NVIDIA Kepler VBIOS mods - Overclocking Editions, modified clocks, voltage tweaks

[deadsmiley]

Yeah I'm not complaining. The heat is not fun though. I didn't use Inspector on the stock vBIOS so I don't know if the temperature target in 92C from the factory or not but even high 80s is a bit much in my opinion.

At the same time though, I was doing some comparisons between the 880M SLI and a single desktop 780 Ti or Titan card and the 880M SLI actually has the leg up on both cards a good bit of the time which is impressive given the limited power and cooling in a mobile graphics chip. I wouldn't worry about it if it didn't make the fans sound like a whining, dying animal... Its a creepy sound when these go on full blast LOL! Of course, I'm worried about long term damage as well but I've seen worse come out of nVidia so who knows.

The 880M is definitely the end of the line for Kepler. It is after all, just an overclocked 780M with more (and better, thankfully) RAM.

After running dozens of test runs in Firestrike and 3DMark 11 I am amazed at how Johnksss kept his borrowed 880M running so fast and cool during his testing.

If you flip over your laptop and look at the bottom cover you will see how restrictive the air inlet vents are for the GPU. This is what prompted me to cut mine out and isolate that airflow from the rest of the system. The CPU fan and other air inlets I didn't modify. I am guessing we need *some* airflow inside the case to cool off other components.

I also blocked off a small section of the fan outlet as it blows into the heat pipe fins. There was a gap about 10mm wide at the corner of my laptop. Yours may be different but it's worth a peek.

[Ethrem]

The 880M is definitely the end of the line for Kepler. It is after all, just an overclocked 780M with more (and better, thankfully) RAM.

After running dozens of test runs in Firestrike and 3DMark 11 I am amazed at how Johnksss kept his borrowed 880M running so fast and cool during his testing.

If you flip over your laptop and look at the bottom cover you will see how restrictive the air inlet vents are for the GPU. This is what prompted me to cut mine out and isolate that airflow from the rest of the system. The CPU fan and other air inlets I didn't modify. I am guessing we need *some* airflow inside the case to cool off other components.

I also blocked off a small section of the fan outlet as it blows into the heat pipe fins. There was a gap about 10mm wide at the corner of my laptop. Yours may be different but it's worth a peek.

Yeah I would like to know how he pulled that off too because 90s seem to be par for the course with this chip. Although Alienwares have always had excellent cooling capabilities (with proper fan tables of course) - my M17x can move some serious air due to the fact that a good 50% of the underside of the machine is a steel mesh for air intake.

It is pretty much the same situation with cooling on my machine as yours although they (either Sager or XOTIC) put in foam pads to fill in that very gap that you have with the fins there.

The good news is that I can adjust the temperature target for most games without negatively impacting performance too much thanks to you pointing out Inspector (I've never had Kepler GPUs before, the 260M was my last mobile and 8800 GTS was my last nVidia desktop) so realistically, heat is only a problem for benchmarking... and Watch Dogs until it gets optimized.

I made a 70C/-12.5mv every day profile, a -12.5mv benchmark profile, and an 85C/-12.5mv gaming profile I'm going to test out on Watch Dogs and see how it runs. I'm going to test them both with and without the memory overclock. I know that what it throttles is the memory but I scored 9031 on Fire Strike using the 70C with +450 as opposed to 8700 without it.

I wish we could make custom fan tables for the machines though. 70C wouldn't be as limited if I could switch the fan settings to be in line with the behavior I want for the temps.

[J95]

[MENTION=9518]Lucifer Nymphetamine[/MENTION]

AW M17xR2/R3/R4 - M18xR1/R2 - GTX 880M

- GeForce 327.23 WHQL

- Extract driver with 7-Zip/Winrar

- Display.Driver folder, Copy/overwrite nvdmn.inf

- Nvidia 327.23...International folder\ run setup.exe

nvdmn.inf_v327.23_AW_GTX_880M.zip

[Ethrem]

Firestrike w/ 87C temperature target, +450 memory: NVIDIA GeForce GTX 880M video card benchmark result - Intel® Core i7-4940MX CPU @ 3.10GHz,Notebook P377SM-A

Not too bad.

I did find out that even -12.5mV is not stable when it comes to Unigine Valley. I really lost the silicon lottery on this one. There's absolutely zero room for a stable undervolt.

87C seems to be the best compromise between heat, system noise, stability, and speed. The fan never hits its maximum which is a small trade off for the performance hit whereas 88C will let one of the GPUs pass 91C and kick the fan to max until well after the benchmark or game is closed. That fire strike is a lot faster than my stock vbios, which was 9131.

3DMark 13: NVIDIA GeForce GTX 880M video card benchmark result - Intel® Core i7-4940MX CPU @ 3.10GHz,Notebook P377SM-A

Again, still faster than my stock 12907.

Thanks again for the vbios slv7! Hopefully I can get my thermals under control and see what kind of performance can really come out of this thing but this is still a sizable increase over stock which is awesome!

EDIT: Catzilla went to Catzilla 3* - 15202

http://www.catzilla.com/showresult?lp=274612

It throttled when it got to the fluid and raymarch benchmarks so it probably would have been 200 or 300 points higher if my temps were better.

[erickgois]

I was going 25mv at a time. I am not an expert... the most I went was 1100mv. I was just producing more heat which was slowing the card down so it wasn't helping beyond that.

- - - Updated - - -

Win 7 Pro

I was referring to hardware. I looked back and saw a 870M, but what laptop are you running it in? (Might help to add your setup to your signature.)

Now I got a signature kkk

[6zumlachen]

@Lucifer Nymphetamine

AW M17xR2/R3/R4 - M18xR1/R2 - GTX 880M

- GeForce 327.23 WHQL

- Extract driver with 7-Zip/Winrar

- Display.Driver folder, Copy/overwrite nvdmn.inf

- Nvidia 327.23...International folder\ run setup.exe

what has this driver that is better than the new one? i got the AW 17 new model can i use it aswell ?

ive got the 880m

[mrm2x]

Tried to run 3DMark Firestrike today.

System-Specs as follows:

M17xR4 i7-3720QM GTX880M 240W PSU

im using the modified vBios + stock clocks (993mhz core, 2500mhz ram), video-driver v337.88 (latest)

no change in any setting whatsoever

as soon as i start the benchmark (about 5s) the card starts to throttle for 2-3sec, than goes back to normal for about 10-15sec, starts to throttle for 2-3sec and so on...

its basically the same behaviour as with the standard-bios, except the throttle-time is a bit shorter (id say up to 6-8sec with standard-bios).

max temperature according to nvidainspector is 71°C

during firestrikes tests, the video-driver crashed and restartet itself about 2-3 times (basicially after each test except physics ofc)

question is obvious: what could be the problem here? video-driver, not enough power, problem with the vbios?

any help is much appreciated

[deadsmiley]

ok, maybe i didn't say it clear enough...

it is not your core that is generating the most heat. at stock or over clocked.

Ok, so what IS generating the heat??? I don't mind doing the work, but I am unsure what to do other than what I have done (hacked the bottom of my case, undervolted, isolated the GPU intake air...) Those are all GPU related and have worked well. Is it the memory cooling? I noticed that the heat pipe for the memory is in front of the heat pipe for the GPU. So the air flowing through the GPU heat pipe fins are pre-heated by the memory heat pipe. Throw me a bone dude!

- - - Updated - - -

Tried to run 3DMark Firestrike today.

System-Specs as follows:

M17xR4 i7-3720QM GTX880M 240W PSU

im using the modified vBios + stock clocks (993mhz core, 2500mhz ram), video-driver v337.88 (latest)

no change in any setting whatsoever

as soon as i start the benchmark (about 5s) the card starts to throttle for 2-3sec, than goes back to normal for about 10-15sec, starts to throttle for 2-3sec and so on...

its basically the same behaviour as with the standard-bios, except the throttle-time is a bit shorter (id say up to 6-8sec with standard-bios).

max temperature according to nvidainspector is 71°C

during firestrikes tests, the video-driver crashed and restartet itself about 2-3 times (basicially after each test except physics ofc)

question is obvious: what could be the problem here? video-driver, not enough power, problem with the vbios?

any help is much appreciated

Since your temps are good you could try bumping up your voltage a bit (25mv steps) to see if it stabilizes.

P.S. Thanks for posting your system specs. Please put that in your signature so we don't have to refer back to this post or ask you again.

[deadsmiley]

+100 core, +480 memory, +12.5mv core (1.0125v)

6300, Graphics Score 7138

Peak of 91c on the core... kinda toasty.

NVIDIA GeForce GTX 880M video card benchmark result - Intel Core i7 4900MQ,Notebook P17SM-A

[Robbo]

Ok, so what IS generating the heat??? I don't mind doing the work, but I am unsure what to do other than what I have done (hacked the bottom of my case, undervolted, isolated the GPU intake air...) Those are all GPU related and have worked well. Is it the memory cooling? I noticed that the heat pipe for the memory is in front of the heat pipe for the GPU. So the air flowing through the GPU heat pipe fins are pre-heated by the memory heat pipe. Throw me a bone dude!

- - - Updated - - -

Since your temps are good you could try bumping up your voltage a bit (25mv steps) to see if it stabilizes.

P.S. Thanks for posting your system specs. Please put that in your signature so we don't have to refer back to this post or ask you again.

He might be referring to the VRM's that generate a lot of heat. Those are the little square chips around the edge of the card I believe, or are they the mosfets, I might be confused. Both of those produce a lot of heat when a lot of current is going through the card - and an 880M running at 993Mhz does have a fair amount of current running through it - 105W I read somewhere. Also, with the unlocked vBIOS and the ability to increase the Power Slider in NVidia Inspector, then this allows for even more current to pass through the card. Your VRM's & mosfets (no temperature sensors) are probably running hotter than your core, but whether they produce more heat is debateable though - I would say your core generates more heat, but the heatsink is able to take that heat away from the core more efficiently than the VRM areas (paste has better conductivity than the pads, also the heatpipes are directly above core, hence the lower temperatures of the core). You've limited your overclock due to your core temperatures, and I think that's a sensible thing to do. I don't think this knowledge will really change what you can do with your overclock though, it's just interesting. (You could re-pad your heatsink using high quality pads with the highest thermal conductivity possible, which would lower the temperature of the VRM's/mosfets, but that wouldn't lower your core temperature - if anything the core temperature would increase as more heat would be transferred from the VRM area through the heatsink & then to your GPU core).

[deadsmiley]

+90 core, +500 memory, +12.5mv core (1.0125v)

3DMark11 P9156, Graphics Score 9838

Peak of 92c on the core.

NVIDIA GeForce GTX 880M video card benchmark result - Intel Core i7 4900MQ,Notebook P17SM-A

- - - Updated - - -

He might be referring to the VRM's that generate a lot of heat. Those are the little square chips around the edge of the card I believe, or are they the mosfets, I might be confused. Both of those produce a lot of heat when a lot of current is going through the card - and an 880M running at 993Mhz does have a fair amount of current running through it - 105W I read somewhere. Also, with the unlocked vBIOS and the ability to increase the Power Slider in NVidia Inspector, then this allows for even more current to pass through the card. Your VRM's & mosfets (no temperature sensors) are probably running hotter than your core, but whether they produce more heat is debateable though - I would say your core generates more heat, but the heatsink is able to take that heat away from the core more efficiently than the VRM areas (paste has better conductivity than the pads, also the heatpipes are directly above core, hence the lower temperatures of the core). You've limited your overclock due to your core temperatures, and I think that's a sensible thing to do. I don't think this knowledge will really change what you can do with your overclock though, it's just interesting. (You could re-pad your heatsink using high quality pads with the highest thermal conductivity possible, which would lower the temperature of the VRM's/mosfets, but that wouldn't lower your core temperature - if anything the core temperature would increase as more heat would be transferred from the VRM area through the heatsink & then to your GPU core).

Ok, thanks for the insight.

[Robbo]

Is that a gaming overclock Deadsmiley? Looks like your peak temperature on that benchmark was only a peak temperature due to the short duration of the benchmark portions - as steep temperature curve with hardly any plateau. If you gamed at that overclock with say Far Cry 3 pushing the GPU to 100% load for say a 20min block, then it would be hotter than 93 degC right? Well, it would throttle at that point wouldn't it, you've probably set it to throttle at about 93 degC?

[svl7]

btw i just got a problem when i plugout my PSU and start the laptop it crashes in windows and asking me to change to my iGPU which i dont have due to 120HZ Display...and after 5 sec and some flickers to black screeen it just freeze... i can see the mouse only but nothing works.. i shutdown and put the psu and all work again.

That's odd... did it switch to 3d clocks? I'm not sure how the current driver handles battery mode, if it switches to 3d then it's possible that your system has a very strict current limit for the battery that would cause issues when the gpu switches to the 3d state in battery mode, at least with the modified vbios.

i save the original bios with gpu z am i able to flash it back or will that brick the card

Yes.

Hello, I have some questions. I have a Sager NP9370 (Clevo P370EM) with a Dual GTX 680M SLi 4GB and want to "flash" it, my questions are:

1. If I flash the GPUs, will I be able to normally update it using the Drivers from Nvidia site?

2. The only way to go back to normal is doing a backup of the default files and putting it in a secure local? I mean, if I lose it, I'm in serious problems, right?

1 - Yes, since your system officially supports the 680m.

2 - Not sure what you mean... Restoring the vbios? Yeah, make a backup and then you can flash it back anytime. If you loose it... well, you'll find a copy of the vbios on the web. Or here, if you ask.

Why is your score with single gtx880m so much higher then mine?

I've also the modded vbios from svl, is this because I only got 8GB Ram and I7 4710MQ ?

Maybe the Intel GFX slows my Graphiccard down?

If you're referring to the post two above this, then it's because your system run at stock and the other was overclocked.

Hey, so I have an ASUS G75VX and I've seen on multiple posts that the vBIOS has been unlocked, but all the links are dead / removed cause apparently the poster would rather not provide the link instead of just posting a disclaimer warning people to not use it if they don't know how to use it properly.

Right, I never uploaded a version for the Asus. Need a copy of your vbios for this.

BIOS stuff is kinda off-topic here.

Flashed my 880m's to SVL7's vbios last night.

Thanks for all the work. It's greatly appreciated.

I feel there's something still hindering me though. Even very minor adjustments to core/memory are resulting in full system crashes when I try to OC. I thought even with stock bios you should be able to do Core +135/Memory +500, no? My performance is still very good stock, but I thought I could at least get some more oomph with the unlocked vbios.

Depends on the card... some might not be able to go much above the stock settings with the voltage set in the mod vbios. Also the new AWs are limited when it comes to overvolting. The mobo doesn't allow much more than the stock voltage. It simply won't provide the required power to the card.

what has this driver that is better than the new one? i got the AW 17 new model can i use it aswell ?

ive got the 880m

Nvidia has quite a mess with the more recent drivers. The newest one is certainly not recommendable imo. The suggested driver above is one that works pretty well, but of course it lacks support for the most recent games, so you might just want to try different drivers and find one that fits your needs best.

Tried to run 3DMark Firestrike today.

System-Specs as follows:

M17xR4 i7-3720QM GTX880M 240W PSU

im using the modified vBios + stock clocks (993mhz core, 2500mhz ram), video-driver v337.88 (latest)

no change in any setting whatsoever

as soon as i start the benchmark (about 5s) the card starts to throttle for 2-3sec, than goes back to normal for about 10-15sec, starts to throttle for 2-3sec and so on...

its basically the same behaviour as with the standard-bios, except the throttle-time is a bit shorter (id say up to 6-8sec with standard-bios).

max temperature according to nvidainspector is 71°C

during firestrikes tests, the video-driver crashed and restartet itself about 2-3 times (basicially after each test except physics ofc)

question is obvious: what could be the problem here? video-driver, not enough power, problem with the vbios?

Drivers. That would be my first guess. Try an older one, e.g. 327.23 if you want to bench.

[deadsmiley]

Is that a gaming overclock Deadsmiley? Looks like your peak temperature on that benchmark was only a peak temperature due to the short duration of the benchmark portions - as steep temperature curve with hardly any plateau. If you gamed at that overclock with say Far Cry 3 pushing the GPU to 100% load for say a 20min block, then it would be hotter than 93 degC right? Well, it would throttle at that point wouldn't it, you've probably set it to throttle at about 93 degC?

This is just an exploratory overclock... finding limits the best way I know how with my limited knowledge. I don't run games at these settings as it would indeed throttle due to temps.

[Robbo]

This is just an exploratory overclock... finding limits the best way I know how with my limited knowledge. I don't run games at these settings as it would indeed throttle due to temps.

Cool, that's what I thought. If your temperatures on your model of laptop are significantly higher than other users with the same GPU and laptop model (that also have the modified vBIOS), then you might want to consider re-pasting your card if you want to game at higher overclocks. I can't speak for your system though, because I'm not familiar with what temperatures you should be seeing. Room temperature is a huge factor in this though, and the maddest of the overclockers, like Mr Fox likes to bench his Alienwares outside on the patio in the depths of Winter (apparently!)! (so try to find out rough room temperatures from people when comparing laptop temperatures).

[deadsmiley]

Cool, that's what I thought. If your temperatures on your model of laptop are significantly higher than other users with the same GPU and laptop model (that also have the modified vBIOS), then you might want to consider re-pasting your card if you want to game at higher overclocks. I can't speak for your system though, because I'm not familiar with what temperatures you should be seeing. Room temperature is a huge factor in this though, and the maddest of the overclockers, like Mr Fox likes to bench his Alienwares outside on the patio in the depths of Winter (apparently!)! (so try to find out rough room temperatures from people when comparing laptop temperatures).

I game at stock clocks, -50mv. The temps are in the mid 80's depending on room temperature, as you noted.

Ok, the card won't clock to 993MHz at -50mv. I set it back to 1000mv and it stays at 993 now. <shrug> Running about 85c at the moment.</shrug>

[Robbo]

I game at stock clocks, -50mv. The temps are in the mid 80's depending on room temperature, as you noted.

Well that seems ok, but if that is underperforming vs identical systems to yours, then I'd re-paste if you want better temperatures and/or higher gaming overclocks.

[deadsmiley]

Well that seems ok, but if that is underperforming vs identical systems to yours, then I'd re-paste if you want better temperatures and/or higher gaming overclocks.

I repasted with AS5 (what I had on hand) 2 days ago. I have MX-4 on the way. This thing was running 90c on stock clocks before I repasted, hacked out big vent holes at the GPU inlet and isolated the inlet air from the rest of the case.

Man... Mechwarrior Online takes up 98% GPU at the match start menu. That's kinda crazy.

[Lucifer Nymphetamine]

327.23-notebook-win7-win8-64bit-international-WHQL with nvdmn.inf_v327.23_AW_GTX_880M

that you sent me the following Same thing with the 337.88 driver,

giving following crash tests in 3Dmark, and on games.

anyway thanks.

Accept any other suggestions .... = /

[svl7]

Try increasing the voltage of your card by like 25mv and run again.

[Lucifer Nymphetamine]

Unexpected error running tests.

Workload Single init returned error message: File: device_resources.cpp

Line: 313

Function: class eva::com_ptr<struct id3d11buffer=""> __cdecl eva::d3d11::device_resources::create_buffer(const struct D3D11_BUFFER_DESC &,const struct D3D11_SUBRESOURCE_DATA *) const

Expression: hr: DX11 call failed [-2005270523].

Hardware device removed.

Tried run trine 2 , game not start,bugfield 4, black screen.... just for examples..

DXGI_ERROR_DEVICE_REMOVED: ID3D11Device::CreateBuffer:

same...</struct>

[J95]

Unexpected error running tests.

Workload Single init returned error message: File: device_resources.cpp

Line: 313

Function: class eva::com_ptr<struct id3d11buffer=""> __cdecl eva::d3d11::device_resources::create_buffer(const struct D3D11_BUFFER_DESC &,const struct D3D11_SUBRESOURCE_DATA *) const

Expression: hr: DX11 call failed [-2005270523].

Hardware device removed.

Tried run trine 2 , game not start,bugfield 4, black screen.... just for examples..

DXGI_ERROR_DEVICE_REMOVED: ID3D11Device::CreateBuffer:

same...</struct>

Try increasing the voltage as svl7 suggested.

Post a screenshot of NV Inspector/settings

PSU Plugged into a Surge Protector / UPS (Estabilizador)?

[Lucifer Nymphetamine]

I have no direct stabilizer is in the making, reinstalled 337.88, because the problem is of course not the drivers tried 3 and all present the same'' locked'' in 3DMark tests and games ...

I will download the nvinspector, because it's all original not changed anything

[J95]

Plug the PSU (330W?) directly to the wall outlet. NV Inspector increase voltage 25mV / default clocks.

[Lucifer Nymphetamine]

337.88

Direct 330w wall

Just open the nvidia inspector and took the picture did not change anything