One card couldnt oc as much as the other
...That's what we've been telling you
Topic: Power Drain amout issues [Closed~] (Read 2151 times)
...That's what we've been telling you
I insist you keep the BTC, Legit. As I said, you've been quite helpful to me.
That's funny. I definitely suggested that first. I'll consider the payment to Legit compensation for advice he gave me in another thread. 
Yep, sorry I didn't notice but you definitely did suggest that particular issue first. (Though not as eloquently, if I may be allowed stroke my own ego
)What is your donation address so I can send the 0.2 BTC he sent me your way?
http://blockexplorer.com/address/1MwsYzs11tUPbu2iqywSjvu5854H1FYABn
That's funny. I definitely suggested that first. I'll consider the payment to Legit compensation for advice he gave me in another thread.
One card couldnt oc as much as the other
To all: I received payment, all is good.
Just FYI.
Just FYI.
Thanks for the validation, it will be my first donation once it arrives. 
Out of curiosity, what did your issue turn out to be?
+1
You are kinda wrong on this. While typically the 2xsha256 kernel could require as few as 4 bytes of __global (video) memory to store kernel results, you are likely utilizing much more than that. The kernel code is initially placed in the video memory and most modern AMD GPUs have just 48KB worth of instruction cache (this might be different on Cayman as far as size is concerned, haven't checked). Any bitcoin kernel binary takes more than 100KB and naturally does not fit there and this leads to performance degradation the bigger kernel size is. This is also beyond what profilers like sprofile can show you.
Out of curiosity, what did your issue turn out to be?
Oh, and Legit strikes again!
Oh, and Legit strikes again!
It's probably not power that is causing the issue, likely its either a buggy card, or the motherboard needs a firmware upgrade.
Test over-clocking with each card individually, see if the freezes occur only with one card, or if they work separately just fine.
To make sure power is not the issue:
You can get a simple multimeter for 10-15 bucks from nearly any hardware store/walmart to test the 12v Rails when at full load, see if they drop amperage significantly, that could also be an issue. Setting the cards on different Rails is also a good idea.
Also check with the manufacturer of the motherboard for a bios update, sometimes that fixes issues. Also in the bios you can sometimes set "constant voltage" or "load spectrum" or similar, that should make over-clocking the GFX more stable.
Adjusting the PCI frequency/HT freq, can help on some AMD mobos.
Test over-clocking with each card individually, see if the freezes occur only with one card, or if they work separately just fine.
To make sure power is not the issue:
You can get a simple multimeter for 10-15 bucks from nearly any hardware store/walmart to test the 12v Rails when at full load, see if they drop amperage significantly, that could also be an issue. Setting the cards on different Rails is also a good idea.
Also check with the manufacturer of the motherboard for a bios update, sometimes that fixes issues. Also in the bios you can sometimes set "constant voltage" or "load spectrum" or similar, that should make over-clocking the GFX more stable.
Adjusting the PCI frequency/HT freq, can help on some AMD mobos.
You win!
I respectfully disagree, Kiwi. Hashing makes minimal use of video memory and extreme use of the GPU. Naturally, this means that a slight flaw in the GPU will be more pronounced in a setup that has the GPU running faster than it ever safely could without the memory downclocked. Ergo, the best stress test for hashing, is hashing.
Although the best test for hashing is hashing, the best test for a truely stable overclock is something that stresses both heavily. I believe this is what he meant, though I may be wrong.
Mining may only use a small amount of ram.. but it does use it, so any instability in the ram will affect mining, despite being less pronounced or even unnoticeable.
Both RAM and Core must both be stable and able to communicate to each-other without collisions or fluctuation.
I respectfully disagree, Kiwi. Hashing makes minimal use of video memory and extreme use of the GPU. Naturally, this means that a slight flaw in the GPU will be more pronounced in a setup that has the GPU running faster than it ever safely could without the memory downclocked. Ergo, the best stress test for hashing, is hashing.
I'd like to second the idea that it is not a problem with the PSU
Your first 6870 that (supposedly) did 1040 core on stock volts is of a very clean and good fabrication batch.
Your second 6870 is more than likely from an "average" or "below average" batch that cannot get high clocks without the volts.
You can overclock each card separately. The good card you can run at 1000 stock volts and you can leave the other one at stock. All this can be accomplished in Afterburner, just make sure each GPU has a separate clock control.
If your second 6870 is of reference design, you can bump the volts to try to get it as high as the other one.
Also, mining and gaming are NOT good stress testers in the first place. Get a real stress tester like Furmark to verify if your overclock is truly stable. Just because you can play games and not crash does not mean your overclock is stable.
Your first 6870 that (supposedly) did 1040 core on stock volts is of a very clean and good fabrication batch.
Your second 6870 is more than likely from an "average" or "below average" batch that cannot get high clocks without the volts.
You can overclock each card separately. The good card you can run at 1000 stock volts and you can leave the other one at stock. All this can be accomplished in Afterburner, just make sure each GPU has a separate clock control.
If your second 6870 is of reference design, you can bump the volts to try to get it as high as the other one.
Also, mining and gaming are NOT good stress testers in the first place. Get a real stress tester like Furmark to verify if your overclock is truly stable. Just because you can play games and not crash does not mean your overclock is stable.
Like the others said, no two gpu's are exactly the same, there are always imperfections in the circuitry that can prevent higher than normal speeds.
Try out what I mentioned, over-clock each card separately, (with only one card in the case at a time) to see which is causing the issue.
A systematic approach to testing is always the most effective.
Try out what I mentioned, over-clock each card separately, (with only one card in the case at a time) to see which is causing the issue.
A systematic approach to testing is always the most effective.
It's probably not power that is causing the issue, likely its either a buggy card, or the motherboard needs a firmware upgrade.
Test over-clocking with each card individually, see if the freezes occur only with one card, or if they work separately just fine.
To make sure power is not the issue:
You can get a simple multimeter for 10-15 bucks from nearly any hardware store/walmart to test the 12v Rails when at full load, see if they drop amperage significantly, that could also be an issue. Setting the cards on different Rails is also a good idea.
Also check with the manufacturer of the motherboard for a bios update, sometimes that fixes issues. Also in the bios you can sometimes set "constant voltage" or "load spectrum" or similar, that should make over-clocking the GFX more stable.
Adjusting the PCI frequency/HT freq, can help on some AMD mobos.
Test over-clocking with each card individually, see if the freezes occur only with one card, or if they work separately just fine.
To make sure power is not the issue:
You can get a simple multimeter for 10-15 bucks from nearly any hardware store/walmart to test the 12v Rails when at full load, see if they drop amperage significantly, that could also be an issue. Setting the cards on different Rails is also a good idea.
Also check with the manufacturer of the motherboard for a bios update, sometimes that fixes issues. Also in the bios you can sometimes set "constant voltage" or "load spectrum" or similar, that should make over-clocking the GFX more stable.
Adjusting the PCI frequency/HT freq, can help on some AMD mobos.
A card can work perfectly fine and still not be able to OC more than a few mHz above stock...
And it can also happen because the actual chip is one of the lesser ones in a batch.I can suggest you try to overvolt the cards a small amount, if you have enough leeway temperature wise, to see if it helps in stability when overclocking.
Also... Raidmax PSUs are really that bad. Never trust their box numbers. They can handle normal usage just fine, generally, but the constant and steady drain taxes them in dangerous ways. If you're using raidmax PSU's, it really is a possibility that they are the problem!
If they indeed are that bad, then you should definitely recommended you use the 1200w PSU in your gaming rig.
Hashing stress and gaming stress are very different things. Case in point, everybody can downclock their Graphics card memory to a ridiculous degree, and this works wonderfully for mining, but try that while trying to fire up Crysis and watch things grind to a screeching halt. Test your cards, one by one, in a live fire exercise - mining under stress.
Also... Raidmax PSUs are really that bad. Never trust their box numbers. They can handle normal usage just fine, generally, but the constant and steady drain taxes them in dangerous ways. If you're using raidmax PSU's, it really is a possibility that they are the problem!
Also... Raidmax PSUs are really that bad. Never trust their box numbers. They can handle normal usage just fine, generally, but the constant and steady drain taxes them in dangerous ways. If you're using raidmax PSU's, it really is a possibility that they are the problem!
The cards are fine, Im currently Mining, Playing DIRT 3, And typing this message (dirt in windowed to proove my point)
Nothing is wrong
Nothing is wrong
A card can work perfectly fine and still not be able to OC more than a few mHz above stock...
The cards are fine, Im currently Mining, Playing DIRT 3, And typing this message (dirt in windowed to proove my point)
Nothing is wrong
Nothing is wrong
It's probably not power that is causing the issue, likely its either a buggy card, or the motherboard needs a firmware upgrade.
Test over-clocking with each card individually, see if the freezes occur only with one card, or if they work separately just fine.
To make sure power is not the issue:
You can get a simple multimeter for 10-15 bucks from nearly any hardware store/walmart to test the 12v Rails when at full load, see if they drop amperage significantly, that could also be an issue. Setting the cards on different Rails is also a good idea.
Also check with the manufacturer of the motherboard for a bios update, sometimes that fixes issues. Also in the bios you can sometimes set "constant voltage" or "load spectrum" or similar, that should make over-clocking the GFX more stable.
Adjusting the PCI frequency/HT freq, can help on some AMD mobos.
Test over-clocking with each card individually, see if the freezes occur only with one card, or if they work separately just fine.
To make sure power is not the issue:
You can get a simple multimeter for 10-15 bucks from nearly any hardware store/walmart to test the 12v Rails when at full load, see if they drop amperage significantly, that could also be an issue. Setting the cards on different Rails is also a good idea.
Also check with the manufacturer of the motherboard for a bios update, sometimes that fixes issues. Also in the bios you can sometimes set "constant voltage" or "load spectrum" or similar, that should make over-clocking the GFX more stable.
Adjusting the PCI frequency/HT freq, can help on some AMD mobos.
AFAIK the 6990 and 5830 will consume more power than 2 6870 so i think you should stick with the 1000W in the gaming computer and first do as Goldenmaw said just to make sure it's not the second 6870 that can't get past stock...
A better way to do so (assuming you're using MSI afterburner) would actually be to go into settings and uncheck the "Synchronize settings for similar graphics processors" and then overclock each card on it's own little by little while checking stability by having a miner run on the card you're testing...
A better way to do so (assuming you're using MSI afterburner) would actually be to go into settings and uncheck the "Synchronize settings for similar graphics processors" and then overclock each card on it's own little by little while checking stability by having a miner run on the card you're testing...
Jump to: