Topic: cgminer and 6990's - want to start tweaking them for max performance (Read 4605 times)
February 01, 2014, 08:29:19 AM
Thread necro detected!
February 01, 2014, 08:28:16 AM
Hi everyone ,
I get stable without hangs with 6990 total hashrate 950khs /s it is amazing card yesterday i've got 500khs - per core but
now it isnt stable and its wave from 430 to 510 on each core . it takes 400W but the speed is amazing
I will post the snapshot from miner later because I'm out of my mining rigs
Cheers
I get stable without hangs with 6990 total hashrate 950khs /s it is amazing card yesterday i've got 500khs - per core but
now it isnt stable and its wave from 430 to 510 on each core . it takes 400W but the speed is amazing
I will post the snapshot from miner later because I'm out of my mining rigs
Cheers
October 31, 2011, 12:37:58 PM
Have you flipped the overclock switch? Im not entirely sure, but if you havent, I think the card will throttle to stay within 375W, potentially rendering your overclock pointless.
yes i have flipped the overclock switch.before i did my Mh/s jumed around a lot.
now its steady.
also waranty is void once you put a waterblock on the card.
but it runs a lot cooler. never goes above 50Deg C
Is it cheaper to cool them with water than air con or other methods people are using?
50c is very low - is it costing more to keep them that low?
well there is the intial cost of the waterblock, pump radiator and fan.
your looking at $120 for water block, $70 for pump, $80 for rad(120x4), and $20-$40 for fans. (depens on what fans you get. I use Scythe Ultra Kaze 120MM fans)(best bang fro buck i think)
you need about 1 fan per GPU, so with the above configuration you could cool 2 6990's (my 50Deg C is colling 1 6990 and 1 5770 video card) ambient temp around 27 deg.
other disadvantage is that any maintance you want to do it could involve draining your loop, and refilling. and if your not careful, posible leaks.
advantages are where you put your rads. I cut a hole in my cold air duct and am not venting the heat into it which heats my house. (sub zero temperatures and i havn't turned on the furnace yet)
I also have 4 6970's watercooled on a single rad(4x120) and GPUs doesn't go above 53 deg. (45-53 deg)
in the summer i plan to just put the rads outside in a box and dump all the heat out there.
I guess your need to factor in your air conditioning costs. how much it takes long term.
then there is setup time. and if you enjoy this. I have been watercooling now for 8 years, and enjoy it as a hobby.
once its set up it just runs.
October 31, 2011, 12:36:46 PM
Have you flipped the overclock switch? Im not entirely sure, but if you havent, I think the card will throttle to stay within 375W, potentially rendering your overclock pointless.
yes i have flipped the overclock switch.before i did my Mh/s jumed around a lot.
now its steady.
also waranty is void once you put a waterblock on the card.
but it runs a lot cooler. never goes above 50Deg C
Is it cheaper to cool them with water than air con or other methods people are using?
50c is very low - is it costing more to keep them that low?
No nothing is cheaper than air. 24 of my GPU are aircooled in my garage. I have another 6 GPU in my workstation in my office which are watercooled.
Watercooling is quiet, very efficient, and allows higher overclocking but the cost of waterblocks make it more expensive than "cheap but noisy air". It is pretty cool though to have >2.5GH quieter than a desk fan. Next summer I am going to rig up some mount for my window so I can mount the radiator outside. Dumping >1KW directly outside and eliminating the AC cost too.
October 31, 2011, 12:25:20 PM
Have you flipped the overclock switch? Im not entirely sure, but if you havent, I think the card will throttle to stay within 375W, potentially rendering your overclock pointless.
yes i have flipped the overclock switch.before i did my Mh/s jumed around a lot.
now its steady.
also waranty is void once you put a waterblock on the card.
but it runs a lot cooler. never goes above 50Deg C
Is it cheaper to cool them with water than air con or other methods people are using?
50c is very low - is it costing more to keep them that low?
October 31, 2011, 11:08:17 AM
Have you flipped the overclock switch? Im not entirely sure, but if you havent, I think the card will throttle to stay within 375W, potentially rendering your overclock pointless.
yes i have flipped the overclock switch.before i did my Mh/s jumed around a lot.
now its steady.
also waranty is void once you put a waterblock on the card.
but it runs a lot cooler. never goes above 50Deg C
Only if you tell them.
Then again I have never had a watercooled GPU die.
October 31, 2011, 10:03:25 AM
Have you flipped the overclock switch? Im not entirely sure, but if you havent, I think the card will throttle to stay within 375W, potentially rendering your overclock pointless.
yes i have flipped the overclock switch.before i did my Mh/s jumed around a lot.
now its steady.
also waranty is void once you put a waterblock on the card.
but it runs a lot cooler. never goes above 50Deg C
October 31, 2011, 06:06:21 AM
Have you flipped the overclock switch? Im not entirely sure, but if you havent, I think the card will throttle to stay within 375W, potentially rendering your overclock pointless.
October 30, 2011, 06:09:15 PM
i have mine watercooled,
it clocked at 960mhz core and 775 memory
i get 429 Mh/s from each core.
Intesity 9
thing to watch when your overclocking is the voltage regulators.
if your doing this in windows... run GPUz
it will show your voltage regulators as well
it clocked at 960mhz core and 775 memory
i get 429 Mh/s from each core.
Intesity 9
thing to watch when your overclocking is the voltage regulators.
if your doing this in windows... run GPUz
it will show your voltage regulators as well
October 30, 2011, 04:23:46 PM
I've tweaked my 2*6990 for several weeks. The best settings I found (w/o hanging my computer) is:
Windows 7 with ATI Catalyst 11.7 AMD APP SDK 2.5
Engine 900 Mhz with memory 775 Mhz, powertune 20%
mining engine - cgminer 2.0.7, Intensity 9, with --auto-fan but do not turn on auto-gpu
which give me ~400 Mhash/s speed.
(I've tried ubuntu as well, but it can only show me around 350 Mhash/s. I've heard of linuxcoin but haven't tried it. Maybe next week.)
Well uping the intensity should help I believe it is mentioned specifically in the readme not to use over 8 unless you have a 6990 so try a 10 for it and see what that gets you maybe even a 25-50mhz bump in the speed of the cards as well. You should be getting better than that rate hell I have a 5830 that does 320mh/s on its own mind you it is running at 1000,300 to do that but even still you should be getting better than you are.
I tried setting the clock to 950 on one of the GPU's - got a hard hang. All the other GPU's slowed down to more or less nothing. Had to go down on reboot the thing - so annoying. cgminer should be switching the clock between 830 and 930 already anyway.
Windows 7 with ATI Catalyst 11.7 AMD APP SDK 2.5
Engine 900 Mhz with memory 775 Mhz, powertune 20%
mining engine - cgminer 2.0.7, Intensity 9, with --auto-fan but do not turn on auto-gpu
which give me ~400 Mhash/s speed.
(I've tried ubuntu as well, but it can only show me around 350 Mhash/s. I've heard of linuxcoin but haven't tried it. Maybe next week.)
Hi,
As per the title I want to start tweaking these machines now that they are all up and running and stable.
I'm using cgminer and auto-fan (with max set to 80), auto-gpu without specifying a clock range. Automatic intensity is on and hovers under 5 predominantly. Average throughput from each GPU is 342MH/s. It seems some people have got up to 400MH/s per gpu or 800 per card.
I want to avoid (like the plague) finding myself in a position where I have to reboot the machine over and over again. I'd nearly just leave them stock rather than the hassle.
I changed the air-con temp from 22c down to 17c as a test to see if cgminer would work the gpu's harder, that was about a half hour ago and no performance increase has been noted as of yet.
As per the title I want to start tweaking these machines now that they are all up and running and stable.
I'm using cgminer and auto-fan (with max set to 80), auto-gpu without specifying a clock range. Automatic intensity is on and hovers under 5 predominantly. Average throughput from each GPU is 342MH/s. It seems some people have got up to 400MH/s per gpu or 800 per card.
I want to avoid (like the plague) finding myself in a position where I have to reboot the machine over and over again. I'd nearly just leave them stock rather than the hassle.
I changed the air-con temp from 22c down to 17c as a test to see if cgminer would work the gpu's harder, that was about a half hour ago and no performance increase has been noted as of yet.
Well uping the intensity should help I believe it is mentioned specifically in the readme not to use over 8 unless you have a 6990 so try a 10 for it and see what that gets you maybe even a 25-50mhz bump in the speed of the cards as well. You should be getting better than that rate hell I have a 5830 that does 320mh/s on its own mind you it is running at 1000,300 to do that but even still you should be getting better than you are.
I tried setting the clock to 950 on one of the GPU's - got a hard hang. All the other GPU's slowed down to more or less nothing. Had to go down on reboot the thing - so annoying. cgminer should be switching the clock between 830 and 930 already anyway.
Nice one dude. That has got my hash rates up to 400, however... with massive variance. It jumps up to 500, even 600, then down to figures like 280 (or less
) and back up. However, the average reading is staying at roughly 344 for each card (which is not brilliant considering I was getting an avg. of 342 before). 
This is with cgminer on Linuxcoin.
October 30, 2011, 07:38:56 AM
I've tweaked my 2*6990 for several weeks. The best settings I found (w/o hanging my computer) is:
Windows 7 with ATI Catalyst 11.7 AMD APP SDK 2.5
Engine 900 Mhz with memory 775 Mhz, powertune 20%
mining engine - cgminer 2.0.7, Intensity 9, with --auto-fan but do not turn on auto-gpu
which give me ~400 Mhash/s speed.
(I've tried ubuntu as well, but it can only show me around 350 Mhash/s. I've heard of linuxcoin but haven't tried it. Maybe next week.)
Well uping the intensity should help I believe it is mentioned specifically in the readme not to use over 8 unless you have a 6990 so try a 10 for it and see what that gets you maybe even a 25-50mhz bump in the speed of the cards as well. You should be getting better than that rate hell I have a 5830 that does 320mh/s on its own mind you it is running at 1000,300 to do that but even still you should be getting better than you are.
I tried setting the clock to 950 on one of the GPU's - got a hard hang. All the other GPU's slowed down to more or less nothing. Had to go down on reboot the thing - so annoying. cgminer should be switching the clock between 830 and 930 already anyway.
Windows 7 with ATI Catalyst 11.7 AMD APP SDK 2.5
Engine 900 Mhz with memory 775 Mhz, powertune 20%
mining engine - cgminer 2.0.7, Intensity 9, with --auto-fan but do not turn on auto-gpu
which give me ~400 Mhash/s speed.
(I've tried ubuntu as well, but it can only show me around 350 Mhash/s. I've heard of linuxcoin but haven't tried it. Maybe next week.)
Hi,
As per the title I want to start tweaking these machines now that they are all up and running and stable.
I'm using cgminer and auto-fan (with max set to 80), auto-gpu without specifying a clock range. Automatic intensity is on and hovers under 5 predominantly. Average throughput from each GPU is 342MH/s. It seems some people have got up to 400MH/s per gpu or 800 per card.
I want to avoid (like the plague) finding myself in a position where I have to reboot the machine over and over again. I'd nearly just leave them stock rather than the hassle.
I changed the air-con temp from 22c down to 17c as a test to see if cgminer would work the gpu's harder, that was about a half hour ago and no performance increase has been noted as of yet.
As per the title I want to start tweaking these machines now that they are all up and running and stable.
I'm using cgminer and auto-fan (with max set to 80), auto-gpu without specifying a clock range. Automatic intensity is on and hovers under 5 predominantly. Average throughput from each GPU is 342MH/s. It seems some people have got up to 400MH/s per gpu or 800 per card.
I want to avoid (like the plague) finding myself in a position where I have to reboot the machine over and over again. I'd nearly just leave them stock rather than the hassle.
I changed the air-con temp from 22c down to 17c as a test to see if cgminer would work the gpu's harder, that was about a half hour ago and no performance increase has been noted as of yet.
Well uping the intensity should help I believe it is mentioned specifically in the readme not to use over 8 unless you have a 6990 so try a 10 for it and see what that gets you maybe even a 25-50mhz bump in the speed of the cards as well. You should be getting better than that rate hell I have a 5830 that does 320mh/s on its own mind you it is running at 1000,300 to do that but even still you should be getting better than you are.
I tried setting the clock to 950 on one of the GPU's - got a hard hang. All the other GPU's slowed down to more or less nothing. Had to go down on reboot the thing - so annoying. cgminer should be switching the clock between 830 and 930 already anyway.
October 29, 2011, 06:27:13 PM
Hi,
As per the title I want to start tweaking these machines now that they are all up and running and stable.
I'm using cgminer and auto-fan (with max set to 80), auto-gpu without specifying a clock range. Automatic intensity is on and hovers under 5 predominantly. Average throughput from each GPU is 342MH/s. It seems some people have got up to 400MH/s per gpu or 800 per card.
I want to avoid (like the plague) finding myself in a position where I have to reboot the machine over and over again. I'd nearly just leave them stock rather than the hassle.
I changed the air-con temp from 22c down to 17c as a test to see if cgminer would work the gpu's harder, that was about a half hour ago and no performance increase has been noted as of yet.
As per the title I want to start tweaking these machines now that they are all up and running and stable.
I'm using cgminer and auto-fan (with max set to 80), auto-gpu without specifying a clock range. Automatic intensity is on and hovers under 5 predominantly. Average throughput from each GPU is 342MH/s. It seems some people have got up to 400MH/s per gpu or 800 per card.
I want to avoid (like the plague) finding myself in a position where I have to reboot the machine over and over again. I'd nearly just leave them stock rather than the hassle.
I changed the air-con temp from 22c down to 17c as a test to see if cgminer would work the gpu's harder, that was about a half hour ago and no performance increase has been noted as of yet.
Well uping the intensity should help I believe it is mentioned specifically in the readme not to use over 8 unless you have a 6990 so try a 10 for it and see what that gets you maybe even a 25-50mhz bump in the speed of the cards as well. You should be getting better than that rate hell I have a 5830 that does 320mh/s on its own mind you it is running at 1000,300 to do that but even still you should be getting better than you are.
I tried setting the clock to 950 on one of the GPU's - got a hard hang. All the other GPU's slowed down to more or less nothing. Had to go down on reboot the thing - so annoying. cgminer should be switching the clock between 830 and 930 already anyway.
October 29, 2011, 04:49:12 PM
Hi,
As per the title I want to start tweaking these machines now that they are all up and running and stable.
I'm using cgminer and auto-fan (with max set to 80), auto-gpu without specifying a clock range. Automatic intensity is on and hovers under 5 predominantly. Average throughput from each GPU is 342MH/s. It seems some people have got up to 400MH/s per gpu or 800 per card.
I want to avoid (like the plague) finding myself in a position where I have to reboot the machine over and over again. I'd nearly just leave them stock rather than the hassle.
I changed the air-con temp from 22c down to 17c as a test to see if cgminer would work the gpu's harder, that was about a half hour ago and no performance increase has been noted as of yet.
As per the title I want to start tweaking these machines now that they are all up and running and stable.
I'm using cgminer and auto-fan (with max set to 80), auto-gpu without specifying a clock range. Automatic intensity is on and hovers under 5 predominantly. Average throughput from each GPU is 342MH/s. It seems some people have got up to 400MH/s per gpu or 800 per card.
I want to avoid (like the plague) finding myself in a position where I have to reboot the machine over and over again. I'd nearly just leave them stock rather than the hassle.
I changed the air-con temp from 22c down to 17c as a test to see if cgminer would work the gpu's harder, that was about a half hour ago and no performance increase has been noted as of yet.
Jump to: