OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 267.

HellDiverUK

hero member

Activity: 1246

Merit: 501

Quote from: Aurum on July 26, 2013, 05:44:52 AM

For 7XXXs I'd recommend uninstalling Catalyst 13.1. Then install Catalyst 13.4. But, 7XXXs run great with the latest drivers so I just stay current.

Disagree. 12.8 are still the most consistently stable version for me. They literally just work. Test on Windows 7 machines with 2x7950, another with a 6970, and with another machine with a 7850 and 7770.

HellDiverUK

hero member

Activity: 1246

Merit: 501

Quote from: Aurum on July 26, 2013, 05:26:21 AM

Every time I go looking for answers I see no end of posts saying you must put these lines in your batch file:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
For Win7 users I think this is nonsense.

As for GPU_MAX_ALLOC_PERCENT 100 it seems to do nothing. Trying running for 24 hours with it set to 100, 40 and then deleting it. I don't see any difference.

I hear what you're saying, but you're wrong.

You may have found that, but I couldn't mine on a machine with 4GB RAM with 2x7950 until I added the setx commands.

Yes, it should be possible to just do it once, but Windows being Windows does like to change stuff (be it some other application, Windows Update or even ATI driver). It does no harm to leave the commands there.

Really, people wouldn't be recommending it, if it didn't work.

Aurum

sr. member

Activity: 453

Merit: 250

dfgfdgfdg

Quote from: San1ty on July 26, 2013, 05:33:39 AM

Quote from: Aurum on July 26, 2013, 05:26:21 AM

Every time I go looking for answers I see no end of posts saying you must put these lines in your batch file:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
For Win7 users I think this is nonsense. I deleted them from my batch file and from Windows Environmental Variables since they stay there forever until deleted. If I open Windows Resource Monitor I use significantly less CPU with a multi-core CPU than if I set GPU_USE_SYNC_OBJECTS 1, which as best as I can tell forces the use of a single core. Aren't CPUs designed to direct traffic? Why not let them do their job? Most of my miners use a low power consumption AMD Sempron 145 CPU which only has one core anyway. Why even tell windows to use only one core when there is only one core?

As for GPU_MAX_ALLOC_PERCENT 100 it seems to do nothing. Trying running for 24 hours with it set to 100, 40 and then deleting it. I don't see any difference.

Thank you very much for your advice, I'll try excluding --auto-gpu (although I think it's a long shot

).
I have set those two user variables in windows 8, but like you said, I'm not sure they are doing much.

Also what I didn't mention yet is my OS Spec:
Windows 8 64-Bit
Catalyst Drivers 13.1

Anything wrong with that?

Hi, Your main problem is twofold: intensity is way too high for 7xxx GPUs, and thread concurrency is too high.
I have 32-bit and 64-bit Windows 7 PCs and have never tried Win8.
For 7XXXs I'd recommend uninstalling Catalyst 13.1. Running AMDs cleanup utility. Running Guru3D Driver Sweeper. Then install Catalyst 13.4. I never tried 13.1 and it may work just fine. But, 7XXXs run great with the latest drivers so I just stay current.

Just my two cents worth.

San1ty

sr. member

Activity: 462

Merit: 250

Quote from: Aurum on July 26, 2013, 05:26:21 AM

Every time I go looking for answers I see no end of posts saying you must put these lines in your batch file:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
For Win7 users I think this is nonsense. I deleted them from my batch file and from Windows Environmental Variables since they stay there forever until deleted. If I open Windows Resource Monitor I use significantly less CPU with a multi-core CPU than if I set GPU_USE_SYNC_OBJECTS 1, which as best as I can tell forces the use of a single core. Aren't CPUs designed to direct traffic? Why not let them do their job? Most of my miners use a low power consumption AMD Sempron 145 CPU which only has one core anyway. Why even tell windows to use only one core when there is only one core?

As for GPU_MAX_ALLOC_PERCENT 100 it seems to do nothing. Trying running for 24 hours with it set to 100, 40 and then deleting it. I don't see any difference.

Thank you very much for your advice, I'll try excluding --auto-gpu (although I think it's a long shot

).
I have set those two user variables in windows 8, but like you said, I'm not sure they are doing much.

Also what I didn't mention yet is my OS Spec:
Windows 8 64-Bit
Catalyst Drivers 13.1

Anything wrong with that?

Aurum

sr. member

Activity: 453

Merit: 250

dfgfdgfdg

Every time I go looking for answers I see no end of posts saying you must put these lines in your batch file:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
For Win7 users I think this is nonsense. I deleted them from my batch file and from Windows Environmental Variables since they stay there forever until deleted. If I open Windows Resource Monitor I use significantly less CPU with a multi-core CPU than if I set GPU_USE_SYNC_OBJECTS 1, which as best as I can tell forces the use of a single core. Aren't CPUs designed to direct traffic? Why not let them do their job? Most of my miners use a low power consumption AMD Sempron 145 CPU which only has one core anyway. Why even tell windows to use only one core when there is only one core?

As for GPU_MAX_ALLOC_PERCENT 100 it seems to do nothing. Trying running for 24 hours with it set to 100, 40 and then deleting it. I don't see any difference.

Aurum

sr. member

Activity: 453

Merit: 250

dfgfdgfdg

Quote from: San1ty on July 26, 2013, 04:03:47 AM

--scrypt --api-listen --api-network --api-port 4028 -I 20,20 -g 1 --thread-concurrency 22400 --lookup-gap 2 --temp-target 75,75 --temp-overheat 80,80 --temp-cutoff 85,85 --gpu-memclock 1250,1250 --gpu-engine 300-1100,300-1100 --auto-fan --auto-gpu

Another thing, I see you're putting all this in a command line. Better to minimize your batch file and edit your conf file. Here's my batch:
timeout /t 20
cgminer.exe -c MultiPool_Miner_Aurum.conf
pause
exit
I make a shortcut and put it in my Startup folder. I set my BIOS so that if the power goes out it restarts. The timeout is to allow my miner to discover a network connection and finish loading stuff before launching cgminer. The pause gives me a chance to read the run summary before closing the cmd window.

Aurum

sr. member

Activity: 453

Merit: 250

dfgfdgfdg

Quote from: San1ty on July 26, 2013, 04:03:47 AM

Hi All,

I'm currently running CGMiner on 2 7950's (NO CROSSFIRE), with the following settings:

--scrypt --api-listen --api-network --api-port 4028 -I 20,20 -g 1 --thread-concurrency 22400 --lookup-gap 2 --temp-target 75,75 --temp-overheat 80,80 --temp-cutoff 85,85 --gpu-memclock 1250,1250 --gpu-engine 300-1100,300-1100 --auto-fan --auto-gpu

I'm getting extremely frustrated with cgminer results as they make no sense to me, for example:

GPU: 1, Rate: 310kh/s, Temp: 75 °C, Fan Percent: 37%, GPU Clock: 1100, Mem Clock: 1250, Intensity 20, HW Errors: 0
GPU: 2, Rate: 660kh/s, Temp: 76 °C, Fan Percent: 100%, GPU Clock: 500, Mem Clock: 1250, Intensity 20, HW Errors: 0

How the hell is it possible that GPU 2 has a higher hashrate with the lowest GPU Clock?
Why is my Fan Percent on GPU 2 100%?
Why isn't my Fan Percent not going up on GPU 1 as temps are about equal?
Why isn't my GPU Clock of GPU 1 throttling down?

This is driving me nuts, I have went through the documentation and numerous forum posts but no help.

Please, Anyone, Help me out!
Cheers,

I don't know the answer to your questions but I never had a lick of luck with auto-gpu. All my 7970s run great on this configuration:
"scrypt" : true,
"intensity" : "13,13",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"thread-concurrency" : "10240,10240",
"shaders" : "2048,2048",
"gpu-engine" : "1050,1050",
"gpu-fan" : "0-85",
"gpu-memclock" : "1650,1650",
"gpu-vddc" : "0,0",
"temp-cutoff" : "95",
"temp-overheat" : "88",
"temp-target" : "75",
"temp-hysteresis" : "9",
"api-port" : "4028",
"auto-fan" : true,
"expiry" : "120",
"scan-time" : "111",
"queue" : "1",
"failover-only" : true,
"gpu-dyninterval" : "77",
"gpu-threads" : "1",
"hotplug" : "0",
"log" : "5",
"shares" : "0",
"kernel-path" : "/usr/local/bin"

San1ty

sr. member

Activity: 462

Merit: 250

Hi All,

I'm currently running CGMiner on 2 7950's (NO CROSSFIRE), with the following settings:

--scrypt --api-listen --api-network --api-port 4028 -I 20,20 -g 1 --thread-concurrency 22400 --lookup-gap 2 --temp-target 75,75 --temp-overheat 80,80 --temp-cutoff 85,85 --gpu-memclock 1250,1250 --gpu-engine 300-1100,300-1100 --auto-fan --auto-gpu

I'm getting extremely frustrated with cgminer results as they make no sense to me, for example:

GPU: 1, Rate: 310kh/s, Temp: 75 °C, Fan Percent: 37%, GPU Clock: 1100, Mem Clock: 1250, Intensity 20, HW Errors: 0
GPU: 2, Rate: 660kh/s, Temp: 76 °C, Fan Percent: 100%, GPU Clock: 500, Mem Clock: 1250, Intensity 20, HW Errors: 0

How the hell is it possible that GPU 2 has a higher hashrate with the lowest GPU Clock?
Why is my Fan Percent on GPU 2 100%?
Why isn't my Fan Percent not going up on GPU 1 as temps are about equal?
Why isn't my GPU Clock of GPU 1 throttling down?

This is driving me nuts, I have went through the documentation and numerous forum posts but no help.

Please, Anyone, Help me out!

Cheers,

kano

legendary

Activity: 4592

Merit: 1851

Linux since 1997 RedHat 4

Quote from: flound1129 on July 25, 2013, 11:19:17 PM

Quote from: kano on July 25, 2013, 10:27:25 PM

Quote from: flound1129 on July 25, 2013, 08:10:16 PM

Why does this happen?

http://i.imgur.com/BPqpMjU.png

In cgminer I have queue="1", why is cgminer continuing to submit difficulty 16 shares after the pool diff changes to 32?

This is in 3.3.1.

Irrelevant of the value of queue, it is not ideal for cgminer to throw away work it is doing (or just done) when diff changes.

Tell the pool to fix their code.

... an issue with stratum that was swept under the carpet with a hack.

I am the pool, and this code isn't live yet. I'm trying to patch stratum to do it the correct way. What do you suggest?

Obviously you can't reject valid work you sent to the miner.
How long will a miner take to mine work? Ignoring crappy slow hardware below 100MH/s ... 100MH/s will take 43 seconds.

The stratum flaw is that difficulty is not part of the work.
Read around here:
https://bitcointalksearch.org/topic/m.1287015
It should be part of that.
The hack is to send them together ... though of course that doesn't guarantee they will arrive together ...

But either way, the point is that when you get work and have a difficulty, any work that reaches that difficulty should be accepted by the pool.
As mentioned in that link, ignoring those shares, e.g. on a PPS pool, is simply the pool ripping off the miners.

flound1129

hero member

Activity: 938

Merit: 1000

www.multipool.us

Quote from: Jcw188 on July 25, 2013, 11:24:33 PM

Quote from: Jcw188 on July 25, 2013, 10:54:10 PM

I have finally gotten some mining going after a lot of fooling around with drivers. I'm running xp 32 bit, Radeon 6670s, sdk 2.5 and catalyst 11.7. I have two machines that I believe are almost identical. However one I have running fine at 94 khash/sec.

The big problem is the other machine that seems to be getting a lot of hardware errors (hw). Anyone hav any ideas what can cause this? I'm just running it standard with I 15 and have tried fooling around with the core
Clock and memory. Thanks for help I just don't understand hat can cause hw errors.

Btw this is scrypt mining, coincidentally at flounds pool multipool.in. An yes this stratum issue is really annoying, hits both my miners to cause often disconnects which I can only assume are slowing my hashing.

I've never gotten good results with I > 13 on any of my miners.

Aurum

sr. member

Activity: 453

Merit: 250

dfgfdgfdg

Quote from: Jcw188 on July 25, 2013, 11:24:33 PM

Quote from: Jcw188 on July 25, 2013, 10:54:10 PM

I have finally gotten some mining going after a lot of fooling around with drivers. I'm running xp 32 bit, Radeon 6670s, sdk 2.5 and catalyst 11.7. I have two machines that I believe are almost identical. However one I have running fine at 94 khash/sec.

The big problem is the other machine that seems to be getting a lot of hardware errors (hw). Anyone hav any ideas what can cause this? I'm just running it standard with I 15 and have tried fooling around with the core
Clock and memory. Thanks for help I just don't understand hat can cause hw errors.

Btw this is scrypt mining, coincidentally at flounds pool multipool.in. An yes this stratum issue is really annoying, hits both my miners to cause often disconnects which I can only assume are slowing my hashing.

I don't have any 6xxxs but my 5970s mine scrypt using Catalyst 12.8 and 7970s using the latest Catalyst 13.4. Don't install SDK separately, just let Catalyst install it. Are you using CGMiner 3.3.1?

Sometimes when a miner is headless the card in slot 1 runs very slow. If I plug in a monitor or a dummy load it ramps up to full speed even if I unplug monitor later.

If you go back a few pages ckolivas was discussing the stratum disconnect issue says there's a fix but it hasn't come out in a release yet. He described a way to pull the fix code out of the git library.

Jcw188

hero member

Activity: 546

Merit: 500

Carpe Diem

Quote from: Jcw188 on July 25, 2013, 10:54:10 PM

I have finally gotten some mining going after a lot of fooling around with drivers. I'm running xp 32 bit, Radeon 6670s, sdk 2.5 and catalyst 11.7. I have two machines that I believe are almost identical. However one I have running fine at 94 khash/sec.

The big problem is the other machine that seems to be getting a lot of hardware errors (hw). Anyone hav any ideas what can cause this? I'm just running it standard with I 15 and have tried fooling around with the core
Clock and memory. Thanks for help I just don't understand hat can cause hw errors.

Btw this is scrypt mining, coincidentally at flounds pool multipool.in. An yes this stratum issue is really annoying, hits both my miners to cause often disconnects which I can only assume are slowing my hashing.

flound1129

hero member

Activity: 938

Merit: 1000

www.multipool.us

Quote from: kano on July 25, 2013, 10:27:25 PM

Quote from: flound1129 on July 25, 2013, 08:10:16 PM

Why does this happen?

http://i.imgur.com/BPqpMjU.png

In cgminer I have queue="1", why is cgminer continuing to submit difficulty 16 shares after the pool diff changes to 32?

This is in 3.3.1.

Irrelevant of the value of queue, it is not ideal for cgminer to throw away work it is doing (or just done) when diff changes.

Tell the pool to fix their code.

... an issue with stratum that was swept under the carpet with a hack.

I am the pool, and this code isn't live yet. I'm trying to patch stratum to do it the correct way. What do you suggest?

Jcw188

hero member

Activity: 546

Merit: 500

Carpe Diem

I have finally gotten some mining going after a lot of fooling around with drivers. I'm running xp 32 bit, Radeon 6670s, sdk 2.5 and catalyst 11.7. I have two machines that I believe are almost identical. However one I have running fine at 94 khash/sec.

The big problem is the other machine that seems to be getting a lot of hardware errors (hw). Anyone hav any ideas what can cause this? I'm just running it standard with I 15 and have tried fooling around with the core
Clock and memory. Thanks for help I just don't understand hat can cause hw errors.

kano

legendary

Activity: 4592

Merit: 1851

Linux since 1997 RedHat 4

Quote from: flound1129 on July 25, 2013, 08:10:16 PM

Why does this happen?

http://i.imgur.com/BPqpMjU.png

In cgminer I have queue="1", why is cgminer continuing to submit difficulty 16 shares after the pool diff changes to 32?

This is in 3.3.1.

Irrelevant of the value of queue, it is not ideal for cgminer to throw away work it is doing (or just done) when diff changes.

Tell the pool to fix their code.

... an issue with stratum that was swept under the carpet with a hack.

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

Quote from: erk on July 25, 2013, 08:27:30 PM

Any plans to support the new Intel® SHA Extensions?

http://software.intel.com/en-us/articles/intel-sha-extensions

No.

erk

hero member

Activity: 826

Merit: 500

Any plans to support the new Intel® SHA Extensions?

http://software.intel.com/en-us/articles/intel-sha-extensions

flound1129

hero member

Activity: 938

Merit: 1000

www.multipool.us

Quote from: crazyates on July 23, 2013, 11:33:05 PM

Anyone have any clue why multiple people have noticed weird hashrate increases a few nights ago while using LTC in cgminer? See the posts here: https://bitcointalksearch.org/topic/m.2789664

Apparently over the night a few nights ago, while tinkering with my 7970 (which was not performing as well as it should be, but it never has), it just decided to work while I was sleeping. Upped the hashrate from 522KH/s to >700KH/s (which it should be getting) for no reason. I checked it's status the next morning, and it instantly decided to go back to it's slow 522KH/s. Havn't been able to get it back up to the faster speeds since.

This wasn't just CGMiner mis-reporting the hashrate, either. The WU: rate was much higher than it had been. And other people have noticed similar issues. I'm not sure if their issue was the same night as mine, as I think our dates were 1 night off.

Any help would be appreciated. Thanks!

flound1129

hero member

Activity: 938

Merit: 1000

www.multipool.us

Why does this happen?

In cgminer I have queue="1", why is cgminer continuing to submit difficulty 16 shares after the pool diff changes to 32?

This is in 3.3.1.

Foss

full member

Activity: 125

Merit: 100

ckolivas
Hi, can you add into cgminer:
1) SHA 3 algorithm for mining CopperLark
2) time counter like in BFGMiner:

Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 267. (Read 5805537 times)