*** The Official 2019 MSI GS65 Stealth with RTX GPUs Owners and Discussions Lounge ***

Jrey, thank you very much for not only the reply but also all the information you've been posting.

The new drivers just showed up in Geforce Experience tonight, so I was able to update. My clock speeds are jumping up and down like crazy. I did the LM paste, so my temps are pretty good maxing out at 79 degrees.

Just ran Firestrike again: 10,014 - lower than before!

I downloaded MSI Afterburner and it looked like my gpu was going in and out on the readings. Something is wrong with the GPU I presume.

 

That sucks, though at least you will get it checked at. Aside from the paste job, the undervolt makes a massive difference. Once you get it back, make sure to do the BIOS modification to have the least voltage use and thus lower temps.

What sort of issues? Clean installing windows itself will bring its own issues, mainly due to drivers. Your scores for CPU might be fine, but you need to check your intel and nvidia drivers for proper usage as well as power profiles for your machine. When the nvidia drivers do not properly install you will consistently have issues with 3dmark.

 

You can try, as a test, to force the GPU in max performance from the nvidia control panel, and if you leave MSI afterburning running as it seems you did, it should force the GPU to always be on. If after testing it this way it still has low scores, there might be a factory hardware issue.

 

We have a lot of time, don't worry. We will eventually nail it. Though I still think a custom vbios with up to 100w might be the best for us. I doubt we could handle the full 150 at all, without the laptop melting

 

You’re crazy! Thanks for stepping up though. What temps did you see with the 90w bios?

 

Use this.
https://forums.guru3d.com/threads/nvidia-driver-slimming-utility.423072/

Download the driver directly from Nvidia. Also, download the latest iGPU driver from intel.
Also, download CCleaner. It’s free.

This is what you do.

Use the driver slimmer to slim down the nvidia driver. Check mark everything in the first “folder”, and also the NGX and USB C boxes.

Now, put your laptop into airplane mode.

Uninstall the control panels first, then the dGPU then iGPU IN THAT ORDER.

Do not restart.

Run DDU, boot into safe mode, and uninstall the dGPU (clean but do not restart) then the iGPU with restart.

Reboot

Run ccleaner and do the general and registry cleaning. (Note, all your temporary internet files and anything in the recycle bin will be deleted). Run both back and forth until nothing is cleaned on the registry side.

Install the intel driver.

Reboot

Install the slimmed down nvidia driver.

Reboot

Install the control panels from the Microsoft store

Reboot

Retest.

Note, if you’re using throttlestop and dragon center, I would disable dragon center in the task scheduler and reboot.

Let us know what happens. Personally, I see slight improvements in performance with a slimmed down driver.

Also, disable GEForce experience for now. Let’s remove all potential variables.

Sent from my iPhone using Tapatalk

 

About a 4c increase on average for a about 5% increase in base performance.

With my fan curve and laptop elevated, I went from 68/69c to ~73c. I have the thermal headroom to keep the bios BUT I’d rather keep fan RPM down as much as possible.


Sent from my iPhone using Tapatalk

 

Ran some more Time Spy tests, looks like the reason my temps are solid at 79 degrees is that I'm hitting a temperature limit. That's why my GPU is throttling up and down. How did the limit get set so low and how do I adjust it?

 

Are you on auto fan?
If you’re using dragon center or silent option, jack the fan up to 100


Sent from my iPhone using Tapatalk

 

I'm on Auto fan, I'll try it on cooler booster.

 

You have to do bios stuff to do an undervolt? I thought it was just through throttlestop?

 

I did the clean install of windows because I couldn't even open Steam nor update any drivers. It looks like my GPU is throttling down at 79 degrees exactly on the benchmarks. However, I was running Company of Heroes 2 (a 7 year old game) and it was even having a hard time running that. All my drivers are up to date.

 

You may have conflicting drivers.
If your scores don’t improve to within 5% of my stock scores with coolerboost on, try my driver install method a few posts back.


Sent from my iPhone using Tapatalk

 

When you undervolt with throttlestop, it is to compensate for the insane extra voltage that runs on these CPUs. Using the BIOS menu and guides we have to the loadline, you will dial back the voltage of the CPU to what it is "supposed" to be. When you run at this "correct" voltage, you hardly can undervolt anymore, and no need to use throttlestop for that specific function.

 

You mention that your GPU is throttling at 79 C (that is somewhat normal) but not how you are experiencing it. What should happen instead is that your clocks slowly fluctuate to keep temps in check. If you are reaching 79C and that is your real temp while gaming and you have very low clocks, then something is wrong with the GPU. What are your idle temps? Monitor your temps and launch something very "light" on the GPU, a simple game, and cap the framerate down to 60 or something. If under very low %usage you have high temps then that might be the culprit.

 

To use Msi afterburner to OC the dGPU without breaking Optimus, you can set up 2 tasks in task scheduler, first one loads msi afterburner 45 seconds after an user is logged in (Not on start up), second one kills the MsiAfterburner.exe using the following command 1 minute after an user is logged in (Not on start up):

C:\Windows\System32\taskkill.exe /f /im "MSIAfterburner.exe"

This way you can still use your favorite V/F curve settings.

Check this post if you do not know how to adjust the voltage range visible to you in a V/F curve Editor:

http://forum.notebookreview.com/thr...cussions-lounge.815216/page-345#post-10868706

May sure neither of the Launch only on AC / Stop the task on Battery are ticked otherwise if you boot up on battery msi afterburner will not load.

 

Can you tell me where is this option? Couldn't find it

 

Just finished the complete uninstall/reinstall process - now Firestrike score of 13,668, so something still off.

 

Post a screenshot of your temps and the firestrike score.


Sent from my iPhone using Tapatalk

 

Definitely something is wrong. My scores (RTX2060):
https://www.3dmark.com/3dm/33866959
https://www.3dmark.com/3dm/33867183

 

Is there a specific minimum value that you are expecting? I too have the RTX 2060 GS65 model. And my scores are actually lower than your's.
My CPU is running at a -160mV V(offset) and my GPU is at stock.

https://www.3dmark.com/3dm/33889102
https://www.3dmark.com/3dm/33889833

In my opinion, the graphics score can be improved by at least 10-15% by modifying the v/f graph. I'll get to that once nvidia drivers are able to install on this system. Can't afford to break optimus again.

 

After my optimus broke I've reinstalled everything with stock GPU drivers and then simply installed 419.17 drivers from nvidia site.
Couldn't go further with 418 (no supported device problem) but 419 just installed without any tricks and with no problems.

 

Did all the tests,

I ran Guacamelee for awhile, did not see high temps and I capped it at 60 fps. I ran a Firestrike this morning after I followed Jrey's driver process, got a 13,668, so still way below the benchmarks. Do you have any other ideas or do I need to send this computer back.

 

I have the 2080 Max Q version with liquid metal repaste and it's running way below the benchmarks.I'm getting about 75% of the average users.

 

Run the firestrike benchmark again with HWinfo on. I want to see your CPU and GPU temps.


Sent from my iPhone using Tapatalk

 

I did this, go the following scores:

Firestrike 13,668
Timespy - 3,963

So something is still not working right. This is the 2080 Max Q with liquid metal repaste so I was expecting to get somewhere near your scores.

 

Please redo the benchmark with HWinfo so we can see what's going on. Take a screenshot and add it to your post showing both the scores (graphics, cpu, combined, overall) and HWinfo.

 

In PEG0 in BIOS disable ASPM, that adds another 300 to 400 to FS.

Some units come BIOS tweaked from the boutique.

Sent from my BLA-L09 using Tapatalk

 

The GPU immediately jumps to 79C and hangs out there. I ran Company of Heroes 2 ( a 7 year old game) for 5 minutes at a locked 60fps, the GPU still heats up to 79C quickly, then thermal throttles the performance so that I cant even maintain 60fps...goes down to like 30fps.

Again, thanks for your help. Let me know if there's any other data you'd like to see. Do you think there's an issue with the LM paste on the GPU?

 

Looks like you have a heat issue. I don’t think your heatsink or your fans is operating properly.


Sent from my iPhone using Tapatalk

 

M

 

I find the 79 figure even intriguing. Assuming something is wrong with the heatsink and/or fans, why would it only cap at that specific number? I would imagine that an issue with cooling would certainly creep temps even more, but perhaps the GPU is aggressive enough to prevent? Even when I tweaked wrong settings with my machine, I easily could go near 100C despite throttling...

Seems like something might be wrong with the hardware. Perhaps it is best to send it back. You already have enough evidence it is not performing as it should no matter what you do.

 

My fans work...very loud on cooler booster and the CPU levels seem to be okay. Yes, I don't think cooling is being applied to the GPU properly because I keep testing with old games and the same thing happens...temps jump and throttling begins, and down come frame rates. I've sent an email to GenTechPC, hopefully they make it right. Thank you for your help.

 

I believe the Max-Q GPU's cap out at 79c before they start throttling. I recall reading an article about it, I need to find it somewhere...

@superstew24 Your CPU scores look about right. I really think your GPU isn't making good contact with the heatsink. I'd RMA it personally.

 

Yup. This works!. Thanks.

Has anyone been able to play with v/f curves in msi afterburner? It is always blank for me. Even after updating to 419 nvidia drivers.

 

So I finally got my GS65 with RTX 2060 from CyberPower UK. Below is a story with a few questions attached and benchmarks as promised To be clear, there is no LM or any other additions to the standard laptop - they don't do this in the UK...

FYI I am not very experience user, but spending a few days over the web I realized sort of what needs to be done - if there are any other benchmark tests I should rung / other things to do to improve my performance - please let me know!

First thing I did was run a few test right form the box - installed Steam, 3DMark & a couple of other programs, but nothing major. Here are the results (fans @ auto):

• Firestrike - 12954 - CPU / GPU temperatures reaching 90 degrees a few times
  • Graphics - 14761
  • Physics - 13051
  • Combined - 6715
• Timespy - 5454 - CPU / GPU temperatures reaching 90 degrees a few times
  • Graphics - 5581
  • CPU - 4836
• Cinebench - scores around 800-900 which were quite low compared to what I have expected...

Second thing I did was to open up my new GS 65 and install the new 1TB SSD in it. However, you should have seen my face when I turned over the motherboard and saw only 1 RAM module of 16 GB, instead of a promised 2x8 GB modules...Boy I was pissed, as obviously I didn't have another 16GB stick in my hand...which meant I will need to open this laptop again and go through another stressful 30 minutes to install another stick of RAM soon...argh!

Third thing I did was to download a few games and try this beast out. I got Far Cry 5, Metro Exodus, and Anthem. To be honest the performance was not satisfactory in any of them (probably due to single channel RAM and thermal throttling - CPU was hitting 90-93 degrees quite often). All games nowhere near the benchmarks I saw online (e.g. Far Cry 5 should have had 100+ average FPS), Anthem should be running 60-70+ FPS on Ultra.

• FarCry 5 (Ultra, 1080) - Mostly 50-60 FPS, but quite a few consistent dips to 40-50 FPS
• Anthem (Ultra, 1080) - dips to 25-30 FPS in some areas, average around 35-40 FPS
• Anthem (High/Ultra, 1080) - dips to 25-30 FPS in some areas, average around 40-60 FPS
• Metro Exodus (Ultra, 1080) - 50-60 FPS

Fourth thing I did was to undervolt the CPU using ThrottleStop I used the following parameters which I mostly picked up around different threads / throttlestop guides:

• SpeedShift - enabled, value of 63 for gaming
• BD Prochot - unchecked
• FIVR:
  • Turbo Ratio Limits - untouched from standard
  • CPU Core - undervolted to -150V (it BSODs on me at -160 on TimeSpy stress test)
  • CPU Cache - undervolted to -150V (it BSODs on me at -160 on TimeSpy stress test)

See pictures here:

• Throttlestop main screen
• FIVR

Final thing I did was re-do the benchmarks (fans @ auto) They have improved across the board as you can see below:

• Firestrike - 13351 (+400 vs base)
  • Graphics - 14765 (+4 vs base)
  • Physics - 15430 (+2400 vs base) - CPU & GPU reached 90 degrees on this test
  • Combined - 6954 (+230 vs base)
• Timespy - 5574 (+120 vs base)
  • Graphics - 5644 (+60 vs base)
  • CPU - 4836 (+380 vs base)
• Cinebench - scores fluctuating from 1050 to 1215 (+150-300 vs base), not sure what causes such a range... but much better scores nonetheless

See pictures here:

• Firestrike
• Timespy

In-game performance has also somewhat improved, but not significantly:

• FarCry 5 (Ultra, 1080) - Mostly 60-80 FPS, but quite a few consistent dips to 50 FPS
• Anthem (Ultra, 1080) - dips to 30-40 FPS in some areas, average around 40-50 FPS
• Anthem (High/Ultra, 1080) - dips to 30-40 FPS in some areas, average around 40-60 FPS

My main questions:

1. Could lackluster game performance and throttling be somehow connected to the fact that I have only single-channel RAM now?
2. Are my benchmarks in expected range? If not, can you point which one exactly and give your hypothesis as to why this could be the case?
3. Is 90 degrees expected after undervolting as much as -150V? Seems pretty hot for me...
4. Which 16GB RAM stick should I use? Mine came with Kingston DDR 2666 Mhz, but I don't know the exact model (suprisingly nither does CyberPower UK...) - should i try to identify it precisely or would any DDR4 2666 Mhz stick do? Any pointers are hugely appreciated - e.g. if you upgraded it yourself, what did you use?
5. Anything else you recommend to change / test in throttlestop to improve CPU performance / heat issues? Maybe check anything else?

Thanks for the read!

 

I would download HWInfo64 https://www.hwinfo.com/download/ with it you can see your ram model, temps, if you’re voltage limiting (performance limits). The performance limit due to MSI dragon overvolting the CPI probably caused your low cinebench score before your undervolt. I have the GS65 and GS75, both came with Samsung 16gb stick. I have noticed once I put in another Samsung 16gb stick both GPUs increased 5-8% in performance, CPU remained about the same for synthetic tests.

 

I opened that case, turned the laptop on it's side and set the fans to coolerbooster. I have visual confirmation that the fans work, so you are likely right that the GPU is not making contact with the heat sink. I will most likely be sending this unit back to GenTechPC. I will post the solution to let anyone else with a similar problem know.

 

Thanks for your reply!

it says the following for the RAM - "DDR4-2666 / PC4-21300 DDR4 SDRAM SO-DIMM" So I guess I should be looking for "DDR4-2666 / PC4-21300" type of module? I put it in amazon and immediately got a few results like corsair, crucial, etc. Should i aim to find same Kingston module or they can be from 2 different manufacturers?

Also, it says that the clock is 1333 MHz, shouldn't it be 2666 MHz? (might be a stupid question, but see here for details)

 

Would you be able to copy and paste the model number of the Kingston ram - I can look up a matching ram module for you? I’ve never had a problem mixing manufacturers for ram as long as the speed and timings were matched.

The ram is dual channel and reports one channel speed. The clock speed will always report half it’s actual speed. It is running at 2666mhz.

 

Unfortunately neither me nor CyberPower UK knows the exact model, but thanks for the offer! I will try to find it online as I remember how it looks like and now that I know what exactly I am looking for in terms of characteristics, it should be easier.

Thanks for this clarification!

Update - I think I identified the right one - here. It also looks exactly the same as the one inside the laptop.

 

That ram module is ECC, Error Correcting - it wouldn’t be a match and I don’t believe the MSI can accept it. I’m looking up on other forums which Kingston stick MSI uses. Once I find it, I’ll update you.

Edit update : the ram stick is https://www.kingstonmemoryshop.co.uk/laptop/msi/gs-series/msi-gs65-stealth-thin-8re-laptop

 

Sounds like a bad LM application.

 

Do you know if the Kingston ram is using hynix micron or Samsung dies?

 

I recall hearing from a Kingston rep, lower frequency they’ll use Micron, Elpida or Hynix and faster frequency they only use Samsung. To know for sure since I believe it’s “valueram”, can you open hwinfo -> Summary-only selection -> on the drop down selection expand memory -> click row: -> in bold you should see “module manufacturer” that should either have a single letter or a full company name.

 

I'm going to be taking apart my laptop soon to add another 16gb dimm > while I have it apart, considering repasting I don't really want to go liquid metal, I have some Noctua NT-H1 that came with my cooler, I've not really had an concerns with heat, if the issue with stock compound that it can be poorly applied, lower quality, both? it would be worth changing?

 

I have no experience with that paste, but if its better than stock, or really a high quality paste then yes its worth it. After a proper paste job and undervolting you should be good to go for both performance and temps.

 

1) With respect to your ram, just look up the CAS Latency timings from HWinfo and match that along with ram frequency with your new ram. It will work fine.
2) Your cinebench score is jumping around because of thermal throttling. Try using cooler boost. I used XTU to undervolt my cpu by 160mV. I also limited my turbo boost short max power to 55W from 90W. With this I'm able to achieve stable 3.9GHz across all cores. I get 1207 approx in my first cinebench and then it stabilizes at 1136 odd. Despite the undervolt, I still face 90 degrees on the CPU while gaming. For normal loads, I get under 80.
I plan on messing with the v/f curve next in order to improve the GPU and CPU temps.

Also, I'm not a big fan of temps or using cooler boost as I'm afraid it will increase the wear and tear of the same. I disable my turbo clocks for most games and I'm able to game with an avg of 66 degrees on both GPU and CPU while comfortably getting 80 odd FPS in BF V, NFS Payback and DotA

PS: I found that limiting the turbo short power max to have the greatest effect on my temps and clock stability.

 

Have you repasted your CPU?
On Cinebench, i'm able to hit 1240+ consistently with back to back runs on my custom fan curve (0, 50, 65, 75, 85, 100) and hit temps on 88c

 

I decided to do not LM re-paste, I am risk-averse in that particular decision.

I will be opening my laptop again anyway to add another RAM module, but if I see that my thermal issues during gaming session have become better (as seen below), I will most like not repaste at all.

I also made a few additional tweaks:

• FIVR - Limited my Turbo Ratio Limits to 3.9 MHz for first 4 cores and 3.8 for the last 2
• TPL - Turbo Boost short max power - limited it from 90 to 60

My Cinebench scores are 1120-1200, but I've also played Metro Exodus on all Ultra + Nvidia hairworks and was able to get 60-90 FPS, with some areas dipping to 50 FPS. What's more important CPU temperatures stayed at 60-80 degrees with very rare spikes to 87 degrees max.