Mining Troubleshooting

[bittenbob]

Hi everyone,

I have just upgraded from a GTS 250 to an HD 6950 for mining and gaming but I cannot seem to get mining to work on the new card.

Every time i launch guiminer it says no OpenCL devices installed. If I check in GPU-Z it does not say OpenCL is supported but according the the specs for the card 1.1 should be.

I have tried installing the latest version of steam SDK and have the latest graphics drivers installed as well (Catalyst 11.10). If it helps I am running Windows 7 Ultimate 64 bit, Phenom II X6 1090T on a Sabertooth 990FX board.

Any help is appreciated as I would like to get mining by the time I go to bed again tonight.

[1714039915]

1714039915

[1714039915]

1714039915

[1714039915]

1714039915

[RyNinDaCleM]

Did you completely remove ALL nVidia drivers using something like driver sweeper or ccleaner?

[bittenbob]

I just tried removing all drivers including the Catalyst 11.10 and installed 11.07.

07 is the last version I see officially supported by SDK on its AMD page and I can once again mine. If/when I update my drivers to 11.10 I suspect it will stop working again. Can anyone confirm SDK works on Catalyst version 11.10?

[bittenbob]

So I am using the default miner on guiminer without any extra flags and I am getting about 310 Mhash/s. I will be primarily mining at night and while I am at work so I have no qualms about my computer being slow to respond. My graphics card is XFX Radeon 6950 ZNDC XXX.

Can anyone give me any tips on how to optimize my mining without damaging my graphics card?
The default clocks for my card are:
Core Clock:   830 MHz
Memory Clock:   5200 MHz

I have read that I can push the core clock to 880Mhz on stock voltages which I would prefer to stay on.

Flags/Reccommended clock speeds, etc would be greatly appreciated.

Also as per my last post if anyone could confirm SDK should work properly with this card with the latest graphics drivers that would be great (11.10).

Update

With Phoenix and "-k phatk2 VECTORS BFI_INT FASTLOOP=false AGGRESSION=11" I am getting 340Mhash/s. Quite an improvement but cannot get deepbit to connect with phoenix.

Update 2

With default miner and "-v -w256" I am getting 330-340... acceptable now
May try pushing hte core speed to 850 tomorrow but im happy with what I am getting, especially since I was getting 35Mhash/s with my GTS250. Still no confirmation on Catalyst 11.10's ability to run SDK properly.

[RyNinDaCleM]

For GUIminer, use these flags:

Code:

-v -w128

cgminer is a faster,  more efficient miner! Set intensity to 8 if you decide to use it.

I run mine at 935/1000 on stock voltage. Set your miners' affinity to just 1 CPU core. And keep cool!

[bittenbob]

Is there a reason I should only set the affinity to 1 core? I have 6 cores and if there is any performance gain I wouldn't mind using them. For polcbm I use all 6 assuming there is some benefit but I havent tweaked it a whole lot. I removed the workload clause form polcbm and it performs the same as with it so it appears to have little to no effect on the 6950. I have a Corsair A50 on my CPU it and I havent even broken 35C on prime95 even so thermals are not an issue for me on the CPU.

 I might try cgminer next and report on my results in this thread.

Update

Cannot get cgminer to extract because every time I do my antivirus detects it as a "trojan". I know its a false positive because some people must be using it in botnets. Just out of curiosity what kind of performance gains should I expect if I use this over polcbm with my current settings?

Update 2

Trying to connect to deepbit to no avail with CGminer. I added an exception so that it would no longer be detected as a false positive. I notice it isnt on their list of supported miners but sometimes it can take about 10 minutes or so to connect normally.

Update 3

CONNECTED!!! Running at 330Mhash/s instead of 330-340 though. polcbm with -v as a flag enabled appears to be faster. Shortly after it dropped to 230Mhash and it doesnt appear to be using any CPU's... Back to guiminer I guess. Do you have any other suggestions? It started out strong (not as strong as polcbm) and then just declined

[Gabi]

I just tried removing all drivers including the Catalyst 11.10 and installed 11.07.

07 is the last version I see officially supported by SDK on its AMD page and I can once again mine. If/when I update my drivers to 11.10 I suspect it will stop working again. Can anyone confirm SDK works on Catalyst version 11.10?

Every driver is fine, sdk doesn't matter. So, install 11.10 and mine (i have 11.10)

Make sure you install the whole package of course.

I use guiminer with -v -w128 (-w256 is fine too). If you want add -f0, it will make the computer a bit more sluggish but should squeeze some mh/s more.

About clocks, underclock your memory as much as possible, so your card will have a much lower temperature and will consume less. I have a 6950 and i mine with 300mhz memory clock

[DeathAndTaxes]

Is there a reason I should only set the affinity to 1 core? I have 6 cores and if there is any performance gain I wouldn't mind using them. For polcbm I use all 6 assuming there is some benefit but I havent tweaked it a whole lot. I removed the workload clause form polcbm and it performs the same as with it so it appears to have little to no effect on the 6950. I have a Corsair A50 on my CPU it and I havent even broken 35C on prime95 even so thermals are not an issue for me on the CPU.

CPU mining is worthless.  The coins aren't worth their electrical cost.  If you are CPU mining just understand you are burning up a huge amount of electricity for a negligible amount of bitcoins.  

Most people only mine w/ GPU (i.e. no CPUs selected in GUI miner).

If you really want to use your CPU then
1) Use only GPU for Bitcoin
2) Get a Litecoin miner (see Alternate currency forum) and use CPU cycles to mine Litecoins.

Don't like Litecoin?  Who cares use an exchange like BTC-E to trade LTC for BTC.  You will earn far more BTC that way.

[Fiyasko]

Is there a reason I should only set the affinity to 1 core? I have 6 cores and if there is any performance gain I wouldn't mind using them. For polcbm I use all 6 assuming there is some benefit but I havent tweaked it a whole lot. I removed the workload clause form polcbm and it performs the same as with it so it appears to have little to no effect on the 6950. I have a Corsair A50 on my CPU it and I havent even broken 35C on prime95 even so thermals are not an issue for me on the CPU.

 I might try cgminer next and report on my results in this thread.

Update

Cannot get cgminer to extract because every time I do my antivirus detects it as a "trojan". I know its a false positive because some people must be using it in botnets. Just out of curiosity what kind of performance gains should I expect if I use this over polcbm with my current settings?

Update 2

Trying to connect to deepbit to no avail with CGminer. I added an exception so that it would no longer be detected as a false positive. I notice it isnt on their list of supported miners but sometimes it can take about 10 minutes or so to connect normally.

Update 3

CONNECTED!!! Running at 330Mhash/s instead of 330-340 though. polcbm with -v as a flag enabled appears to be faster. Shortly after it dropped to 230Mhash and it doesnt appear to be using any CPU's... Back to guiminer I guess. Do you have any other suggestions? It started out strong (not as strong as polcbm) and then just declined

Use GUIminer with MSIafter burner to do your clocking and mining
The whole "set cpu core to 1 affinity" Is for people with More than One GPU, If you dont have more than one, Then just put this thought out of your head.
I reccommend using (flags) -v -w128 -f5 for overnight mining with your GPU CORE at about 880mhz, And your GPU Memory core down at about 575mhz,  Dropping the memory clock will save you about 4-12°C in heat,

When Using your comp, have the -f flag at -f250, This will allow you to still play majority of games at a high frame rate while still mining.
Do NOT EVER overclock your memory, IMO it's always the reason WHY someones card broke

Use the Lastest version of catalyst, There are better effiecencys with the ATi Stream tech, that give higher m/hash

If you've not use MSIafter burner, Then keep in mind that your going to need to set a few things in the .cfg to 'clock' 'properly'
LOL i've got one of my 6970's running at 1050core 300mem, And it's only running at 81° with fans at 85%!.

MINE SOME BITCOINS #LIKEABOSS

[bittenbob]

Thanks for all the help guys. Electricity cost isnt a problem for me at all since I rent. My landlord decided to jack up my rent $50 by waiting until january instead of the $8 that would be if they would have done it normally. As a result I have no qualms about using extra electricity to mine coins inefficiently.

I will try updating my drivers again tonight and run with f- 5 tonight when I do to bed.

The -w flag did not seem to make any difference whether at 128 256 or just not included at all. It appears the vectoring made a huge difference. I havent installed MSI afterburner yet but i hear good things. Someone at work told me that if overdrive in catalyst wont let me make changes that afterburner wouldnt though. Overdrive is enabled by default (my card is OC'd stock) but will not let me change any clock speeds.

Update:

Updated to 11.10 and it works fine now. Must have been a driver interaction issue with Nvidia. Thank you to whoever suggested this first. It boosted my mining rate to 340.6 steady. Overdrive lets me change the clocks up to 840Mhz for the core now but does not let me adjust the memory speed below 1300Mhz. This boosted my mining to 345 without specifying the -f flag. I tried running afterburner but it still will not let me clock past 840Mhz. That is fine and I will keep that setting for the forseeable future. I tried underclocking the memory but even that would not allow me to push the core speed any faster.

Update 2:

Tried the -f flag and it seems to make no difference. I have it set at f -2 and it performs the same as not being specified. I think its safe to say that with my 6950 the only flag that improves hash rate is the vector one.

[Fiyasko]

Thanks for all the help guys. Electricity cost isnt a problem for me at all since I rent. My landlord decided to jack up my rent $50 by waiting until january instead of the $8 that would be if they would have done it normally. As a result I have no qualms about using extra electricity to mine coins inefficiently.

I will try updating my drivers again tonight and run with f- 5 tonight when I do to bed.

The -w flag did not seem to make any difference whether at 128 256 or just not included at all. It appears the vectoring made a huge difference. I havent installed MSI afterburner yet but i hear good things. Someone at work told me that if overdrive in catalyst wont let me make changes that afterburner wouldnt though. Overdrive is enabled by default (my card is OC'd stock) but will not let me change any clock speeds.

Update:

Updated to 11.10 and it works fine now. Must have been a driver interaction issue with Nvidia. Thank you to whoever suggested this first. It boosted my mining rate to 340.6 steady. Overdrive lets me change the clocks up to 840Mhz for the core now but does not let me adjust the memory speed below 1300Mhz. This boosted my mining to 345 without specifying the -f flag. I tried running afterburner but it still will not let me clock past 840Mhz. That is fine and I will keep that setting for the forseeable future. I tried underclocking the memory but even that would not allow me to push the core speed any faster.

Update 2:

Tried the -f flag and it seems to make no difference. I have it set at f -2 and it performs the same as not being specified. I think its safe to say that with my 6950 the only flag that improves hash rate is the vector one.

Using MSIa is quite complex
You need to go into the MSIafterburner.cfg
and set these parameters:
[ATIADLHAL]
UnofficialOverclockingEULA   = I confirm that I am aware of unofficial overclocking limitations and fully understand that MSI will not provide me any support on it
UnofficialOverclockingMode   = 2
AccessibilityCheckingPeriod   = 0
-----------------------------------------
Now the magic starts to happen,
When you drag a slider Allllllll the way over to one side, (such a the memclock) Then hit apply, Then, CLOSE MSIa
Reload it, And Voila That "minimum" point is now the "halfway" point, You can do this repeatedly
My msia lets my 6870's cores goto 1400max and memcore 255min

[Gabi]

As jackrabbit said

First, modify the .cfg file in that way

Then, change clocks!

Pay attention to core clock, keep an eye on temperature.

As for memclock, go happily down. At first you will be able to set it at 625mhz. Save and close msi afterburner. Reopen and it will let you move the slider down at 310mhz. Save and you are ok. 300 or 310 is like the same thing. If you want you can close and reopen again and go down to like 150 but i never tried that and i do not guarantee that it will work and/or it's safe.

Underclocking memory is very helpful, the card will heat MUCH less and that's a good thing

[bittenbob]

Thanks again for the help. I was considering uninstalling MSI afterburner but will try running at 880Mhz core and 500Mhz ram speed now. I can try running lower settings but I think it should be alright with those. I will let you know the new hash speed in an update to this post.

Update

Memory clock will not go below 650 and set core to 881 without chaning stock voltage. I am mining at 360Mhash/s now = w00t. Was running at 84C on average but am getting 81-82C with these settings which is acceptable for this card I read. I could crank the fan up Im sure by changing the profile but this is acceptable from what I have read.

Thanks everyone

Update 2

83C at 899Mhz core clock. 368 Mhash/s. Not sure I want to push it a whole lot faster than that because I do want to use this card for gaming lol. If someone can verify that they have been running faster settings on a 6950 of comparible specs (see above) for the last 6 months I will try that. I read that someone else could only push theirs to 880Mhz so I am grateful to get ~900

[bittenbob]

Alright so against my better judgement as per my last post I tweaked the card some more to see what I could get out of it. I significantly improved performance but also hit some limits. It seems anything above 960Mhz causes the ATI drivers to crash after a few minutes of mining. The memory speed has a small effect on mining speed but it does seem to make a difference. The "sweet spot" I have been running in now is:

Core Voltage - 1.2 V (+0.1V)
Core Clock - 957Mhz
Memory Clock - 400 Mhz
Fan 80%
Temp 81C
Mining Speed 389

This appears to be stable but I am worried about the speed drop. Can anyone comment on whether or not this will be safe for long term use?

Edit

Found how to program the fan profile which is very nice. I have it configured to run 80% at 80C, 90% at 90C and 100% at 100C (hope I never hit that). This makes me feel a little bit better than a fixed fan speed which could potentially allow overheating. I also disabled ATI overdrive which other threads have suggested to prevent the risk of it interfering.

[DeathAndTaxes]

Most people push the 6950 faster by unlocking the 15% extra shaders that the 6970 has.  The 6950/6970 use the same GPU.  AMD just turns off 15% of the shaders in the 6950.  With shader unlock you can get over 400MH/s.

If you have a 2 GB reference model it is very easy just download the 6970 2GB bios and flash it.  The computer will think it has a 6970 (which is 99% accurate just remember the real 6970 has beefier VRM to handle higher current so don't go too crazy).

If you have a non-reference design or 1 GB model you can't do a straight bios flash but people have made unlock utilities which modify the bios to tell AMD drivers it has the full number of SP.

You likely want you memory speeds lower.  It cuts thermal output (which reduces life of card), and lowers your electric bill.  Try playing around with setting from 300MHz to 400Mhz.  It has been a while since I used a 6950/6970 but IIRC there is a sweet spot in that range.  When trying a mem setting let it run for a couple hours to and record average hash rate.  Remember hash rate varies so give it a little while before decided you found the fastest speed.

[bittenbob]

Most people push the 6950 faster by unlocking the 15% extra shaders that the 6970 has.  The 6950/6970 use the same GPU.  AMD just turns off 15% of the shaders in the 6950.  With shader unlock you can get over 400MH/s.

If you have a 2 GB reference model it is very easy just download the 6970 2GB bios and flash it.  The computer will think it has a 6970 (which is 99% accurate just remember the real 6970 has beefier VRM to handle higher current so don't go too crazy).

If you have a non-reference design or 1 GB model you can't do a straight bios flash but people have made unlock utilities which modify the bios to tell AMD drivers it has the full number of SP.

You likely want you memory speeds lower.  It cuts thermal output (which reduces life of card), and lowers your electric bill.  Try playing around with setting from 300MHz to 400Mhz.  It has been a while since I used a 6950/6970 but IIRC there is a sweet spot in that range.  When trying a mem setting let it run for a couple hours to and record average hash rate.  Remember hash rate varies so give it a little while before decided you found the fastest speed.

The card is non-reference design and I heard the ones that could be flashed have a bios switch which I was unable to locate. I tried running the mem clock as low as 300 but at 400 I can push about an extra 2 Mhash out of the card. I let the card run for about 20 minutes at each setting keeping a running average of the hash speeds. This is how I discovered that anything above this caused crashes anywhere from 2-10 minutes after running at full load. I have not tried unlocking the extra shaders but I did have it running briefly at 407 Mhash/s then crash. The fan profile I have created helps to keep the thermals down. Whats interesting is that even though afterburner says the voltage is at 1.2V HWMonitor is still saying 1.10V. My power supply might not handle a 6970 too well because its a 600W thermaltake with 36A on 2 12V rails. This makes me think that I am probably pushing the max safe Mhash/s out of my card.

Thanks again for the suggestions.


Update

So after several hours of mining I had a black screen crash. This was either due to the mining or poor coding of a DNS analysis tool I was running. Going to run these settings overnight and see if it happens again.

[DeathAndTaxes]

I would recommend not running two voltage monitoring utilities at the same time.  I recall a thread about that having the potential to damage some VRMs.

Even if you get 0.5% higher throughput at 400Mhz mem you likely save 10W per card at 300Mhz.  If you are paying for electricity what matters is net generation (gross revenue - electrical cost). A kWh saved is a kWh earned.  If you get a Kill-A-Watt you can see for yourself.  If you have free electricity well then it doesn't really matter.

[bittenbob]

I would recommend not running two voltage monitoring utilities at the same time.  I recall a thread about that having the potential to damage some VRMs.

Even if you get 0.5% higher throughput at 400Mhz mem you likely save 10W per card at 300Mhz.  If you are paying for electricity what matters is net generation (gross revenue - electrical cost). A kWh saved is a kWh earned.  If you get a Kill-A-Watt you can see for yourself.  If you have free electricity well then it doesn't really matter.

I live in an apartment so I do not pay for electricity. The landlord recently decided to jack my rent up $50 instead of the $8 that would normally be allowed so I have no qualms about using extra electricity. My only concern would be overloading my 600W power supply.

Update

Some famous last words. It died last night in violent popping death. Was planning on buying a new one anyways