°C. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
jsm Send message Joined: 1 Oct 16 Posts: 124 Credit: 51,135,572 RAC: 298 |
I think you are still using 18.04. As 19.04 is stable is there a reason for not moving on and up? jsm |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13750 Credit: 208,696,464 RAC: 304 |
As 19.04 is stable is there a reason for not moving on and up? If it isn't broken, then don't fix it. Basically if something's working, why tempt fate by fiddling with it? Grant Darwin NT |
jsm Send message Joined: 1 Oct 16 Posts: 124 Credit: 51,135,572 RAC: 298 |
Logic suggests that the regularity of the drop off is such that this is proactive rather than reactive to a temp increase. The suggestion of simply sending the processor RMA seemed to me to be 'I have no idea so assume it is defective". jsm |
MarkJ Send message Joined: 17 Feb 08 Posts: 1139 Credit: 80,854,192 RAC: 5 |
Back to the X570 boards, apparently the chipset won't work with 1st gen Ryzens. It only supports 2nd gen and 3rd gen. Not that I could think of any reason why someone would buy a new motherboard and stick a 1st gen CPU on it. BOINC blog |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
Back to the X570 boards, apparently the chipset won't work with 1st gen Ryzens. It only supports 2nd gen and 3rd gen. Not that I could think of any reason why someone would buy a new motherboard and stick a 1st gen CPU on it. Yupper. Apparently these models will be expensive enough that the only reason to buy one is if you are upgrading to a Ryzen 3000 and really, really need the Pcie gen4 and other goodies. 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 think you are still using 18.04. As 19.04 is stable is there a reason for not moving on and up? Each of the interim releases is only good for 9 months before being retired. I prefer to stay on LTS releases good for 6 years of support. I will upgrade next year at the 20.04 LTS release. Looking forward to that as the new 5.00 kernels are FAST. As Grant states, if it ain't broke don't fix it. 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 |
Logic suggests that the regularity of the drop off is such that this is proactive rather than reactive to a temp increase. Sure you can to do that and take a chance that they would accept the RMA as defective and replace the cpu. But then, then they also could say that there is nothing wrong with the cpu and then you have to buy a new cpu. Or they could then just send you back your original and you are out shipping both ways for no gain. Ask Zalster how that went with his recent RMA of his i7-6950K. He was told, nope not defective, you broke it. You have to purchase a new one. The one time I RMA'd a graphics card turned out there was nothing wrong with the card and I was sent a replacement B-stock item in return that had nowhere near the performance of my original. I lost out on that exchange. I still maintain you are seeing the natural behavior of the cpu at load with XFR2 and PBO in action. Did you ever try turning all those off, reduce down to locked base clocks and then observe the behavior? Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
jsm Send message Joined: 1 Oct 16 Posts: 124 Credit: 51,135,572 RAC: 298 |
I have absolutely no intention of returning the cpu unless it dies. I am only too aware of what you describe. However I am suspicious of the regularity. I cannot believe that the temp jumps up exactly on the same interval time after time. Thus any suggestion that it is a reactive process of whatever kind is difficult to fathom. It seems much like a process which has been triggered and continues for ever. I suppose that triggering could be because of a temp limit being breached but then surely it should be stopped when the temp was reduced. Also I have obtained sensor input through the drop off regularly and the temps have been in the low 60s throughout, Its interesting if a bit confusing. No I haven't followed the off with XFR2 etc yet as I have concentrated on getting the new GPU installed and working correctly. I will give that a week to settle down and then do a bit more experimenting. However I may upgrade to 19.04 as it does have better support of AMD and I dont mind that it only has a 9 month support life as I would go to the next LTS as soon as it became available anyway. jsm |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
If you have no issues with the interim release life, then I say you should update to 19.04. I have that on my test partition. It is using the 5.0 kernel. Which is fast. It is using GNOME 3.32. Which is faster. It leads to a perception that the OS is much faster in context switching, user UI response and Desktop redraws. It supposedly has better frequency governor and kernel improvements for the newest processors. I gather my knowledge from reading the daily news articles at Phoronix.com. That site does the best job in posting news about Linux developments in my opinion. 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 |
If you consider an Epyc 7601 cpu as a stand-in for the Threadripper 2990WX, then you might be interested in the charts comparing the progression of performance from the kernel of Ubuntu 17.04 to the latest kernel 5.2 with znver1 optimizations and with Spectre mitigations turned off in the compiler. The article is at Phoronix. https://www.phoronix.com/vr.php?view=27960 The summation at the end of the article nicely illustrates the advantages of using a newer kernel compared to the Ubuntu 18.04 kernel. These performance improvements are a culmination of both AMD/Zen-specific improvements throughout the stack as well as general improvements we've been covering each kernel cycle as well as further enhancements to the GCC toolchain and elsewhere in the stack. An 11% improvement over the past two years is nice especially considering that Intel systems in affected workloads generally see around an 18% performance hit overall from the combined Spectre/Meltdown/L1TF/Zombieload mitigations to date. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
jsm Send message Joined: 1 Oct 16 Posts: 124 Credit: 51,135,572 RAC: 298 |
Now running 18.10 for a day to ensure stable on this 2990. Will up to 19.04 tomorrow night if OK. jsm |
jsm Send message Joined: 1 Oct 16 Posts: 124 Credit: 51,135,572 RAC: 298 |
Confirm upgrade to 19.04 without drama. GPU tasks now taking between 7 and 9 minutes but cpu still well over an hour and the 1 second in 20 drop off is still occurring. jsm |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Again, I maintain what you are observing is natural for modern AMD processors. Nothing to be concerned with. You should get this utility and run it it help illustrate what I am speaking of. https://www.reddit.com/r/linux/comments/9djs3b/monitoring_the_power_consumption_of_amd_ryzen_and/ I put in on my daily driver and it illustrates quite nicely the power used by the cores and where the loads are. You can see the loads very nicely walking around the dies. All normal. 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 |
And another "leaked slide" surfaces today showing the Ryzen 3 R9-3950X back in the SKU mix. But I don't believe it is real because the TDP remains at 105W even though it supposedly boost clocks another 100Mhz higher than the R9-3900X part which is also rated at 105W TDP. https://videocardz.com/newz/amd-ryzen-9-3950x-to-become-worlds-first-16-core-gaming-cpu 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 |
And another "leaked slide" surfaces today showing the Ryzen 3 R9-3950X back in the SKU mix. But I don't believe it is real because the TDP remains at 105W even though it supposedly boost clocks another 100Mhz higher than the R9-3900X part which is also rated at 105W TDP. It would be amazing at that TDP. I think I would be tempted because I could run it on one of my current AM4 MB's. But it would be more believable at the higher TDP you posted earlier. If I am hearing it right. We get more info today. At least we are promised more Navi info. I don't know if we are promised more AM4 info. 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 still will wait and see before believing. It is possible I guess that it might be rated at 105W TDP, I don't know how 7nm scales with frequency. Notice the base clock dropped 300Mhz compared to the 12 core part. I still think someone is fudging numbers. 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.