Message boards :
Number crunching :
ubuntu Install
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · Next
Author | Message |
---|---|
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Hi Grant, sorry for mis- appellation. Yes, I concur the AVX app is most definitely faster under Windows. Not sure that is the case under Linux. So far, Rick's tests have put the AVX way down on the bottom in performance compared to everything else other than SSE2 on his 1950X system. I have to assume that a Ryzen system would perform similarly since the 1950X is just two 1800X dies with some special sauce thrown on it. I found the SSE4.1 app faster than the AVX app under Linux on the FX processor. As I stated, I am going to get a week's baseline performance with the SSE4.1 app then switch back to the AVX app. I have my suspicions that the AVX app WILL be faster on the Ryzen system under Linux too. But I want to test anyway instead of just taking a leap of faith. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
The SSE 4.1 version is faster on Linux with a Ryzen. I`m running it myself since i installed Mint 18.2. With each crime and every kindness we birth our future. |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
Thanks for the continued testing Rick. I have just now brought the Linux machine back online after the Ryzen 1800X upgrade. The hardware upgrade went fairly smoothly other than putting in a duff AIO that I had my suspicions about anyway. It never worked correctly in Windows but I thought the issue was a Windows thing. Turned out Windows had nothing to do with the problem, it was the hardware. Pump to be more specific. Fell back on the original AIO that was in place when it was a FX system. That is working fine now. Have to RMA the H110i on Monday. Hi Keith, I also had to work through a cooling problem over the weekend. I found my Windows 1950X system throttled at 1.5GHz @68.5C. Inspection found the coolant lines out of the CPU block had almost no coolant flowing. Ended up being gummed up coolant in the CPU block. Looks like the green Mayhems dye that I added to PrimoChill True coolant wasn't compatible. No problems in Mayhems coolant. I posted some photos on instagram. It just took me the evening to get the system back up, but temps look great now. 53C when fully loaded. I agree with the concern on system monitoring software in Linux. I am using lm-sensors. Reports back good GPU temps, but CPU temps and Vcore don't look right. So far have not found a good solution. I am setting a Vcore (not offset) in BIOS, so there should be no issue with it running at some other voltage, but it would be great to have software confirmation. I have switched Eos over to sse4.1 based on initial data. The 14 WU experiment is still running. Should be done tomorrow. I will post the results when complete. GitHub: Ricks-Lab Instagram: ricks_labs |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
I agree with the concern on system monitoring software in Linux. I am using lm-sensors. Reports back good GPU temps, but CPU temps and Vcore don't look right. So far have not found a good solution. I am setting a Vcore (not offset) in BIOS, so there should be no issue with it running at some other voltage, but it would be great to have software confirmation. I don`t get temps for my CPU with lm-sensors ? With each crime and every kindness we birth our future. |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
I agree with the concern on system monitoring software in Linux. I am using lm-sensors. Reports back good GPU temps, but CPU temps and Vcore don't look right. So far have not found a good solution. I am setting a Vcore (not offset) in BIOS, so there should be no issue with it running at some other voltage, but it would be great to have software confirmation. Here is my output of sensors command of lm-sensors: amdgpu-pci-4700 Adapter: PCI adapter fan1: N/A temp1: +48.0°C (crit = +0.0°C, hyst = +0.0°C) amdgpu-pci-4300 Adapter: PCI adapter fan1: N/A temp1: +44.0°C (crit = +0.0°C, hyst = +0.0°C) amdgpu-pci-0b00 Adapter: PCI adapter fan1: N/A temp1: +46.0°C (crit = +0.0°C, hyst = +0.0°C) amdgpu-pci-4800 Adapter: PCI adapter fan1: N/A temp1: +48.0°C (crit = +0.0°C, hyst = +0.0°C) amdgpu-pci-4400 Adapter: PCI adapter fan1: N/A temp1: +46.0°C (crit = +0.0°C, hyst = +0.0°C) amdgpu-pci-0c00 Adapter: PCI adapter fan1: N/A temp1: +48.0°C (crit = +0.0°C, hyst = +0.0°C) nct6779-isa-0290 Adapter: ISA adapter Vcore: +0.57 V (min = +0.00 V, max = +1.74 V) in1: +1.09 V (min = +0.00 V, max = +0.00 V) ALARM AVCC: +3.28 V (min = +2.98 V, max = +3.63 V) +3.3V: +3.28 V (min = +2.98 V, max = +3.63 V) in4: +1.80 V (min = +0.00 V, max = +0.00 V) ALARM in5: +0.93 V (min = +0.00 V, max = +0.00 V) ALARM in6: +1.21 V (min = +0.00 V, max = +0.00 V) ALARM 3VSB: +3.46 V (min = +2.98 V, max = +3.63 V) Vbat: +3.28 V (min = +2.70 V, max = +3.63 V) in9: +0.00 V (min = +0.00 V, max = +0.00 V) in10: +0.81 V (min = +0.00 V, max = +0.00 V) ALARM in11: +0.85 V (min = +0.00 V, max = +0.00 V) ALARM in12: +1.65 V (min = +0.00 V, max = +0.00 V) ALARM in13: +0.92 V (min = +0.00 V, max = +0.00 V) ALARM in14: +0.86 V (min = +0.00 V, max = +0.00 V) ALARM fan1: 1953 RPM (min = 0 RPM) fan2: 0 RPM (min = 0 RPM) fan3: 1890 RPM (min = 0 RPM) fan4: 1896 RPM (min = 0 RPM) fan5: 0 RPM (min = 0 RPM) SYSTIN: +31.0°C (high = +0.0°C, hyst = +0.0°C) ALARM sensor = thermistor CPUTIN: +43.5°C (high = +80.0°C, hyst = +75.0°C) sensor = thermistor AUXTIN0: +15.0°C sensor = thermistor AUXTIN1: +36.0°C sensor = thermistor AUXTIN2: +34.0°C sensor = thermistor AUXTIN3: +33.0°C sensor = thermistor SMBUSMASTER 0: +84.5°C PCH_CHIP_CPU_MAX_TEMP: +0.0°C PCH_CHIP_TEMP: +0.0°C PCH_CPU_TEMP: +0.0°C intrusion0: ALARM intrusion1: ALARM beep_enable: disabled GitHub: Ricks-Lab Instagram: ricks_labs |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Rick, so you know what your SIO chip is on your motherboard? Lm-sensors does not recognize my ITE 8665 chip at all. It supposedly can if you recompile the lm-sensors module with a configuration file for the IT87 and should then pick up the motherboard sensors on my ASUS Prime X370 Pro motherboard. But I have not been able to compile the code so far. Did you find a version of LM-sensors that worked with your motherboard at first shot? Or did you compile a new module like I am trying to do? This is the resource I am trying to use. monitor-amd-ryzen-temperatures-in-linux-with-latest-kernel-modules#h10-build-lm-sensors 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 |
Thanks for the post Rick. Looks like you have the NCT6779 SIO chip. I am assuming you built a new lm-sensors module to incorporate that chip by the process in my link. Was that so? If so, did you run into any issues compiling the module and how did you resolve them? Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
Thanks for the post Rick. Looks like you have the NCT6779 SIO chip. I am assuming you built a new lm-sensors module to incorporate that chip by the process in my link. Was that so? If so, did you run into any issues compiling the module and how did you resolve them? Hi Keith, No, I did not do a custom build. Just the standard download and answered all questions "Y" when doing the detection. I have the ASRock X399 Professional MB. |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Hi Rick, thanks for the reply. I guess you got lucky with lm-sensors already having your SIO chip configured and I am unlucky in that it doesn't have support for my ITE8665 chip. If I can't figure out how to compile the lm-sensor module for my chip, I guess I will have to do without or until the lm-sensors maintainers do the work for me in future releases. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
It took 40 hours to run, but finally have AVX vs SSE4.1 results on 14 WU running in Linux. Same as before, my setup is a Linux VM on my Win10 machine Nemesis. The VM is setup with 1 core and 8GB of RAM. The main machine is fully loaded (93%) with SETI CPU/GPU work. The VM runs the benchmarks while the main machine continues to run SETI. This way I can achieve a consistently loaded environment during the test. Here are my results: |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Thanks for the update Rick. The results are trending along with your previous tests it seems. A general consensus is that the AVX app is faster on BLC tasks and the SSE4.1 app is faster on Arecibo tasks. The question now should be ...... for how long are we going to receive Arecibo tasks and is it worthwhile to stay on SSE4.1 until they are no more and we only receive BLC tasks. And further out is the question, if and when we start getting Parkes data, what app is fastest for them. Back to other news. I have been successful in getting the ASUS Prime X370 Pro motherboard sensors to show up. I now have the CPU and VRM temps showing in GKrellm. I also have the CPU Fan, Fan1 and Fan2 showing. There are some voltages showing but they are nonsense until I can figure out how to scale them. VBat and 3VSB are showing correctly though. Turns out sensors-detect does not detect the SIO chip. And I also found out the chip is NOT the ITE-8665E as far as any software is concerned. The system identifies the chip as the ITE-8620. Once I did a modprobe at the correct address, sensors in terminal shows the chip output. I then added the module to my boot modules and now I have permanent sensor readout. modprobe it87 force_id=0x8620 keith@Darksider:~$ sensors asus-isa-0000 Adapter: ISA adapter cpu_fan: 0 RPM it8620-isa-0290 Adapter: ISA adapter in0: +1.48 V (min = +1.61 V, max = +2.29 V) in1: +2.75 V (min = +1.44 V, max = +0.66 V) in2: +2.15 V (min = +1.32 V, max = +1.42 V) in3: +2.18 V (min = +2.65 V, max = +2.84 V) in4: +0.04 V (min = +1.48 V, max = +0.68 V) in5: +0.04 V (min = +2.96 V, max = +3.02 V) in6: +0.04 V (min = +0.52 V, max = +2.87 V) 3VSB: +3.62 V (min = +0.31 V, max = +0.65 V) Vbat: +3.50 V fan1: 1430 RPM (min = 15 RPM) fan3: 873 RPM (min = 10 RPM) fan5: 726 RPM (min = -1 RPM) ALARM temp1: +40.0°C (low = -86.0°C, high = -51.0°C) sensor = thermistor temp2: +35.0°C (low = +58.0°C, high = -25.0°C) sensor = thermistor temp3: +30.0°C (low = +127.0°C, high = +126.0°C) sensor = thermistor temp4: +30.0°C temp5: +30.0°C intrusion0: ALARM keith@Darksider:~$ So for Mike and anyone else running the Crosshair VI or Prime X370 motherboard, try to modprobe the it87 module and see if you can find yours. If that doesn't produce any output you will have to probe some other alternate addresses such as 0x8622, 0x8607, 0x8628 and 0x8632. It looks like the ROG Crosshair VI Hero is at 0x8665. The driver is it87.ko and is in the: /lib/modules/4.10.0-38-generic/kernel/drivers/gpio/gpio-it87.ko /lib/modules/4.10.0-38-generic/kernel/drivers/hwmon/it87.ko directories or your equivalent distributions. Hope this helps others running a Ryzen under Linux. [Edit] A quick check with lsmod will print out all your current loaded modules. Look for the it87 module. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
I have observed what looks like a new problem with my Ubuntu 16.04 machine. Within a few days, it looks like all available memory will be in use. The top consumer of memory is Xorg root 1436 38.5 84.5 28801344 27798864 tty7 Rsl+ Nov-24 947:18:00 /usr/lib/xorg/XorgIf I run xrestop, Compiz is on the top of the list with Pxm_mem at 24220K. Anyone have experience with this? free -m total used free shared buff/cache available Mem: 32099 8831 21598 34 1668 22551 Swap: 65417 0 65417 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 15023 14985 34 2090 16269 Swap: 65417 0 65417 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 16046 13946 34 2105 15233 Swap: 65417 0 65417 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 16517 13472 34 2108 14760 Swap: 65417 0 65417 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 17448 12528 34 2121 13817 Swap: 65417 0 65417 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 17977 11993 34 2127 13283 Swap: 65417 0 65417 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 20703 9238 34 2157 10526 Swap: 65417 0 65417 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 20847 9104 34 2147 10379 Swap: 65417 0 65417 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 23277 7980 27 841 7950 Swap: 65417 517 64900 rick@Eos:~$ free -m total used free shared buff/cache available Mem: 32099 30990 257 27 850 150 Swap: 65417 719 64698 GitHub: Ricks-Lab Instagram: ricks_labs |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
I noticed that a while ago. It also happened on Ubuntu 14.04. With each crime and every kindness we birth our future. |
ML1 Send message Joined: 25 Nov 01 Posts: 20310 Credit: 7,508,002 RAC: 20 |
This completely normal and shows that maximum use is usefully being made of the system RAM. Note the cache/buffers numbers. The one to watch is if your swap space starts to become heavily used... All by good design :-) Happy fast crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
Hi Mike, ML1, Thanks for your confirmation that this is not an issue. I can confirm that I noticed no degradation in system performance. GitHub: Ricks-Lab Instagram: ricks_labs |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
Since I recently found out about the availability of Linux AVX2 apps, I decided to revisit the work I did in this posting on performance comparisons of the different app versions and testing methods. Here are the new results. Previous results are lower in this thread. My approach was to use the same single core Linux VM on my 1950X Win10 desktop that I had previously used. Only difference is that this time the CPU is only 70% loaded. This showed AVX2 was best, but strangely invalidated my previous conclusion that AVX was better than SSE41. I suspect this is the result running with the system 70% utilized. I got similar results using the new Win10 linux subsystem. I released AVX2 on Eos and found that processing times increased about 10%. Then I decided to test directly on Eos, my main contributer to SETI. I set CPU usage to 97% which left only 1 thread idle. I then ran one of the original test WUs and found that the results were not that far off from Nemesis, with AVX2 app significantly faster than AVX. I then used one the current WUs (Second column of data) and found that AVX2 was worse and SSE42 was best. I released SSE42 on my system and processing times decreased by 10%. Seems like WUs coming in now are a bit different from what we received previously and previous optimizations don't apply. GitHub: Ricks-Lab Instagram: ricks_labs |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Good new info, Rick. Yes the current batch of BLC05 tasks run particularly faster than anything the project has released before. Seems to work well with any AVX variant. Since the return of Arecibo data other than resends is unlikely, i would stick with the AVX2 app. As TBar and Raistmer have explained numerous times, the base code is the same across all the flavors, the only difference is in the optimization for each type by the compilers. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
From my experience it depends on the tasks and the CPU load. On my benches AVX2 was slower than SSE4.1 in most cases on my Ryzen 1800X. With each crime and every kindness we birth our future. |
bluestar Send message Joined: 5 Sep 12 Posts: 7032 Credit: 2,084,789 RAC: 3 |
Hi RueiKe! Message 1896117 on the previous page in this thread. Perhaps you meant to say "but I can not get to advanced view" here? Except for that, becomes the same word twice, but still perhaps not the same. |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
From my experience it depends on the tasks and the CPU load. Guess it's time to run the benchmarks on the BLC05 cpu tasks with the AVX, AVX2 and SSE4.1 apps. 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.