GTX 980M Overclocking

Same symptoms of crashing when I was overclocking too far on the 970m. So I would chalk that up to an unstable overclock. For me to overcome I could just up the voltage a bit.

 

i noticed on mine, that +62.5mV is the max voltage i can apply when ocing both vram and core. when only core is oced then up to +100mV is fine. i recommend ocing both memory and core and find a good balance between the two for the largest performance boost. ocing only core or only mem brings only half the benefits, even if ocing both restricts the clocks a bit dure to the power limit.

Sent from my Nexus 5 using Tapatalk

 

Yeah, when I didn't use +25mv when OCing. I was unstable really soon with the same symptoms, but I figured I was stable. It sucks heh. Until I flash and unlock, I will need to keep it down.

 

Some fire strike extreme.

Total score 9775
Graphics Score 11021
Physics Score 16108
Combined Score 4010

4930K @ 4.3ghz
980M SLI @ 1436/1500

http://www.3dmark.com/fs/4306829

I am testing out 1315/1500 @ 1.1v to see if it will run 24/7 at this point.

 

how do u test for gpu oc stability? what are your satisfactory conditions to call an oc stable for 24/7?

Sent from my Nexus 5 using Tapatalk

 

Benchmark first, then games such as star citizen, mech warrior.

I know it can benchmark at 1365mhz at this voltage so I have a fair margin I believe.

 

Very nice. I am hoping for similar OC capability. How's your ASIC quality btw? Trying to see if there is any correlation.

 

No actually the stock voltage is 1.0500V for the 4GB 980M.

 

depends. clevo mxm cards have 1.0625V as stock. msi and/or asus have lower stock voltages though... dont remember which

 

Clevo non-mxm is 1.0500V

 

good to know

 

My test for overclocking has been Crysis 3. Crysis 3 will cause my OC to be proven unstable when it's stable on Heaven/Valley so far. VRM is less stable than core right now. 175/175 @ + .25mv

 

therell always be other applications that will prove an overclock unstable even though uve seen it stable in crysis 3 in the end, whats important is that ur clocks are stable in the applications that you personally use

 

How strange... I have noticed the exact opposite!

 

New drivers... Max I can get at stock voltage is 1321MHz on my master card. About to find out if the slave can do it too.

 

yupyup, time to give those new drivers a test drive(r)

 

Slave couldn't do 1321 but it did do 1316 but only in 3DMark 11 performance. I had to drop to 1298 for the rest of the tests but I don't necessarily think ~1300 core is bad for stock voltage... its certainly better than the 1222 wall I was stuck at previously!

3DMark 11 P (1321)- http://www.3dmark.com/3dm11/9569354
3DMark FS - http://www.3dmark.com/3dm/6267192?
Catzilla - http://www.catzilla.com/showresult?lp=454237
Bioshock Infinite DX11 Ultra + DDOF preset - Average, Min, Max - 185.76, 13.97, 487.01

3DMark Vantage and the extreme presets on both 3DMark FS and 11 crashed the driver @ 1298 and FEAR crashed as well.

Think I'm done testing for the day

Slave card is the unstable one and bumping voltage by +10mv did nothing to improve stability in any scenario, don't have the thermal headroom to push the voltage higher but the results are much better than I expected anyway.

 

Well done buddy!nice scores you have there!!
So coclusion is that the new drivers are veru ocable than previous drivers? right?

 

They are for me. 344.75 went unstable at higher overclocks immediately. Nothing else has changed on my laptop that would account for gaining so much more room without touching the voltage.

 

I've been trying to see what OC I can get on stock voltage [Edit: Also stock vBIOS "84.04.22.00.0A"]. I got a few green/gray screen of death and they always recover to desktop. Now I'm having sporadic startup issues on cold starts. More often if I had a g/gsod in the last session. [I didn't keep a record, but it could be cause & effect].


<Provider Name="nvlddmkm" />

<EventID Qualifiers="49322">14</EventID>


<Provider Name="Microsoft-Windows-Kernel-Power" Guid='{331C3B3A-2005-44C2-AC5E-77220C37D6B4}' />

<EventID>41</EventID>


<Provider Name="Microsoft-Windows-WER-SystemErrorReporting" Guid='{ABCE23E7-DE45-4366-8631-84FA6C525952}' EventSourceName="BugCheck" />

<EventID Qualifiers="16384">1001</EventID>

This is very much like a driver issue I had from trying to rollback from 347.52 a few months ago, but this time it is recovering and rebooting into widows, probably the bugcheck since that didn't happen before.

I thought I had a "good OC" stability-wise +119/+150 using afterburner, but I now think it depended on what power profile the laptop was using at the time. Process Lasso sometimes changes the profile on me and I haven't figured out how to stop it 100% of the time.

FYI:
-OC is not set for startup and I return to stock drivers before shutdown
-With DDU, it said to not delete nvidia file ...but I did anyway because I used to on the old G73. I don't see why that would matter

Q: Is the GSOD'ing corrupting the driver?

If I restart once in between, it doesn't seem to happen. That's why I think it is cause & effect. With my G73, a driver I had a while would go into 8-bit or 16-bit color if I closed the lid in a game and stayed that way. This feels like the same thing with different symptoms.

Q: If anyone else is OCing on stock voltages, what are you getting? Thanks in advance.

 

i get +150 stable on stock, whereas +200 needs voltage increase by 25mV. havent checked in between yet though, but its somewhere between those two values

Sent from my Nexus 5 using Tapatalk

 

I tried Unigine Valley yesterday at the stock voltage of 1.0500V and clocks of +125/+100 (1250/1300) and it started crashing on me. I'm almost at the point where I want to uninstall the nefarious Valley as it always seems determined to try to completely dash any dreams of mild overclocking on stock voltage that I have. It seems determined to be unstable on anything but completely stock clocks. Party pooper. I have 69.9% ASIC quality for crying out loud.

 

New 347.78 driver has been a blessing in disguise... My max Firestrike GPU score went up with less voltage

@HTWingNut, start benching ... This was my 970M on +430/520 @1.162V... 64% ASIC I believe..

http://www.3dmark.com/fs/4349123

 

Nice scores

 

What about your 970M scores? How much do ur cards do?

 

check his sig

 

Old benchmarks.. want to see his benchmarks with this driver.. Also what his max OC is on each card

 

Man, Defense Grid 2 seems to be unstable even with +100mhz memory. Every other game, crysis 2, Devil may cry, benchmarks etc, are fully stable at +300mhz memory. I wonder why? The game itself is a very light load.

Slightly related, since I am running my monitor at 96hz, I can finally see the benefit of extra framerate, and the performance SLI brings

 

lucky bastard my previous rig had a better panel in terms of overclocking ability @95 Hz. still, wouldnt trade it back, the higher 3K res and more intense colors make up for the meager 65 Hz i can squeeze out of it

 

Well now im at the airport,travel back to greece
I will test them when i get home...
This benches on my sig is not the max i can get with the old drivers of course.
This clocks are only 1320/1480 with 1.05v.
And i can play stabe 24/7 games such bf4 multi.
If i only bench i can go even more up...
My asic is on both cards 71.3%-71.2%.
Is i weird??lol

 

Is there any way to push past the +135 core limit without modding the vbios?

 

NO... It's just not possible.. On a hardware/software level.. You need to have a modded vBIOS...

 

so I just test them and I got for 1.137v=1438/1478
http://www.3dmark.com/3dm/6332281
but now go to sleep now....
tomorrow is a new day for benching...lol

 

what about single gpu score ocd?

 

well I don't have enough time to bench single to single card
this is with 1.16v and 1485/1475
http://www.3dmark.com/3dm/6338141

 

Single is easy, doesn't even need a reboot. Just edit the app profile in nVidia control panel to set single GPU and restart 3dmark.

 

Monster performance!

 

I replaced all the thermal pads on my 980M and now running with +475 on memory at stock voltage. Maybe will help with you too. I think it was a thermal issue for me on memory.

 

here is my score with a basic overclock of +135/+200

http://www.3dmark.com/3dm/6351688?

 

Try to ov this beast

 

dont think i can, they said the new alienwares are impossible to OV, dont know if this is still true?

 

Oh its 980m with 4gb on board ah?
Svl didnt make a vbios for that versions?

 

No he has not, he did mention its possible though

 

You get a decent boost from it at 135. Also beyond 135, you probably need to do an overvolt. My max without overvolt is 1275. I'm staying away from overvolting.

 

I think that if you ov at least on our cards at 1.05v its good and safe.
now at your cards stock is 1.05v right?
if you go a little up I don't think it will be big deal.

 

Hey all,

Just got a NP8278-S (P170SM-A) a couple weeks ago, with the 8gb card. I'm running 1080p only, always with all settings maxed out.

Yesterday, I took the plunge on Prema's BIOS and VBIOS mods. It's stable so far, but I haven't taxed it with Crysis 3, Metro LL or other really demanding games. Been playing AC3 (I know), Dishonored, and a heavily modded Skyrim. Max temperatures with the internal fan at max speed and an additional cooler are never higher than 75 C. This is my benchmark score post-OC: http://www.3dmark.com/fs/4399791

Here are my previous scores, with the stock BIOS and VBIOS:
Stock clocks: http://www.3dmark.com/fs/4351941
Slightly unstable overclock: http://www.3dmark.com/fs/4352026 (I finally settled on +135 Core, +333 Memory pre-mod; don't have a saved score for that, but I think it was about 8600-700)

I'm looking into overvolting slightly, but haven't done it before. I have played with undervolting on my old PC, with a 545M. As I understand, the dangers are higher temperatures and more power consumption, correct?

Also, I can't seem to get the voltage slider unlocked in Afterburner, do people just use Nvidia Inspector instead? And if I overvolt by say 60mv, would there be a noticeable boost in OC potential?

Sorry for the long post, and the lack of potatoes...

Thanks in advance.

 

@shaparo Most 980Ms don't go a whole lot higher than what you have now so its debatable whether or not the added voltage and heat is worth it for another 20-30MHz stable core. I mean for benchmarking, you can push them a bit further but it's not likely to be stable. Keep in mind that performance you're at now is just over what a non-throttling 880M SLI system will pull on stock clocks lol

 

well, it depends how far u want to push it stock voltage gives me something around +150-200 Mhz on core, whereas overvolting allows me to go up to +290-300 Mhz on core, thats 45-100% more OC

 

o_0 I knew the 980M was a big step up, but that big?!

So I upped voltage by 75 mV, stuck to my current stable OC of +195/+303, and in FireStrike the temperature went up to 90 C - throttling the core down from 1300+ MHz boost to about 1230. That's with max fans plus my three-fan cooler, when pre-OV it went up to about 82 or so. GPU-Z put my ASIC quality at 65%, which isn't too great as I understand.

Could I ask how much you have overvolted by, and what temperatures you are getting in FireStrike?

 

ouch, those temps are definitely not nice... at +250 im stable at +62.5mV, +290-300 mhz needs +82.5mV/+100mV. max. temps are rarely in the 70s, highest ive seen thus far was mid 70s after an hour of unigine valley loop with max. fans...

based on user reports, the cooling on the ZM series seems to be significantly better than on SM-A machines (both single and dual gpu models)

Sent from my Nexus 5 using Tapatalk