Continuing To Stress Ryzen

[Naib]

am one of the (lucky?) few who post in these such threads (especially on the gentoo forum) who does not have any issues. I do run each test I come across that "causes this" I have just finished rebuilding php with pcntl support and have started this command so far so good

 [jrb:~/tmp/phoronix-test-suite] 14s $ TS_CONCURRENT_TEST_RUNS=4 TOTAL_LOOP_TIME=60 ./phoronix-test-suite stress-run build-linux-kernel build-php build-apache build-imagemagick  STRESS-RUN ENVIRONMENT VARIABLES:  PTS_CONCURRENT_TEST_RUNS: Set the PTS_CONCURRENT_TEST_RUNS environment variable to specify how many tests should be run concurrently during the stress-run process. If not specified, defaults to 2.   TOTAL_LOOP_TIME set; running tests for 60 minutes  SYSTEM INFORMATION:     PROCESSOR:          AMD Ryzen 5 1600 Six-Core @ 3.20GHz (12 Cores)     Core Count:       6     Thread Count:     12     Extensions:       SSE 4.2 + AVX2 + AVX + RDRAND + FSGSBASE     Cache Size:       512 KB     Microcode:        0x8001126     Scaling Driver:   acpi-cpufreq schedutil    GRAPHICS:           NVIDIA GeForce GTX 970 4096MB (1151/3505MHz)     OpenGL:           4.5.0     Vulkan:           1.0.46     Display Driver:   NVIDIA 384.59     Screen:           1280x1024    MOTHERBOARD:        MSI X370 GAMING PRO CARBON (MS-7A32) v1.0     Memory:           16384MB    DISK:               256GB OCZ VERTEX4 + 1000GB Samsung SSD 960 EVO 1TB     File-System:      ext4     Mount Options:    data=ordered noatime nodiratime rw     Disk Scheduler:   NONE    OPERATING SYSTEM:   Gentoo 2.4.1     Kernel:           4.12.4-gentoo (x86_64)     Desktop:          GNOME     Display Server:   X Server 1.19.3     Compiler:         GCC 7.1.0 + LLVM 4.0.1   pts/build-apache-1.5.1      : Test Started pts/build-php-1.3.1         : Test Started pts/build-php-1.3.1         : Average: 29.72 Seconds pts/build-php-1.3.1         : Test Ended pts/build-apache-1.5.1      : Average: 42.62 Seconds pts/build-apache-1.5.1      : Test Ended pts/build-php-1.3.1         : Test Started pts/build-linux-kernel-1.7.0: Test Started pts/build-linux-kernel-1.7.0: Average: 126.79 Seconds pts/build-linux-kernel-1.7.0: Test Ended pts/build-php-1.3.1         : Average: 76.76 Seconds pts/build-php-1.3.1         : Test Ended pts/build-imagemagick-1.7.2 : Test Started pts/build-apache-1.5.1      : Test Started

[Serafean]

AFAIK conftest segfaults are pretty normal. You probably should count only segfaults of bash, gcc itself, etc..

I'd second that. I see them quite often while updating my Gentoo machine. Although nothing close to Michael's frequency...

[foppe]

Thanks for the exhaustive testing, Michael . Can you check how (the idea behind) this 'save ryzen' variant of the script affects gcc performance? https://github.com/hayamdk/ryzen_segv_test/pull/3

[Naib]

am one of the (lucky?) few who post in these such threads (especially on the gentoo forum) who does not have any issues. I do run each test I come across that "causes this" I have just finished rebuilding php with pcntl support and have started this command so far so good

[jrb:~/tmp/phoronix-test-suite] 14s $ TS_CONCURRENT_TEST_RUNS=4 TOTAL_LOOP_TIME=60 ./phoronix-test-suite stress-run build-linux-kernel build-php build-apache build-imagemagick STRESS-RUN ENVIRONMENT VARIABLES: PTS_CONCURRENT_TEST_RUNS: Set the PTS_CONCURRENT_TEST_RUNS environment variable to specify how many tests should be run concurrently during the stress-run process. If not specified, defaults to 2. TOTAL_LOOP_TIME set; running tests for 60 minutes SYSTEM INFORMATION: PROCESSOR: AMD Ryzen 5 1600 Six-Core @ 3.20GHz (12 Cores) Core Count: 6 Thread Count: 12 Extensions: SSE 4.2 + AVX2 + AVX + RDRAND + FSGSBASE Cache Size: 512 KB Microcode: 0x8001126 Scaling Driver: acpi-cpufreq schedutil GRAPHICS: NVIDIA GeForce GTX 970 4096MB (1151/3505MHz) OpenGL: 4.5.0 Vulkan: 1.0.46 Display Driver: NVIDIA 384.59 Screen: 1280x1024 MOTHERBOARD: MSI X370 GAMING PRO CARBON (MS-7A32) v1.0 Memory: 16384MB DISK: 256GB OCZ VERTEX4 + 1000GB Samsung SSD 960 EVO 1TB File-System: ext4 Mount Options: data=ordered noatime nodiratime rw Disk Scheduler: NONE OPERATING SYSTEM: Gentoo 2.4.1 Kernel: 4.12.4-gentoo (x86_64) Desktop: GNOME Display Server: X Server 1.19.3 Compiler: GCC 7.1.0 + LLVM 4.0.1 pts/build-apache-1.5.1 : Test Started pts/build-php-1.3.1 : Test Started pts/build-php-1.3.1 : Average: 29.72 Seconds pts/build-php-1.3.1 : Test Ended pts/build-apache-1.5.1 : Average: 42.62 Seconds pts/build-apache-1.5.1 : Test Ended pts/build-php-1.3.1 : Test Started pts/build-linux-kernel-1.7.0: Test Started pts/build-linux-kernel-1.7.0: Average: 126.79 Seconds pts/build-linux-kernel-1.7.0: Test Ended pts/build-php-1.3.1 : Average: 76.76 Seconds pts/build-php-1.3.1 : Test Ended pts/build-imagemagick-1.7.2 : Test Started pts/build-apache-1.5.1 : Test Started ###### STRESS RUN INTERIM REPORT #### AUGUST 5 16:53 BST START TIME: August 5 16:48 BST ELAPSED TIME: 5 Minutes, 3 Seconds TIME REMAINING: 54 Minutes, 57 Seconds SYSTEM IP: 192.168.0.33 HOSTNAME: fluidmotion # OF CONCURRENT TESTS: 2 TESTS CURRENTLY ACTIVE: - pts/build-imagemagick-1.7.2 [PID: 1735] - pts/build-apache-1.5.1 [PID: 1736] TESTS IN RUN QUEUE: TIMED APACHE COMPILATION [pts/build-apache-1.5.1] TIMED IMAGEMAGICK COMPILATION [pts/build-imagemagick-1.7.2] TIMED LINUX KERNEL COMPILATION [pts/build-linux-kernel-1.7.0] TIMED PHP COMPILATION [pts/build-php-1.3.1] SYSTEM INFORMATION: Processor: AMD Ryzen 5 1600 Six-Core @ 3.20GHz (12 Cores) Motherboard: MSI X370 GAMING PRO CARBON (MS-7A32) v1.0 Memory: 16384MB Disk: 256GB OCZ VERTEX4 + 1000GB Samsung SSD 960 EVO 1TB Graphics: NVIDIA GeForce GTX 970 4096MB (1151/3505MHz) Audio: NVIDIA Generic OS: Gentoo 2.4.1 Kernel: 4.12.4-gentoo (x86_64) Desktop: GNOME Display Server: X Server 1.19.3 Display Driver: NVIDIA 384.59 OpenGL: 4.5.0 Vulkan: 1.0.46 Compiler: GCC 7.1.0 + LLVM 4.0.1 File-System: ext4 Screen Resolution: 1280x1024 TESTS EXECUTED TIMES CALLED pts/build-apache-1.5.1: 2 pts/build-imagemagick-1.7.2: 1 pts/build-linux-kernel-1.7.0: 1 pts/build-php-1.3.1: 2 SENSOR DATA: SENSOR MIN AVG MAX CPU Usage (CPU0): 6.00 66.03 100.00 Percent CPU Usage (CPU1): 0.00 74.09 100.00 Percent CPU Usage (CPU2): 7.69 70.84 100.00 Percent CPU Usage (CPU3): 0.00 69.61 100.00 Percent CPU Usage (CPU4): 6.12 76.82 100.00 Percent CPU Usage (CPU5): 10.00 77.41 100.00 Percent CPU Usage (CPU6): 6.00 75.58 100.00 Percent CPU Usage (CPU7): 12.00 76.63 100.00 Percent CPU Usage (CPU8): 0.00 66.23 100.00 Percent CPU Usage (CPU9): 4.00 75.47 100.00 Percent CPU Usage (CPU10): 2.00 71.41 100.00 Percent CPU Usage (CPU11): 6.00 73.96 100.00 Percent CPU Usage (Summary): 9.03 67.11 100.00 Percent GPU Temperature: 35.00 36.06 39.00 Celsius GPU Usage: 0.00 5.56 35.00 Percent Drive Read Speed (nvme0n1): 0.00 0.01 0.11 MB/s Drive Write Speed (nvme0n1): 0.00 6.19 90.82 MB/s Memory Usage: 1992.00 2551.06 3535.00 Megabytes System Temperature: 35.00 35.11 36.00 Celsius ###### pts/build-apache-1.5.1 : Average: 64.44 Seconds pts/build-apache-1.5.1 : Test Ended pts/build-imagemagick-1.7.2 : Average: 70.01 Seconds pts/build-imagemagick-1.7.2 : Test Ended pts/build-linux-kernel-1.7.0: Test Started pts/build-php-1.3.1 : Test Started

--edit 6 segfaults they seem to align when the test suite starts to execute kernel or php build [Sat Aug 5 16:48:54 2017] conftest[4564]: segfault at 0 ip 00000000004005e0 sp 00007fffed3c10d0 error 4 in conftest[400000+1000] [Sat Aug 5 16:48:55 2017] conftest[4817]: segfault at 0 ip 00007fb80c9a10c6 sp 00007ffe72b5a8a8 error 4 in libc-2.24.so[7fb80c877000+192000] [Sat Aug 5 16:49:59 2017] conftest[4887]: segfault at 0 ip 00000000004005e0 sp 00007fffb93595d0 error 4 in conftest[400000+1000] [Sat Aug 5 16:49:59 2017] conftest[4965]: segfault at 0 ip 00007f600bd030c6 sp 00007fffb5e32228 error 4 in libc-2.24.so[7f600bbd9000+192000] [Sat Aug 5 16:54:20 2017] conftest[22456]: segfault at 0 ip 00000000004005e0 sp 00007ffd3cc2fca0 error 4 in conftest[400000+1000] [Sat Aug 5 16:54:20 2017] conftest[22523]: segfault at 0 ip 00007f750fe4e0c6 sp 00007fff5ad2cdc8 error 4 in libc-2.24.so[7f750fd24000+192000]

[Shevchen]

Thanks for the test. The more you test this issue, the better - esp. clearing the things that don't work. Thats why I subscribed.

[boxie]

wow - that's a massive effort Michael - well done on the research and testing!

[satai]

Just out of curiosity - don't ECC DIMMs improve the situation?

[thechef]

Is it voltage related or not? Didn't quite understand that since you keep researching in all directions?

[pjssilva]

Just out of curiosity - don't ECC DIMMs improve the situation?

There are people in the AMD Technical Support forum using ECC and facing the bug.

[worohregger]

Can someone please confirm that the "conftest" segfaults are not the ones we should be looking for?
So far these are the only ones I'm seeing here (Arch Linux, 16 GB RAM, 1700x OC to 3.7 GHz, VCore set manually to 1.375 V, VSoc set to something between 1 and 1.1 V).