overclocking the 980 Ti Strix

I have an i7 [email protected] GHz, 16 GB RAM DDR4@2800 MHz and the screen is 1440p@144Hz.

The other day I decided to attempt to overclock my 980 Ti strix with msi afterburner.
With no OC, I got 97 fps score 4088 on the valley benchmark. (1080p ultra). I use valley because heaven crashes a lot.

With an overclock of +400 mhz memory clock and +100 mhz core clock and no voltage increase, the fps increased to 103 and the score was 4332. The fans were at 100% and the temps didn't go above 75C (air cooled).

However, when  I opened Witcher 3, I noticed "artifacting" and graphical glitches and the temps started going into the 80s.

Does anyone have good experience with OCing this card while air cooled? How far can it really go? Can I achieve at least a 10 fps increase in Witcher 3?

I heard somewhere that just running the fans at 100% constantly with no OC, and aiming to keep the temps below 70C will help fps a lot.

If you say his name three times he might show up...

Who? The self-described guru of all things OC ...

All together now: "Meller, Meller, Meller!"

the "Strix" 980ti is already factory overclocked.  the reference card core clocks are: (base) 1000MHz/(boost) 1075MHz.  the strix are: 1190MHz/1216MHz in the default "Gaming Mode" respectively.  

when you say "+400MHz on the memory and +100MHz on the core", do you mean you added 100MHz to the boost clock in the default mode?  did you increase the power target? 

i use Heaven, so i can't compare programs, but i get 101FPS/2544 score on my benchmark.  my GPU-Z sensors register 1430MHz on the core (boost)@1.18V VDDC; memory clock to 7600MHz.  my settings (using Asus GPU TweakII software) are: 1291MHz on the boost clock; 1.163V; 110% power target; my cards are water-cooled using DS thermal control board running on "auto mode".

the Heaven benchmark software reported gpu clock at 1531MHz and memory clock at 3800MHz (x2 =7600) during the test.  

at any rate, why the artifacts in Witcher3?  were you using any of the "game works effects" settings in game?  could be drivers or game optimization issues if your oc is stable.  any artifacts with other games?  Witcher 3 is a pretty demanding title.


Edited by db188 - 04 May 2016 at 3:26pm

if Heaven crashes as soon as you start it, that means your oc is really unstable.  if it crashes or freezes during the running of the program further down the line, that means you have some minor tweaking (either add more voltage or pull back on the frequency) to do.  

you should be able to get a good 10% extra (in game) performance from this card with a proper oc.  the ASIC quality of the card i was testing was MEH (75.8) and i can get a stable oc with the numbers i posted earlier.  

here's my suggestion:  set everything back to the card's default settings.  run your tests and find out where your base clock/boost clock/memory clock is at.  i'm still not clear on whether you added +100MHz to the (already) factory oc'd Strix card or if you started with the clocks at stock reference settings.  if you did, that can explain why you need to either back off 20-50MHz on the frequency or add some voltage (10mV increments at a time) until stable.  once you have the settings at default, run Heaven and see if it freezes or crashes.  if not, proceed to add frequency to the boost clock (it might say core clock on afterburner, but it's really the offset for nvidia cards) until you do get a freeze or crash.  then, back off 20-50 MHz on the frequency and test it.  if you pass, you can then decide whether to stick with that or add voltage+frequency for a higher oc.  true frequency and voltage in game will be higher that what you see in afterburner, because the bios makes dynamic changes as needed while under load.  you need to monitor with GPU-Z.  watch the VDDC and make sure it doesn't rise over 1.23V.  if it does, i would suggest backing off the voltage some.  

you can oc the memory after you get the gpu clock stable.  i settled on 7600MHz for my oc, but you should be able to push 8k.  tinker with it some to see where your card is happy.

your temps (on air) should be in the low to mid 70's with fans at 100% under load.  overall case thermals/airflow can cause this number to vary.  as long as you aren't seeing temps hit over 85C for extended time there's no need to worry.  leave your fans on auto and set the thermal threshold to 90C (you won't hit that really) and power target to 110%.  the card will adjust itself to those values. 




Edited by db188 - 04 May 2016 at 4:10pm

One more thing, do you have a liquid gfx cooler and what kind? I think i can handle the technicals of installing one on my 980 Ti. I would much rather see temps of 40 rather than 80, and fan RPM of 1000 rather than 4000.

Success! my first stable overclock is +90 MHz on the core clock and +5mv. heaven runs smoothly with better fps and witcher 3 runs well with no glitches.

gpu-z readings were:
73C
103% TDP
1.1990V
1518MHz on the core clock and
1811MHz on the memory clock even though I didn't touch the memory in msi afterburner.

with the EK unit you have to buy the pre-filled gpu block w/quick disconnect fittings separately (so it fits your specific gpu) and pair with the Predator cooling system.

the only problem might be the fact that you can't just use any 980ti water block if you have a Strix version.  i know that EK has water blocks for it (i have it on my pair), but i don't know if these are the same as the "pre-filled" units.  my advice is to get into contact with their sales team.    

not unless it's an expandable system (which i doubt).  there are only a couple of these "hybrid" AIO expandable systems currently available (and those from the vendors i already linked to you).  however, you can replace the cooler you have with, say, a 360mm rad Predator QDC system and have them add in the pre-filled gpu water block when you order it from EK.  it's something to think about, because it's not cheap (although cheaper than a complete custom loop).  

you can't mix and match from an AIO unit like that.  that's one of the limitations of those types of coolers.  that's why we are starting to see these "hybrid" units from vendors for people who don't want to invest in an all out "custom" water loop.  the hybrids fall in the middle between basic, non-expansion, AIO water coolers that are comparable in price to the very best high end air coolers, and full custom jobs that are prohibitively expensive. 

bottom line:  you'll have to completely swap out your entire cooling loop for the "expandable" unit, which starts out with either 240mm or 360mm rad+cpu block+fittings and tubing.  you can then have EK add their pre-filled, gpu block (if they have one for your particular card).  i looked on their site, and even though they advertise their new pre-filled gpu blocks, i didn't see where you can click and order one.  i advise calling them to check availability.  have your exact gpu model, mobo and cpu specs as well ready for them.  

and yes, a single 360mm rad can adequately cool both the cpu and single graphics card.  the pre-filled graphics water block comes with its own quick disconnect fittings and tubing to be added into the same loop as the cpu. 




Edited by db188 - 06 May 2016 at 12:27am

Originally posted by Snaike If you say his name three times he might show up... Who? The self-described guru of all things OC ... All together now: "Meller, Meller, Meller!"

You called.

Originally posted by Meller Don't abuse the beetle juice power if you're not going to use it properly, Snaike.

Well, see, there's the rub. Had the call be answered in a timely manner there would have been specific questions for the "Meller-juice".

But another font of knowledge was quick to respond. (And, if I may, by the lack of rebuttal seems to have handled himself more than adequately.)

Therefore, while the power was used properly, it seems the response time is lacking.     

I installed the full predator 360 water cooling system, i have both the cpu and gpu hooked up to a 360mm radiator. Although I stayed up for 24 hours straight figuring it out(not kidding, it was great!), I finally have a full water cooling system in my chassis.

Only problem is, when I turn on the computer, there's no video signal. I can tell the pumps and water are working, the keyboard and mouse are flashing. I don't know what's wrong. Arghh. Could this be another stupid video card malfunction thing where I have to take out the video card and start the system without it?

At the end of my 32 hour awake journey, I hope I did not accidentally screw up something on my video card, that would really suck.(although I doubt its possible I permanently screwed anything up, i followed the instructions exactly in the manual, as well as youtube videos on installing water coolers.)

I'm pretty sure all the connections are secure, I know everything outside the computer is intact, so it has to be something in the case thats causing the problem.
The hard drives are all plugged in.

I've reseated all the components on the mb several times. I've even tried removing the video card and booting from mb video, still no luck.

When I was installing the gpu water block, I accidentally made a small scratch on the very edge of the video card, near a screw hole. Could this possibly wreck the whole card and make it unusable?

There's no beep code on startup. I can't even get any video on the screen at all, that is my problem.

how do i clear cmos? I remember something from a long time ago called "flashing the bios" where you remove the silver circle battery on the mb for a while.

I can't get into the bios at all, there is nothing on my screen except for "no signal"

Clearing the cmos did not appear to do anything, unless I did it wrong; I removed the battery and touched a screwdriver to the two pins for 10 seconds. Then i turned it back on and held down f2, I tried DEL as well. the bios still does not appear.

I think I will leave the battery out for an hour just to make sure it is clearing the cmos.

Edited by funeralpyre88 - 14 May 2016 at 8:03pm

If you have no display at all, then no way to know if you're getting into Bios.   A small scratch on the edge of the video card board shouldn't be the problem unless it's where the tracks are that carry signals, meaning the foil type lines .

Strange you can't get the igpu, mobo video, to work. Something is missing. Either a plug is off, or a component blew.

If it was me, I'd get a cheap video card that needs a pciex16 slot and try it.

crap, now I see some small scratches on the far left side of the mb, right under where the metal outer bracket for the gpu rests. only one of the scratches crosses the gold line and goes into the area with the metal lines and metal domes and stuff. Could this be causing the problem?

Those scratches do cut across the tracks/traces where signals travel... Can't tell by the pic how deep, but any break in the track spells doom.  If you had the know-how with a soldering gun  and nothing else was blown, you could repair the track yourself, I have, but its an iffy situation. 

If anther video card is still a no-go, chances are the mobo is defective, assuming all the wiring is correct.