Average 980m overclock?

My ASIC is around 66 and 68%. We use the 100/100 OC because the GPUs have clockspeeds when doing demanding tasks and then turbo clocks when it has headroom. Using the overclock 100/100, the GPUs tend to run at their max turbo clocks constantly without throttle (unless you hit temps etc etc).

ASIC mostly determines the voltage needed to maintain certain clocks.

As for overclocking, when you overclock even more, it doesn't mean you are actually running at your max clocks, plus you need higher voltage to be stable, and the more you OC, the more you require power, and when you reach the limit of your PSU, your computer can shutdown. Overclocking is not just about temps.

 

Can I +100/+100 on stack voltage?

 

unless you unlocked it first.

 

In a complete sentence?

 

Not sure, what is your stock voltage? With prema's, it is defaulted to 1.062v, and for me that is stable.

 

you can OC on a stock vbios before the lockbit was implemented. better if it was using a modded vbios as you have more control.

 

I thought they took that lockbit off the drivers after the back lash?

 

they also lockbit the vbios and for those cards that have the locked vbios you would need to flash a modded vbios to overclock.

 

ah really how do you tell if you got locked vbios? thats a dirty trick

 

no idea but it should be wiped never to exist again. i downloaded the stock vbios from prema's site and i think that might be pre-lock bit.

EDIT: as a general thing, if you're posting you're OC amounts (like +132/+100) please provide a frame of reference as to which vbios you're using as there's 3 vbios's: stock, prema, svl7.

 

Core is at 1260 (+135)
Mem is at 3000mhz(+500)
Driver 353.30
Temps 87C max

Cpu is at only 3.0 ghz tho.

I have yet to repaste.

Sent from my GT-I9300

 

Mine:
Core 1260 (+135)
Memory only 2786 (+280). Very low, but does it matter? I heard Maxwell memory OC is pointless.
Change termopads could help with memory OC?

ASIC 79%, voltage 1.018 on stress.
G751JY.

 

hmmm....+500 on the vram (1500) hard freezes my PC. 1400 for memory is still functional.

will test +134/+295 with the heaven benchmark.

 

+135/+350 - if I push into 360 it crashes in games, 380 crashes in benchmarking.
Top Firestrike with 4710HQ was 9151
http://www.3dmark.com/fs/5189969

 

hmmm...the best i can do now is +134/+295/+12.5 with prema's vbios and that's fine by me. in my 3 hour stress test its max temp was 75 deg C with it sometimes dropping to 72 deg C.

updating sig!

 

What is the +12.5 part? And what stress test are u using?


Sent from my iPhone using Tapatalk

 

Voltage. I use 3dmark11 and the heaven benchmark.

Sent from my HTC One using Tapatalk

 

I was just curious which u ran for 3hrs


Sent from my iPhone using Tapatalk

 

I'm getting +170 / +320 so far on stock voltage.

 

the heaven benchmark i run for 3 hours.

 

Memory usually isn't heat as much as it is voltage and there's no way to modify voltage.

 

Well not without breaking out the hardware volt mods

But to be honest we are more at the limit of the bus and trace interference than anything.

 

Can get up to +170/+450 at stock voltage (1.0625).

 

nvidia inspector seems to be remembering old clock settings

EDIT: Nvidia inspector doesn't like to load on startup so i've switched back to svl7....tired of flashing so many times

 

interesting: when i have my core clock set to 1261MHz @ 1.075v the actual clock is 1260.5MHz (with the stock clock clocking at 1201.9MHz for the svl7 vbios). though when i change it to 1.081v the stock clock is 1202.9MHz but my OC is 1260.1MHz.

not sure what that means or which is the correct clock.

EDIT: also, if i set the gpu to prefer maxium performance the gpu doesn't downclock when sitting on the desktop.

 

The clocks are on a set divider... I don't have the exact values available...

As for the GPU not downclocking, that's the same issue I had with svl7 and 880M. Disabling Shadow Play fixed that issue for me... But others weren't so lucky. Nothing against svl7 but @Prema actually supports his work and responds to questions... I'd highly recommend that you use his vbios.

To give my impartiality as a mod, I will say it has been since November that svl7 has responded to his own mod thread on his own forum while I have maintained communication with @Prema since the beginning, first as a tester, then a special needs customer, and I can't support someone that will leave you hanging. Prema takes his time... Because he does it right. That 1v 1200MHz clock? You know how few chips can do that? From what I've seen... 10, maybe 20%. Prema bases his vbios voltage on average Clevo... 1.062v. No out of the box instability... And you won't wait months hearing nothing if you ask him a question. It would be irresponsible of me to recommend a vbios that few cards can support and the maker doesn't support at all.

 

oh ok, i am still trying to find out why nvidia inspector won't set the clocks i want on start up

 

Anyone tried premamod v2 of the vbios for the 980m?

 

what? i don't see it, i only see v1.1.

EDIT: finally figured out how to get nvidia inspector to load at start - created a scheduled task for when a user logs in.

 

nvidia inspector multi display power saver (right click on overclocking button)

 

What the eff?

Sent from my HTC One using Tapatalk

 

My new 980M seems to be a better overclocker than my old one (at least on the core).

My previous 980M had ASIC of 69.9% and stock voltage of 1.0500V under load. New 980M has ASIC of 73.6% and stock voltage of 1.0620V under load.

Whereas previously with the old 980M, anything over +125MHz on the core would cause Unigine Valley to crash, I managed to run +180MHz on the new 980M with no crashes after two whole loops of Valley. Obviously longer testing is needed to confirm the stability properly, but it is immediately clear that it's going to be more capable than what I had before (stock voltages)

Now to move on to get preliminary VRAM clocks... *fingers crossed!*

EDIT: yup, VRAM appears to be able to go higher too (previous limit in Valley was +110MHz) - now to see how high it'll go!

2nd edit: VRAM limit appears to be somewhere between +300-350(ish). More testing needed. +400 and up definitely produced artifacts. +350 I might have seen an artifact or two but will have to repeat, to make sure that my eyes weren't playing tricks on me!

 

damn! i want a higher ASIC! #GPUenvy

 

I sympathise, I know that feeling man! I've been there with my lemon 980M (remember only +110 VRAM was my max stable!).

Turns out there VRAM's possibly even better than I thought. I must have imagined the artifacts at +350 because I've been testing at +395 VRAM and haven't seen any artifacts yet. +400 and up definitely produced artifacts.

Also, I have had several driver crashes at +180 core after turning on 8x MSAA in Valley and Heaven (before I was using no AA). +175 might end up being the final stable core clock as that has not crashed once yet. For stock voltage, that's not bad though!

I also notice that surprisingly the total system power consumption is slightly higher when using no AA compared to 8x AA. I can only assume that the higher framerate of the non-AA run means that the CPU had to work harder.

I may even have a little room to overvolt ever so slightly if this theory proves correct as I always use VSYNC in every game I own, therefore limiting power consumption in any game that's not GPU bottlenecked.

More testing ahoy!

 

Are your clock increases based on the stock clocks?

Sent from my HTC One using Tapatalk

 

Yeah, I'm currently testing 1300/1450 (+175/+395).

 

Damn, I can't even get mine to 1300. I even tried putting the vram to 1500 and it hard locked so reverted to lower clocks...boo!

Sent from my HTC One using Tapatalk

 

forces p8 state when gpu is idle

 

Trust me... Its the Silicon not the ASIC... My 78 can't do more than +165 without the full voltage boost to 1.2v and I have a hard enough time cooling it anyway... ASIC has nothing to do with overclock capability, it just means the chip uses a lower voltage...

There is an expense though... Low voltage high ASIC chips tend to actually run hotter because the electron leakage is higher. That's why you never see a 90% chip in a laptop... The electron migration exponentially increases heat.

 

I really don't get ASIC.

> My old 980M had 69.9% and voltage of 1.0500 under load.
> New 980M in replacement laptop has 73.6% and yet runs at 1.0620 under load.

Higher ASIC and yet higher voltage? Doesn't make sense to me.

I'm not complaining though, it's definitely a better overclocker than what I had before!

Today I'll do some longer, more thorough testing.

 

About identical to what I can do.

There is no direct correlation between ASIC and overclocking or even voltage. It's simply a measure of *average* silicon power leakage. So you can still have a larger variation that can prevent it from overclocking without needing a decent bump in voltage.

 

According to @Prema this is not correct... The vbios looks at the ASIC to determine voltage... Since @Cakefish has the 1.062v he is probably running the Prema Mod Clevo 980M vbios that pushes 1.062v because the average Clevo chip is around 68? Or somewhere around there... Let's put it this way... Prema asked me to send him a GPU-Z of my 78... Because it was the highest Clevo he had ever heard of... Both of my cards were probably cherry picked by Sager.

 

Yeah, Prema VBIOS v1.11 here.

+225/+395 seems stable when +37.5mV is applied (1.1V). But power consumption is quite a bit higher. Not sure if I want to risk it, considering this is only a 180W PSU adapter.

Metro Last Light benchmark sends power consumption to a brief peak of 215W so when the 85-87% efficiency is applied that's right on the borderline. It hovers between 170-200W through most of the test.

The Witcher 3 uses ~190W under the same conditions. Paring it back to +175/+395 with stock voltage sees it go down to ~170W.

I haven't tried lowering voltage on that +225 core clock yet, so I can try that, see if it's stable at either +25mV or +12.5mV.

Edit: I think the max I'm willing to overvolt is +12.5mV. I can't risk going higher. Already had to replace this laptop (not due to overclocking but something else), can't afford to do that again! I'll see where that gets me but I'm not going to 1.1V. 180W PSU is not enough to safely power that level of voltage.

 

I don't know. I've seen cards all over the place regardless of ASIC. I'd like to see a database of correlation between overclock and ASIC. I haven't seen anything yet other than some random sampling from some forums and nothing is concrete or definite.

 

probably time to upgrade to a 330W PSU?

 

It has nothing to do with overclock, its about the voltage and leakage balance. High ASIC cards run lower voltage but the cost being heat. I don't entirely understand the phenomenon myself but from my limited understanding, high ASIC cards are prone to electron migration which drives up the heat... So you could have a 60 percent running the same clicks a 70 percent did with the 60 percent running a higher voltage and actually have the 70 percent run hotter. That's why high ASIC is great for water and LN2 but would be horrible in a laptop... Because electrons can migrate and heat up other components on the card and cause a real big mess if that heat can't be handled which is why Clevo heatsinks are trash.

 

So, in essence, low ASIC quality chips are desirable for our laptops? At least from thermal and OC perspective?

 

It doesn't mean anything in either case. Just a dumb number you shouldn't read into.

 

According to GPU-z it does have an effect on OC:



But I have not seen any correlation between voltage and ASIC and ability to OC.

I just don't think you can read too much into that number is all I'm saying. There's more factors involved. Kind of like Windows WEI numbers.

 

must have something to do with thermo dynamics?