[BRAG] Highest Stable Card Clocks, Show Proof

[NLA]

I think it's interesting that there is a thread for posting pictures of mining rigs, but no thread for posting (read: bragging) about highest core clocks managed on cards, and the configuration that went into it, etc. What works for one person with one card might help another person with the same card, etc. I'll start.



GFX Card: Sapphire 6970
Core: 1030MHz
Mem: 150MHz
VCore: 1225mV
Fan Speed: 100%
Consistent Mining Temp: 69.5C
Stable for at least: 24 hours

This card is an absolute champ, has always run cooler than the others no matter what slot I put it in. (I have one of these, http://www.drillspot.com/pimages/213/21333_300.jpg, directly blowing into the case.) Looking to up its VCore a bit further so I can increase the core clock a bit more and get the temperature in line with the others. My philosophy: push it as hard as possible until the temps are about the same for all cards, and stay below 80C if possible!

[jjiimm_64]

OK, so you have a high core clock.  How long does it stay stable?,  and how much electricity does it use?

I believe the winning miners will be the ones that can produce the most btc for the least amount of electricity in the long run.


I have 2 (soon to be 3)  rigs that get 2900 Mhash on 1180 watts.   

4x5970,  msi gd70, linuxcoin, clocked at 820/420

pics:  https://picasaweb.google.com/112408294399222065988/October12011

[NLA]

OK, so you have a high core clock.  How long does it stay stable?,  and how much electricity does it use?

Stable for at least: 24 hours

I know that its good for at least 24 hours (I've been playing with clock speeds on the cards in the rig, stopping mining, starting again, etc), and I'm fairly certain it uses <300W, but I don't have any concrete measurements yet.

[ssateneth]

24/7 stable over 500 mhash. plummeting btc prices have forced a more conservative and efficient overclock since i pay for electricity (1010 @ stock volts)

[Starlightbreaker]

uh.
and how much power does an overvolted 5870 use?


better get a kill-a-watt.

[m3sSh3aD]

uh.
and how much power does an overvolted 5870 use?


better get a kill-a-watt.

Shapphire xtreme 5850's run 115-120 amp @ 1000/300 so the 5870 would be higher. No PSU's supply this ampage so you will be hurting the hardware with power fluctuations. This is why we need the new 7 series to lower these figures and have some serious m/hashing

[ssateneth]

uh.
and how much power does an overvolted 5870 use?


better get a kill-a-watt.

Have one. Don't want to turn off my PC as well as disassemble it to get baseline numbers.

TDP of 5870 is 188 watts. Stock frequency is 850 MHz. Stock volts seem to be 1.163. OC'd frequency = 1080. OC'd voltage = 1.238

188 watts*(1080/850)*((1.238/1.163)^2)= 270.67 watts.

This is best guess. There is no way to get solid numbers without pulling apart my computer as well as getting PSU efficiency since wall socket power won't be the same power draw as inside the computer.

[AssemblY]

I think it's interesting that there is a thread for posting pictures of mining rigs, but no thread for posting (read: bragging) about highest core clocks managed on cards, and the configuration that went into it, etc. What works for one person with one card might help another person with the same card, etc. I'll start.



GFX Card: Sapphire 6970
Core: 1030MHz
Mem: 150MHz
VCore: 1225mV
Fan Speed: 100%
Consistent Mining Temp: 69.5C
Stable for at least: 24 hours

This card is an absolute champ, has always run cooler than the others no matter what slot I put it in. (I have one of these, http://www.drillspot.com/pimages/213/21333_300.jpg, directly blowing into the case.) Looking to up its VCore a bit further so I can increase the core clock a bit more and get the temperature in line with the others. My philosophy: push it as hard as possible until the temps are about the same for all cards, and stay below 80C if possible!

Which tool used to make the overvolt?

[d.james]

I can do 820mh on a single 5970.
no proof you just have to believe me. 

[ssateneth]

I can do 820mh on a single 5970.
no proof you just have to believe me. 

cant printscreen and save jpg on tinypic or something?

[sadpandatech]

 Sapphire 5830 Extreme. Usually run the fan a lot lower but I was playing. This is stable 24/7/2 months now. Hash goes from what is seen in screenie at 97% GPU load up to 380.3 when it peaks at 99%. It will ramp up to 1040 stable at this voltage but the amperage on the 1.156v goes up to 95+ and hash only jumps another few points.  For those that can't add the amps, volts, etc. I noticed one guy above noted the 90+ amp and said something about PSU exploding. hehe, yeppers it sure would. But, you are not drawing those amps from PSU. The vregs(VRMs) on the card drop the 12v down to the 1.156, in my case, @ 84a which = 97.104 watt(not counting slot draw of course). This is only ~7.92 AMP draw on the PSU at 12.25V. Add to that a very high end PC Power and Cooling MK II 950 Silver Cert(one that we need a new cert above the 80 + for since its 88% at the temps I'm keeping it. *shrug*) and you're gold. literally. This is card 3 of 4 in this rig. Just happened to be the one up at the time when i decided to snap a shot. All cards pull the same numbers within 1Mhash of each other. The other key is custom cooling layouts for the cards. The thermal compund has been replaced on all with ceramique II and additional air is forced from back of cards through custom made housings for each card. I played around a lot with the dimensions and fan layouts, intake locations, etc of the enclosures until I got what seemed like the best temps with the least amount of added amps for air flow..

  And to anyone posting crap about elec cost, etc. Bugger off or post some pics. that is what the topic is about after all, isn't it?   

[TurboK]

I wouldn't brag about those, the fans are running at 3000-4000 rpm. That shit sounds like a vacuum cleaner.

[m3sSh3aD]

I wouldn't brag about those, the fans are running at 3000-4000 rpm. That shit sounds like a vacuum cleaner.

Wrong, I got a shapphire 5870 vapour X here that runs 3000+ rpm and its noisy. I have loads of shapphire 5850 extremes and they are all set to 100% @ 4400RPM and they are WAY QUIETER than the single 5870 doind 4000RPM. 2 rigs of 4 all on 100% and there quiet. YOu forgot that theres many types/makes/qualities of parts, and fans are a gulf of different ideas.

On topic, the 5870 does 1010/300(no volt mod) and all 5850's reach 1000/300@ 1.93V except 1 which is VERY close. THe best does 1015/300 @ stock! Its a crazy card but it does just that. Highering the volyage on it doesnt make much difference. Seems 1015 is the ceiling and upping the V doesnt matter. Still, 1015/300 stock i have them all on 965/300 @1.87V mainly, so they last! lol

1015 out of a 5830 is very impressive

[sadpandatech]

I wouldn't brag about those, the fans are running at 3000-4000 rpm. That shit sounds like a vacuum cleaner.

  I wouldn't brag about your reading skills either. But haters gonna hate. 

 

     Usually run the fan a lot lower but I was playing

     cards through custom made housings for each card.


  And to anyone posting crap about elec cost, etc. Bugger off or post some pics. that is what the topic is about after all, isn't it?   

  I usually run them at 45% at under 70c, spanx ya. And as the guy above pointed out they are very quiet. And, unlike what some others may have for noise req's I do not as I have this in one of my offices at one of my shops........

[sadpandatech]

I wouldn't brag about those, the fans are running at 3000-4000 rpm. That shit sounds like a vacuum cleaner.

Wrong, I got a shapphire 5870 vapour X here that runs 3000+ rpm and its noisy. I have loads of shapphire 5850 extremes and they are all set to 100% @ 4400RPM and they are WAY QUIETER than the single 5870 doind 4000RPM. 2 rigs of 4 all on 100% and there quiet. YOu forgot that theres many types/makes/qualities of parts, and fans are a gulf of different ideas.

On topic, the 5870 does 1010/300(no volt mod) and all 5850's reach 1000/300@ 1.93V except 1 which is VERY close. THe best does 1015/300 @ stock! Its a crazy card but it does just that. Highering the volyage on it doesnt make much difference. Seems 1015 is the ceiling and upping the V doesnt matter. Still, 1015/300 stock i have them all on 965/300 @1.87V mainly, so they last! lol

1015 out of a 5830 is very impressive

Messhead - am I reading you right here??? 5850 extremes @ 1.93V or 1.87V ??

You know I've got the same cards as you, we had that thread with yours and my pictures of the two different types of 5850 Extreme (i.e. top placement / side placement of power sockets etc.). We bought them at around the same time.

I get around 375 MH/sec at a conservative 900 MHz core clock. But I run stock voltage, and AMDOverdriveCtrl tells me the stock voltage is 1.088V. So in my 'overclock' settings file, I set the core voltage to 1.088V.

This suggests that you are over volting your cards by around 0.8V - for the mathematically disadvantaged (not you) that is getting on for *nearly doubling* the voltage.

I'm not calling you a liar. I'm asking WTF is going on. Doubling the voltage usually causes destruction - so is your voltage reading wrong, or mine, or does the Extreme *really* let you up the voltage by 0.8V ? Seriously mate - I may not be getting over 400 MH/sec from my 5850 Extremes (as you claim they 'all' can) but at 0.8V less per card, I must be using a shitload less electricity, surely?

And whilst I'd like to test out your theory of the >400 MH/sec by upping the voltage, I'm a bit nervous about whacking 1.87V into a card that has been hashing away at 1.088V since I bought it.....

WTF is going on mate??? Sounds like one of our tools is fecked - I run Linux, IIRC you're a Windowz chap so the tools are different, but the APIs should be the same inside FFS - this is weird...

Anyone else with a 3-month-old 5850 Extreme bought from OCUK here? Please can you verify the *standard* voltage the cards run at?

I don't think I've uncovered some awesome new undervolting trick to save power - my shelf rig is eating exactly 2000W at the moment, but that's 12 cards and 3833 MH/sec, so only 1.92 MH/W. Not bad for a bag-o-bones, but not brilliant. I haven't done any further undervolting or overclocking yet though - the 5830s I *know* will hit 300 MH/s, and the 5850s *definitely* do 375, so I can add 25 to every one of these cards, but at what power cost? Best find out

  Aye, I wanna know too. He refers to the 1.87 as stock on certain cards. Sounds very, very high.  The 5830 extremes I am running now came 1.178 stock, so maybe he left out the 1, I lower mine with a bios mode to lock them down to 1.165, which tricks trixx into dropping them another few points when I lower it to trixx lowest of 1.163.(Note the 1.156 in screenie) Should be 1.187 for 5850 and 1.193 for 5870.  All will vary a bit depending on brand and whether or not the Manu. OC'd or customzied the bios, volts, etc, etc.


  For your cards, Catfish, I always recommend NOT upping the volts as you are testing OC. Most of them ATI and Nvidia alike seem to be able to handle much higher clocks at their stock volts than what they would if we were using them for gaming. I.e. maxing out the vrams, which would most certainly need more juice over the GPU reqs. I up the clocks as far as they will go stable at stock volts, carefully monitoring temps and screen artifacts. I have found in my own toying that once you reach screen artifacts all the extra volts in the world will not make the GPU go any faster. Though, with lowering the volts at that point I have been able to sneak in a few more Hz, but not with stable mining and the amps jump up like nuts. All of that will vary by card brand, cooling layout, PSU quality, OS, ICH efficiency, etc, etc, etc..  And, yea the 58xx serious is definetly a rarity, atlerast for the model lines most of us are using in their clocking abilities. I hear rumor the last run of 58xx chips made recently are not doing nearly as well.

  In fact you can likely lower your voltage a good bit since you are not OCing or are OCing very little. It should improve your MH/W ratio a bit.

  One last thing, the cards you 2 were comparing with the power inputs on different locations.  Generally, top-in is reference or atleast reference design and the back-in ones are custom board layouts from whoever actually manufactured them. Reference usually beats the pants off of the others in terms of durability, OCability, etc.

   Cheers

[sadpandatech]

  Just wanted to drop a line for others trying to OC or trying to cut juice, etc.

   COOLING COOLING COOLING


   Without breaking out the thermo dynamics book, since I havn't taken the course yet and frankly that shit is boring as hell, lets see if we can explain this in an easy manner.  Think of the Vregs on your vid vards, thats the part that drops the 12v down to 1.xxxv for the GPU and ram. Picture them almost as resevoirs of energy. They hold it pretty well, convert it and send it on in a 'smaller' pipe to the other components. Now, as they heat up some of that energy escapes in the form of heat, up into your musty basement or other domicile of choice, before it is able to be sent on to the other components. As these vregs heat up more, the energy finds it much easier to escape along what is now a lesser path of resistance due to a bunch of fancy terminologies that I will not proport to be able to convey here in any meaningful manner.


  In summary; COOL Vregs = higher % of energy making it to where you want it, GPU in this case.
                    HOT Vregs = higher % of energy warming your balls.....  Which will happen anyhows, but it much better that the GPU does it rather than the components on the path before it......

[sadpandatech]

24/7 stable over 500 mhash. plummeting btc prices have forced a more conservative and efficient overclock since i pay for electricity (1010 @ stock volts)

Very sexy, btw!  Have you checked for stability at lower volts?  1.238 is not bad at all, but am curious if that beefy 5870 will do it any lower at that OC.

[m3sSh3aD]

Whoops, 1.187 and 1.193.... :/

Sorry about that I have 2 seperate rigs with 4 of 1 kind in 1 and 4 in another. 2 got replaced with the new ones meaning i had 4 of each kind there all 965/300 @ 1.193V, not 1.92V :/ i know a friend that runs one at 1.247 24/7 but i dubious about that and he doesnt get much more out of it. 1020/300. VRM's will be straming there and there life will be damaged in time, Im sure of it. 1.2 is the highest ill go but 1.25 isnt uncommon as you would think in the gaming circles

here.... 2 inner windows are the 2x rigs, and the other is the comp im on (vapour X 5870) on intensity 5 in CGminer. I can use comp and believe it or not play games, even rage without any problems. Obviously effects your hashing rate but it runs 24/7 and is usable

http://imageshack.us/photo/my-images/26/btcu.jpg/

[sadpandatech]

  Heya, mesh. Very sexy. Thanks for the clearification on the volts. What brand 5850's are those?  I assume that GPU 2 on the left group is being mounted(terminology?) as the display over team viewer? Or is just hotter naturally or not getting good air flow? Wondering why it is so much hotter than the others and slightly lower hash.? =)  And, have you tried dropping the volts any lower for either of your 5850 groups? If so did it become unstable for you?

  That bottom 5870 stat is very sexy! Love seeing the low volts on it. ;p

[m3sSh3aD]

 Heya, mesh. Very sexy. Thanks for the clearification on the volts. What brand 5850's are those?  I assume that GPU 2 on the left group is being mounted(terminology?) as the display over team viewer? Or is just hotter naturally or not getting good air flow? Wondering why it is so much hotter than the others and slightly lower hash.? =)  And, have you tried dropping the volts any lower for either of your 5850 groups? If so did it become unstable for you?

  That bottom 5870 stat is very sexy! Love seeing the low volts on it. ;p

Missing a 1 here and there and you have blown hardware I use a program called teamviewer. It works on android to so i can control everything from my phone wherever i am in the world or any computer of cause

1 rig (a wooden table) doesn't have as good air flow. THe converted METAL table can release heat UP also, whereas my first rig cant. WHy i thought catfish's new setup is great, i just think hes blowing the hot air down? i know mine suck in from the front and side and blow out the back (the front been where the ports are. Seem to anyways

Bit rig 1

http://imageshack.us/photo/my-images/710/imag0068d.jpg/

Bitcoin rig 2

http://imageshack.us/photo/my-images/14/imag0069f.jpg/

pics big so just linked I got the fans blowing across the cards going forward in them pics, the metal table also as a few blowing up behind it and generally more fans I use a 12" fan/filter etc to expel all the air out the room in less than 10 seconds which draws air from outside through some windows. I recon this setup will handle 30-40+ cards before heat becomes a major issue. THen theres always HVAC

ARS down so mining to bitcoins.lc ARS is amazing compared to all the others. Guy uses amazon for servers, think he needs a new provider