Need some help OCing my 970

so i finally got my hands on a liquid cooler (kraken g10 bracket w/ corsair h90 push pull) for my evga superclocked 970. I didnt bother trying to OC it before hand coz i knew i would redo it after i got the cooler. The card is rated for 1317 boost clock, but thanks to gpu boost 2.0 it ran 1354 out of the box. the very first problem is that most reviews ive seen claimed these cards pushing over 1400 out of the box. the second is that from my understanding gpu boost 2.0 should take advantage of the cooler temps (50% fan speed has it under 50 celcius under load), but it hasnt changed; its still 1354. i tried for a 1400mhz clock speed, but even with the temp target maxed, power limits maxed, and an extra 37mv (the most precision x will allow on this card) i cant get it to be stable. i dont get any artifacting it just dies. my screen goes black, and i need to manually restart my pc. i also have the fans and pump running off a fan controller so theres no power draw from the card for cooling. i dont want to think that i got a dud overclocker since that already happened with my cpu (my 4770k refuses to go over 4.3). also i should point out that unigine valley and heaven both report my card as running 1506mhz. i know thats a problem with how unigine deals with gpu boost 2.0 but it makes me think theres a lot of untapped power in my card. can anyone help with this? maybe im missing something since im used to radeon cards.

Sounds like a power issue to me.. What PSU do you have in the rig?

If its not power then its really one or two things..
One: You got a shitty binned GPU and your just SOL.. sorry bud..
Or two: You have done something wrong in your overclock.. I do know that on the 970 Reducing your memory speed by a bit does help stability with a high core clock.. I also know GPU boost is finiky.. (way to go NVIDIA).. If everything Fails.. RMA it.. it really could be a bad card... If unigine reports that mhz then it really is at that mghz since its communicating directly with the driver..

Check power and how well your card is seated in its slot.. you never know.. things do come loose

Oh forgot to add. TRY THE BETA NVIDIA DRIVERS>> They are way better at OCing.. at least the current betas are from what I here.... :D and DEFINITLEY clean install the drivers...

Also before I forget... disable power savings on the PCIE settings in windows Power settings.. its in the advanced area... those can sometimes interfere with OCing

Thanks a lot for the suggestions. im gonna go see what i can do. also im using an evga supernova 650watt 80+gold psu. my system draws around 370 watts at the moment so my psu should be more than capable. il post back with the results
p.s unigine reports 1506 but everything else ive tried (msi afterburner, and evga precision x) report 1354. from what ive read this is because unigine reads gpu boost wrong and will just report the highest gpu boost step with whatever offset you try.

still having the same problem. whats really wierding me out though is that any time i have overclocked a graphics card in the past ive seen artifacting before crashing. every. single. one. with this card though no artifacting just crashing, and to make things worse i can push it up to 1600mhz and get through any of my benchmarks (unigine heaven/valley, and firestrike) just fine. as soon as i try playing a real game though regardless of how small the overclock is (the lowest i tried was a 10mhz increase) it will crash within 5 minutes. i know that benchmarks tend to be easier on your overclock than games but if my card cant handle 10mhz in a game i would imagine it wouldnt be able to handle 15 times that on a benchmark.

Its not the cardmthen.. I owned a low wattage supernova like yours.. Had the same symptoms.. I would suggest get another power supply or test with a friends you know is good.. If you find it to be faulty.. Get a seasonic or corsair.. Evga sucks no offence and your really do get what you pay for

i didnt think it would be the psu since my system uses less than 400 watts under load. though il take your advice and try my old cooler master silent pro m2 850 watt and post the results

Thats what i thought until i rmad it and got my money back... They admitted it was faulty and apologized if it caused any other damages in a support email.. Evga is pretty much terrible in terms of quality control.. If it isnt the psu.. Then rma the card it shouldnt crash under normal use +/-20 mhz

Alright so i bit the imaginary bullet and returned my card and bought a gigabyte g1 gaming 980. i can guarantee its stable at 1525mhz, and im currently testing it at 1550mhz

Unigine valley is useless as a benchmark. It reports the wrong numbers. It insist my two GTX670PE's are running at 1305 mhz and I know for a fact that they do not.

Boost 2.0 does not really take into account temperatures below a set throttle point.

Also even though the GPU on the 970 uses less power than those of the 700 series, the VRM's still run hot, and the VRM protection will limit clock speeds if they are hitting about 100C. Do not use something like the G10 unless you can properly cool the VRM's (they still need to have heatsinks (the RAM also needs active cooling)

With a stock bios, and a good non reference cooler, there is not much of a need for liquid cooling. Furthermore for most non reference cards, the build in cooling if often more than enough to reach the max overclock with the stock bios and voltage limits. (they simply do not allow you to use a voltage that will stress the cooling of the card.

Many people who get the G10, often fail to get better overclocks, and in some cases, even lower overclocks due to the VRM protection kicking in. If possible, return it and the H90 for refund, especially if your card does not use a separate dedicated heatsink for the VRM's.

Basically an easy way to tell if your card can work with a G10, is to examine the stock cooler. Does the stock cooler only touch the GPU, or does it touch the GPU in addition to using thermal pads to cool the memory, and other components on the card.

PS if you want to see your GPU boost info, is t use GPUz to grab the bios from the card, then load it into maxwell bios tweaker (these sample bios files come with the maxwell bios tweaker application

You will see that the only time the boost clock is impacted by the temperature, is when the viceocard is hot enough to throttle. (also the boos state is impacted by the base clock offset, so you will not hit the max boost clock in that list unless you are have the base clock overclocked.

Ive had the card at about 1540mhz with the stock cooler for several hours now. its gone through the full 3dmark benchmark suite, soaked in unigine heaven for about two hours, and three 1000 ticket rounds of TDM in bf3. Even though ive allowed the card a power limit target of 122% its only taking at most 75% during firestrike runs. it is however hitting the max temp ive allowed for it at 75 degrees. im fairly certain the only thing holding this card back at this point is the stock cooler. the kraken g10 bracket has a fan for VRM cooling, and after testing it on the 970 it did keep the VRMs cooler than the stock cooler, and the chip was 15-20 degrees cooler.

Your overclocking ability will not really take a hit until you start getting in the 90C range.

For your 970, did the VRM's have their own separate heatsink, or were they cooled via thermal pads connected to the main stock heatsink?

there were thermal pads directly attached to the heatsink. this area was directly under the fan on the kraken g10. i dont want my card to hit 90C at all so i have the temp limit set to 75. if the kraken effects my 980 the same way it did the 970 it should take the temperature down 20C. more room for overclocking. theres also custom bios roms available for this card that will expand the power limits.

The issue is many of the bios will allow a higher power limit, but not allow a higher voltage.
For nvidia, after the fermi GPU's, they took many steps into locking down the voltages, thus it is harder to increase the voltage limits. With the stock voltage limits, at max overclock, you will likely only get around a 5C increase in temperature over stock settings.

For chips such as VRM's and linear regulators, if they were attached to a heatsink via a thermal pad, then it means that they lack the surface area on their own to effectively be cooled by air. e.g., you aim a leaf blower a 65 watt CPU that is missing the heatsink, and even with 250MPH wind, it will still overheat. If you manage to unlock the voltage on the card, and start pushing the clockspeeds and voltages higher, then the card will use considerably more power, and the VRM's will heat up. So far, I don't think anyone is running into power limit issues (unless they have one of the reference based cards with a 110% limit that they need to increase in the bios.

Overall if you have a thermal probe or IR thermometer, then carefully measure the VRM temperatures with the card under load.

an ir thermometer is how i checked the VRMs on my 970. i havnt checked the 980 yet because its still on air. i will be checking them when i put the water cooler on. i dont expect to have to flash the bios, but i am aiming for 1600mhz. so if i cant hit that without flashing the bios i will. right now on air it seems maxed out at 1540mhz (+173mhz) and 7298mhz (+288mhz) on the memory.

dont feel too bad man,
also, dont just crank up that power bar,
sometimes the more power you throw at the core the MORE unstable it can get

start off small
set everything to stock
and ONLY bump up the core 10 mhz at a time , boot somthing up and try it,
then bump it up 10 more
rinse and repeat a while till it crashes
when it crashes back it off like 15 or 20 mhz

save that profile,
then add a small volatage bump
just to see if that helps at all
ive read that people cant get to 1600 because of power delivery or something, my card is an evga on water, but i can get to 1590mhz with no changes in voltage or current. so just keep trying.

but i have to ask
what the hell are you playing that you need that much power?
i can play skyrim modded at 4k with my 970 (500 mhz Under stock speed)
i can play older games like dishonored at 144mhz core clock

well i have a 1080p 144hz monitor so i like to try to push games like bf3/4 as fast as it will go with max settings. i use a lot of landscape overhauls in my skyrim mods list. plus a full enb. i havnt tried the 980 on it yet, but the 970 would drop to 45ish fps while in the open world. caves and building interiors would be over 100fps however with all the grass, trees and enb effects the frame rate would tank. i also like dayz, but that games frame rates are all over the place regardless of which card i use. i honestly dont need this much power. i didnt even NEED the 970. hell my old 280x would still be way better than a console. i like having this kind of GPU power though.

geeze, my rooomate has those asus 144hz screens as well,
i play synched at 60 and ive never felt the need for anything more

honestly i dont care if im at 30fps, SOLONG as my fps does not fluctuate UP or DOWN
if its a Constant speed then i really cant tell, cos for me its the stutter that upsets me
(granted, yes 60 is totally a more enjoyable experience just go to youtube and find a 1080 @ 60fps and tell me you cant see the diff, lol)

for the bethesda games try this,
Optimizer textures at skyrim nexus
Takes a while to run
And pegs your cpu all along the way
But with impressive result
Also works on fallout
It finds corrects and compresses files into a better format
so the gpu memory isnt abused as bad

This is one of those tools that you run then go have a break
Takes a while
And with the cpu @ 100% you wouldnt be doing much else for a while anyways
if you just target the textures folder, it usually squeezes files down about half size, so they suck up less ram in the gpu

- decreases some stuttering
- fixes crash/CTD caused by bursting of VRAM
- free up hard disk space / reduces the size of the MOD
if you ever go past that 3,5gb ram limit on the 970 youll get a good hit to performance too

but i dont have a 970 anymore. i have a g1 gaming 980