Repository navigation
Having issues w/ Rejected benchmarks #2280
Description
Activity
This is probably a sign of instability. Try to run offline benchmark and see if you get green or red hash in the end.
Feeling like a cuda problem. --cuda red, w/o green.
Cuda version above. Nvidia Driver, 466.11, 4/14/2021
Rebuilt xmrig-cuda 6.5.0 from scratch using latest CUDA 11.3
Still NO love. (red hashcode) Also tried going backwards with other combinations too, Cuda 11.0 and 6.4.1. No luck either.Open to other suggestions.
PS W:\projects\xmrig\xmrig-6.11.2> .\benchmark_1M.cmd
- ABOUT XMRig/6.11.2 gcc/10.1.0
- LIBS libuv/1.41.0 OpenSSL/1.1.1j hwloc/2.4.1
- HUGE PAGES permission granted
- 1GB PAGES unavailable
- CPU AMD Ryzen 9 3950X 16-Core Processor (1) 64-bit AES
L2:8.0 MB L3:64.0 MB 16C/32T NUMA:1 - MEMORY 15.9/127.9 GB (12%)
DIMM_A1: 32 GB DDR4 @ 2666 MHz F4-3200C16-32GTRS
DIMM_A2: 32 GB DDR4 @ 2666 MHz F4-3200C16-32GTRS
DIMM_B1: 32 GB DDR4 @ 2666 MHz F4-3200C16-32GTRS
DIMM_B2: 32 GB DDR4 @ 2666 MHz F4-3200C16-32GTRS - MOTHERBOARD ASUSTeK COMPUTER INC. - ROG CROSSHAIR VIII HERO
- DONATE 0%
- ASSEMBLY auto:ryzen
- POOL 128tx exploit #1 benchmark algo auto
- COMMANDS hashrate, pause, resume, results, connection
- OPENCL disabled
- CUDA 11.3/11.3/6.5.0
- NVML 11.466.11/466.11 press e for health report
- CUDA GPU #0 0a:00.0 NVIDIA GeForce RTX 3090 1800/9751 MHz smx:82 arch:86 mem:23328/24576 MB
[2021-04-19 16:41:21.119] bench start benchmark hashes 1M algo rx/0
[2021-04-19 16:41:21.120] cpu use argon2 implementation AVX2
[2021-04-19 16:41:21.667] msr register values for "ryzen_17h" preset have been set successfully (547 ms)
[2021-04-19 16:41:21.668] randomx init dataset algo rx/0 (32 threads) seed 0000000000000000...
[2021-04-19 16:41:21.670] randomx allocated 2336 MB (2080+256) huge pages 100% 1168/1168 +JIT (0 ms)
[2021-04-19 16:41:23.123] randomx dataset ready (1452 ms)
[2021-04-19 16:41:23.123] cpu use profile rx (32 threads) scratchpad 2048 KB
[2021-04-19 16:41:23.127] nvidia use profile rx (1 thread) scratchpad 2048 KB
| # | GPU | BUS ID | INTENSITY | THREADS | BLOCKS | BF | BS | MEMORY | NAME
| 0 | 0 | 0a:00.0 | 5248 | 32 | 164 | 6 | 25 | 10496 | NVIDIA GeForce RTX 3090
[2021-04-19 16:41:23.150] cpu READY threads 32/32 (32) huge pages 100% 32/32 memory 65536 KB (25 ms)
[2021-04-19 16:41:23.541] nvidia READY threads 1/1 (414 ms)
[2021-04-19 16:42:23.063] bench benchmark finished in 59.937 seconds (16684.2 h/s) hash sum = 5721BF496B4C2AFA
[2021-04-19 16:42:23.064] bench press Ctrl+C to exit
[2021-04-19 16:42:23.494] cpu stopped (1 ms)
[2021-04-19 16:42:25.680] nvidia stopped (2186 ms)
[2021-04-19 16:42:25.682] nvidia #0 0a:00.0 289W 66C 1920/9501 MHz fan0:60% fan1:60%
[2021-04-19 16:42:37.270] signal Ctrl+C received, exiting
Press any key to continue . . .
Terminate batch job (Y/N)?
^C
PS W:\projects\xmrig\xmrig-6.11.2>
Reacted by peme14kI'm starting to wonder if the card is mining-crippled.
Wait, are you trying to run benchmark? This is only for CPU.
Benchmarks where GPUs are enabled will always be rejected, as benchmarking is only available for CPUs
In my benchmark_1m file.
xmrig.exe --algo=cryptonight-heavy --donate-level=0 --print-time=5 --bench=1M --cudaSo @SChernykh yes to GPU.
@Spudz76 I just need to confirm the GPU hash sum is correct. I don't necessarily need to submit the benchmark.
Well it's ignoring your algo and doing an rx/0 benchmark, mostly because benchmark mode only tests rx/0 or rx/wow
Well I was trying different combinations as the original of just removing --submit and adding --cuda always comes back with a red checksum.
xmrig.exe --bench=1m --> returns green hash sum
xmrig.exe --bench=1m --cuda --> returns red hash sumI've tried all combinations of xmrig 6.11.2, 6.10.0, xmrig-cuda 6.5.0, 6.4.2-dev, 6.4.1 both binaries and compiling from scratch, different CUDA versions 11.0, 11.0.2, 11.1, 11.1.1, 11.3 and various command line flags just to test if it was some configuration issue.
I can't get any --cuda hash sum to come back as green. That's the issue I'm trying to figure out.
See side by side:
https://gofile.io/d/FjEPOkBenchmark was not designed to run with GPU enabled, this is a bug in a sense that it should ignore any GPU settings.
@SChernykh Thank you. I get it now.
What's the best way to confirm that the GPU is working correctly?
Best way is probably to start mining and see if you get all green "accepted" in the log.
Reacted by gbolcerI'll close this issue out, so thank you for the help. One last question though, even with the log colors:true, the log file seems to not generating colors. Suggestions?
- added a commit that references this issue
on Apr 25, 2021
Any advice where to start looking appreciated.
Cuda 11.1 (Can use 11.0, 11.1, 11.2)
Win10 Pro 20H2 19042.928
xmrig-cuda-6.5.0-cuda11_1-win64
nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2020 NVIDIA Corporation
Built on Mon_Oct_12_20:54:10_Pacific_Daylight_Time_2020
Cuda compilation tools, release 11.1, V11.1.105
Build cuda_11.1.relgpu_drvr455TC455_06.29190527_0
.\xmrig.exe --version
XMRig 6.11.2
built on Apr 11 2021 with GCC 10.1.0
features: 64-bit AES
libuv/1.41.0
OpenSSL/1.1.1j
hwloc/2.4.1
https://xmrig.com/benchmark/3S5qfS