GTX 980M Benchmarks thread

I agree with Mr. Fox. I've always been an advocate of 3DMark 11. Let's see some of those!

 

Notebook Check : NVIDIA-GeForce-GTX-980M

3DMark 11 - Performance GPU 1280x720 = 11817
3DMark 11 - Performance 1280x720 = 10575
(MSI GT72 - 4710HQ - 1038 / 2500MHz - 8192MB GDDR5 - ForcWare 344.00)

 

So, it performs like a heavily overclocked 780M? Doesn't sound right. I think those results are too early for accuracy. Drivers have improved, etc.

The 780M scores 7.5k~ stock and more than 10k overclocked. Mr. Fox had over 17k with 780M SLI.

 

Yeah, that's weird. Which driver are you using? I'm on the latest Nvidia ones. That could be the issue.

 

I updated stock one to newest, but it was 2 weeks ago ... Dunno if nvidia released newer version already

 

Because 11 is cpu bound and running at 1280x720.

 

^^ It shines much more on extreme preset or FireStrike.

 

Then we're on the same one. Weird. Do you have the i7-4810qm? Because if so, why the hell am I getting nearly 10% slower speeds than you?

 

It was on i7-4860hq ... That's weird as soon as i get a new one i will test it again

 

Hmm, that seems like it would have an effect but not 7-9%.

How about the RAM? Same as my sig?

 

Yes ,exactly the same ...

 

Windows 8.1 or 7?

 

as soon as i have my laptop from scan i will be benching the hell out of it! 980m SLI and i74810 cpu.. they are saying begining of december now before they ship tho...

have clevo had a delay on the new systems with the 980 sli?

 

It's finally here! Clevo P170SM-A!

Any requests? I'll try to fill them. So far temps have been AWESOME.

 

holy crap no offense but 3dmark 11

 

P10789
with Generic VGA(1x) and Intel Core i7-4700MQ
Graphics Score
12614
Physics Score
7581
Combined Score
7440


FWIW: my i7 is overclocked +200 mhz via Intel Extreme Tuning Utility

 

So how come it's same as my 3630QM?

Hint: Your CPU is underperforming. You should be getting 8000+ Physics Score.

 

Not sure. I actually checked Intel Tuning Utility and it seems that the CPU is having a hard time hitting the 3.4 all core 3.6 single overclocks...


But- it seems that it hasn't peaked over 75 degrees on GPU or CPU.


Processor

Processor
Intel Core i7-4700MQ
Reported stock core clock
2,400 MHz
Maximum turbo core clock
3,491 MHz
Physical / logical processors
1 / 8
# of cores
4
Package
PGA946
Manufacturing process
22 nm
TDP
47 W

 

GG Haswell.

 

It seems so, but since the cores are low temp so far, I want to see if I can get a i7-4910mq... Wonder how that would run.

 

Check your power and current limits if overclocking haswell. You can easily be limited there. I could not hit a full overclock on the 4810 without drastically increasing both.

I can't remember the specifics, but I am sure others here can help. My laptop is in for repairs right now.

Edit - sounds like 980m owners are quite happy. Lots of performance and not much heat. All without shrinking the process.

 

I don't remember the specifics, but I recall needing in excess of 120A to push 4.2GHz on my 4900MQ, otherwise it simply doesn't even boost to that speed. Another thing I noticed was even at stock speeds, clamping the current limit to 71A from 95A resulted in a 5% reduction in XTU benchmark score. The difference is pretty reproducible so it's definitely not a one-off random chance thing.

 

can you do firestrike.
some gaming FPS? shadow of mordor, watch dogs, ryse, etc...
I almost convinced that sager's gpu chips are not up to par with Asus so far, even though it has double the VRAM

 

Like others have mentioned, looks like you need to increase the allowable current limit, and/or get rid of the limited time constraints allowed spent at max boost. I don't know if it's the same for the new gens of Intel CPUs or whether its relevant to your model of laptop, but I used Intel XTU and Throttlestop together to allow max boost on my CPU up to 56W indefinitely (no time limit constraints spent at max boost). Here's the link to the process I used, as I said I don't know if some of it will be relevant to you (some of it definitely isn't!), but might give you some pointers - Throttlestop might be the key for you:
http://forum.notebookreview.com/ali...-2960xm-cpus-m17x-r3-without-modded-bios.html
(Section 1 of the first post in that link)

 

I ordered a i7-4900MQ last night, so I'll hold off on the OC stuff until I get that- but I do plan on trying to get 4GHz 4 cores on the 4900 so I'll be using this- thanks!

 

What makes you convinced of that?

 

Kroez and Vitors benchmarks kinda showed me that.

Also I had my hands on a G750 870m 3gb and Sager NP8278s 870m 6gb. and the G750 just blew the sager away like how kroez 980m 4gb beat out vitors 980m 8gb on shadow of mordors.

Not sure why the case is, but to me, it just seems that asus makes better quality chips, as to why they only sell half the amount of Vram on their units.

But I'm hoping that youre able to fix that power draw problem and prove that Clevo chips are as good as Asus, if not better.

 

I did a TR bench yesterday, maxed out but with FXAA not MSAA (will rectify that later). Ended up with an average of around 60FPS. I'll upload the screenshot from my home PC later.

I am concerned about the same thing too. As long as the difference is only 5% as suggested in SoM, that's fine, but any more and I'd be pretty annoyed.

Ryse ran between 50-60FPS maxed out for me which is about double the 880m performance which is nice.

 

Are you sure you only got 60fps on your Tomb Raider benchmark? I get avg 67fps with my sig in my system even, and that's with everything maxed (including shadows - shadows don't get maxed on the Ultra preset - you have to set them manually in Advanced Settings to change Shadows to 'Ultimate'), except for FXAA only (like you). Mine's at 1600x900, but you're not running it a 3 or 4K right?!

EDIT: Apologies, I forgot about TressFX - I had that disabled, and you probably have that enabled - that's why you 'only' got 60fps then!

 

XTU will allow you to set the 4900MQ to 4ghz across 4 cores, but it won't run at those speeds. Under load the CPU runs out of the power needed to hold those speeds. The CPU has unlocked multis but not the necessary unlocked power controls to set it high enough to reach those speeds. At least this is what I have found and I've had my 4900MQ for over a year and haven't been able to ever get it working at those speeds under load.

 

Set your Dynamic CPU Voltage Offset to -90 or -95mV and also overclock. Depending on your CPU you should be able to hold the +200mhz overclock and also run undervolted. It will allow you to reach higher clocked speeds before hitting the power limits. Then you should be able to get 3.4ghz X4 all day long. I was able to do it with my old 4700MQ when I had it.

Your mileage may vary, and you may need to run it at a slightly higher voltage than -90mV. I run my 4900MQ all day at -90mV stable. It shaved off a few degrees off hot Haswell.

 

My 4900MQ disagrees with you

To be fair this seems to be dependent upon the laptop manufacturer, and whether they allow users to override the default turbo power lock. My Clevo never came with a power lock it would appear, so my 4900MQ can turbo indefinitely at my chosen speed as long as I supply enough current and power, and don't run into thermal throttling.

 

Yeah, running 1080 and with TressFX on. Everything else set to ultra manually. I think it was 60. Can't remember. The bench dropped to 48 when the camera pulled in close but also went up to like 75 at points. Still seems worse than benchmarks I've seen for the 980m in other places though.

I mean, I can finally play the game with TressFX on and at 60 FPS (which I've been waiting to do), but it's a shame that it still dips.

EDIT: OK, what the actual f*ck, scroll to the Tomb Raider bench: https://www.youtube.com/watch?v=fl62HC83rqE#t=92

980m inn the MSI never drops below 62FPS. My Sager was at 48FPS at its lowest point on the same bench. Averaged out to 60 but still, what gives? I'm losing 10-20% performance at least.

 

Throttlestop?

 

Don't forget to monitor your CPU & GPU frequencies when you do this kind of testing - it's what dictates the performance. GPUz for graphing history of the GPU (sensor window can be enlarged to fill screen to show a lot of history, but only can be resized when you actually have the Sensor tab selected - bizarre!). HWMonitor (or was it HWMonitor Pro) can log your CPU (& GPU) frequencies and even produce graphs automatically.

 

Guess I'll do that first thing once I'm home. Appreciate the help.



Above is the Tomb Raider bench. Lowest FPS was 46, average of 62.7. All settings maxed, including Tress FX, FXAA. GPU Load was at peak, don't know why it's not showing, wavered between 98 and 100%.

EDIT: Doesn't seem like it's an issue with the clocks. GPU load is completely f*cked in Ryse now. Jumps from 60% to 90% for no reason. It stuck to 99% earlier.

 

Hmm, I don't like this...
344.00 drivers first, 344.48 second

Per Scene Stats:
Scene Duration (seconds) Average FPS Min FPS Max FPS Scene Name
32.46 78.4 31 116.26 Welcome Center
7.09 77.33 25.2 107.44 Scene Change: Disregard Performance In This Section
21.72 91.5 42.34 114.06 Town Center
8.08 90.5 42.39 118.48 Raffle
9.1 121.02 59.05 148.67 Monument Island
3.53 127.61 111.77 135.07 Benchmark Finished: Disregard Performance In This Section
81.97 89.84 25.2 148.67 Overall


Per Scene Stats:
Scene Duration (seconds) Average FPS Min FPS Max FPS Scene Name
32.61 70.9 18.14 113.51 Welcome Center
6.64 57.53 10.3 75.49 Scene Change: Disregard Performance In This Section
22.27 71.8 8.72 100.4 Town Center
8.14 72.16 19.6 122.4 Raffle
9.12 99.05 22.37 131.9 Monument Island
3.03 104.39 77 145.81 Benchmark Finished: Disregard Performance In This Section
81.8 74.59 8.72 145.81 Overall


High performance, only thing that changed was driver.

 

Doesn't Ryse have optimisation issues anyway which would account for your GPU usage - vague recollections of that being talked about - pretty sure notebookcheck.net did a benchmark review of Ryse where they make some conclusions if you wanted to google that.

Clocks look good for your Tomb Raider benchmark. 60fps does seem low, I just ran a test at the same settings you ran at (but 1600x900 resolution), and I got 44fps avg with 33 minimum, so you only got 36% higher framerate - maybe Tomb Raider doesn't perform well with Maxwell, try running without TressFX because I remember notebookcheck.net got 107fps when they did so with the 980M (it would be a comparison test for you).

 

I'll check out TR without TressFX now.

Also, here's the thing about Ryse - it was getting a constant 50-60FPS at sitting at 99% load yesterday. Now it's f'ed. Nothing's changed. In fact I tried reinstalling drivers with a full wipe and it's still a problem.

EDIT: Well, 114 is my max FPS for TR. Average is 92.3. Still over a 10% deficit if you're talking about average FPS compared to the 104 listed on Notebookcheck at equivalent settings.

 

Ah, 92fps, that tallies again with your previous result and mine again - still just about 36% more performance than mine. Did you monitor your GPU clock rates again during that benchmark run? One reason you might be getting significantly less than the 104 listed on notebookcheck is that they run it at the Ultra Quality Preset, the Ultra Preset doesn't fully max out the shadows - I'd just check you ran at the basic Ultra Preset (no custom settings). Use the latest NVidia drivers, but I think you mentioned you were using the latest ones? If you find you're still getting consistently lower performance that the other 980M machines out there, can you return your laptop for a refund & buy a different model or make? You've tried uinstalling the graphics drivers, using DDU to clean the remnants of drivers, and then reinstalled the NVidia drivers - you've tried that?

 

Oh damn, I totally had shadows set on Ultra. That explains it then - it's a big performance hit. I'll try again once I'm home but I imagine it'll close the gap.

And yes, latest drivers installed through DDU and removing stuff in safe mode etc. Seems to fix my HDMI issues but Ryse is still being weird. Which is a shame, because I don't want to replay it but it's doing the same thing that BF4 did. And if every AMD-preferred game does that, well, that's less than ideal.

I should try out Crysis 3 again. Both it and Ryse had been going through the GPU load madness until I thought I'd fixed it with a driver reinstall. Now that Ryse has gone crazy again, I should try out another Cryengine game.

 

I googled Ryse, and there's loads of people on Steam forums talking about low GPU usage, it seems like it's a game issue & not a problem with your GPU. If you run Tomb Raider again with the shadows at a lower setting (just the Ultra Preset), then see if you get close to the 104 fps; if you do, then I don't think you have a problem with your GPU or system. When you graphed your GPU frequencies you were at full boost the whole time, that's good!

 

Here's the thing though, it was running fine two days ago. On the same PC. With no settings changed. It was the first game I tested and was at 99-98% usage constantly. Same with Crysis 3. Then, yesterday, it tanked. I imagine Crysis has done the same but I forgot to check. I could get 54 FPS more or less maxed out at 1080 and Ryse looked and ran beautifully. No idea what's up now.

But yes, the GPU frequencies were running at constant lock during TR so I imagine it's not too bad.

 

If you want to get to the bottom of the differences with Ryse you'll have to investigate further I guess. Were you running the Ryse benchmark, or just playing the game when comparing? The benchmark is reproducible, the gameplay might not be.

 

There's a benchmark?

And I'm reproducing the first 30 seconds on each level. Easy to do. I loaded up the Britain level when I first got my 980m and it was 54-60FPS locked, full GPU load. Played through it for 10 minutes. Didn't deviate.

Did the same thing yesterday, GPU load would frequently drop down to 60% and would yo-yo up to 99% before dropping. FPS would go from 40FPS to 60FPS when standing still and the camera wasn't moving. Super uneven.

There's no reason for that amount of variation. I'm just spinning the camera around on the level starting up and walking forwards.

 

I'm with you. I got really wild variances on Bioshock as I posted before- only change being a driver...
Now I'm wondering about my card, and why. I'll wait to start really tearing into the benchmarks once my 4900mq arrives. I feel like the 4700mq (especially without it even hitting normal clocks) has something to do with it. I just really don't want Clevo's 980's to be less powerful since I don't want a soldered CPU.

That new GT80 though...

 

Any of you guys with 980m or 970m able to test Evil Within for me? Free demo is on Steam now.

 

I didn't even change driver though. This happened from one day to the next with no settings being changed on my PC. It was weird, Done a clean reinstall of my drivers and that hasn't helped.

 

Edit: I see what your saying. Derp on my part.

 

Re-installed Windows, fresh, back on the provided 344.00 driver (No OC on 4700 this time)

Per Scene Stats:
Scene Duration (seconds), Average FPS, Min FPS, Max FPS, Scene Name
32.33, 108.42, 24.69, 222.72, Welcome Center
7.08, 92.10, 19.01, 136.89, Scene Change: Disregard Performance In This Section
22.18, 144.44, 34.07, 216.20, Town Center
8.06, 140.18, 26.91, 212.14, Raffle
9.05, 187.78, 60.14, 253.62, Monument Island
3.02, 196.43, 149.23, 273.33, Benchmark Finished: Disregard Performance In This Section
81.72, 132.01, 19.01, 273.33, Overall

Scores are better, but still a bit troublesome. Is Infinite that CPU bound? Or is the 980M GTX?
Some of the scenes still have quite a bit of choppy frames to them. Tearing in the middle of the screen.