[ANN] sph-sgminer: multi-coin multi-algorithm GPU miner | added MaruCoin - page 35.

CryptoGretzky

sr. member

Activity: 322

Merit: 250

I ran aticonfig --odgc --adapter=all for the rig that drops after a min.

On boot, it's still running at 2.5-2.6 Mh, all 4 GPU's data are the same. eg. core, mem are 1025,1400, bus speed is 5000:

After a few mins it started dropping down to 2.3 Mh. ALL GPU's data remained the same. eg. core, mem are 1025,1400, bus speed is 5000:

So, nothing really changed at all from the GPU's setting perspective....

On my other rig that held steady at 2.77 Mh/s now over 15 hours at this speed, core, mem is the same as my other rig, BUT bus speed is only 2500

Any one know how to set bus speed independently so I can test further to see if that has anything to do with it??

CryptoGretzky

sr. member

Activity: 322

Merit: 250

Quote from: janos666 on March 30, 2014, 12:05:54 PM

Quote from: CryptoGretzky on March 29, 2014, 11:03:38 PM

How do I set the clock outside of sgminer in Linux?

aticonfig --adapter=all --odsc=1000,1450

You might need to be root. The clocks are [GPU,VRAM]. You can set unique clocks for every cards (starting from adapter=0).

aticonfig --adapter=all --odcc

This applies the earlier "scratchpad" settings (makes them permanent through reboots).

Use can issue this command to read your actual clocks (and limits + current GPU load percentage which can also be very useful!):

aticonfig --adapter=all --odgc

Thanks for the info. I could run --odgc fine, but when I use --odsc, it ask me to also set powerstate and performance level. I tried --lsp, --list-powerstate, but somehow it's not recognizing the command even though they are clearly shown in man for aticonfig.... I did get some more info that I am posting in another post now though.

janos666

hero member

Activity: 588

Merit: 500

Quote from: CryptoGretzky on March 29, 2014, 11:03:38 PM

How do I set the clock outside of sgminer in Linux?

aticonfig --adapter=all --odsc=1000,1450

You might need to be root. The clocks are [GPU,VRAM]. You can set unique clocks for every cards (starting from adapter=0).

aticonfig --adapter=all --odcc

This applies the earlier "scratchpad" settings (makes them permanent through reboots).

Use can issue this command to read your actual clocks (and limits + current GPU load percentage which can also be very useful!):

aticonfig --adapter=all --odgc

flipme

hero member

Activity: 2170

Merit: 640

Undeads.com - P2E Runner Game

Quote from: CryptoGretzky on March 29, 2014, 06:27:56 PM

Quote from: boubou on March 29, 2014, 06:23:54 PM

Quote from: CryptoGretzky on March 29, 2014, 06:03:24 PM

So... this is on my Asus R9 290x Elphida flashed with the Stilt Bios running on PiMP 1.1 beta (64 bit BAMT)... Motherboard is a Gigabyte GA-EP45-UD3P.

Don't know if it's coincidence or what yet, I will test it on another rig, but using PiMP, but so far, it hasn't dropped after many hours running on it.... getting around 2.78 Mh/s

Oh... forgot to mention that I already tested this version of sgminer 4.1.0-96-g983e on BAMT 1.6 running on Asrock H81 Pro BTC with, but that dropped after a few mins still. Will post again after I test PiMP on another rig.

Have you got a link to your bamt image that you are using?

http://getpimp.org/

It's by the same guys (Bee) that made the older 1.3-1.5 Bamt https://litecointalk.org/index.php?topic=2924.0

Just downloading it. Since it's from Bee, does it have "donation mining" enabled, like the other distros he made?

boubou

full member

Activity: 196

Merit: 100

Quote from: CryptoGretzky on March 29, 2014, 08:03:32 PM

Quote from: CryptoGretzky on March 29, 2014, 06:58:37 PM

Ok, on my other rig, I got PiMP running and this is the result after a few mins (all are Sapphire Tri-X 290x on Asrock H81 Pro BTC):

GPU 0, 1, 2 are from a newer batch. GPU 3 is an older one where I couldn't get it stable until I flashed it with the bios that I ripped from the newer GPU.

Noticed the first 2 that the temp/fan speed can be read dropped after a few mins back to 2.4 Mh?

GPU 2,3 temp couldn't be read for some reason and they stayed up at 2.78-2.8 Mh.

So... I am thinking that auto-gpu is throttling it back. Temperature is quite low though... my temp-target is 75. Let me test temp-target with 80.

After I did a sudo aticonfig --adapter=all --initial -f, it recogznied all 4 adapters.

I also changed auto-gpu to false and temp-target to 80. On reboot, it started out at 2.6 Mh/s.

Once it hit temp-target, it dropped ALL GPU down to 2.38 Mh... I don't know what else I can do to get it to hold at that initial hash. Anyone have any ideas?

This is the best clue we had for a month...you were able to make the bug happened from a working config.

So if I sumarize it:
You had 4 290: 2 that has a drop and that has temp and fan listed, 2 others that did not have the drop and did not have temp and fan listed.
You do some aticonfig, and then all 4 has a drop and all 4 has temp and fan listed.

Is this correct?
If yes we could probably try to get back to the working config, then reproduce.

boubou

full member

Activity: 196

Merit: 100

Quote from: CryptoGretzky on March 29, 2014, 11:03:38 PM

Quote from: badman74 on March 29, 2014, 10:42:09 PM

Quote from: CryptoGretzky on March 29, 2014, 10:19:15 PM

Quote from: badman74 on March 29, 2014, 09:59:03 PM

Quote from: CryptoGretzky on March 29, 2014, 09:03:17 PM

Quote from: badman74 on March 29, 2014, 09:00:26 PM

try auto-fan
if that don't work try setting your fanspeed so it doesn't reach the temp target
for some reason it seems like disabling auto-gpu doesn't work right

I already have my auto-fan as true. Do you mean set it to false?

This is what I have as my conf already:

"api-allow" : "W:127.0.0.1,W:192.168.1.2",
"api-listen" : true,
"failover-only" : true,
"log" : "5",
"no-submit-stale": false,
"queue" : "0",
"scan-time" : "1",
"expiry" : "12",
"auto-fan" : true,
"auto-gpu" : false,
"gpu-threads" : "1",
"gpu-engine" : "1025,1025,1025,1025,1025",
"gpu-fan" : "35-100",
"gpu-memclock" : "1400,1400,1400,1400,1400",
"gpu-powertune" : "20",
"gpu-vddc" : "1.20",
"intensity" : "22",
"temp-target" : "80",
"temp-overheat" : "89",
"temp-cutoff" : "94",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"shaders" : "2816",
"thread-concurrency" : "27000",

maybe try "gpu-fan" : "50-100",
or higher...
just to keep from hitting temp-target

I set auto-fan to false, gpu-fan to 80-100. It never hit the temp-target, but it STILL dropped in hash after a few mins. argh. The only way to make it not drop was when my gpu wasn't detected... this is insanity....

well... you could try setting your clocks with sapphire trixx or the equivalent then running with --no-adl
you can also start it normally top set the clock then restart it with --no-adl

How do I set the clock outside of sgminer in Linux? Or, did you mean run sgminer, let it set the OC, then stop it, and restart with --no-adl. I did that, now none of the temp and fan are showing. But, even I did that, after a few mins, all the GPU still dropped from 2.6 back down to 2.3... argh...

It's probably not a clock issue since egineclock memoryclock stay steady, before and after the drop in hashrate.

janos666

hero member

Activity: 588

Merit: 500

I have the same problem with my 290(X)s and DarkCoin: unstable hashrate with a dropping average.

I tried to set gpu-threads from 1 to 8 and it's doesn't actually help. The 1s average seems to be significantly higher, but the real average remains the same and WU seems to drop!
Also, starting with 10 threads causes instant system-wide crash (no reboot via ssh, I have to manually flip the power switch in the garage).
And 4 or 2 threads doesn't make any real (measurable/significant) difference compared to 1 (it never did for my 290(X)s with any miner/algo/bios/clocks/etc, may be caused a slightly lower WU).

Setting lookup-gap to 1 from 2 doesn't cause any significant difference (it almost halves my hashrate with scrypt).

Setting thread-concurrency to 0.5x or 2x (the latter requires lookup-gap >=2) of my default 20480 doesn't help either.

By the way. I saw many people using worksize 512 even though it's actually maxed at 256 in every cgminer variants I know.
May be somebody should actually make that usable.
If it's as simple as changing the limit in the source then could you please point to the right file and the line number?

CryptoGretzky

sr. member

Activity: 322

Merit: 250

Quote from: badman74 on March 29, 2014, 11:15:03 PM

Quote from: CryptoGretzky on March 29, 2014, 11:03:38 PM

How do I set the clock outside of sgminer in Linux? Or, did you mean run sgminer, let it set the OC, then stop it, and restart with --no-adl. I did that, now none of the temp and fan are showing. But, even I did that, after a few mins, all the GPU still dropped from 2.6 back down to 2.3... argh...

i am pretty much out of ideas now....

One interesting note is that for that earlier config where it couldn't detect temperature, but could detect fan, the setting in the xorg.conf had this following section (only 3 GPU detected):

Section "Device"
   Identifier "aticonfig-Device[0]-0"
   Driver "fglrx"
   BusID "PCI:1:0:0"
EndSection

Section "Device"
   Identifier "aticonfig-Device[1]-0"
   Driver "fglrx"
   BusID "PCI:2:0:0"
EndSection

Section "Device"
   Identifier "aticonfig-Device[2]-0"
   Driver "fglrx"
   BusID "PCI:7:0:0"
EndSection

Well, my cards weren't plugged into the PCI:7:0:0. They were plugged into the 1, 2, 5, 6. So, somehow when it detected it as PCI:7:0:0, it gave me fan detection, but not temp, but it could hash at 2.8 Mh/s without issue and without dropping.

badman74

hero member

Activity: 658

Merit: 500

Quote from: CryptoGretzky on March 29, 2014, 11:03:38 PM

Quote from: badman74 on March 29, 2014, 10:42:09 PM

Quote from: CryptoGretzky on March 29, 2014, 10:19:15 PM

Quote from: badman74 on March 29, 2014, 09:59:03 PM

Quote from: CryptoGretzky on March 29, 2014, 09:03:17 PM

Quote from: badman74 on March 29, 2014, 09:00:26 PM

try auto-fan
if that don't work try setting your fanspeed so it doesn't reach the temp target
for some reason it seems like disabling auto-gpu doesn't work right

I already have my auto-fan as true. Do you mean set it to false?

This is what I have as my conf already:

"api-allow" : "W:127.0.0.1,W:192.168.1.2",
"api-listen" : true,
"failover-only" : true,
"log" : "5",
"no-submit-stale": false,
"queue" : "0",
"scan-time" : "1",
"expiry" : "12",
"auto-fan" : true,
"auto-gpu" : false,
"gpu-threads" : "1",
"gpu-engine" : "1025,1025,1025,1025,1025",
"gpu-fan" : "35-100",
"gpu-memclock" : "1400,1400,1400,1400,1400",
"gpu-powertune" : "20",
"gpu-vddc" : "1.20",
"intensity" : "22",
"temp-target" : "80",
"temp-overheat" : "89",
"temp-cutoff" : "94",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"shaders" : "2816",
"thread-concurrency" : "27000",

maybe try "gpu-fan" : "50-100",
or higher...
just to keep from hitting temp-target

I set auto-fan to false, gpu-fan to 80-100. It never hit the temp-target, but it STILL dropped in hash after a few mins. argh. The only way to make it not drop was when my gpu wasn't detected... this is insanity....

well... you could try setting your clocks with sapphire trixx or the equivalent then running with --no-adl
you can also start it normally top set the clock then restart it with --no-adl

How do I set the clock outside of sgminer in Linux? Or, did you mean run sgminer, let it set the OC, then stop it, and restart with --no-adl. I did that, now none of the temp and fan are showing. But, even I did that, after a few mins, all the GPU still dropped from 2.6 back down to 2.3... argh...

i am pretty much out of ideas now....

CryptoGretzky

sr. member

Activity: 322

Merit: 250

Quote from: badman74 on March 29, 2014, 10:42:09 PM

Quote from: CryptoGretzky on March 29, 2014, 10:19:15 PM

Quote from: badman74 on March 29, 2014, 09:59:03 PM

Quote from: CryptoGretzky on March 29, 2014, 09:03:17 PM

Quote from: badman74 on March 29, 2014, 09:00:26 PM

try auto-fan
if that don't work try setting your fanspeed so it doesn't reach the temp target
for some reason it seems like disabling auto-gpu doesn't work right

I already have my auto-fan as true. Do you mean set it to false?

This is what I have as my conf already:

"api-allow" : "W:127.0.0.1,W:192.168.1.2",
"api-listen" : true,
"failover-only" : true,
"log" : "5",
"no-submit-stale": false,
"queue" : "0",
"scan-time" : "1",
"expiry" : "12",
"auto-fan" : true,
"auto-gpu" : false,
"gpu-threads" : "1",
"gpu-engine" : "1025,1025,1025,1025,1025",
"gpu-fan" : "35-100",
"gpu-memclock" : "1400,1400,1400,1400,1400",
"gpu-powertune" : "20",
"gpu-vddc" : "1.20",
"intensity" : "22",
"temp-target" : "80",
"temp-overheat" : "89",
"temp-cutoff" : "94",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"shaders" : "2816",
"thread-concurrency" : "27000",

maybe try "gpu-fan" : "50-100",
or higher...
just to keep from hitting temp-target

I set auto-fan to false, gpu-fan to 80-100. It never hit the temp-target, but it STILL dropped in hash after a few mins. argh. The only way to make it not drop was when my gpu wasn't detected... this is insanity....

well... you could try setting your clocks with sapphire trixx or the equivalent then running with --no-adl
you can also start it normally top set the clock then restart it with --no-adl

How do I set the clock outside of sgminer in Linux? Or, did you mean run sgminer, let it set the OC, then stop it, and restart with --no-adl. I did that, now none of the temp and fan are showing. But, even I did that, after a few mins, all the GPU still dropped from 2.6 back down to 2.3... argh...

badman74

hero member

Activity: 658

Merit: 500

Quote from: CryptoGretzky on March 29, 2014, 10:19:15 PM

Quote from: badman74 on March 29, 2014, 09:59:03 PM

Quote from: CryptoGretzky on March 29, 2014, 09:03:17 PM

Quote from: badman74 on March 29, 2014, 09:00:26 PM

try auto-fan
if that don't work try setting your fanspeed so it doesn't reach the temp target
for some reason it seems like disabling auto-gpu doesn't work right

I already have my auto-fan as true. Do you mean set it to false?

This is what I have as my conf already:

"api-allow" : "W:127.0.0.1,W:192.168.1.2",
"api-listen" : true,
"failover-only" : true,
"log" : "5",
"no-submit-stale": false,
"queue" : "0",
"scan-time" : "1",
"expiry" : "12",
"auto-fan" : true,
"auto-gpu" : false,
"gpu-threads" : "1",
"gpu-engine" : "1025,1025,1025,1025,1025",
"gpu-fan" : "35-100",
"gpu-memclock" : "1400,1400,1400,1400,1400",
"gpu-powertune" : "20",
"gpu-vddc" : "1.20",
"intensity" : "22",
"temp-target" : "80",
"temp-overheat" : "89",
"temp-cutoff" : "94",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"shaders" : "2816",
"thread-concurrency" : "27000",

maybe try "gpu-fan" : "50-100",
or higher...
just to keep from hitting temp-target

I set auto-fan to false, gpu-fan to 80-100. It never hit the temp-target, but it STILL dropped in hash after a few mins. argh. The only way to make it not drop was when my gpu wasn't detected... this is insanity....

well... you could try setting your clocks with sapphire trixx or the equivalent then running with --no-adl
you can also start it normally top set the clock then restart it with --no-adl

CryptoGretzky

sr. member

Activity: 322

Merit: 250

Quote from: badman74 on March 29, 2014, 09:59:03 PM

Quote from: CryptoGretzky on March 29, 2014, 09:03:17 PM

Quote from: badman74 on March 29, 2014, 09:00:26 PM

try auto-fan
if that don't work try setting your fanspeed so it doesn't reach the temp target
for some reason it seems like disabling auto-gpu doesn't work right

I already have my auto-fan as true. Do you mean set it to false?

This is what I have as my conf already:

"api-allow" : "W:127.0.0.1,W:192.168.1.2",
"api-listen" : true,
"failover-only" : true,
"log" : "5",
"no-submit-stale": false,
"queue" : "0",
"scan-time" : "1",
"expiry" : "12",
"auto-fan" : true,
"auto-gpu" : false,
"gpu-threads" : "1",
"gpu-engine" : "1025,1025,1025,1025,1025",
"gpu-fan" : "35-100",
"gpu-memclock" : "1400,1400,1400,1400,1400",
"gpu-powertune" : "20",
"gpu-vddc" : "1.20",
"intensity" : "22",
"temp-target" : "80",
"temp-overheat" : "89",
"temp-cutoff" : "94",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"shaders" : "2816",
"thread-concurrency" : "27000",

maybe try "gpu-fan" : "50-100",
or higher...
just to keep from hitting temp-target

I set auto-fan to false, gpu-fan to 80-100. It never hit the temp-target, but it STILL dropped in hash after a few mins. argh. The only way to make it not drop was when my gpu wasn't detected... this is insanity....

DemetriusAstroBlack

sr. member

Activity: 338

Merit: 250

I had weird issues with sgh sgminer and temps/fans as well. I had to set fan range at 20-85 auto-gpu true, and change the temp target to 62. My cofig file is as follows:

{
"pools" : [
   {
      "url" : "stratum+tcp://stratum1.suchpool.pw:3335",
      "user" : "EDITED",
      "pass" : "x"
   },
   {
      "url" : "stratum+tcp://stratum2.suchpool.pw:3335",
      "user" : "EDITED",
      "pass" : "x"
   },
   {
      "url" : "stratum+tcp://stratum.forkpool.com:6347",
      "user" : "EDITED",
      "pass" : "x"
   }
]
,
"intensity" : "13,13,18",
"worksize" : "256,256,256",
"gpu-threads" : "2,2,1",
"thread-concurrency" : "8192,8192,15232",
"kernel" : "darkcoin,darkcoin,darkcoin",
"gpu-engine" : "500-1075,500-1075,500-1150",
"gpu-fan" : "20-85,20-85,20-85",
"gpu-memclock" : "1450,1450,1500",
"auto-gpu" : true,
"auto-fan" : true,
"temp-cutoff" : "85,85,85",
"temp-overheat" : "80,80,80",
"temp-target" : "62,62,62",

"kernel-path" : "/usr/local/bin"
}

badman74

hero member

Activity: 658

Merit: 500

Quote from: CryptoGretzky on March 29, 2014, 09:03:17 PM

Quote from: badman74 on March 29, 2014, 09:00:26 PM

try auto-fan
if that don't work try setting your fanspeed so it doesn't reach the temp target
for some reason it seems like disabling auto-gpu doesn't work right

I already have my auto-fan as true. Do you mean set it to false?

This is what I have as my conf already:

"api-allow" : "W:127.0.0.1,W:192.168.1.2",
"api-listen" : true,
"failover-only" : true,
"log" : "5",
"no-submit-stale": false,
"queue" : "0",
"scan-time" : "1",
"expiry" : "12",
"auto-fan" : true,
"auto-gpu" : false,
"gpu-threads" : "1",
"gpu-engine" : "1025,1025,1025,1025,1025",
"gpu-fan" : "35-100",
"gpu-memclock" : "1400,1400,1400,1400,1400",
"gpu-powertune" : "20",
"gpu-vddc" : "1.20",
"intensity" : "22",
"temp-target" : "80",
"temp-overheat" : "89",
"temp-cutoff" : "94",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"shaders" : "2816",
"thread-concurrency" : "27000",

maybe try "gpu-fan" : "50-100",
or higher...
just to keep from hitting temp-target

HooDude

newbie

Activity: 55

Merit: 0

Im finding now that if I have the threads up at 10,10 or even 8,8 on my pair of R9 290x I get a ton of stale shares. Anyone else having this?

EnTiTy77

newbie

Activity: 50

Merit: 0

Could be your vddc. try setting it to 0.000
and see if that makes a difference.

CryptoGretzky

sr. member

Activity: 322

Merit: 250

Quote from: badman74 on March 29, 2014, 09:00:26 PM

try auto-fan
if that don't work try setting your fanspeed so it doesn't reach the temp target
for some reason it seems like disabling auto-gpu doesn't work right

I already have my auto-fan as true. Do you mean set it to false?

This is what I have as my conf already:

"api-allow" : "W:127.0.0.1,W:192.168.1.2",
"api-listen" : true,
"failover-only" : true,
"log" : "5",
"no-submit-stale": false,
"queue" : "0",
"scan-time" : "1",
"expiry" : "12",
"auto-fan" : true,
"auto-gpu" : false,
"gpu-threads" : "1",
"gpu-engine" : "1025,1025,1025,1025,1025",
"gpu-fan" : "35-100",
"gpu-memclock" : "1400,1400,1400,1400,1400",
"gpu-powertune" : "20",
"gpu-vddc" : "1.20",
"intensity" : "22",
"temp-target" : "80",
"temp-overheat" : "89",
"temp-cutoff" : "94",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"shaders" : "2816",
"thread-concurrency" : "27000",

badman74

hero member

Activity: 658

Merit: 500

Quote from: CryptoGretzky on March 29, 2014, 08:03:32 PM

Quote from: CryptoGretzky on March 29, 2014, 06:58:37 PM

Ok, on my other rig, I got PiMP running and this is the result after a few mins (all are Sapphire Tri-X 290x on Asrock H81 Pro BTC):

GPU 0, 1, 2 are from a newer batch. GPU 3 is an older one where I couldn't get it stable until I flashed it with the bios that I ripped from the newer GPU.

Noticed the first 2 that the temp/fan speed can be read dropped after a few mins back to 2.4 Mh?

GPU 2,3 temp couldn't be read for some reason and they stayed up at 2.78-2.8 Mh.

So... I am thinking that auto-gpu is throttling it back. Temperature is quite low though... my temp-target is 75. Let me test temp-target with 80.

After I did a sudo aticonfig --adapter=all --initial -f, it recogznied all 4 adapters.

I also changed auto-gpu to false and temp-target to 80. On reboot, it started out at 2.6 Mh/s.

Once it hit temp-target, it dropped ALL GPU down to 2.38 Mh... I don't know what else I can do to get it to hold at that initial hash. Anyone have any ideas?

try auto-fan
if that don't work try setting your fanspeed so it doesn't reach the temp target
for some reason it seems like disabling auto-gpu doesn't work right

EnTiTy77

newbie

Activity: 50

Merit: 0

Quote from: CryptoGretzky on March 29, 2014, 07:18:10 PM

Quote from: EnTiTy77 on March 29, 2014, 07:09:03 PM

I know 12 threads aren't being engaged.

But it is stabilizing the hashrate

The code shows that your thread is actually back to 1 if you set it to an invalid gpu-thread count. So, if you set your thread back to 1, you will have exactly the same result.

Yup, your right Tongue

Best I leave the advising to the professionals lol

Still... stops the drop on my rig and with the low memory usage the engine clock seems to have a bit more room for tweaking. At 1060 on the core 290's. Couldn't even get to 1000 on scrypt.

mfqrs3

sr. member

Activity: 419

Merit: 250

Quote from: EnTiTy77 on March 29, 2014, 07:09:03 PM

I know 12 threads aren't being engaged.

But it is stabilizing the hashrate

I have tyed that, and hashrate went from 2.4 to almost 3, BUT avg/max hashrate doesn't change, also pool is not registering more hashes.

Topic: [ANN] sph-sgminer: multi-coin multi-algorithm GPU miner | added MaruCoin - page 35. (Read 515713 times)