Closed quydau35 closed 2 years ago
Confirmed on my K1100M
Can't seem to make it work. It didn't fail launch when using:
"astrobwt": [
{
"index": 0,
"threads": 32,
"blocks": 3,
"bfactor": 8,
"bsleep": 25,
"affinity": -1
}
],
But it never got a hashrate either and more or less hung my system (Linux doesn't actually use bsleep). Had to ssh in and kill it after while.
AstroBWT already didn't compile at all with CUDA 8 or Fermi perhaps there are other issues with Kepler for some reason even though it compiles (but won't launch).
So it seems there are some combinations of thread/block that will launch but most will hit invalid config arg.
I had some "normal" Desktop Kepler 35's (GTX770) but they are not currently online or testable. It could just be the Mobile Kepler 30 is deficient somehow compared to higher 3x archs.
In general I quit mining the GPU on this Lenovo W540 laptop because the shared cooling solution is already undersized with just the CPU making heat. CPU at 90C makes the GPU 90C real fast and then everything dies (poweroff for thermal protect). So perhaps even if it worked it would just overheat (cup half full?)
Okay I merged the newest src/3rdparty/cub
and now it works with 32/3
See also #130 if you can build from that and test with 32/3 (or other combos) on your own K1100M to verify it at least works for you.
* CUDA 10.2/11.2/6.15.2-dev
* NVML 11.460.91.03/460.91.03 press e for health report
* CUDA GPU #0 01:00.0 Quadro K1100M 705/1400 MHz smx:2 arch:30 mem:1275/1999 MB
[2021-10-04 10:23:41.590] net use pool gulf.moneroocean.stream:20004 TLSv1.3 54.188.223.206
[2021-10-04 10:23:41.590] net new job from gulf.moneroocean.stream:20004 diff 454 algo astrobwt height 6214889
[2021-10-04 10:23:41.590] nvidia use profile astrobwt (1 thread) scratchpad 20480 KB
| # | GPU | BUS ID | INTENSITY | THREADS | BLOCKS | BF | BS | MEMORY | NAME
| 0 | 0 | 01:00.0 | 96 | 32 | 3 | 8 | 0 | 892 | Quadro K1100M
[2021-10-04 10:23:41.678] nvidia READY threads 1/1 (87 ms)
[2021-10-04 10:23:45.914] nvidia accepted (1/0) diff 454 (515 ms)
[2021-10-04 10:23:50.924] nvidia accepted (2/0) diff 454 (516 ms)
[2021-10-04 10:23:52.152] nvidia accepted (3/0) diff 454 (519 ms)
[2021-10-04 10:23:55.917] nvidia accepted (4/0) diff 454 (516 ms)
[2021-10-04 10:23:58.392] nvidia accepted (5/0) diff 454 (517 ms)
[2021-10-04 10:24:01.068] net new job from gulf.moneroocean.stream:20004 diff 3510 algo astrobwt height 6214889
[2021-10-04 10:24:05.744] net new job from gulf.moneroocean.stream:20004 diff 3510 algo astrobwt height 6214890
[2021-10-04 10:24:09.978] net new job from gulf.moneroocean.stream:20004 diff 3510 algo astrobwt height 6214891
[2021-10-04 10:24:17.153] nvidia accepted (6/0) diff 3510 (516 ms)
[2021-10-04 10:24:45.826] nvidia #0 01:00.0 0W 63C 705/1400 MHz
[2021-10-04 10:24:45.826] miner speed 10s/60s/15m 78.83 n/a n/a H/s max 84.27 H/s
[2021-10-04 10:24:58.264] nvidia #0 01:00.0 0W 63C 705/1400 MHz
| CUDA # | AFFINITY | 10s H/s | 60s H/s | 15m H/s |
| 0 | -1 | 84.82 | 80.40 | n/a | #0 01:00.0 Quadro K1100M
| - | - | 78.62 | 79.36 | n/a |
[2021-10-04 10:25:00.141] miner speed 10s/60s/15m 78.62 79.36 n/a H/s max 84.27 H/s
[2021-10-04 10:25:07.349] signal Ctrl+C received, exiting
[2021-10-04 10:25:07.637] nvidia stopped (287 ms)
Also my nvidia-smi for reference:
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.91.03 Driver Version: 460.91.03 CUDA Version: 11.2 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 Quadro K1100M Off | 00000000:01:00.0 Off | N/A |
| N/A 57C P0 N/A / N/A | 777MiB / 1999MiB | 16% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
Used toolkit 10.2 against somewhat recent driver on Linux. Did not test toolkit 10.1 or similar revision driver as you are running on Windows, but expect it to still work.
Hi, I try build #130 and got few error due to the lack of agent/agent_radix_sort_histogram.cuh
. This is simply solved by provide the original cub
library and correct the path and recompile it.
And run with 32 threads/3blocks solves all the issue! Thanks for your helps! I built with cuda 10.1, test with K1100M and also K600 and it runs like charm!
You're correct that running on mobile GPU is not a good idea and its temperature raise up to its critical temp really fast.
For reference, my K600 got around 50H/s and my K1100 got around 78H/s.
Looking forward for this pull request to be merged into master and everybody could enjoy mining with old Kepler cards.
Okay I merged the newest
src/3rdparty/cub
and now it works with 32/3See also #130 if you can build from that and test with 32/3 (or other combos) on your own K1100M to verify it at least works for you.
* CUDA 10.2/11.2/6.15.2-dev * NVML 11.460.91.03/460.91.03 press e for health report * CUDA GPU #0 01:00.0 Quadro K1100M 705/1400 MHz smx:2 arch:30 mem:1275/1999 MB [2021-10-04 10:23:41.590] net use pool gulf.moneroocean.stream:20004 TLSv1.3 54.188.223.206 [2021-10-04 10:23:41.590] net new job from gulf.moneroocean.stream:20004 diff 454 algo astrobwt height 6214889 [2021-10-04 10:23:41.590] nvidia use profile astrobwt (1 thread) scratchpad 20480 KB | # | GPU | BUS ID | INTENSITY | THREADS | BLOCKS | BF | BS | MEMORY | NAME | 0 | 0 | 01:00.0 | 96 | 32 | 3 | 8 | 0 | 892 | Quadro K1100M [2021-10-04 10:23:41.678] nvidia READY threads 1/1 (87 ms) [2021-10-04 10:23:45.914] nvidia accepted (1/0) diff 454 (515 ms) [2021-10-04 10:23:50.924] nvidia accepted (2/0) diff 454 (516 ms) [2021-10-04 10:23:52.152] nvidia accepted (3/0) diff 454 (519 ms) [2021-10-04 10:23:55.917] nvidia accepted (4/0) diff 454 (516 ms) [2021-10-04 10:23:58.392] nvidia accepted (5/0) diff 454 (517 ms) [2021-10-04 10:24:01.068] net new job from gulf.moneroocean.stream:20004 diff 3510 algo astrobwt height 6214889 [2021-10-04 10:24:05.744] net new job from gulf.moneroocean.stream:20004 diff 3510 algo astrobwt height 6214890 [2021-10-04 10:24:09.978] net new job from gulf.moneroocean.stream:20004 diff 3510 algo astrobwt height 6214891 [2021-10-04 10:24:17.153] nvidia accepted (6/0) diff 3510 (516 ms) [2021-10-04 10:24:45.826] nvidia #0 01:00.0 0W 63C 705/1400 MHz [2021-10-04 10:24:45.826] miner speed 10s/60s/15m 78.83 n/a n/a H/s max 84.27 H/s [2021-10-04 10:24:58.264] nvidia #0 01:00.0 0W 63C 705/1400 MHz | CUDA # | AFFINITY | 10s H/s | 60s H/s | 15m H/s | | 0 | -1 | 84.82 | 80.40 | n/a | #0 01:00.0 Quadro K1100M | - | - | 78.62 | 79.36 | n/a | [2021-10-04 10:25:00.141] miner speed 10s/60s/15m 78.62 79.36 n/a H/s max 84.27 H/s [2021-10-04 10:25:07.349] signal Ctrl+C received, exiting [2021-10-04 10:25:07.637] nvidia stopped (287 ms)
Fixed the missing files, forgot to add previously nonexistent ones to the commit.
I got the Invalid configuration argument on all of my Kepler-based with compute capability 3.0 cards
nvidia-smi
outputs:and never got this error when I run with maxwell cards (with compute capability > 5.0) like quadro K2200 etc,... I use the xmrig-6.15.1 and xmrig-cuda 6.15.1 built with cuda toolkit 10.1. Does anyone have any idea why this is happen? and how to fix it? I tried find namespace
AstroBWT_Dero
insrc/AstroBWT/dero/AstroBWT.cu
but doesn't see any abnormal configuration. Here is my configuration file with deleted wallet address: