Worried about Over Powering - 6950 unlock+20%+ULPS+OC+CPU VOLTS

likewise

Distinguished
Jun 3, 2011
9
0
18,510
Hey, So I know there's info everywhere for safe Volts & Watts on a 6950 (2-2.5v) but what im wondering is this volt limit is still the same, safe ceiling if:

The 6950 2gb + Unlocked Shaders + 20% Pwr in CCC + Disabling ULPS + Upping the volts on the Gpu up to 1.2v (not at that just limit I feel comfortable as a ceiling)... + HyperTrans O.C. to 2424 + Fsb 2424 + 1.4v added to CPU (955xII BE OC 3.93ghz)...

Will watts on the GPU be effected by the Volts on the CPU or the HT/FSB Speed increase? I have a 850HTX Corsair PSU. Im not talking about drawing too much from the PSU, I mean the GPU getting overpowered with these settings... and the volts on the CPU affecting anything cumulatively...

I Have a 790fxta-ud5 MB and have oc the Fsb + HT to 2424mhz to open the pipeline for the ddr3 1600 ram (3xFsb), (Ive read the HT doesn't really need to be that high and can have adverse effects on GPU but figured they matched at 2000mhz stock so to do the same on the oc. Also it's what was done in tutorial I watched on a similar board for my first OC. Im up for suggestions I just want to make sure when Im reading "safe volts for 6950 is 2-2.5" and wattage limits etc, are not going to be surpassed with this setup, especially with the CPU volts at 1.4v and HT/FSB being OC... What is safe to overclock the GPU to in this scenario including the Vram... Just wanna b sure... seems like a case fan died after my OC could just be coincidence... Sorry if its many questions in one thread but Im OC my GPU today and wanna play safe but would like 900-940 core on the 6950 and squeeze in the 1300-1400 on the VRAM. Thanks in advance.

- Oh and btw its a the shader unlock using my own bios not a 6970 bios flash
 

michaeljhuman

Distinguished
Jul 28, 2012
100
0
18,690
Don't know for a fact, but I presume video card needs an on card voltage regulator. It's plugged into the PCIe bus, so has that as a power source, along with power connections to the PSU.

As you can adjust voltage for many cards in software just for the GPU, that implies an onboard voltage regulator.

The CPU is supplied by a voltage regulator on the moo.
 

likewise

Distinguished
Jun 3, 2011
9
0
18,510
thanks for the reply... so specific to this card when you unlock the shaders and boost the power to 20% in CCC and turn off the ULPS (power regulator) ... is it still safe to go up to 1.2v-1.25v on the card? Im worried about upping the vram past 1350 too... I just dont know if everyone that comments its safe voltage for this card is taken into account the powerplay turned off, the unlock shaders, and the extra 20% power control, and is overclocking like a regular 6950... My goal is for a safe worthy O.C. for 24/7 use.
 
As mentioned, GPU volts are completely different from CPU volts, although they both are going to hit the power supply pretty hard. If the PSU isn't able to supply a stable voltage, these increases can cause more damage than they otherwise might. It's like a car that starts swerving - it's ok up to a certain point, and then the tail end starts sliding out and the car fishtails more and more until it crashes.

Any reason you disabled ULPS? I've read about people doing it but it also seems like there's no purpose to it these days, as that was a limitation with drivers or something. Just curious if you needed to do it, or just did it for fun.

As far as a "safe" OC goes, you're well past that. Any increase in voltage is not safe, but that doesn't necessarily mean you're going to fry it in a matter of weeks. It might well last a few years over volted. Have you used GPUZ to check the GPU, VRAM, and VRM temps? It should have sensor readings on all of them. On my GPUs when I overvolt, my VRM temps shoot up a lot, I've crashed at 120C. Besides that, increasing the card's power cap, and disabling the ULPS is a potentially dangerous combination. As far as the shader unlocking goes, if it's stable it's fine. Yeah it increases the power used, but the card should have all the same hardware as a 6970 so I don't see that as an issue.

As far as frequency goes, bear in mind that frequency increases power used, and volts increase it at a squared rate (not exactly this equation but basically V^2*Freq) and the more watts, the more possible damage.

As for 2 to 2.5V on the 6950... that would easily destroy the card. 1.2V is ok, 1.3V would be getting pretty high, anything more is quite dangerous. I'd also be surprised if it didn't overheat at that point anyway.

VRAM speed can be increased as much as you want, and it is unaffected by your core voltage. Most GPUS can't adjust the VRAM voltage. With GDDR5, it's important to bench the results of VRAM OCing, easily done with Furmark at 4xAA. As you increase VRAM frequency it will speed up, but at a certain point the score will start to go back down and shortly after with more speed increases it will crash. Basically there's a fault protection where GDDR5 resends data until it is received - that slows it down. So you need to keep the speed under the point where it has these issues.

Finally, I highly recommend you see what you can do on stock volts and bench it. Then see what gains you're actually making with the over volting. In my experience, my 5850s go from 700/1000 stock to 870/1200 without more voltage, and 980/1200 with ~1.2V. The gains in benchmarks is pretty small over 870 core, and in gaming it's not noticeable (maybe 1-2fps?), so why risk blowing the card? With your 6950 shader unlock, a little OC at stock volts will probably be a very large gain over the stock card already, with very diminshed returns above that.
 

likewise

Distinguished
Jun 3, 2011
9
0
18,510


Thanks, I dont know if it was misread or if I miss-typed lol but im am 1.165v was asking if 1.2-1.25 was ok... ya 2-2.5v the thing would prolly explode lol... haha I have been getting good results during my benches just dont want to push it too far... I appreciate you helping me out... currently im at 1.165v 900mhz 1350vram... and set a pretty agressive fan at 54%... I dont really see much improvement past that... The psu 850hx corsair and seems to be doing fine... temps dont ever really get passed the mid 60s... I took the ULPS off just for benching purposes, I dont want to upgrade my ati drivers yet because they have been rock stable for me and I read all kinds of problems with the later drivers... im on 12.1 CCC... at that setup im getting about 1870-1880 scores extreme on 3dmark11... and about 47-48fps avg on crisis1 bench fully maxed including resolution... The OC to the 955x2BE/fsb/HT did really helped out my physics scores a TON... currently 1.4v on the 955BE @ 3.93ghz, fsb 2424, HT 2424, 1600ddr3ram 8-8-8-24/1.63v is super stable and not too bad at 51-53c full load on Prime95x2hrs blend and small ftt. With the ULPS off i wont open furmark for stability yet... been sticking to Heaven, 3dmark, Crisis... Now that I found a pretty stable at 900/1.165v... should i bring the volts down by about .02v until unstable and then bump it back up .02-.04 or just leave it if stable?
 
Sounds good. Can't guarantee anything, of course, but it should be fine for a while.

You should definitely update your drivers if you're playing any games released this year. It's pretty rare for anything to screw up if you do it right. Instead of downloading and installing over top of old drivers, you should follow this procedure:
http://wolframpc.blogspot.ca/2012/01/how-to-install-crossfire-sli-or-new-gpu.html
It's pretty quick and guarantees a good install.
 

likewise

Distinguished
Jun 3, 2011
9
0
18,510


Wolfram I think you are 100% accurate about mixing the disabled ULPS and the additional power to the card... I bumped up from 1.185v to 1.190v at the same clock of 920mhz/1350 ram and boom Hard Crash BSOD in 3dMark11... the card was pretty stable at the 1.185 so going up in .005v shouldnt of crashed me like that so I know its gotta be the watts from mixing the disabled ULPS and adding the 20%... I know with the shader unlock though it would be dangerous to not add some extra power and 20% in CCC was recommended... but at the same time I don't wanna get throttled by powerplay before its necessary either (if that really occurs in some games)... So which route should I go? Leave ULPS disabled and bring the extra power down to +0% in CCC (i would prefer if possible) or should I slap the Powerplay Back on and leave the extra 20%??? Is there anyway to check how much Wattage is hitting the card with software? I haven't even touched 1.200v starting this process and was thinking I should be well in the clear over the past 2 days benching and finding a sweet spot... If the temps have been fine the whole time am I ok? Maybe the card just wasn't stable yet at the original 1.185? Im hoping thats the case... Back at stock settings now looking forward to your reply... Thanks again.



 
As far as I know, Power Play is basically increasing the ULPS limit. Also, I think it's partially temperature related, although I might be wrong about that.

Personally I'd try it at +0% with ULPS enabled, bench it, then try +20% and bench it. I doubt there will be much difference.

As for the hard crash, I'd guess that's a heat issue. Have you run GPUZ on the Sensors tab? You might be surprised that VRM and VRAM get way hotter than the GPU. Most cooling solutions only focus on the GPU core, while the VRM are where all that power is getting man handled.
 

likewise

Distinguished
Jun 3, 2011
9
0
18,510
ya gpuz was showing cores 2-3 70c... gpu 65c MAX... Card is staying cool... I know the ram can be a lot hotter then you think when you only monitor GPU so thats the first thing I checked (had gpu-z running in the background...) only thing is It was specific on TechPowerUp to add the 20% after doing the shadder unlock, and then I was noticing it throttling which led me to turn powerplay off... ehhh :/ prolly will just slap the ULPS back on and play it safe...
 

likewise

Distinguished
Jun 3, 2011
9
0
18,510



Oh wait after just re-reading this I just noticed something... I though Powerplay and ULPS were the same exact thing :eek: I guess I have some more reading to do... Just so you know only reason Y im even messing with the thing is because afterburner unoffical overclocking is needed to get passed 840hz on the card... so I added the EULA and then choose "2" for overclocking mode "traditionally turn off power play" and then I was instructed to go into the registry and turn off every instance of ULPS... but was also instructed to choose mode "1" in afterburner which left power play on... I turned them both off thinking they were the same thing to prevent any software conflicts with Afterburner having it on and ULPS being disabled... I guess I was mistaken and they are actually 2 different things... one effects the other...