OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 728.

DeathAndTaxes

donator

Activity: 1218

Merit: 1079

Gerald Davis

Quote from: Proofer on December 23, 2011, 04:59:25 PM

Should I even be thinking about overclocking? This is with spec/stock clocking on my 3x5970s:

I note that the first card, GPUs 0/1, has temps 75-80 with fan 85%. (BTW, how is the 100% value determined?)

0% to 100% is based on the cards max speed so it is card dependent. Graphics cards control fans via PWM so 85% means 85% of peak current is being supplied to the fan motor.

BEFORE overclocking you need to get that memspeed way down. Running mem @ full speed consumes a LOT of power and thus a lot of heat. High memspeed doesn't help hashing it just increases power and heat.I run @ 190 Mhz on memclock. You may need to experiment to find what results in best temp/speed combo. Also before overclocking run the system for 24 hours just to ensure it is rock solid.

Then start adjusting clocks for max speed that is stable. If this is a closed system you need some serious airflow is you are going to overclock.

BkkCoins

hero member

Activity: 784

Merit: 1009

firstbits:1MinerQ

Quote from: -ck on December 23, 2011, 06:09:35 PM

Quote from: BkkCoins on December 23, 2011, 02:00:41 PM

I use auto-fan to keep the card temp steady at 70C. I've noticed that cgminer sets the fan high when it starts (for safety?) and then slowly works it down to the speed that maintains temp. I don't like that as it's really noisy for a while. I modified my code here to set the fan to moderate 55% value at start because that's not noisy but high enough that the gpu doesn't start hot, and it's close to what I end up running long term - usually 55-65% depending on how hot a day it is here.

You know that if you specify a single fanspeed and --auto-fan, then it will start at the fanspeed you set already instead of 85% ?

Nope, had no idea. Ha ha. I'll give that a try! Thx.

Edit: I tried that but the problem with a single value is that it acts as an upper limit. If I set 50% then it starts at 50%, and will drop when cool at first, but after a while of heating up it won't go above 50% and it lets the temp rise beyond the target. It seems to treat a single value the same as 0-50 and uses it as an upper limit.

-ck

legendary

Activity: 4088

Merit: 1631

Ruu \o/

Quote from: BkkCoins on December 23, 2011, 02:00:41 PM

I use auto-fan to keep the card temp steady at 70C. I've noticed that cgminer sets the fan high when it starts (for safety?) and then slowly works it down to the speed that maintains temp. I don't like that as it's really noisy for a while. I modified my code here to set the fan to moderate 55% value at start because that's not noisy but high enough that the gpu doesn't start hot, and it's close to what I end up running long term - usually 55-65% depending on how hot a day it is here.

You know that if you specify a single fanspeed and --auto-fan, then it will start at the fanspeed you set already instead of 85% ?

gnar1ta$

donator

Activity: 798

Merit: 500

Quote from: Proofer on December 23, 2011, 04:59:25 PM

Should I even be thinking about overclocking? This is with spec/stock clocking on my 3x5970s:

It's your money...

Code:

GPU 1: 72.0C 3239RPM | 391.7/389.2Mh/s | A:18724 R:87 HW:0 U:5.35/m I:8
GPU 2: 70.5C | 387.0/380.0Mh/s | A:18137 R:86 HW:0 U:5.18/m I:8

I like more Mh/s for more BTC. Don't forget lowering your mem clock can save on power costs.

Proofer

member

Activity: 266

Merit: 36

Should I even be thinking about overclocking? This is with spec/stock clocking on my 3x5970s:

Code:


 cgminer version 2.0.7 - Started: [2011-12-23 09:48:21]
--------------------------------------------------------------------------------
 (5s):1968.9 (avg):1751.2 Mh/s | Q:3517  A:5930  R:29  HW:0  E:169%  U:24.07/m
 TQ: 11  ST: 13  SS: 1  DW: 182  NB: 20  LW: 13057  GF: 1  RF: 21
 Connected to http://us.eclipsemc.com:8337 with LP as user Redacted_321
 Block: 0000067a09a776bc6493cfd6fe308c8b...  Started: [13:41:44]
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit
 GPU 0: 75.5C 4535RPM | 330.9/289.9Mh/s | A:995 R:6 HW:0 U:4.04/m I:8
 GPU 1: 79.5C | 329.1/292.8Mh/s | A:998 R:1 HW:0 U:4.05/m I:8
 GPU 2: 74.0C 4045RPM | 328.9/292.6Mh/s | A:939 R:8 HW:0 U:3.81/m I:8
 GPU 3: 69.0C | 325.4/292.3Mh/s | A:993 R:4 HW:0 U:4.03/m I:8
 GPU 4: 74.5C 3498RPM | 328.1/292.1Mh/s | A:1004 R:6 HW:0 U:4.08/m I:8
 GPU 5: 78.5C | 327.0/291.9Mh/s | A:1002 R:4 HW:0 U:4.07/m I:8
--------------------------------------------------------------------------------

GPU 0: 324.2 / 290.0 Mh/s | A:995  R:6  HW:0  U:4.04/m  I:8
[2011-12-23 13:54:54] Pool 0 communication resumed, submitting work
75.5 C  F: 85% (4535 RPM)  E: 725 MHz  M: 1000 Mhz  V: 1.050V  A: 98% P: 0%
Last initialised: [2011-12-23 09:48:25]
Intensity: 8
Thread 0: 176.5 Mh/s Enabled ALIVE
Thread 6: 158.0 Mh/s Enabled ALIVE

GPU 1: 325.4 / 292.9 Mh/s | A:998  R:1  HW:0  U:4.05/m  I:8
79.5 C  E: 725 MHz  M: 1000 Mhz  V: 1.050V  A: 97% P: 0%
Last initialised: [2011-12-23 09:48:26]
Intensity: 8
Thread 1: 164.7 Mh/s Enabled ALIVE
Thread 7: 147.3 Mh/s Enabled ALIVE

GPU 2: 327.7 / 292.7 Mh/s | A:939  R:8  HW:0  U:3.81/m  I:8
74.0 C  F: 69% (4037 RPM)  E: 725 MHz  M: 1000 Mhz  V: 1.050V  A: 97% P: 0%
Last initialised: [2011-12-23 09:48:27]
Intensity: 8
Thread 2: 163.3 Mh/s Enabled ALIVE
Thread 8: 163.1 Mh/s Enabled ALIVE

GPU 3: 328.9 / 292.3 Mh/s | A:993  R:4  HW:0  U:4.03/m  I:8
68.5 C  E: 725 MHz  M: 1000 Mhz  V: 1.050V  A: 97% P: 0%
Last initialised: [2011-12-23 09:48:28]
Intensity: 8
Thread 3: 149.4 Mh/s Enabled ALIVE
Thread 9: 164.7 Mh/s Enabled ALIVE

GPU 4: 326.9 / 292.1 Mh/s | A:1004  R:6  HW:0  U:4.07/m  I:8
74.5 C  F: 56% (3492 RPM)  E: 725 MHz  M: 1000 Mhz  V: 1.050V  A: 97% P: 0%
Last initialised: [2011-12-23 09:48:29]
Intensity: 8
Thread 4: 165.6 Mh/s Enabled ALIVE
Thread 10: 160.2 Mh/s Enabled ALIVE

GPU 5: 323.4 / 291.9 Mh/s | A:1003  R:4  HW:0  U:4.07/m  I:8
78.5 C  E: 725 MHz  M: 1000 Mhz  V: 1.050V  A: 96% P: 0%
Last initialised: [2011-12-23 09:48:30]
Intensity: 8
Thread 5: 146.8 Mh/s Enabled ALIVE
Thread 11: 148.0 Mh/s Enabled ALIVE

I note that the first card, GPUs 0/1, has temps 75-80 with fan 85%. (BTW, how is the 100% value determined?)

Proofer

member

Activity: 266

Merit: 36

Quote from: DeathAndTaxes on December 23, 2011, 01:39:44 PM

Brief explanation of intensity. The way OpenCL works is once it starts working it can't stop. So you give it a job, it starts working and is completely unresponsive until it finishes. A nonce range is 2^32 hashes so even a 400MH/s GPU would take about 10 seconds. Most computers would die if the graphics card was unresponsive for 10 seconds so miners calculate hashes in batches. There is some overhead in setting up a batch so longer batches (higher intensity) are slightly more efficient. If intensity is too high it can make the system unresponsive because the video card is unavailable while hashing and that can "freak out" the OS.

Interesting! On all six I changed from dynamic, which was reporting I=3 or 4, mostly 4, to 7, which increased the hash rate by about 14% and power consumption by 7%. I wonder how much we are losing in order to avoid "freaking out" an OS that is designed to work with spreadsheets, video games, and email.

jjiimm_64

legendary

Activity: 1876

Merit: 1000

I also use the auto fan, because in a cool room, the card temps stay below 75 and the fan goes down significantly, lowering the overall electrical usage

BkkCoins

hero member

Activity: 784

Merit: 1009

firstbits:1MinerQ

Quote from: DeathAndTaxes on December 23, 2011, 01:39:44 PM

Quote from: Proofer on December 23, 2011, 12:40:36 PM

In my initial run, the fan speed of the card containing device 0 was much lower (20%) than that of the unused cards (85%); its temps, by the time I quit, were only in the upper 50s. Did cgminer set the fan speeds of the unused cards to 85%?

No idea. I never use auto-fan. I have every fan set @ 75%.

I use auto-fan to keep the card temp steady at 70C. I've noticed that cgminer sets the fan high when it starts (for safety?) and then slowly works it down to the speed that maintains temp. I don't like that as it's really noisy for a while. I modified my code here to set the fan to moderate 55% value at start because that's not noisy but high enough that the gpu doesn't start hot, and it's close to what I end up running long term - usually 55-65% depending on how hot a day it is here.

I think what you saw was the unused devices not getting throttled back down but I don't know why that would be unless you didn't leave it run for long as it seems to take a few minutes to come back down.

DeathAndTaxes

donator

Activity: 1218

Merit: 1079

Gerald Davis

Quote from: Proofer on December 23, 2011, 12:40:36 PM

In my initial run, the fan speed of the card containing device 0 was much lower (20%) than that of the unused cards (85%); its temps, by the time I quit, were only in the upper 50s. Did cgminer set the fan speeds of the unused cards to 85%?

No idea. I never use auto-fan. I have every fan set @ 75%.

Quote

I omitted "intensity" so got the default. Intensity is defined as "Intensity of GPU scanning (d or -10 -> 10, default: d to maintain desktop interactivity)". I don't understand what intensity really means: what is GPU "scanning"? (Is there something I should read to get me up to speed?) I notice a "scan-time" setting -- is that the same scanning activity that intensity controls?

If you omit intensity you get DYNAMIC which means cgminer adjusts intensive based on other computer workload. If you have no monitors connected to any cards (i.e. using SSH remote) then for 5970s I have found intensive 8 or 9 is best.

Brief explanation of intensity. The way OpenCL works is once it starts working it can't stop. So you give it a job, it starts working and is completely unresponsive until it finishes. A nonce range is 2^32 hashes so even a 400MH/s GPU would take about 10 seconds. Most computers would die if the graphics card was unresponsive for 10 seconds so miners calculate hashes in batches. There is some overhead in setting up a batch so longer batches (higher intensity) are slightly more efficient. If intensity is too high it can make the system unresponsive because the video card is unavailable while hashing and that can "freak out" the OS.

Simple version:
For 5870/5970/6970/6990 try intensity 8 or 9
For lower end cards try intensity 7 or 8
If you have a monitor connected try using dynamic (or low intensity) for just that GPU.

If the system seems "laggy" or unstable try a lower intensity.

Quote

gpu-threads -- I don't know how this affects things, or what the multi-thread logic looks like or is intended to accomplish. (I'm familiar with the uses of multi-threading in conventional CPU programming.)

The miner uses a thread to request work and keep data ready for the miner. cgminer miner seems to gain a tiny boost in performance by using two threads which alternate access to the GPU. I have seen no reason to every change this higher or lower.

DeathAndTaxes

donator

Activity: 1218

Merit: 1079

Gerald Davis

Quote from: Proofer on December 23, 2011, 12:52:53 PM

Quote from: README

Q: How does the donation feature work and how does it affect my shares?
A: The donation feature is disabled by default and only does anything once
enabled. It queries the author's website for login credentials and contributes
up to a proportion of work to the author's account. While the overall
accepted/rejected rates will include this work, none of these will appear in
your own accounts. On exiting, the summary will tell you how many shares were
contributed to the author.

Note: this is a technical question, not about whether or not to donate!

I don't yet really understand "shares". I can probably find material to read on that. My question is, what login credentials? To the pool I'm using? That would assume that CK is a member of any pool that a cgminer user accesses. If not, then what?

Shares are just a mechanism used by pools to track how many hashes each miner has attempted in a fair and cheat proof method.

A share is a difficulty 1 hash. Current difficulty for a block solution is 1.15 million so solving a block is 1.15 million times harder than finding a share.

Only a share which is valid for the full difficulty (~1.15 mil) solves the block but your miner submits any shares which is valid for difficulty 1 or higher to the pool. The pool keeps track of those and uses the count to fairly distribute block rewards when one miner does solve the block.

Simple version: shares are just low difficulty hashes uses to keep miners honest and provide a mechanism to fairly divided the rewards.

Full version:
http://bitcoin.stackexchange.com/q/1505/307

It is important to understand a share has no "real value" they are just an accounting mechanism.
Any hash larger than the target for full block difficulty is completely worthless.

On donation:

If you use the donation option some % (set by you) of your hashing power will work on blocks for the authors account at the pool he uses. He only has one account at one pool. It doesn't matter what pool you use. If you have donation set a 1% then 99% of the time it will hash for your account at your pool. 1% of the time it will hash for the authors account at the pool used by the author. If you have 1000 MH/s of hashing power than each second the miner will perform 990MH for you and 10 MH for the author.

BkkCoins

hero member

Activity: 784

Merit: 1009

firstbits:1MinerQ

Quote from: Proofer on December 23, 2011, 12:52:53 PM

Quote from: README

Q: How does the donation feature work and how does it affect my shares?
A: The donation feature is disabled by default and only does anything once
enabled. It queries the author's website for login credentials and contributes
up to a proportion of work to the author's account. While the overall
accepted/rejected rates will include this work, none of these will appear in
your own accounts. On exiting, the summary will tell you how many shares were
contributed to the author.

Note: this is a technical question, not about whether or not to donate!

I don't yet really understand "shares". I can probably find material to read on that. My question is, what login credentials? To the pool I'm using? That would assume that CK is a member of any pool that a cgminer user accesses. If not, then what?

cgminer will contact the developer's web site and get the info it needs for the pool/user/pwd such that cgminer can send some shares on his behalf.

A share is a portion of work. cgminer requests work from the pool and when completed it sends proof back to the pool as a share. Pools count how many shares you do to determine your pay. If the result of your work matches specific criteria then the share also indicates a "found" block and triggers the pool to announce the result to the network. This is how the pool gets paid and in turn it splits it according to all shares contributed.

cgminer is capable of managing work and shares to multiple pools at once so it has no problem doing some portion of work for another pool than the one you're using. In fact one option for pool strategy is "load balance" where it spreads your shares across multiple pools.

Proofer

member

Activity: 266

Merit: 36

Quote from: README

Q: How does the donation feature work and how does it affect my shares?
A: The donation feature is disabled by default and only does anything once
enabled. It queries the author's website for login credentials and contributes
up to a proportion of work to the author's account. While the overall
accepted/rejected rates will include this work, none of these will appear in
your own accounts. On exiting, the summary will tell you how many shares were
contributed to the author.

Note: this is a technical question, not about whether or not to donate!

I don't yet really understand "shares". I can probably find material to read on that. My question is, what login credentials? To the pool I'm using? That would assume that CK is a member of any pool that a cgminer user accesses. If not, then what?

Proofer

member

Activity: 266

Merit: 36

Quote from: jjiimm_64 on December 23, 2011, 12:09:55 PM

Quote from: Proofer on December 23, 2011, 10:00:54 AM

(2) Until I'm back to three cards, run at stock speeds (except maybe fan?). What should the relevant cgminer.conf file entries look like for that?

I have 21 5970's. i run them all with the following
{
   "_comment1" : "Any long-format command line argument ",
   "_comment2" : "may be used in this JSON configuration file",

   "url" : "http://pool:8337",
   "user" : "username.minerNumber",
   "pass" : "password",
   "gpu-engine":   "820",
   "gpu-memclock":   "420",
   "auto-fan":   true,
   "intensity":   "7"

}

Here's the relevant portion (without url, etc.) of the config for my initial run, which was a success (where success is defined as running until I enter Q, with no hanging/freezing, smoke, or menacing sounds):

"device" : "0",

"gpu-engine" : "725",
"gpu-memclock" : "1000",
"gpu-fan" : "0-85",
"temp-cutoff" : "95",
"temp-overheat" : "85",
"temp-target" : "75",
"temp-hysteresis" : "3",
"auto-fan" : true,

"expiry" : "120",
"gpu-threads" : "2",
"log" : "5",
"queue" : "1",
"retry-pause" : "5",
"scan-time" : "60",
"worksize" : "0",

Questions:

I thought I'd start with stock gpu-engine/memclock values, and later start to overclock, but let cgminer handle the fan. OK?
In my initial run, the fan speed of the card containing device 0 was much lower (20%) than that of the unused cards (85%); its temps, by the time I quit, were only in the upper 50s. Did cgminer set the fan speeds of the unused cards to 85%?

I omitted "intensity" so got the default. Intensity is defined as "Intensity of GPU scanning (d or -10 -> 10, default: d to maintain desktop interactivity)". I don't understand what intensity really means: what is GPU "scanning"? (Is there something I should read to get me up to speed?) I notice a "scan-time" setting -- is that the same scanning activity that intensity controls?

gpu-threads -- I don't know how this affects things, or what the multi-thread logic looks like or is intended to accomplish. (I'm familiar with the uses of multi-threading in conventional CPU programming.)

jjiimm_64

legendary

Activity: 1876

Merit: 1000

Quote from: Proofer on December 23, 2011, 10:00:54 AM

(2) Until I'm back to three cards, run at stock speeds (except maybe fan?). What should the relevant cgminer.conf file entries look like for that?

I have 21 5970's. i run them all with the following
{
   "_comment1" : "Any long-format command line argument ",
   "_comment2" : "may be used in this JSON configuration file",

   "url" : "http://pool:8337",
   "user" : "username.minerNumber",
   "pass" : "password",
   "gpu-engine":   "820",
   "gpu-memclock":   "420",
   "auto-fan":   true,
   "intensity":   "7"

}

Proofer

member

Activity: 266

Merit: 36

Quote from: DeathAndTaxes on December 23, 2011, 09:09:33 AM

850 is still pretty hot for 5970s if you don't have sufficient airflow. The larger issue is why are your overclocking an unstable system?

"My system is unstable let me try to overclock it less" .... how about "My system is unstable let me see if it is stable at STOCK"?

The smaller issue is some of you guys might be overlooking or forgetting what a bumbling newbie I am.

I didn't even know what stock is until just now (5970: 725MHz engine, 1GHz memory).

Without advice to the contrary, I want to do two things:

(1) Use only one card (of the three I have), then when that one's checked out, add one more, etc. Does it matter if I do this -- remove and add cards -- by (a) editing /etc/X11/xorg.conf and rebooting; (b) physically removing/adding, rebooting, doing aticonfig --adapter=all --initial, rebooting again; or (c) using cgminer's --device|-d option?

(2) Until I'm back to three cards, run at stock speeds (except maybe fan?). What should the relevant cgminer.conf file entries look like for that?

DeathAndTaxes

donator

Activity: 1218

Merit: 1079

Gerald Davis

Quote from: Proofer on December 23, 2011, 02:01:54 AM

I changed gpu-engine from "0-950" to "850" and removed both "auto" settings. That caused it to go a few more seconds -- to the extent of outputing three "Accepted" messages that it hadn't previously...

[2011-12-22 22:38:28] Long-polling activated for http://us.eclipsemc.com:8337/LP
[2011-12-22 22:38:32] Accepted 00000000.943f188b.123aad4f GPU 2 thread 2 pool 0
GPU2 47.5C 4504RPM | (5s):326.7 (avg):4026.5 Mh/s | A:1 R:0 HW:0 U:600.00/m I:8
[2011-12-22 22:38:33] Accepted 00000000.ddd1a27e.08200eba GPU 1 thread 1 pool 0
GPU1 52.5C | (5s):363.6 (avg):222.7 Mh/s | A:1 R:0 HW:0 U:11.99/m I:5
[2011-12-22 22:38:34] Accepted 00000000.307282d1.c779af85 GPU 2 thread 2 pool 0
GPU2 48.0C 4519RPM | (5s):333.0 (avg):238.0 Mh/s | A:2 R:0 HW:0 U:23.98/m I:8

...but then the system was frozen.

So I think we might be on a productive track. The above was still with three cards. Next, I'll remove two. Any other ideas for hardware-related experiments, including .conf values, welcome.

850 is still pretty hot for 5970s if you don't have sufficient airflow. The larger issue is why are your overclocking an unstable system?

"My system is unstable let me try to overclock it less" .... how about "My system is unstable let me see if it is stable at STOCK"?

Fiyasko

legendary

Activity: 1428

Merit: 1001

Okey Dokey Lokey

Quote from: ancow on December 23, 2011, 02:13:49 AM

Quote from: Proofer on December 23, 2011, 02:01:54 AM

GPU2 47.5C 4504RPM | (5s):326.7 (avg):4026.5 Mh/s | A:1 R:0 HW:0 U:600.00/m I:8
[2011-12-22 22:38:33] Accepted 00000000.ddd1a27e.08200eba GPU 1 thread 1 pool 0
GPU1 52.5C | (5s):363.6 (avg):222.7 Mh/s | A:1 R:0 HW:0 U:11.99/m I:5

You could try lowering intensity and not using the dynamic intensity setting.

Im with the Dynamic intensity setting beinf a factor, Think about it "GPU runs at 98% when mining" "Okay, Raise intensity by 1!" "k!" "100%GPU" "Hey thats too high, Lower it by 1!" "----------------------------" "hey dude you there?" "---------" "Sh!t he's at 100% and cant respond to me telling him to calm down!"

ancow

full member

Activity: 373

Merit: 100

Quote from: Proofer on December 23, 2011, 02:01:54 AM

GPU2 47.5C 4504RPM | (5s):326.7 (avg):4026.5 Mh/s | A:1 R:0 HW:0 U:600.00/m I:8
[2011-12-22 22:38:33] Accepted 00000000.ddd1a27e.08200eba GPU 1 thread 1 pool 0
GPU1 52.5C | (5s):363.6 (avg):222.7 Mh/s | A:1 R:0 HW:0 U:11.99/m I:5

You could try lowering intensity and not using the dynamic intensity setting.

kano

legendary

Activity: 4634

Merit: 1851

Linux since 1997 RedHat 4

Quote from: Proofer on December 23, 2011, 02:01:54 AM

As some of you kind and helpful folks will undoubtedly recall, I haven't been able to get cgminer running on my new 3x5970 rig; it hangs, or it freezes the system tight, after a few seconds. Over at the Pools/Eclipse forum, I started to ask about configuring apps such as poclbm and Phoenix, but we got sidetracked into my can't-run-cgminer problem. BkkCoins had a possibly useful insight, and as a result I'm coming back here to pursue it:

Quote from: BkkCoins on December 23, 2011, 01:19:30 AM

... The only time I've ever had lock ups is when gpu settings were pushed too far. Engine 950 is not that high but some cards will choke on even that. I'd definitely start really conservative at 850 and default mem speed and then see how stability is before altering any gpu settings.

Cards behave very differently even from the same mfr. batch. I have 2 5830s bought together and one craps out above 970 while the other is happy at 1020. If you got a fussy card then anything higher than 850 could be causing failures. If cgminer starts and outputs init messages then I'd say software is likely ok and you should debug hardware issues and really back off to ultra conservative values, and no "auto" settings that may be causing cgminer to adjust things. Also start with one card and add cards as stability is verified.

I changed gpu-engine from "0-950" to "850" and removed both "auto" settings. That caused it to go a few more seconds -- to the extent of outputing three "Accepted" messages that it hadn't previously...

[2011-12-22 22:38:28] Long-polling activated for http://us.eclipsemc.com:8337/LP
[2011-12-22 22:38:32] Accepted 00000000.943f188b.123aad4f GPU 2 thread 2 pool 0
GPU2 47.5C 4504RPM | (5s):326.7 (avg):4026.5 Mh/s | A:1 R:0 HW:0 U:600.00/m I:8
[2011-12-22 22:38:33] Accepted 00000000.ddd1a27e.08200eba GPU 1 thread 1 pool 0
GPU1 52.5C | (5s):363.6 (avg):222.7 Mh/s | A:1 R:0 HW:0 U:11.99/m I:5
[2011-12-22 22:38:34] Accepted 00000000.307282d1.c779af85 GPU 2 thread 2 pool 0
GPU2 48.0C 4519RPM | (5s):333.0 (avg):238.0 Mh/s | A:2 R:0 HW:0 U:23.98/m I:8

...but then the system was frozen.

So I think we might be on a productive track. The above was still with three cards. Next, I'll remove two. Any other ideas for hardware-related experiments, including .conf values, welcome.

Lower the settings until that doesn't happen ...

Proofer

member

Activity: 266

Merit: 36

As some of you kind and helpful folks will undoubtedly recall, I haven't been able to get cgminer running on my new 3x5970 rig; it hangs, or it freezes the system tight, after a few seconds. Over at the Pools/Eclipse forum, I started to ask about configuring apps such as poclbm and Phoenix, but we got sidetracked into my can't-run-cgminer problem. BkkCoins had a possibly useful insight, and as a result I'm coming back here to pursue it:

Quote from: BkkCoins on December 23, 2011, 01:19:30 AM

... The only time I've ever had lock ups is when gpu settings were pushed too far. Engine 950 is not that high but some cards will choke on even that. I'd definitely start really conservative at 850 and default mem speed and then see how stability is before altering any gpu settings.

Cards behave very differently even from the same mfr. batch. I have 2 5830s bought together and one craps out above 970 while the other is happy at 1020. If you got a fussy card then anything higher than 850 could be causing failures. If cgminer starts and outputs init messages then I'd say software is likely ok and you should debug hardware issues and really back off to ultra conservative values, and no "auto" settings that may be causing cgminer to adjust things. Also start with one card and add cards as stability is verified.

I changed gpu-engine from "0-950" to "850" and removed both "auto" settings. That caused it to go a few more seconds -- to the extent of outputing three "Accepted" messages that it hadn't previously...

[2011-12-22 22:38:28] Long-polling activated for http://us.eclipsemc.com:8337/LP
[2011-12-22 22:38:32] Accepted 00000000.943f188b.123aad4f GPU 2 thread 2 pool 0
GPU2 47.5C 4504RPM | (5s):326.7 (avg):4026.5 Mh/s | A:1 R:0 HW:0 U:600.00/m I:8
[2011-12-22 22:38:33] Accepted 00000000.ddd1a27e.08200eba GPU 1 thread 1 pool 0
GPU1 52.5C | (5s):363.6 (avg):222.7 Mh/s | A:1 R:0 HW:0 U:11.99/m I:5
[2011-12-22 22:38:34] Accepted 00000000.307282d1.c779af85 GPU 2 thread 2 pool 0
GPU2 48.0C 4519RPM | (5s):333.0 (avg):238.0 Mh/s | A:2 R:0 HW:0 U:23.98/m I:8

...but then the system was frozen.

So I think we might be on a productive track. The above was still with three cards. Next, I'll remove two. Any other ideas for hardware-related experiments, including .conf values, welcome.

Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 728. (Read 5805728 times)