Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 763. (Read 5805728 times)

Thanking you muchly!

1. Well that's very unlucky. Not sure what I can do about it though, software shouldn't be able to take out your entire machine (unless you're doing dangerous clocks and shit).
2. Some hardware really doesn't like the memory being much slower than the gpu clock speed. 69x0 cards in particular usually ignore lower settings to avoid that from happening.
3. The 6990 has dissociated sensors on the 2nd core so it really doesn't know what the fan speed is. Furthermore, the ATI drivers and the ATI display library do NOT have a way of agreeing on what the device numbers are, so we're left to *hoping* they coincide in order. However, the temperature and fan speeds reported by ADL are coming from the same device so even if your GPU number listed doesn't correspond with the fanspeed you think, the temperature/fanspeed combination listed for one card is definitely the temperature and fanspeed for that core. Bear in mind, though, that the 2nd core of 6990s doesnt have a fanspeed and is just sharing the fanspeed from the 1st core. Thus any fanspeed it's reporting will likely be a percentage value that is being ignored and the fan of its corresponding partner core in RPM is the real fanspeed.

Switched on 0.5% donation on my 4gh/s farm, hardly felt the difference. Hope that makes the difference for you, Con! Keep up the good work. Thanks!

Yay, finally got a real one (pool of course, not solo) Block 147825 (Ars)
[2011-10-03 10:12:03] Accepted 00000000.00000943.a6f1622c BLOCK! GPU 0 thread 0 pool 1
(That's Aus EDT 35 minutes ago)

Hi there. I'm running a Windows 7 x64 machine and dual 6990's. I just switched to cgminer after having used poclbm directly launched from the command line for about 3 months without issues. Since switching, I have three problems:

1) Without fail, every time I [Q]uit the miner, it causes a BSOD. The driver reporting a failure is atikmdag.sys. I was trying Catalyst 11.5 when it started happening. Updated to Catalyst 11.9 and the behavior is the same. The only way for me to safely exit the program is to KILL the process, and then I use another program to bring the fan speeds down to normal, or reboot.

2) Attempting to set the memory clock speed to anything below 950Mhz causes instability. Anything below 900Mhz and the system locks up instantly. I've found the best course of action is to leave it at the default setting.

3) Fan speed. NO MATTER WHAT I DO, one of the fans eventually reaches 100%. Ambient temperature in the room is 68º. I never had this problem using poclbm directly. Here are my settings:
--auto-fan --auto-gpu --gpu-engine 800-900 --gpu-fan 60-75 --temp-target 85 --temp-overheat 90

I believe what is happening with the fan speed problem is that cgminer is attributing each fan to the wrong GPUs. Invariably, GPU 0 and GPU 2 are cool, at around 75º, while GPU1 and GPU3 are hot at around 89-91º. My guess is that 0+2 is one card, and 1+3 is the second card, but cgminer is thinks that the cards are 0+1 and 2+3 respectively. Therefore, it is not associating the GPU temperatures with the correct GPU fan speed settings. In fact, 0+2 used to run hotter than 1+3, because the card is physically on top of the other card in the machine. It is only so cool now because the fan is blowing so hard.

Yes that's right, the "donated" power so far is amassing to between 100-200 Mhash at the moment. The bithopper app which uses a 1% donation model is getting ~2.4Ghash from "donations".

Mhash/s

Just out of curiosity and because I couldn't find this looking at the website or MOTD, what unit would the returned value be in?

For those that dont have an irc client go to http://ozco.in and click the "WebIrc Chat" link

Thanks for using Ozcoin ckolivas
And thanks once again for an awesome piece of mining sofware - all my miners are set to 1% - be interesting to see what would happen if a few hundred of the thousands of ppl profiting from useing cgminer set 1% 
It would still be small reward for the hours of time, effort and support contributed to the Bitcoin community by ckolivas (conman in IRC)

Thanking you very much If you drop into #ozcoin and type "!worker ckolivas.donation" anyone can see the total hash power contributed to donations at this time.

Correct. I went out of my way to make it as unobtrusive as possible. Therefore if it can't connect to my account it will just quietly stop getting work from there. It will not roll work, and it will only get *up to* the donation amount of work to do. Therefore if you set it to 1% it almost never will actually amount to 1%. Submitted donation shares will show on the screen as "donate" instead of "pool x". You can tell how many shares are submitted at the end in the summary (unless you are on windows and have the binary which crashes on every exit and get no summary).

That is my understand.

For example if you donated 0.5%
99.5% of the time the miner will work on your pool
0.5% of the time it works on the developer's pool.

How do donations work? Does it allocate a percentage of hashing power to the owners pool or something?

Been running the 2.0.5 with --donation 1 turned on.. works great.. keep up the good work!

Can you copy and paste your AMDOverDriver configuration file, with overvolt, for your 5970?!

Thanks,
Thiago

as a long time folding@home addict (we run 100% GPU load 24/7 too), I can assure you prolonged running at 100% fan is NOT good for your card. some odd non ref coolers are OK with it (ones with multiple lower speed 92's say) but with the bulk of the cards out there its a good way to kill the fan in 6 months or so. it may still turn at reduced RPM (and/or increased noise) but the bearings will be pretty much be shot.

60-70% is probably the best compromise speed for cooling vs long life. I have gotten 3+ years of 24/7 use from 5 or 6 video cards at 65-70% fan.

Hmm running the fans constantly at 100% ...
Isn't that considered a bad idea since the fans will die much faster than expected?
That's one of the things that cgminer tries to avoid for that specific reason.
I guess it may not be considered bad on some cards?
(I don't know)

It depends... for example, I'm running three 5970s in one machine under Ubuntu 11.04 64-bit.  memclocks are all at 180.  core clocks on one card are 820,810 , another 850,840, and another 870, 860.  All at stock voltages.  I arrived at those core clocks by watching which gpu's got sick and lowering them 10 mhz at a time until they didn't get sick very often.  I use -I 9 on the command line.

At 870 mhz, I average 400 MH/s, at 850 its 390 MH/s, and at 810 its 370 Mh/s.  If you can overvolt, you can probably get higher performance.

I keep the fans at 100% and temps are all in the 70-75C range in a warm-ish room with 16 rigs.

I don't use cgminer to set the clocks or fans (yet), I still use AMDOverdriveCtrl in a bash script because that's what I had been doing before cgminer came along.  I originally had higher core clocks under phoenix, but get as good or better performance from cgminer with fewer crashes.

Anyone running 5970's with cgminer?  What settings are best and what hash rates should I be getting?

Yeah - it will have 5 zeros in the second 8 nibbles.
I actually have exactly the expected number of 3 zeros and even the expected number of 4 zeros, but no 5 zeros yet
Each one will of course be 16 times less likely than the previous.

Or I should say, to be exactly correct, with the current difficulty it will be less than or equal to 00000000.000009ee.5d000000
(thus to have an a->f after the zero is also not low enough ...)