Interpreting error 3090

[exploitpanda]

I've briefly read thru some discussions but made my own to see if anyone has some input.

memtest_vulkan.log

[galkinvv]

Error found. Mode INITIAL_READ, total errors 0x1 out of 0x3A000000 (0.00000010%)
Errors address range: 0x2475614AC..=0x2475614AF  iteration:1472
values range: 0x6E6E7DDB..=0x6E6E7DDB   FFFFFFFF-like count:0    bit-level stats table:
         0x0 0x1  0x2 0x3| 0x4 0x5  0x6 0x7| 0x8 0x9  0xA 0xB| 0xC 0xD  0xE 0xF
SinglIdx                 |                 |            1    |                 
TogglCnt       1         |                 |                 |                 
1sInValu                 |                 |                 |                 
   0x1?                  |            1    |                 |                 

   1815 iteration. Passed 30.0438 seconds  written:10331.2GB 732.2GB/sec        checked:12397.5GB 778.1GB/sec
Tester console logging started at 2025-06-29T19:12:49.404959Z

1: Bus=0x01:00 DevId=0x2204   24GB NVIDIA GeForce RTX 3090
Tester worker logging started at 2025-06-29T19:12:50.017590Z
Standard 5-minute test of 1: Bus=0x01:00 DevId=0x2204   24GB NVIDIA GeForce RTX 3090
      1 iteration. Passed  0.0509 seconds  written:   18.1GB 747.1GB/sec        checked:   21.8GB 815.7GB/sec
     21 iteration. Passed  1.0483 seconds  written:  362.5GB 734.9GB/sec        checked:  435.0GB 783.7GB/sec
Error found. Mode INITIAL_READ, total errors 0x1 out of 0x3A000000 (0.00000010%)
Errors address range: 0x262B614AC..=0x262B614AF  iteration:96
values range: 0x6F467555..=0x6F467555   FFFFFFFF-like count:0    bit-level stats table:
         0x0 0x1  0x2 0x3| 0x4 0x5  0x6 0x7| 0x8 0x9  0xA 0xB| 0xC 0xD  0xE 0xF
SinglIdx                 |                 |            1    |                 
TogglCnt       1         |                 |                 |                 
1sInValu                 |                 |                 |                 
   0x1?             1    |                 |                 |                 

    117 iteration. Passed  5.0085 seconds  written: 1740.0GB 739.0GB/sec        checked: 2088.0GB 786.7GB/sec
    696 iteration. Passed 30.0396 seconds  written:10494.4GB 743.9GB/sec        checked:12593.2GB 790.4GB/sec
   1270 iteration. Passed 30.0526 seconds  written:10403.8GB 737.5GB/sec        checked:12484.5GB 782.9GB/sec
Error found. Mode INITIAL_READ, total errors 0x1 out of 0x3A000000 (0.00000010%)
Errors address range: 0x262B614AC..=0x262B614AF  iteration:1512
values range: 0xCCE5E5A0..=0xCCE5E5A0   FFFFFFFF-like count:0    bit-level stats table:
         0x0 0x1  0x2 0x3| 0x4 0x5  0x6 0x7| 0x8 0x9  0xA 0xB| 0xC 0xD  0xE 0xF
SinglIdx                 |                 |            1    |                 
TogglCnt       1         |                 |                 |                 
1sInValu                 |                 |                 |                 
   0x1?    1             |                 |                 |                 

   1852 iteration. Passed 30.0381 seconds  written:10548.8GB 748.2GB/sec        checked:12658.5GB 794.2GB/sec
   2431 iteration. Passed 30.0312 seconds  written:10494.4GB 745.0GB/sec        checked:12593.2GB 789.8GB/sec
Error found. Mode INITIAL_READ, total errors 0x1 out of 0x3A000000 (0.00000010%)
Errors address range: 0x262B614AC..=0x262B614AF  iteration:2512
values range: 0xDE6B7F06..=0xDE6B7F06   FFFFFFFF-like count:0    bit-level stats table:
         0x0 0x1  0x2 0x3| 0x4 0x5  0x6 0x7| 0x8 0x9  0xA 0xB| 0xC 0xD  0xE 0xF
SinglIdx                 |                 |            1    |                 
TogglCnt       1         |                 |                 |                 
1sInValu                 |                 |                 |                 
   0x1?                  |   1             |                 |                 

All your errors are at addresses like 2...614AC never others. So a single bit in one your VRAM ICs is buggy. Do you have any other user-visible errors? If not - depending on your workload this may be problem or not. For many workloads the chance that 1 bit out of 24GB is buggy may be acceptable chance of error.

[exploitpanda]

I have not seen any user visible errors. I am selling the GPU and they are planning to use it for AI. Should I run the test for longer?

[galkinvv]

I haven't enough AI experience to tell if this actually matters or not for AI uasge.

I don't think that running test longer would produce different errors, so 5 minutes are enough