GTX 980M Overclocking

OK I found someone who has the same problem as me here: http://forum.notebookreview.com/threads/how-accurate-is-the-gpu-z-clock-gpu-clock-too-high.737742/

It seems like it's caused by the combination of Windows 8.1 and Nvidia driver. Basically it creates 2 sets of defaults and switch between the 2. It's really weird .... one set of default has 1038 GPU clock & 0.85 GPU voltage but another set has 1126 GPU clock & 1.06 GPU voltage. By default, Nvidia Inspector reads the 2nd set and GPU-Z reads the 1st set. But when I open both programs at the same time, Inspector then changes its values to match GPU-Z. BUT when I open Heaven Benchmark, both Inspector & GPU-Z will match the 2nd set. I mean, I understand GPU can change by itself, but voltage shouldn't jump from 0.85 to 1.06 by itself right? And now when I try to change anything (GPU clock, memory, or voltage), it just won't apply and I am guessing that's because it doesn't know to apply to the 1st or 2nd set. Before when I used MSI Afterburner the values applied because it was only reading the 1st set. I think the 2nd set of default was introduced when I started OC using PrecisionX or Inspector.

Anyway I'm gonna try installing different drivers to see if I can get all the programs to only read 1 set of values. If all else fails maybe I'll have to re-install Windows or something.

Btw, what led me to that thread above was when I tried 3DMark 11 to benchmark and found out that my physics score is extremely low. Did some search about it people said it's probably caused by GPU voltage set too high causing CPU to underclock itself. So I checked my GPU voltage with Inspector and GPU-Z and realized I got 2 different values ...........

-----

UPDATE: Problem solved. It was CPU throttling in the end. I guess I got high benchmark at first because I had just installed my 980M but after awhile it started to affect my CPU. In 3DMark combined test my CPU clock dropped to 1200 (min). Using ThrottleStop now it is running smoothly again. =D

As for OC not applying, it's because I was using the driver 350.12 which disables OC. I now switched to 353.00 which enables OC again. =D

 

just did a quick check: at 4.8 ghz and ddr3-2400 im getting 13618, so thats 10.6% advantage for you

 

I either have a crappy 980M or the PSU can't keep up but i'm only stable at +160 core with Prema bios running at max overvolt setting. Asus G751JY.. The PSU is 230watt.

 

Where can one find an unlocked BGA 980m 8gb VBIOS for my Gigabyte P35xV5?

 

Try lower overvolt settings, you shouldn't need maximum, that adds too much heat.

Always start at stock, then find max OC at that voltage, and use the next up unstable GPU/Mem Clock setting as the guide for finding the next up voltage bump - which should be made in very small increments.

It takes a while to get a feel for it. I do recall thinking early on that max voltage is the place to start - but it really isn't - the best OC will be found well under max voltage.

And, +160 isn't crappy, under +135 on stock voltage is crappy

 

True true.. thanks for the tips, i'm doing some testing now at stock voltage and will test the ceiling there and then increment voltage slowly after that. Appreciate it!

Edit: You were right.. i'm still +160 core stable at stock clocks.. and holding 72C at ultra Heaven bench.. Bet I can finally go higher.. My system was black screening, spinning fans to max then shutting off.. Wonder if I was triggering thermal shutdown!

 

Too much voltage will make things unstable before thermal monitoring is triggered - though really internally that is what is going on - too quickly to register.

The goal is to run as high of clocks as you can with the lowest voltage - stable of course.

Another tip, run GPU MEM OC at mimimum while finding GPU core clock max.

The memory needs to eventually OC to track core OC for best performance, but you need to tune for one then the other.

You want max GPU core first, then follow with the memory to OC as far as stable after that.

Remember you likely don't have independant temp monitoring or voltage control for Memory, so don't worry if it doesn't seem to clock as high as you like - that's often the way it is.

Don't fight trying to get / keep Memory OC after OC'ing core for Maximum.

The GPU card is running with reduced resources when you Max out the Core - the Memory OC max will likely be less at max Core GPU OC.

 

Yep yep, I'm actually only testing GPU core clock maxes right now leaving memory at default... Thanks again for the help! I'm up to +180 stable so far.. +180 crashed at stock voltage. Bumped it to +12.5mv and that made it stable! Didn't think such small voltage increments would help with stability.. I always went for max voltage to avoid having to spend the time to figure out my max core OC but again, you were right, it apparently doesn't work that way to any advantages.. I learned a bit.. On my liquid cooled TItan X SLI on my desktop, I can actually run max voltage + max core +500 on each card with no crashing.. Well it's more like +450.. After that I crash.

 

Yup, that's how most people run, but taking that extra time to walk up the voltage only as necessary really cuts down on the heat, extending the OC and extending the life of the CPU/GPU.

It's likely why my 24/7 OC's never die, and why some others seem to cook theirs on a regular basis. That's where OC'ing gets a bad reputation.

Try doing the same on your desktop CPU/GPU, run at minimum voltage for max OC - you might even get more OC out of it.

I have noticed that high voltage GPU/CPU OC's can get cooked/slotted to be only stable after a while at those high voltages. You need to uncook them - if possible - by reducing gradually the voltage over a few days / weeks to recondition the parts. Sometimes that works, sometimes they are too far gone.

It's all a learning experience, that's why these forums are so good, there are a lot of experienced people with expertise and thinking patterns differing from our own, we all learn together

 

You're right.. I'm going to revisit my Titan Xs and see if I can eek out anymore more.. You never know! Thanks again, you've been a great help.. I'm up to 200mhz on the core now and Heaven hasn't crashed yet after 10+ minutes.

 

Personally, I don't trust Valley for stability anymore. Used to be that I could do +200 on my old card and be stable in valley, but not in Guild Wars 2. For me, the only stability test is real world usage.

 

tru that, best stability testing is done on the games and apps u regularly use still, it helps to get a feel of new hardware using standard tools such as valley.

Sent from my Huawei Mate 8 NXT-AL10 using Tapatalk

 

Hi everyone. I have asus g752 vy and i did try to oc my gtx 980m. (=100 mhz core clock and +100 memory clock).
Temps are good ,no artifact, but when I open gpu z and open sensor tab there is PerfCap: Limited by gpu power. Should I worried about this? I did not change voltage. On my previous laptop (g750jm) with gtx 860m i didnt have any problems with oc.
Can you help me?

 

thats the normal power throttling u get when on the stock 980M vbios premamod should help, although im not sure which version would be compatible to your gpu

Sent from my Huawei Mate 8 NXT-AL10 using Tapatalk

 

After oc my gpu i get quite a nice kick of performance. Fire strike from 8340 to 8900. Also i get about +5 fps in the devision game But this power throtling on stock bios is annoying Is there any danger when I ignore this PerfCap?

 

You would be in danger of enjoying the full performance the card has to offer keep an eye on temps. Make sure they don't get above 85C for prolonged periods.

 

So in other words when I will see this info in gpu z (PerfCap:limited by gpu power) I have to put lower value in evga precision? Temeratures are fine GPU 74, CPU 84

 

Temps will rise when I increase voltage. When I set core clock +100 and mem clock +100 temps are not rise. I have only this info in gpuz limited by gpu power

 

Now I figure out there is irony what a shame

 

Lol

Sent from my SM-A500FU using Tapatalk

 

haswell can certainly hold its own against skylake, no problem just need to get the thermals under control

 

Thanks, I registered to techinferno but can't download the rom file since I need to make 5 quality posts and wait a week to be promoted. Can you send the file to me? I understand the rule but this file probably won't work for me so it doesn't make sense to wait a week just to try it or pay for membership. email is [email protected]

 

been reading this thread for a while now, my g751jy 980m 4gb, ocd 1380/1500 @1.18v voltground vbios v.2 and voltground fancontroller profile, firestrike graphics 11792, stable on gta v ultra/high grass, fxaa on, no msaa, advance graphics on/max xcept viewing distance set at 50%, 60fps 95% of d time anywhere, vsync on(less gpu load/less heat) will crash on witcher 3 in 3mins! my question is why i only get max of 79.2 tdp on gpuz while doing bench on firestrike,gtav,witcher3,etc.

 

have u tried HWInfo64? go ahead and check the TDP numbers there, maybe ull get something different. also, not all tools / games / programs put the same load onto the GPU

btw, welcome to NBR, just saw that this was your first post

 

1. @Jaybee hi there, i reverted back to v1 vbios, (i dnt like the preset core/mem/voltage of v2) set mv to +81(1.1310v) +225/+496, max tdp 69(gpuz) 69.5(hwmonitor) cpu max tdp 39(undervolted/ocd) firestrike graphics 11679

 

adjusted mv to +75 +199/+496, +81mv will simply crash on witcher 3 under 99℅load@81c with slight artifacts in 3mins, now at +75mv +199/+496 it took 20mins to reach 87c/99℅load no artifacts, didnt crash, no throtling, and the temp stayed at 87c for another 30mins before i quit the game, pretty stable since i wont be using that 99℅load all d time, cap the game at 60fps n monitor to 60hz, 72℅tdp(witcher3) at that overvolt, this is my sweet spot atm, almost same/at par performance on stock 1060 laptop, firestrike graphics score 11529, firestrike 9729

 

87c for a Maxwell GPU is very high

 

yep, with that oc and no way to add external cooler(useless to rog g751), ill just avoid 99%gpu load, eg. 60fps cap, no msaa, no hairworks etc. only witcher 3 stresses the gpu to 99% -100% try to turn shadow to ultra... in my case. got no prob in gta v, 70-75c only max 80% load most of the time(forest) maxed out, grass high, fxaa only, advance graphics all on, maxedout, and 50% xtended distance scaling, 60fps 95% of the time, 4give my english guys

 

Thats way too much voltage and low clockspeed for said overclock. I would dial down the overclocks considerably to something more manageable. You are running into stability wall with little improvement.

Use something more conservative when gaming, that preferably requires no extra voltage.

 

back to stock vbios, +135core/+253mem, 10700 firestrike graphics

 

Great thread you guys!

Keep the scores coming.

My 680M is nothing compared to your 980M's... but I am going to try my luck after the repaste, re-cool and new thermal pads job.

I might be able to get 60% of your score... ish

 

keep us updated on your repaste / repad with before vs. after temps and scores and no worries, our 980Ms cant compete with pascal either, but still fun to play around with the hardware

Sent from my Huawei Mate 8 NXT-AL10

 

I'm on prema bios on my 970m. I get a graphics score of 9421 with +260 core +400 memory and a +50mV. I think it still has more in it because I only hit 68C during the benchmark.

 

try playing witcher 3 on ultra, if it doesnt crash for 30mins, then there is still a room for ocing, watch out for temp, 85-86c is d most stable temp i got

 

980m 4gb gsync g751jy my latest stable oc, using ver.2 vbios voltground, firestrike graphics 11461 @ 1319core(boosted)/1500mem, 1.1060v, max temp 82c (witcher 3 ultra after 1hr/99% load)

update! reached throttle temp limit of 89c after 2hrs of 99% load... damn.

 

finally! tamed 980m g751, +181core +496mem +56mV using v1 vbios, artifacts free(+50mV causes very minute artifacts) 11250 firestrike graphics, super stable max temp of 86c after 2hrs witcher 3 ultra 99%load (gtav doesnt even reach that temp/load), running around fields n forest, also i7 4720hq undervolted and multiplier x37 x36 x35 x35 via xtu, i can play with peace of mind now!

 

Ugh I used v2 bios for my p770dm-g 980m. No matter what I did temperature would always get >80 and computer goes black with beeping noise during benchmarking. So I reset settings back to default but I am still getting high temps and black screen after gaming for 30 or so minutes. I think I messed up and fried the GPU...

 

If it works at all it's probably not fried. Did you check the paste quality lately?

 

Hi guys, great thread you have going on here, my 980m is a little bit a of a crappy one, since i can only get +150 on core with +62.5 mv :O and +500 on memory... I was wondering do you guys consider this safe as a 24/7 OC? Oh and the max temps are 74C.

 

Crappy one? It is really good.

If that is stable, turn everything 5-10% (I'd actually go down to 20-25% of your max OC) down of your current OC and that should be considered "safe".

 

Well okay maybe I exaggerated by saying crappy... It's just that the first one I had was such a champ but anw i will do just that then! Thank you
P.S. I wonder why this card stays so cool? I mean even by pushing 1.125V it doesn't go over 74C and the first one I had would go all the way up to 84C with around 1.1V. Maybe it's the ASIC quality this one is 64% and the first one I had was 76%...

 

this high voltage on heavy load would make it power throttle more often if not most of the time if you are on a stock vBios. this would explain the low temps.

 

You would be right, but you can't ov past +25mv on stock, I ran and run prema on both In any case I will not complain I was just curious

 

409/5000
Hello good intentions to modify the BIOS of my 980m for better performance e seen that improves with respect to original but I can not download the bios I am already registered and nothing the page is this
Https://www.techinferno.com/index.p...vbios-mods-900m-series-overclocking-versions/
I could have passed by mail
[email protected]
Or upload it
thank you very much

 

You need to have 5 posts on Techinferno before you can download.

 

Hello, I do not understand you with the 5 entries in the post.

 

you don't get to download from Techninferno till you have posted 5 times on their forums, what is there not to understand ?

 

Ok now if it was google translator problem thank you

 

Hi, I'm already good and I can get the bios of the 980m Nvidia GTX 980m - 'OC edition' rev0 - >> 8 GB Version << - modified by @johnksss & svl7 -> see attachments the truth that has increased performance and but E sometimes I get the black screen while game I have the drivers 384.92 any solution? Thank you

 

Check vrm thermal pads for the blackscreens