Graphics Card upgrade for M6800 ?

Would a bad vbios cause CPU throttle as well? I have never been able to get the CPU to throttle once in three and a half years with my m6100 so I'm surprised it is doing so now.
What vbios do you recommend?

 

I know I'm late to the party - but today I finally upgraded to GTX 980M, I'm using it with Intel GPU enabled.

What was done:
1) Modding the cooler, got regular NV cooler for M6800, local locksmith got the job done in like 2 minutes with a dremel after he saw GPU and heatsink and what I wanted
2) Played around with thermal pads to get proper coverage, everything has proper contact, used Grizly Kryonout for GPU DIE, no LM TIM for me for the time being
3) Modified .inf files

Temperatures (30 min FireStrike loop):
On first run, temperatures got to around 74C, which is not that hot, FirePro M6100 got even hotter 76-77, but what surprised me was fan speed which went full Master Blaster, did not even know M6800 fans can spin up that loud/fast.

My conclusion was that there was proper heatsink contact due to „acceptable“ core temperature and fan noise can be attributed to fact that M6800 isn’t thermally designed for stock 980M (100-140w of heat) as some forum members said card can draw

Undervolting:
So off to undervolting we go. Extracted bios via GPU-Z, download Maxwell II bios Tweaker and went to Voltage Tables and reduced max voltages from CLK 34 to CLK 60 to 0.975v (Asic quality for the card is 77% and people with 69% reported stabile card on 0.981v so I decided to go a bit lower for the first go). I’ve also edited Power Table and reduced Max value to 147% (125000), left Def at 85000.

Flashing was done using “premas” nvflash which was found here as a part of an 980M_UV.rom upload at someones AW 18 topic post. Other versions of nvflash I found gave all sorts of errors when I attempted to flash.

I've attached flasher I've used, as well as my own edited vBIOS, if you decide to use it, use it at your own risk.
Also are included in spoiler are pictures of what I've changed using Maxwell Bios Tweaker

Spoiler

Stabile undervolt and underclock
So my GPU ended up not being stable while undervolted to 0.95V. Altho Metro Last Light and FireStrike on loop were stable when I was running them for few hours, using multiple monitors (3x FHD) ended up crashing drivers once per hour. Bumping up voltage to 0.962 made it all go away.

I wanted to keep that 0.95 voltage due to overall noise reduction and system temperatures so I've ended up limiting turbo boost to base clock of 1038MHz. If I want I can use MSI Aferburner to bump up the core clock, but with 0.95V max stabile clock is 1100MHz, after that, driver crashes start to happen.

I've lost some performance I know, but temperatures are now under 70, all the time, which is not that much of an issue, but the fan noise was. Now I'm back to levels of noise i had with K4100M when it was overclocked to 940MHz for the core.


Option I changed when reducing the maximum core clock.

Spoiler

Results:
Altho core temps didn’t change that much, drop from 74 to under 68ish in “same” conditions, noise levels dropped considerably, motherboard temperature also went down during sustained load for around 5ish C. And last but not least, average GPU clock went from 1GHz-ish to just under 1.1GHz (talking about sustained loads, used HWInfo to get average reading)

Did some light gaming, I’m mostly playing WoT and WoW and card behaved excellent during my 2h gaming session, temperatures were hoovering around 60C, with FPS set to 60 on 1080p monitor and details finally cranked to the max. Doom and Metro LL are downloading as I’m writing this, as is Witcher series so I can enjoy new games properly. That 3 games together with WoT and WoW will keep me busy for at least a year.

Exact numbers and testing done for temperatures

Spoiler

Firestrike before UV: Total: 8129 GPU:9590
Firestrike after UV: Total: 8238 GPU: 9673
Firestrike after 0.950V: Total: 8318 GPU: 9736 Not stabile when using multiple monitors, upped voltage to 0.962

Temperature testing was done by running Valley on loop in extreme preset with prime95 on 4 threads for 30-35 minutes. I like to combine this two to get a bit more heat over real world scenarios, without resorting to power virus that is Furmark. Also i like to load both CPU and GPU to see what will happen and how entire system will heat up.

Temperature readings were reset 2 minute into the test and then test would run for 30-35 minutes.

CPU is undervolted by 30mV and max multi for 4 cores raised to 35.
Measurements taken by HWInfo

Stock vBios:
Max GPU temperature 75C - avg 74C
Max CPU temperature 97C - avg 92C
Max "ambient" temperature (i belive this is PCH) 83C - avg 75

Fan speeds were maximum, after 3-4 minutes of the test
that means max 4900, avg 4800

GPU had dips all the way to 700ish MHz avg was around 1000ish MHz


Undervolted vBIOS
Max GPU temperature 71C - avg 68C
Max CPU temperature 87C - avg 85C
Max "ambient" temperature (i belive this is PCH) 69C - avg 65

Fan speeds were running avg 3500 rpm which is on par with FirePro M6100

GPU clock was stable at 1126MHz

Optimus and .inf modding:
So after I got noise under control, it was time to see that all aspects of the card work properly. Since I often take this notebook to the office with me and at least few times a month I need battery to last 2+ hours I was imperative I get Optimus working properly.

Since everything was working well I was bummed to see that nothing gets rendered by 980M on internal screen. Thanks to posts from @ Aaron44126 culprit was quickly identified, my .inf mod pointed to [SectionXXX] that did not contain "NvSupportMSHybrid = 1" and/or “NVSupportOptimus = 1”. After quick change and driver reinstall, all was well and soon internal screen was running all the cool things I was able to do on external.

Optimus and battery
Last thing that left me baffled is shutting of 980M on battery. AFAIK my old Quadro K4100M, when on battery and not used would shut off and GPUZ/HWInfo would report 0MHz for GPU and Memory + that little tray icon would go gray and tooltip would be “GPU Inactive”.
Whatever I did to .inf of last available driver (419.xx) when modding almost all entries for 13D7 (I have actually added my line and just rotated section prior to every install, instead of rewriting existing HW ID’s) and whichever 13D7 entry I took i did not end up with that tray icon going gray and tooltip saying “gpu inactive”. I even vent with edited quadro drivers and pointing to M5000M, K5100M, 970M, did edit or two on clevo .inf and all to no avail.

Then it go to me that I had same issue with M6100 which this notebook had when I first got it and due to it was delaying upgrade to W10 from W7 for quite some time due to it as it would take me at least 2h of battery life for nothing. When digging around I found out it was due to WDDM version of Intel drivers, current W10 version, as almost each W10 version bumped WDDM revision which was not always followed by drivers and since Optimus is not the thing used anymore but rather “MS Hybrid Graphics” it caused a lot of issues to a lot of people during early days of W10.

So I went the way DDU-ing both Intel and nVidia drivers, then updating system Bios to lastest version (which actually has a line about updating iGPU UEFI code) Downloaded lastest Intel drivers from Dell M6800 site and installed them. As I was typing this Windows Update offered me some older 980M drivers (from 2017) which I decided to install for lulz and voila now on battery icon is gray, discharge rate is around 20W which means 980M finally turned off on battery and the fu**ing icon is finally gray.

Link to Windows Update drivers that do not require inf modding to instal, take note drivers are from December 2017

Microsoft catalog drivers for 980M with HW ID: PCI\VEN_10DE&DEV_13D7&SUBSYS_15CD1028

Few things I’ll have to investigate and report bac as it might help others with similar systems.
1) Was the system Bios + Intel drivers upgrade the reason that Optimus finally turns off 980M
2) was it older nVidia drivers.
3) Was it the fact that I used too old iGPU drivers with newest nVidia drivers and WDDM disparity was causing me issues.

All in all, it works as it should now, will test with new drivers and for now I’m off to build my Lego notebook stand as I’ve found that when using M6800 LCD as secondary display it sits to low and that actually causes strain on my neck. Will also add few 120MM fans to it to help it cool bottom side of the laptop and get that fresh air in.

 

You are right, something strange is going on if the CPU is throttling. Not really sure what to suggest there...

 

I rather suspect DarkyDark has written a step by step guide to solving your issues Epsilon .......

 

@epsilon72
I've had weird throttling issue when I first turned notebook on after installing 980M - CPU was stuck at 0.36GHz. Turned out my brick was plugged in into cord, that was plugged into another cord with various protection and that was plugged into another cord and then into wall socket. Since I had notifications about power adapter not being recognized turned off i took me awhile to find out whats what and to plug power brick into the first cord. I guess that multiple cords caused voltage drop big enough for adapter to send wrong power output reading to notebook and EC stepped in and locked it down.

Also if you don't use battery for M6800/M4800 (one with the gray sticker) you can get weird CPU throttling and get it stuck at 37W, happened to me when I had M4700 and M6800 disassembled and mistakenly swapped batteries which got my 4710MQ stuck at 37W for a month and I've almost lost my ****

Also when looking back, I believe that during 30 minutes initial Firestrike loop with stock 980M clocks even had dips down to 700ish MHz. Didn't game prior to undervolt to see if clock dips would manifest as stuttering. Noise was just unbearable and just wanted to make sure card is working as intended. I believe this was my FireStrike score before undervolt and this was after undervolt and Power Table max lowering. So, undervolting got me a better score, just a tiny bit, much lower noise and heat.

One thing to know about M6800 design - our PCH is located directly under the graphics card and its temperature can dictate power delivery to rest of the system. One thing i observed on the first summer I encountered with M6800 was that if PCH gets hot, I believe threshold was around 85ish, entire system would stutter. There would be less turbo on CPU and weird clock dips on M6100. Solved this by propping the back of the notebook with legos (2x4 single brick for each bottom side) to get more airflow around the bottom and undervolted vbios for M6100 someone made on this forum.

So 980M can dissapate much more heat than a punny M6100 and seeing as 980M even had throttling problems in various Clevo and AW 18 behemoths on stock vBIOS its no wonder you get it in M6800, they are not designed to tackle power of heat in excess of 100ish W and 980M is known to even go up to 140W in short bursts and that is when the fire nation attacks!

So I'd make a stretch here and say that even tho the GPU heatsink can dissipate enough heat to keep GTX 980M at reasonable temperature level (in my case 74C) PCH was getting hot and both fans went berzerk mode. I did not check PCH temperature during my first loop with stock 980M.

 

Those are some good points and I think what you are describing about the PCH getting hot might have been part of what I got hit with yesterday. I also noticed this morning that the blue LED on my power adapter was not on, and I remembered that one time in the past few years the power plug got particularly hot (since it is right next to the GPU vent) and switched off the power adapter. That may have happened intermittently last night when I was testing it.

Anyways, even if I was able to deal with the heat issues this is still a 4GB card instead of the advertised 8GB so I think I'll still return it.

(unplugging my power adapter for a little bit and plugging it back in fixed it)

Edit: I just looked up the part number for the Samsung memory on the board, and it is indeed a 512MB chip (8 of them).

 

In your place I'd try undervolting, since you already have it assembled. Nothing to loose and you will get proper information about cooling capability.

Will post pictures in my original post of vbios values and end result.

My tip would be to edit yor stock vbios and not to use someone else's.

Looking at the power 980m has, does 4 vs 8gb means that much or is it a matter of principle?

I've got new thermal pads inc next week and will redo my paste job and revisit my temp testing.

Sent from my FIG-LX1 using Tapatalk

 

Reading your 980m walkthrough again above makes me want to try it, even though I've already re-installed my m6100
Do you have a link for the version of nvflash that you used?
If I can get temps down and clocks stable, maybe I'll see if the seller can give me a discount or something. The 980m won't be running any games at 4k but some games like DOOM and Apex have texture settings that need 6GB or more of vram.

 

Editor and example of my bios are in my upgrade post now, as well as some numbers and pictures
TLR Undervolt made it from unusable to usable

I almost never turn on AA in games as I don't like the effect that it does. Image looks only blurry to me after that so i recon I could make do even with 4GB model, as I'm only playing on 1080p/60Hz

 

Normally these cards will have 16 RAM chips (eight on top, eight on bottom) — does yours only have eight? If so, something to definitely look at if you purchase another card.