Average 980m overclock?

Alrighty, more testing done and +395 on the VRAM seems a bit much. It was fine first of all, but I began getting artifacts in 3DMark FireStrike after a while (I assume as the memory was heating up). So I'm scaling it back a bit to +375 and see where that takes me.

On the plus side, a mere +12.5mV nets me a seemingly stable +215 on the core. That's a nice jump over the +175 with stock voltage (1.0625V). So I'll be keeping that I think. +37.5mV can get me to +250, but the power consumption gets too close to the 180W PSU limit for comfort at that voltage (1.1000V) in certain games, like Metro 2033 for example.

So in conclusion, the new clocks I'm going to be testing from now onwards are +215/+375 (1340/1440) with a +12.5mV boost (1.0750V).

 

do you have battlefield 4? how many fps extra do you get from stock to overclock when playing at 4k? would you say medium to high settings at 4k looks better than ultra settings at 1080p?

 

so then med-high at 1080p is fine!

 

The *only* thing that is guaranteed by higher ASIC is lower voltage. My master, my 78, runs at 1.018v... To give you an idea Prema has his vbios at the Clevo standard average of 1.062v.

There is zero correlation between ASIC and overclocking in any way. The only thing that has been shown is that high clock + low voltage chips run hot. Look at the 880M for an example... Lowest ASIC I ever saw on those was like... 65 percent or so? I don't completely understand the concept but my limited understanding says that higher ASIC is more prone to electron migration (thus heat) which is why laptops generally get 60-70 ASIC quality chips... Its intentional. Desktops can dissipate the heat but laptops can not. They only rarely have active direct cooling (old school Alienware) while everyone else uses inefficient heat pipes that rely on a fan to draw the heat away. Alienware used to pit the fans right underneath the cards and use really short heat pipes. Highly efficient but they had to be cleaned a lot because they would suck up the dust...

Anyway, ASIC sets the voltage of the stock card... Anything else is guessing... (Yes I'm saying GPU-Z is *wrong*) ASIC Quality determines the optimum voltage for the chip... The voltage needed actually drops a lot the higher the ASIC... My master which is 78 pulls 1.018v while my slave at 73 pulls 1.043v...

 

Good info Ethrem, and based on what you've said it's not directly related to overclocking at all.

But I think heat-pipes are actually fairly efficient. You may only lose a few percent compared with direct cooling. Copper is an amazing conductor of heat, and thankfully fairly cheap too. And the gas/fluid exchange in a closed system is usually pretty efficient too.

 

+135/+500 on mine

Sent from my GT-I9300

 

B-b-b-ut;

Old - 1.0500V - 69.9%
New - 1.0625V - 73.6%

My new 980M defies the rules. It is a rebel! I don't understand why.

 

My GTX980M arrived today, it's already up and running in my P170EM. I will soon share my overclocking experiences with you guys. ASIC quality is only 64.4%. I am running Prema's modded vBIOS v1.1. It's a night and day difference when compared to my OCed 7970M ( @1000/1500) already!

 

Wait, my card has a lower ASIC score and it has a lower voltage of 1.050v on stock vbios. How does that work?

Sent from my HTC One using Tapatalk

 

Are you running Prema's vbios? Because that would explain it.

 

Modest OC +135/+200 @ 1.075v using Prema's vBIOS mod v1.1:
http://www.3dmark.com/3dm/7801219
Highest temperature 84C.
Using 353.12 driver.

My OCed (1000/1500) 7970M scored ~5500 in FireStrike 1.1.

 

Stable overclocks are really hard to pinpoint.

Previous clocks mentioned appear to be unstable.

Edit: Stable core clocks with corresponding voltages;

+170 @ 1.0625V (stock voltage)
+190 @ 1.0750V
+225 @ 1.1000V

The question becomes; which to use as my daily overclock?

 

The question is rather "do you need the overclock daily"? I rather simply double click a NVI profile on my desktop to apply a clockspeed if I'm playing a game that needs the boost and can stay cool with the boost, and keep it stock all the time. Of course cooling is less of a factor for you I suppose, but no need to OC if you don't need it for the most part.

 

So I've done some extended testing in Unigine Valley and I'm fairly certain that I've finally found my 980M's stable clocks at various voltages (determined by leaving Valley looping for 20-30mins with no sign of any driver crash and recovery - clocks monitored via GPU-Z).

- 15% core overclock -> 1297MHz (+170) @1.0625V (stock voltage)
- 17% core overclock -> 1317MHz (+190) @1.0750V (+1.1% overvolt)
- 20% core overclock -> 1352MHz (+225) @1.1000V (+3.5% overvolt)

There's definitely plenty of further headroom but I'm not willing to go over 1.1V to get there. Temps were under control (never once exceeding 80°C with max fan FN+F1 profile activated, peaking in the mid 70's for most of the tests).

Power consumption in taxing games such as The Witcher 3, Dragon Age Inquisition, Metro Redux and Crysis 3 typically hovers between 170-190W at the most aggressive overclock/overvolt. This would fall to roughly 160-180W at stock voltage overclock. Metro Redux benchmark (the most power hungry benchmark that I could find) managed to get the peak power draw to ~215W for a very brief time, but typically drew between 180-210W. This did not change much at the lower voltages (seems to be more dependant on pure clockrates).

I have a 180W adapter with an efficiency rating of 87%. So in these games at max GPU overclock/overvolt, the PSU is working between 80-90% of it's maximum rated capacity.

Based upon the 180W PSU and your own experiences with overclocking/overvolting, which would be the wisest choice?

Yeah man, that's what I mean by 'daily' overclock (poor choice of words). Only in games where it makes sense to have the extra power (i.e. those that run under 60fps). And only via NVI software, manually selecting the profile before launching the game. That's how I've always overclocked my GPUs.

 

Well, 215W is ~187W or so with that 87% efficiency (though I'm not sure how you tested that) so a max power draw from the wall for you is still pretty decent, especially if you're under 200W most of the time. It turned out I've never run out of power with my overclocks, but rather my VRMs or something would get too hot in my system, so I never ran heavy OCs too much. But if your temps are fine, then it's up to you more than anything else.

If there's no downside at all to the boosted clocks, then just use the highest one. If you ever get a single crash, then either slightly drop the speed or pump the voltage. Up to you. I can safely say that the middle clockspeed is worthless to you. It's not nearly enough of a bump from the 15% boost to warrant even thinking about I'd say. Again: if there's no downsides, then why not get the most out of your card?

 

Yeah the middling voltages are kinda weird in terms of stability. I didn't even bother to list +25mV (1.0875V) as it only allows up to +200 - just a 10MHz boost over 1.0750mV! Even more pathetic! Then you just add +12.5mV more and bam - +225MHz becomes stable (seemingly so far).

The efficiency rating is on the PSU itself - category 'V' apparently. So I Googled it and 87% was the answer I received.

The Metro Redux benchmarks are the only way I've managed to make power consumption exceed 200W so far during testing. Every actual game I've tested so far has peaked at 195W (195*0.87 = 170W) or below.

 

Yeah it feels like definitive diminishing returns on that voltage : clockspeed thing. But either way, 1297 vs 1352 is a decent enough jump to make a difference in games, so that's something to consider. But of course, boost clocks take preference too. If you can get 1297 boosting 24/7 and 1352 doesn't boost sometimes... then the former may be the better choice. Who knows, I guess. But yeah we can safely rule out the middle option.

 

1352 boosts solidly. Solid as a rock. Thanks to Prema's modded VBIOS!

 

Sounds great then. I'll probably have to keep the OCs somewhere within the 1200-1250 range if I ever get those cards though; or else I'll be busted for heat XD. Especially since if I do get them, I'll be using them with the slave GPU having worse cooling.

 

i have mine clocked to1260 on the core.

 

Well I just tried out The Witcher 3 for a bit. 100% stable but temps hit 86°C with max fans on. That's the highest I've ever seen on the P651SG!

It's really hot here tonight - 16°C outside, but that's not really to blame. It can get hotter on a sunny day.

 

16C is "really hot", are you joking me? It's the middle of summer. I live in the middle of California which is some of the most moderate weather in the world and it's over 30C today.

 

you think 30C is "hot" then you haven't lived in Australia: it gets up for into the 40C in Sydney.

 

Never said it's hot. It's moderate.

 

Here in south west Britain 16°C during the night is deemed hot hot hot! It's 1AM here (still awake though!).

86°C is a bit warm for a GPU though isn't it?

 

You gotta improve cooling or dial back the OC then. GPU would be over 90C where I live.

 

the max temps i see with the heaven bench is about 75C and games it's much less with witcher 3 (albeit at 30fps) goes between 50c - 63C. though most games will stay under 70C.

 

With Witcher 3, no OC, with my current 35C ambient temps, It pushes one of my GPUs to 84 and the other to 78. My CPU is at 96 tops. I don't use max fans though. Maybe max fans will knock off 1 or 2 degrees.

But yeah, 86C on such a low ambient temp of 16 is a bit worrisome.

 

"It's really hot" and "16°C" do not go in the same sentence together.

Your wut. Try max fans and let me know. I couldn't hit 96c if I OC'd by 300MHz and overvolted my CPU for stability and livestreamed in the daytime where I am. If not, I need to redact my statement about the GT80's cooling ASAP.

I think 30°C is "cool" for daytime.

 

Well it's not 16°C ambient temp inside the house that's for sure! I got heat rash all over me because last night was so warm but tonight is even warmer - sounds like some of you would be hypothermic in such conditions though

I'll test again tomorrow (later today) and see what temperature I get. If it gets anywhere near that again I'll just stick to stock voltages from here on out. Or I'll go down to +12.5mV if it's cool enough.

 

seriously need to check the fans are pulling enough heat away from the heatsink.

 

Have you propped up the machine though? Just curious.

Also, please remember that every program uses the GPU differently and heats it up differently. The most dangerous thing heat-wise for me (besides furmark) is actually GTA V in heists planning mode. Since I don't vsync the game it hits ~180fps in heist planning mode (in multiplayer) and my main GPU happily hits the thermal threshold with an OC. My slave is a champ and doesn't pass 82c even with a +62.5mV and +157/500 OV/OC in that situation, but man does my main GPU get hot (it even uses less voltage; only +37.5mV is necessary for it to be stable at that OC). Even other really hot games don't do that kind of stuff to me like GTA's planning

 

You need to redact about the GT80 cooling indeed then. It is a working solution, but it is not a secret that CPU temperatures are high and hot on most, if not all, GT80s. Granted, I am running on a very hot day that even I am sweating just by being near the computer playing.

I played a mission of witcher 3 just now, max fans. Max CPU temps were ranging between 89 to 94. GPU1 went to 82, and GPU2 67. CPU's max utilization was around 85% and GPU's around 98%.

In general, the GPUs are cool no matter what, in games. Even though I am not overclocking, my GPUs have higher voltage and no power regulation due to Prema's enhanced vbios. On stock vbios, my GPU's usually got around 78 for max temp on GPU1.

The problem is that the GT80 is only dual fan, and the CPU shares exhausts with each GPU. And it is a bit too separated from each fan too. Only witcher 3 has pushed me to this temp though, I generally see games push me to the 80s, but 90s tends to be games that push the CPU.

All in all, I should most likely repaste and see if anything is clogged. I haven't checked in a while. But for now, I will go for a cold drink, I'm melting here with little to no ventilation.

 

It's pretty scary that a weak i7-4720Hq gets so hot in Gt80. Heat Sink / pipes design can not be good on this model. Never been a fan of the design that CPU and GPU share same heatsink / pipes. I see the same problem in several laptop models with similar cooling design ..

 

The 4720hq is hardly weak and before tdp regulation it can consume upwards to 60watts which is normal and even lower than what some MX users end up consuming. 4980hq can consume up to 80watts getting near desktop cpu realm of consumption even if only for a minute.

The CPU cooling indeed is the weak link. I dont mind it because right now i am at the peak hotter ambient temps of my city, and even then it doesnt hold me back with thermal throttle. I do plan on doing something about it because as you said, the vast majority of laptops have inadequate cooling for these cpus. You could say im in the worst possible scenario and still at full performance.

A future revision of the titan should add a third dedicated fan for the cpu, which should eliminate this issue.

Sent from my SM-G925I using Tapatalk

 

Welp, I guess the GT72 is the only decent gaming laptop from MSI then. I just refuse to say "good" until I see a CPU that's fully customizable and a BIOS that lets it run free properly, not sorry. If it were a $2500 desktop it'd never sell under those conditions.

That's... no. NO. My CPU heatsink is a joke and I'm at a higher clockspeed and I can't MAKE my CPU cross 90 even in like 36c days in my humid, stale-air house unless I run linpack or overclock/overvolt and livestream... far less just playing a game. And your GPU? Even the P650SG doesn't get that hot far less the ZM models O_O. Either your GPU1 is really badly pasted, or that cooling is awfully broken.

If you're not overclocking, a 980M shouldn't nearly hit that level of heat. I don't think anyone besides Ethrem (and we know his system is broken) has experienced such heat on 900M series... especially at stock speeds and a massive fan.

Prema's vBIOS fixes the voltage so that it's much more stable than stock, and fixes the power draw that was potentially a problem. Without that vBIOS you might be a bit cooler, but you have no idea how much better your stability could be. I still think such heat is awful.

Yeah, I hope it works out.

57W

at 3.8GHz and stock voltage I can easily consume up to 60W with gaming and streaming. MX users might consume the same or more, far less with an overclock. And that is an enthusiast laptop costing well over $3000... it's still just not RIGHT, even if you don't run into the limitation with what you're doing.

That would require a redesign of the board and entire chassis, so it's very unlikely.

 

Oh yes, I have the notebook on coolers always so there's always plenty of airflow.

Tried again today and 87°C. Wowzers. I'm gonna test the less aggressive overclock without overvolt and see the difference.

EDIT: Temps after about 30 mins of The Witcher 3:

20% OC -> 87°C peak temp at 1352MHz (1.1000V)
17% OC -> 83°C peak temp at 1317MHz (1.0750V)
15% OC -> 80°C peak temp at 1297Mhz (1.0625V)

OK, so which do you think I should go with for long term stability? I'm leaning towards the 15% OC at stock voltage.

 

Yeah, 15% is your best bet, at least until you manage an epic repaste or something like that.

 

Yeah, I haven't repasted once yet. It's just the standard Arctic Silver 5 paste that Scan includes with all orders.

Will keep it to 15% for now. Still better than my previous 980M which was stuck at 11% (1252MHz) at stock voltage and with a significantly lower maximum VRAM overclock to boot.

Thanks for the advice. Gotta love this forum's community. Always so helpful

 

Yeah, ICD or Gelid should definitely boost the cooling. AS5 is okay but not the best. Or you could go balls to the wall and use some CLU on it =D.

 

even an 11% boost is still good

 

NEVER
25% BOOST OR GTFO
*campaigns for OCable nVidia cards*
*campaigns for moving myself to USA*
*campaigns for 3D license and gsync license for 980Ms*
*meows repeatedly*

 

AS5 is okay? HAHA! No it isn't okay and this is the worst choice you can make.

-Gelid gc-extreme
-Prolimatech pk-3
-mx4
-mx2
-ICD7/24(have been hearing it's good but i didn't test it yet)
or for adv users
-Coollaboratory Liquid Pro/Ultra
But not AS5 for God's sake.

 

It's far from the worst: http://i272.photobucket.com/albums/...i7-3770K/MasterChartforbare-dieTIMtesting.png
MX-4 is terrible for the low mounting pressure in laptops
MX-2 is as above
Both MX-2 and MX-4 are worse than AS5 in the context of gaming laptops.

 

I absolutely agree, MX-2 and MX-4 are utter crap, their performance degrades extremely rapidly.

 

I didn't have a choice. That's the default and only option available from Scan!

 

Aware I am. But you no repaste, why I know not.

 

Tell it to my msi gx740 and clevop170em. I tested these pastes in both laptops and AS5 performed the worst. In msi temperature of as5 in the worst case was about 13 degrees higher than the other pastes. I do not want to force anyone to purchase a particular paste, describes just how things were in my case, after that test i never bought as5 again.. I'am currently using gelid and at least once on every two months "repasting" because it loses its performance but beyond that it performs well. peace

 

Repasting once every two months? Isn't that a bit too excessive? You don't even need to repaste in a year, unless it was a bad job or the quality of the paste is bad enough.

 

Maybe your AS5 was bad, or you applied it bad? I couldn't tell you. Gelid shouldn't degrade in 2 months. ICD lasts years.