SILENTARMY v5: Zcash miner, 115 sol/s on R9 Nano, 70 sol/s on GTX 1070

[agis6]

Can u add values to your considerations ?
It would be even better
Thank you

Sure, here you go:

A. 1 x R9 380
Ubuntu 14.04.4 desktop, fglrx
Was 18-21 sol/s, now is 25-26 sol/s

B. 3 x R9 380X
Ubuntu 16.04.1 desktop, amdgpu-pro
Was 84-89 sol/s, now is 92-93 sol/s

C. 5 x RX470 + 1 x RX480
Ubuntu 16.04.1 server, amdgpu-pro
Was 135-147 sol/s, now is 145-148 sol/s

Setting OPTIM_SIMPLIFY_ROUND to 1 had the following results:
A. 29-31 sol/s
B. 92-97 sol/s
C. no change

[osnwt]

C. 5 x RX470 + 1 x RX480
Ubuntu 16.04.1 server, amdgpu-pro
Was 135-147 sol/s, now is 145-148 sol/s

Just wanted to ask about CPU model and usage. It seems that the performance highly depends on the CPU.

Running 470x2 on old single core gave me 7 per thread on 470.
Slightly better mobo + CPU gave 14/thread (28 per 470) - almost your result, see below.
Claymore's gives 45 Sol/s per 470 on Windows - on the 1st single core machine with 1GB of RAM (45 vs 14 Sols/s of this miner on the same mobo under Linux).

The results with single card and few cards on the same machine are also quite different. I use G3240 dual core CPUs on my rigs and it is not enough to run this at full speed.

So as for me, moving some filtering from CPU/bus into cards should be considered as a priority task if we want to outperform Claymore's miner. I've read that it is ALREADY faster. But trying on my cards (7950, 7970, 390, 390X, 470) - all with G3240 CPUs - I was not able to get the same results. Was close to with a single RX470 (Elpida RAM, 1500 straps, 2000 mem/1260 GPU - 46-47 Sol/s, Ubuntu 16.04 + amdgpu-pro 16.40) but without latest eXtremal patch.

So I do not understand how do you get a half of that only (148/6=25 Sol/s) on you rigs, but almost sure that the CPU usage is the reason. Have no other ideas.

[nerdralph]

Can u add values to your considerations ?
It would be even better
Thank you

Sure, here you go:

A. 1 x R9 380
Ubuntu 14.04.4 desktop, fglrx
Was 18-21 sol/s, now is 25-26 sol/s

I was getting over 30 on v3 using tighter memory timings and a 980/1550 clock.
http://nerdralph.blogspot.ca/2016/09/advanced-tonga-bios-editing.html

Still have to merge/build v4 to test it.

[nerdralph]

Just tested out v4 on Ubuntu/fglrx and a couple R9 380 cards.  No material change.
Reducing MAX_SOLS from 2000 to 500 reduces the amount of data xfer over the PCI-E bus, and gives me about a 3-5% speed improvement running sa-solver.
In my fork I have OPTIM_SIMPLIFY_ROUND turned on, and the MAX_SOLS reduced.
https://github.com/nerdralph/sa-nr

[mrb]

Just tested out v4 on Ubuntu/fglrx and a couple R9 380 cards.  No material change.
Reducing MAX_SOLS from 2000 to 500 reduces the amount of data xfer over the PCI-E bus, and gives me about a 3-5% speed improvement running sa-solver.
In my fork I have OPTIM_SIMPLIFY_ROUND turned on, and the MAX_SOLS reduced.
https://github.com/nerdralph/sa-nr

Watch out you are missing solutions if you reduce MAX_SOLS. It might be ok if it correspondingly reduces the Equihash runtime, but you have to look at how many solutions you lose on a good sample >1k, such as "./sa-solver --nonces 1000"

[Streetlight9]

Just tested out v4 on Ubuntu/fglrx and a couple R9 380 cards.  No material change.
Reducing MAX_SOLS from 2000 to 500 reduces the amount of data xfer over the PCI-E bus, and gives me about a 3-5% speed improvement running sa-solver.
In my fork I have OPTIM_SIMPLIFY_ROUND turned on, and the MAX_SOLS reduced.
https://github.com/nerdralph/sa-nr

Watch out you are missing solutions if you reduce MAX_SOLS. It might be ok if it correspondingly reduces the Equihash runtime, but you have to look at how many solutions you lose on a good sample >1k, such as "./sa-solver --nonces 1000"

that good or bad, as I have just updated this miner.  

seems like its good.

[eXtremal]

My latest kernel results (http://coinsforall.io/distr/input.cl), first row - original SA kernel, second - patched.

Ubuntu 13.10, Catalyst 14.4, Radeon R9 290 900/1250 (downclocked)
Total 29.1 sol/s [dev0 30.2] 4 shares
Total 41.1 sol/s [dev0 42.0] 2 shares +40%

Ubuntu 16.04, NVidia 367, GeForce GTX1070
Total 196 solutions in 6588.2 ms (29.8 Sol/s)
Total 196 solutions in 5334.1 ms (36.7 Sol/s)  +20%

Ubuntu 16.04, amdgpu-pro 16.30, Radeon RX480
Total 50.4 sol/s [dev0 51.0] 4 shares
Total 53.1 sol/s [dev0 53.2] 14 shares +4%

[NewbieTreider]

v4 tested and running in ubuntu 16.04 with amd-gpu-pro 16.40! ... seems a bit more solutions per thread!

Good work @mrb!

[ioglnx]

Is there a Windows release or not?!
Genoil's link gives a source only ?

It is a Python script + exe.

1. Download as ZIP on https://github.com/Genoil/silentarmy/tree/windows and unzip
2. Install Python 3.5+ for Win x64: https://www.python.org/downloads/windows/ (allow installer add python to PATH)
3. Launch .bat file (don't edit  )

You last built solvers are causing hard freezes..sa_solver.exe crashes system on exit while the other is not doing anything just opens..and you can't close if you want to reboot system it locks the system and you need to power it off.
Something strange happens with the last build ..the build before was working fine.

[mrb]

My latest kernel results (http://coinsforall.io/distr/input.cl), first row - original SA kernel, second - patched.

Ubuntu 13.10, Catalyst 14.4, Radeon R9 290 900/1250 (downclocked)
Total 29.1 sol/s [dev0 30.2] 4 shares
Total 41.1 sol/s [dev0 42.0] 2 shares +40%

Ubuntu 16.04, NVidia 367, GeForce GTX1070
Total 196 solutions in 6588.2 ms (29.8 Sol/s)
Total 196 solutions in 5334.1 ms (36.7 Sol/s)  +20%

Ubuntu 16.04, amdgpu-pro 16.30, Radeon RX480
Total 50.4 sol/s [dev0 51.0] 4 shares
Total 53.1 sol/s [dev0 53.2] 14 shares +4%

FWIW, exact same silentarmy code running on the same machine on an R9 Nano, dual booting into 2 OSes:
* 33.2 sol/s with fglrx 2:15.201-0ubuntu0.14.04.1 on Ubuntu 14.04
* 47.4 sol/s with amdgpu-pro 16.40 on Ubuntu 16.04

So yeah, a +40% difference just by changing drivers... No wonder you found a way to rework the OpenCL code to get a +40% on fglrx, and it gets you only +4% on amdgpu-pro.

amdgpu-pro compiles the OpenCL code just really, really well on its own, with almost no needs for manual tweaks.

[ioglnx]

But also +4% is something it just shows that optimizations can still be in cooperated

[hacko.bg]

I have a few rigs with 8 x R7 265. Starts mining successfully, but in mining screen I see devX up to 6, meaning giving display of hashrate up to the 6-th card. Possible to make it 8 at least?

[eXtremal]

But also +4% is something it just shows that optimizations can still be in cooperated

Need tests on R9 2xx/3xx cards with new drivers. May be this patch takes effect only on NVidia and old flgrx drivers.

[mrb]

But also +4% is something it just shows that optimizations can still be in cooperated

Absolutely. I even care about +1%, so I'll incorporate the changes!

[krnlx]

Finally fixed cpu load.

libtime.c:

Code:

#include <stdlib.h>
#include <unistd.h>
#include <stdio.h>
#include <dlfcn.h>
#include <assert.h>
#include <time.h>

/*
Temprorary fix for silentarmy - nvidia
The MIT License (MIT) Copyright (c) 2016 krnlx, kernelx at me.com
*/


int inited=0;

void *libc = NULL;

int (*libc_clock_gettime)(clockid_t clk_id, struct timespec *tp) = NULL;

static void __attribute__ ((constructor)) lib_init(void)  {
  if(inited)
    return;

  libc = dlopen("libc.so.6", RTLD_LAZY);
  assert(libc);

  libc_clock_gettime = dlsym(libc, "clock_gettime");
  assert(libc_clock_gettime);
  inited++;
}


useconds_t sleep_time = 1200;


int clock_gettime(clockid_t clk_id, struct timespec *tp){
  lib_init();
  //printf(".");
  usleep(sleep_time);
  int r = (*libc_clock_gettime)(clk_id, tp);
  return r;
}

Code:

gcc -O2 -fPIC -shared -Wl,-soname,libtime.so -o libtime.so libtime.c

Code:

LD_PRELOAD="./libtime.so" ./silentarmy --instances=2 --use=0,1,2,3,4,5
Connecting to us1-zcash.flypool.org:3333
Stratum server sent us the first job
Mining on 6 devices
Total 0.0 sol/s [dev0 0.0, dev1 0.0, dev2 0.0, dev3 0.0, dev4 0.0, dev5 0.0] 0 shares
Total 243.5 sol/s [dev0 36.8, dev1 39.7, dev2 38.8, dev3 43.7, dev4 50.7, dev5 33.8] 1 share
Total 249.5 sol/s [dev0 39.6, dev1 38.1, dev2 39.6, dev3 40.6, dev4 51.5, dev5 40.1] 1 share
Total 254.8 sol/s [dev0 35.7, dev1 41.6, dev2 43.3, dev3 41.0, dev4 50.2, dev5 43.0] 1 share
Total 253.3 sol/s [dev0 31.2, dev1 41.1, dev2 44.6, dev3 42.6, dev4 51.0, dev5 42.8] 1 share
Total 247.6 sol/s [dev0 31.9, dev1 38.5, dev2 44.8, dev3 40.8, dev4 51.5, dev5 40.0] 1 share
Total 239.8 sol/s [dev0 31.2, dev1 38.0, dev2 43.6, dev3 39.3, dev4 47.6, dev5 40.0] 2 shares
Total 244.4 sol/s [dev0 33.0, dev1 39.7, dev2 42.4, dev3 40.4, dev4 46.9, dev5 42.1] 2 shares
Total 242.6 sol/s [dev0 32.2, dev1 40.1, dev2 42.9, dev3 38.1, dev4 47.1, dev5 42.2] 3 shares
Total 241.7 sol/s [dev0 32.2, dev1 39.6, dev2 42.8, dev3 39.7, dev4 46.6, dev5 40.9] 5 shares
Total 238.2 sol/s [dev0 32.9, dev1 39.6, dev2 42.8, dev3 37.7, dev4 45.8, dev5 39.4] 5 shares
Total 238.4 sol/s [dev0 32.8, dev1 39.7, dev2 42.1, dev3 38.2, dev4 45.0, dev5 40.1] 6 shares
Total 238.1 sol/s [dev0 29.4, dev1 40.2, dev2 43.9, dev3 39.2, dev4 44.1, dev5 39.0] 7 shares
Total 235.0 sol/s [dev0 29.0, dev1 38.0, dev2 44.1, dev3 38.3, dev4 41.7, dev5 38.0] 8 shares
Total 236.0 sol/s [dev0 32.1, dev1 38.1, dev2 43.4, dev3 37.7, dev4 40.6, dev5 37.2] 8 shares
Total 238.3 sol/s [dev0 31.5, dev1 40.5, dev2 44.4, dev3 39.0, dev4 40.4, dev5 37.9] 8 shares
Total 238.5 sol/s [dev0 33.1, dev1 41.8, dev2 43.1, dev3 38.8, dev4 42.5, dev5 38.5] 8 shares
Total 239.7 sol/s [dev0 33.6, dev1 40.5, dev2 44.0, dev3 37.8, dev4 42.5, dev5 38.1] 8 shares
Total 240.3 sol/s [dev0 33.7, dev1 40.9, dev2 45.6, dev3 39.3, dev4 42.3, dev5 36.6] 8 shares
Total 238.7 sol/s [dev0 32.9, dev1 41.7, dev2 44.7, dev3 38.0, dev4 41.0, dev5 37.8] 8 shares
Total 238.3 sol/s [dev0 32.5, dev1 41.7, dev2 44.8, dev3 39.6, dev4 40.1, dev5 39.7] 8 shares
Total 237.5 sol/s [dev0 32.5, dev1 41.0, dev2 45.2, dev3 39.1, dev4 38.9, dev5 39.8] 8 shares
Total 239.0 sol/s [dev0 34.9, dev1 41.5, dev2 43.7, dev3 39.2, dev4 39.7, dev5 41.1] 11 shares
Total 238.7 sol/s [dev0 35.5, dev1 43.1, dev2 42.5, dev3 39.1, dev4 41.7, dev5 41.5] 11 shares
Total 237.0 sol/s [dev0 32.6, dev1 44.1, dev2 41.8, dev3 38.3, dev4 40.0, dev5 41.5] 11 shares
Total 237.2 sol/s [dev0 32.6, dev1 43.7, dev2 39.9, dev3 38.6, dev4 39.9, dev5 40.7] 12 shares
Total 239.3 sol/s [dev0 33.0, dev1 43.3, dev2 41.6, dev3 40.2, dev4 41.4, dev5 41.0] 13 shares

6 x 1070

CPU load on celeron  1840

Code:

top - 01:20:37 up 18:38,  3 users,  load average: 3.66, 1.75, 0.91
Tasks: 151 total,  10 running, 141 sleeping,   0 stopped,   0 zombie
%Cpu0  : 62.5 us,  1.0 sy,  0.0 ni, 36.5 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
%Cpu1  : 59.2 us,  4.4 sy,  0.0 ni, 36.4 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem :  3988068 total,  1317788 free,  1465128 used,  1205152 buff/cache
KiB Swap:        0 total,        0 free,        0 used.  2204660 avail Mem 

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                                                                                         
15266 krnl      20   0 29.619g 126804  94508 R  10.9  3.2   0:09.40 sa-solver                                                                                                       
15273 krnl      20   0 29.614g 149356 111812 R  10.9  3.7   0:10.22 sa-solver                                                                                                       
15269 krnl      20   0 29.599g 109488  90176 R  10.6  2.7   0:09.09 sa-solver                                                                                                       
15271 krnl      20   0 29.599g 112464  90116 R  10.6  2.8   0:09.23 sa-solver                                                                                                       
15275 krnl      20   0 29.610g 116104  94132 S  10.6  2.9   0:09.25 sa-solver                                                                                                       
15268 krnl      20   0 29.599g 111812  90404 R  10.3  2.8   0:09.27 sa-solver                                                                                                       
15265 krnl      20   0 29.609g 116228  94232 R   9.9  2.9   0:09.28 sa-solver                                                                                                       
15272 krnl      20   0 29.610g 113848  94216 R   9.9  2.9   0:09.37 sa-solver                                                                                                       
15263 krnl      20   0 29.599g 110692  90236 R   9.6  2.8   0:08.42 sa-solver                                                                                                       
15264 krnl      20   0 29.599g 110464  90004 S   9.6  2.8   0:08.18 sa-solver                                                                                                       
15274 krnl      20   0 29.599g 109500  90116 S   9.6  2.7   0:09.14 sa-solver                                                                                                       
15267 krnl      20   0 29.599g 111504  90152 R   9.3  2.8   0:09.37 sa-solver                                                                                                      

You can adjust sleep time by changing the useconds_t sleep_time value

[liomojo1]

Any news for windows release supporting nvidia?

[krnlx]

with 3 instances getting 240-250 s/s on 6 x 1070 (eXtremal's input.cl and my cpu load workaround)

Code:

Total 243.2 sol/s [dev0 32.1, dev1 40.7, dev2 42.4, dev3 41.3, dev4 47.3, dev5 43.1] 20 shares
Total 244.0 sol/s [dev0 33.9, dev1 42.6, dev2 43.1, dev3 37.9, dev4 44.9, dev5 42.5] 21 shares
Total 247.0 sol/s [dev0 34.2, dev1 42.7, dev2 43.1, dev3 37.2, dev4 44.9, dev5 43.5] 21 shares
Total 245.4 sol/s [dev0 35.1, dev1 43.3, dev2 41.0, dev3 37.3, dev4 43.7, dev5 43.7] 21 shares
Total 244.8 sol/s [dev0 34.4, dev1 44.6, dev2 39.9, dev3 37.6, dev4 43.5, dev5 44.5] 21 shares
Total 243.8 sol/s [dev0 34.0, dev1 43.7, dev2 38.0, dev3 36.6, dev4 42.5, dev5 43.0] 21 shares
Total 244.1 sol/s [dev0 35.1, dev1 45.1, dev2 38.9, dev3 37.2, dev4 42.9, dev5 44.0] 21 shares
Total 244.5 sol/s [dev0 36.9, dev1 44.3, dev2 41.0, dev3 38.7, dev4 41.6, dev5 46.1] 21 shares
Total 244.8 sol/s [dev0 37.2, dev1 43.9, dev2 39.9, dev3 38.0, dev4 42.8, dev5 43.8] 22 shares
Total 243.4 sol/s [dev0 36.0, dev1 43.8, dev2 39.1, dev3 37.3, dev4 42.6, dev5 42.3] 23 shares
Total 243.5 sol/s [dev0 35.8, dev1 43.9, dev2 38.0, dev3 38.2, dev4 41.7, dev5 42.5] 24 shares
Total 243.4 sol/s [dev0 34.4, dev1 41.8, dev2 38.5, dev3 41.0, dev4 42.1, dev5 41.4] 26 shares
Total 243.0 sol/s [dev0 34.5, dev1 42.3, dev2 38.9, dev3 42.0, dev4 41.0, dev5 40.5] 26 shares
Total 243.0 sol/s [dev0 33.9, dev1 42.2, dev2 39.4, dev3 43.1, dev4 40.2, dev5 40.7] 26 shares
Total 243.6 sol/s [dev0 34.5, dev1 42.7, dev2 41.8, dev3 43.6, dev4 40.7, dev5 40.8] 26 shares
Total 243.5 sol/s [dev0 35.1, dev1 45.2, dev2 41.7, dev3 44.7, dev4 42.0, dev5 41.7] 28 shares
Total 244.4 sol/s [dev0 34.7, dev1 45.2, dev2 41.8, dev3 44.5, dev4 42.0, dev5 39.8] 30 shares
Total 243.0 sol/s [dev0 33.9, dev1 44.5, dev2 41.9, dev3 45.2, dev4 40.9, dev5 38.1] 30 shares
Total 244.3 sol/s [dev0 34.5, dev1 43.4, dev2 42.5, dev3 45.9, dev4 40.0, dev5 40.6] 30 shares
Total 245.6 sol/s [dev0 37.2, dev1 46.0, dev2 43.7, dev3 46.0, dev4 38.6, dev5 41.8] 30 shares
Total 247.3 sol/s [dev0 36.9, dev1 48.5, dev2 44.1, dev3 45.5, dev4 40.0, dev5 39.8] 31 shares
Total 245.2 sol/s [dev0 36.2, dev1 47.4, dev2 42.4, dev3 43.6, dev4 40.4, dev5 41.6] 31 shares
Total 245.3 sol/s [dev0 36.4, dev1 45.6, dev2 43.2, dev3 44.3, dev4 40.0, dev5 41.5] 33 shares
Total 244.9 sol/s [dev0 37.7, dev1 45.8, dev2 42.7, dev3 41.7, dev4 40.5, dev5 42.6] 35 shares
Total 246.4 sol/s [dev0 37.4, dev1 46.1, dev2 42.1, dev3 40.8, dev4 41.4, dev5 43.1] 37 shares

[antantti]

with 3 instances getting 240-250 s/s on 6 x 1070 (eXtremal's input.cl and my cpu load workaround)

Could you try with only one gpu? And 1,2,3,4 instances?

[krnlx]

with 3 instances getting 240-250 s/s on 6 x 1070 (eXtremal's input.cl and my cpu load workaround)

Could you try with only one gpu? And 1,2,3,4 instances?

hmm.. I really don't know why my gpu0 card runs slower o_O

Code:

$ LD_PRELOAD="./libtime.so" ./silentarmy --instances=3 --use=0
Connecting to us1-zcash.flypool.org:3333
Stratum server sent us the first job
Mining on 1 device
Total 0.0 sol/s [dev0 0.0] 0 shares
Total 39.6 sol/s [dev0 39.6] 0 shares
Total 38.2 sol/s [dev0 38.2] 0 shares
Total 32.4 sol/s [dev0 32.4] 0 shares
Total 33.2 sol/s [dev0 33.2] 0 shares
Total 37.1 sol/s [dev0 37.1] 0 shares
Total 37.0 sol/s [dev0 37.0] 0 shares
Total 34.7 sol/s [dev0 34.7] 0 shares
Total 34.6 sol/s [dev0 34.6] 0 shares
Total 35.6 sol/s [dev0 35.6] 0 shares

but gpu1 and others

Code:

LD_PRELOAD="./libtime.so" ./silentarmy --instances=3 --use=1
Connecting to us1-zcash.flypool.org:3333
Stratum server sent us the first job
Mining on 1 device
Total 0.0 sol/s [dev1 0.0] 0 shares
Total 45.6 sol/s [dev1 45.6] 0 shares
Total 53.0 sol/s [dev1 53.0] 0 shares
Total 47.9 sol/s [dev1 47.9] 0 shares
Total 45.8 sol/s [dev1 45.8] 0 shares
Total 42.6 sol/s [dev1 42.6] 0 shares
Total 43.5 sol/s [dev1 43.5] 0 shares
Total 42.5 sol/s [dev1 42.5] 0 shares
Total 42.6 sol/s [dev1 42.6] 0 shares
Total 43.3 sol/s [dev1 43.3] 0 shares

Maybe because Xorg is running on it ? On windows 10 with Claymore's miner and rx470 cards, I have same issue - card, connected to monitor, runs slower

Code:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 367.27                 Driver Version: 367.27                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 1070    On   | 0000:01:00.0      On |                  N/A |
|100%   32C    P0    40W / 195W |     15MiB /  8113MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX 1070    On   | 0000:03:00.0     Off |                  N/A |
|100%   32C    P0    42W / 195W |      6MiB /  8113MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   2  GeForce GTX 1070    On   | 0000:04:00.0     Off |                  N/A |
|100%   30C    P0    41W / 195W |      6MiB /  8113MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   3  GeForce GTX 1070    On   | 0000:05:00.0     Off |                  N/A |
|100%   31C    P0    41W / 195W |      6MiB /  8113MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   4  GeForce GTX 1070    On   | 0000:06:00.0     Off |                  N/A |
|100%   28C    P0    41W / 195W |      6MiB /  8113MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   5  GeForce GTX 1070    On   | 0000:07:00.0     Off |                  N/A |
|100%   28C    P0    41W / 195W |      6MiB /  8113MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0      1751    G   /usr/lib/xorg/Xorg                              13MiB |
|    1      1751    G   /usr/lib/xorg/Xorg                               6MiB |
|    2      1751    G   /usr/lib/xorg/Xorg                               6MiB |
|    3      1751    G   /usr/lib/xorg/Xorg                               6MiB |
|    4      1751    G   /usr/lib/xorg/Xorg                               6MiB |
|    5      1751    G   /usr/lib/xorg/Xorg                               6MiB |
+-----------------------------------------------------------------------------+

[krnlx]

Tried without Xorg(no overclock), gpu0 slower, don't know why. Cards overclocked by memory(+1000  = +500 on win), getting 41-42, no overclock 35-38. Shitty nvidia-setting can't run without X ;(