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

There are many apps that by default try to use dGPU. Just check one by one and switch them to use Intel. At least for me Dragon Center was one of them.

 

I have always wanted the GS65 since it was released and finally pulled the trigger on an open box unit from Best Buy for $840 yesterday. It is the RTX 2060 version with I7-8750H and seems to be in perfect working order. I think it had been sitting around for a while since I thought the battery was never going to charge. Eventually it started charging and I calibrated it last night. I think Best Buy did a clean install also since none of the MSI programs were installed. The only one I think I really need is the Steel Series application. Great size, great screen, great speakers...Really liking it so far.

 

Amazing deal, you practically stole it. Nice find. I made nearly 600 more for the same specs.

 

Amazing price dude!

 

Best Buy has lowered the open box ones to this price. A few others said they got this price and I just got lucky finding one locally. I was a little hesitant updating the BIOS through device manager but it seemed to work ok. I have never had a laptop with a screen like this one. Makes a big difference...I am not a huge gamer but do like playing RPGs like Grim Dawn and Divinity Original Sin 2.

 

Hi folks... just picked up a new GS65 9SD (9750h w/1660Ti) over the weekend deal for $300 off (not as good deal as BigCTMs ) and I have some questions.

I'm not a stranger to heat issues since I'm coming from Razer Blade 7700HQ/1060 (which just died 2.5 years into service mind you, I'm done with that brand) but GS65 seems to run very hot when running benchmarks - with or without undervolt I will hit 95C and throttle with prime95 unless I restrict TDP or turbo frequency. Just gaming, I hit ~72C GPU and max 85-87C CPU with undervolt. Just want to see if everyone else has the throttling persistent with 6 cores on prime95 or Aida64.

CPU undervolt wise I'm currently at -145mV. I can go to -165mV and not crash running prime95 but TS bench errors out until I dial it back down to -145mV (both core and cache).

I've also undervolted GPU as well, running stock 1590 MHz flat-curve starting at 725mV and that helps with heat.

I know someone will tell me to enable cooler boost on but it is not realistic to have to running all the time as they are loud and I'm afraid it will burn the fans out. Also, it looks like people recommend here to uninstall Dragon Center anyway, so I might not have access to cooler boost once I uninstall it.

I can't believe Intel had the audacity to actually list this CPU @ 45W TDP - when I actually limit it to 45W, it can only run at 3.5-3.6GHz!

 

Ok, so I've settled on the following settings (i7 9750H and 1660 Ti). Can anyone let me know if any of these thermal numbers look way off?

-CPU undervolted to -150mV, 6 core turbo boost restricted to 3.6 GHz, PL2 = 55W (was at 90W)
-CPU fan profile at 45, 70, 85, 100, 125, 150%. Maxes out at around 6K rpm (Using Silent option, so don't have access to cooler boost but I don't know if I want to run that anyway for a long time)
-GPU undervolted to 725mV @ 1590MHz (I think it was at 818mV or something at the same frequency). Recorded 71C / 69W running Time spy, happy to be 10W under the advertised 80W load.
-GPU fan profile at 0, 60, 70, 80, 90, 100%

Running sustained benchmarks like Cinebench R20 I'm now topping out at 89C on the CPU and using around 52.5W while maintaining 3.6 GHz flat on all six cores. Running prime95 will still throttle at power at 55W and throttle down to 3.2GHz. (TS Bench and Cinebench R15 only seems to use ~46W and hit 85C.)
If I increase the clocks to 3.7 GHz I see the power limit hitting 55W and CPU reaches 92C.
On Cinebench R20 3.7GHz gives me measly 30-40 points vs 3.6 GHz boost so I don't think it is worth the trade off.

Obviously I'm sacrificing some performance but outside of benchmarks I don't feel anything is running slower.


ETA - sorry, these lower numbers were from Cinebench R15, R20 seems to use more power. Updated with R20 numbers...

 

Hi xLima does optimus still work with the 90w bios?

 

Yes, the only thing that doesn't is mini DP port


Sent from my BLA-L09 using Tapatalk

 

I recently got the same (i7 9750H and 1660 Ti) model as well.

My CPU is at -185mV core and -150mV cache. In case you don't know this already you can push core offset a bit lower than cache and still maintain stability.

My GPU is at 1860MHz @ 800mV which is the highest I can get it to without it power limit throttling. My GPU memory is stable at +850MHz.

I'm using stock fan and tdp limit settings though i'll probably explore these later.

After testing multiple 3 hour+ gaming sessions I never hit any thermal or power limits. CPU never went above 89C and GPU maxed out at 75C. I think the 17% increase in GPU performance on my settings is worth the extra 10W of power/heat. As for CPU thermals, I do get thermal throttling under synthetic loads but not below 3.6 GHz with stock fans.

Timespy

 

Razer's 90w vBIOS still works with the mini dp

 

Awesome, thanks! If it isn't too much trouble, please link me which version you used so I can update post and also install it myself.

Sent from my BLA-L09 using Tapatalk

 

https://www.techpowerup.com/vgabios/208278/razer-rtx2060mobile-6144-190108

That's one for the 2060. I personally have the 2070 model and I use this vBIOS.

https://www.techpowerup.com/vgabios/212115/212115

 

Nice info! I think I have the same BIOS as xLima in my GS65 (ASUS 90w maybe) and the minidp doesn't work!
But in the link provided you I see this:
Booster Clock: 1200 MHz

And the 90w BIOS I have has a 1300MHz Boost clock. Same as all 2060RTX 90w, its strange.
If you try this BIOS, please let me know if it works xLima.

Thanks!

 

With Turing the listed boost clock doesn't really matter. I use that BIOS and it says a much lower clock than what I actually reach- about 1600MHz (this is on a 2070 so it may be different clock speed but same point). It's more dependent on how cool you can keep the chip.

 

Thanks for the reply.

For the CPU undervolt, I usually keep the core and cache the same. I know I can go much lower on the core (even -250mV) but at some point it starts hurting performance even if you don't crash.
Was the 185mV on the core sweet spot where you felt it was just starting to have issues delivering full performance? At -250mV I can see the Cinebench scores suffering.

For the GPU, I'm very happy where it is, because I never exceed 72C even at 100% usage w/o using cooler boost. I really wish MSI put more shared heat pipes between CPU and GPU, so CPU could have benefited from better cooling that GPU is getting.

Yes, same for me too - I don't hit thermal throttling while gaming or just using the laptop with just the undervolt, but it will when running synthetic benchmarks on CPU unless I cap the TDP or max clock frequency.

I was tempted to do liquid metal repaste for just the CPU but I don't know if it is worth it to do that just for synthetic benchmarks... I think this is where it starts becoming what is realistic usage vs. just for numbers with benchmarks...

 

Edit: Apparently the information I had about locked core and cached offset voltages only applied to Intel XTU. Thanks kshnandi for pointing this out.

I recently went further into exploring a separate core value offset. There is not a lot of information on this ̶a̶s̶ ̶a̶p̶p̶a̶r̶e̶n̶t̶l̶y̶ ̶p̶r̶e̶v̶i̶o̶u̶s̶ ̶g̶e̶n̶e̶r̶a̶t̶i̶o̶n̶s̶ ̶l̶o̶c̶k̶e̶d̶ ̶t̶h̶e̶ ̶c̶o̶r̶e̶ ̶a̶n̶d̶ ̶c̶a̶c̶h̶e̶ ̶o̶f̶f̶s̶e̶t̶s̶ ̶t̶o̶g̶e̶t̶h̶e̶r̶.̶

I also noticed that in throttlestop at a point around -250mV core offset the setting stop affecting actual voltage, at least according to HWINFO, a -400mV offset preforms exactly the same as -250mV for me. I don't get any obvious stability issues for either but they do preform worse than less aggressive undervolt settings.

Here's my process for finding the best preforming sweat spot core clock undervolt:

1. Run a stress test (I used aida64) and wait until temps stabilize and short turbo max time runs out. Keep it running.
   
2. Run HWINFO64 note the average core clock value rows in the sensors window.
3. Decrease the core voltage offset and reset the average measurement in HWINFO
4. After a minute or two and read the average core clock value. The longer the more accurate but try to keep it consistent.
   
5. If this average is higher than the last value, repeat step 3; if it is lower or if there are any signs of instability revert to the last voltage offset value and end

I ended up with a -225mV core and -150mV cache. With cooler boost this ends up maintaining a sustained 3.9GHz with my 55W TDP limit. I get my highest Cinebench R20 score of 3077 with these settings.

N̶o̶w̶ ̶I̶ ̶b̶e̶l̶i̶e̶v̶e̶ ̶h̶a̶v̶i̶n̶g̶ ̶a̶n̶ ̶u̶n̶l̶o̶c̶k̶e̶d̶ ̶v̶o̶l̶t̶a̶g̶e̶ ̶c̶o̶r̶e̶ ̶o̶f̶f̶s̶e̶t̶ ̶f̶r̶o̶m̶ ̶c̶a̶c̶h̶e̶ ̶o̶n̶ ̶t̶h̶e̶ ̶i̶7̶-̶9̶7̶5̶0̶H̶ ̶i̶s̶ ̶t̶h̶e̶ ̶m̶o̶s̶t̶ ̶i̶m̶p̶o̶r̶t̶a̶n̶t̶ ̶a̶d̶v̶a̶n̶t̶a̶g̶e̶ ̶o̶v̶e̶r̶ ̶a̶ ̶i̶7̶-̶8̶7̶5̶0̶H̶ ̶w̶i̶t̶h̶ ̶t̶h̶e̶ ̶s̶a̶m̶e̶ ̶T̶D̶P̶ ̶l̶i̶m̶i̶t̶.̶

I'm considering a repast as well. With my new undervolt and under full load I sustain 3.9GHz on coolerboost and 3.6GHz on stock fans. I hope repasting will bring my non-coolerboosted levels up to 3.9GHz as well.

 

Great, thanks - that looks like a good method to further undervolt just the core. I will try that tonight.
It is impressive that you are able to maintain 3.9GHz (almost full perf) with 55W TDP. Running Cinebench R20, I'm already at 52.5W at only 3.6 GHz with -155mV on both core and cache.

My previous laptop was i7 7700HQ which I undervolted both core and cache together, so I'm still in that mindset.
I used to undervolt both core and cache together until I hit BSOD, then I went to the other extreme and started using TS bench to detect instability.
TS bench seems to be a bit too sensitive though - on my laptop anything under -140mV throws errors even though I have no issue running Cinebench / prime95 or just regular usage.

 

I just rechecked this, It seems that Aida64 is a bit less demanding than repeated Cinebench runs. Ten minute average clock (not including short turbo boost) for Aida64 was 3.87GHz. For Cinebench it was 3.79Ghz. These values are the average of the 6 average core clock readouts HWINFO provideds rather than the highest one. Some programs give you the highest core clock when displaying a single value rather than the average of the 6. I was running cooler boost and in an air conditioned room to ensure that all throttling is power related. Either way I think I got lucky with the silicon lottery.

I choose to stay 5mV above where TS Bench gives me errors. I occasionally have to run taxing simulations and data analytics stuff for school and I rather stay way clear of any potential for errors.

What timespy / cinebench scores are you getting btw? It's hard to compare online as there are other manufactures that allow their 9750Hs with power limit higher than 55W and 1660Ti's with limits beyond 80W.

 

Actually, this can be done even in the 8750h using throttlestop. I have mine at -250mV and -150mV. I still run quite hot though. I think my highest cinebench is around 2.8K (Not sure). The clocks start off with 3.9 during short time turbo boost window and then they settle at 3.4-3.5GHz.

 

Personally I use prime95

 

So I had to relax the cache UV to -140mV (from previous -155) as it would BSOD as soon as my core UV went below -165mV.
I think the sweet spot on the core for me is between -200mV to -220mV so far, with TDP capped at 55W. After -230mV it doesn't seem to be able to keep full turbo clocks. I'm still not using cooler boost as that use case is unrealistic for me, but the fans are maxed out (6k RPM) using silent option at high CPU temps.

Running Cinebench R20, if I had the speedstep set to 128 (default) then it would hit 55W and throttle back to 50W, stabilizing at ~3.7 GHz and CPU max temp would be at around 89C - with this setting the score is 2900~2920.

If I change the speedstep to 0 (max perf) then it still hits 55W but then it sits at 55W all the way through benchmark, stabilizing at ~3.8 GHz and CPU will max out at 93-94C. In this case, I get high 2980-2990s, but one time it managed to score 3003. Pretty happy with this as it was done within the boundaries of 55W TDP, which I personally think is the maximum CPU thermal capacity that this laptop chassis can dissipate w/o using cooler boost and be on the hairy edge of 95C (Both backed by our experimental data and MSI setting the default PL1 to 55W...don't think that is coincidence)

I might lower the max TDP to 50W, just so that the CPU doesn't go above 90C in ANY case. I was burned from running Razer Blade 7700HQ hot so I'd actually prefer below 80C if possible but that means either low performance or liquid metal repaste.


I will run Timespy again, but I think mine was around 5700 last time, which is quite a bit below yours but I'm also only running at 1590MHz GPU clock.

 

Seanwee, I tried the same bios CPU power limit bypass you did but only got 3190 Cinebench R20, and 7101 TIme Spy CPU, well below your 3229 and 7505 despite maintaining the maximum 4Ghz all core clock. Did you do any other major tweaks?

I ended up undoing the change as when not using cooler boost / ramped up fans I ended up with about the same scores as without the power limit modified due to thermal limits

Here are my final tuning results; all on the more realistic stock fan settings this time:

Code:

MSI GS65 Stealth-296 (I7-9750H and GTX 1660 ti) with stock fan settings
I7-9750H undervolted -225mV Core, -150 Cache. GTX 1660 ti tuned to max of 1860MHz @ 800mV.

                        Stock       Tuned       Improvement

Cinebench R20           2486        3000        20.7%
Timespy CPU             5780        6611        14.3%

Timespy Graphics        5625        6105         8.5%
Heaven 4.0 Extreme      2215        2393         8.0%

Refresh rate OC         144Hz       159Hz       10.4%

CPU-wise am practically at both power and thermal limits. I will need to both repaste with liquid metal and do the TDP limit bypass trick in the bios to see any improvement here. Even then any improvement would be limited as I am already close to the 4.0GHz all core clock maximum on this locked processor.

GPU-wise I'm at power limit but nowhere near thermal limits. If a way to bypass the 80W hard limit is found I should be able to significantly increase my gains here.

 

These are great results just undervolting CPU and OC'ing GPU - not bad from a laptop that is ~0.7 inch thick and 4 lb.

MSI did a great job with cooling GPU on this laptop. I never go above 72C at max usage, and I believe you said you were at 75C but that is with hefty 270MHz+ overclock. (Actually I am quite impressed it is only 3C trade off for massive boost in GPU performance. I believe you should be close to RTX 2070 Max Q levels? Got too excited, but you are definitely over RTX 2060 performance).
Plenty of room to push until you hit that 87C on GPU!

I wish they designed the laptop with more shared heat pipes between CPU and GPU, so CPU could have benefited a bit from the extra cooling on the GPU side...

 

One thing that I don't understand is how much speed shift plays into this.

Using the same settings below
*Core UV to -220mV
*Cache UV to -145mV (on the hairy edge for me)
*TDP set to 55W on both PL1/PL2

Setting speed shift to 0 will pull all available 55W TDP, maintain 3.8GHz and net ~3000 on Cinebench R20. If It set speed shift to 96 it will slightly drop to ~2980 points. Setting it to 112 will result in ~2965, so not much lower.

However, setting speed shift to default 128 will only pull ~32W and it will stabilize at 3.1GHz, netting ~2474 points.

Can someone explain why there is so much difference? Why wouldn't lower speed shift setting utilize the rest of available TDP at 128?

I've noticed something similar while playing certain FPS games - where if I set the speed shift to 128 it won't work the CPU hard enough and give m only 60~70 FPS where as only dropping it down to 112 will give me FPS boost up to 80-90 easily and it is clearly reflected here with Cinebench R20 scores. Why so much jump between 112 and 128?

 

Well, I am using liquid metal on my gs75 so no problems with thermal throttling there. Im also running overclocked ram so that should give a small boost as well.

I think that the current configuration is fine. The gpu should get the brunt of the cooling so that it stays at a lower temperature to get better performance (clocks increase as temps drop). As for the cpu, as long as its not thermal throttling it's fine.

 

It's by design. An easy way to reduce heat and power consumption. http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-1041#post-10943453

 

I use 63 when plugged in and it gives max performance when I need it and will clock down to 800 when running light tasks. Seems like a good balance to me.

 

I was facing the same problem until I manually enabled speed shift in the bios. I could tell that speed shift wasn't enabled properly because the green sst logo wasn't showing up in throttlestop but it was locking my clock speed to low values like you describe. I originally thought that It was my fault as I was messing with the bios earlier and thought that I accidentally disabled it. I just checked the defaults now and the default is actually disabled. I'm not sure if this should be disabled and throttle stop is just failing to properly enable it in software or if this bios setting is just normally enabled in other bioses in general but just for whatever reason disabled in our model.

Either way if you want to manual enable it the setting is in the hidden 'Power and Performance' bios submenu in the advanced tab. If you don't know already you can access this menu by pressing left alt + right crtl + right shift + F2 in the bios.

I have the same setting on AC. I use 191 for my battery profile so that clocks don't jump up from idle 900Mhz to turbo or near turbo every time I do any small action like open a new tab. That being said even at 191 epp is still ramps up to full clock speed, as it should, a few seconds into a long taxing action.

 

Prime95 is a great choice.

 

Not really. That is a very standard 2060 performance. Mine does slightly better. Folks like @xLima are the one who have extraordinary numbers.

 

I didn't mention in that post but I did further up the thread that this with my GTX 1660 ti model. Competitive with 2060 performance is great for a laptop I got for around $480 cheaper than the cheapest 2060 equipped model at the time .

 

Nice results and even better price.

In regards to me scores it took a lot of hours and tweaks. My highest Cinebench R20 is 3000, but I have only been able to hit that once with the 8750H. As far as GPU scores, I had to flash 90w VBIOS.

I did IMON Tweak, ICCMAX, AC/DC Loadline, Native ASPM and ASPM off, UV is only -68 due to other mod, RAM is just around 3100 MHz (with mismatching RAM- ADATA and Kinston ValueRAM). Custom power plan, SGX disabled and I am sure I missed a few others.

So just consider all the mods when trying to compare benchmarks.


Sent from my BLA-L09 using Tapatalk

 

Well, thats a deal!

 

What is Native ASPM? I have ASPM disabled in Advance BIOS under PEG PORT
Did you try the Razer 90w BIOS to check if the minidp works?

Thanks!

 

Best to disable ASPM as this is a power saving feature.
https://en.wikipedia.org/wiki/Active_State_Power_Management

 

There is 2 areas in the BIOS where ASPM shows up. This may have been changed since more recent bios updates.

One is PEG0 and i forget where the other is.

Regarding the minidp, I have not tried but one user fixed his miniDP with razer 90W on the RTX2070 version.



Sent from my BLA-L09 using Tapatalk

 

I found one in advanced > RC ACPI but it was off by default (ge75)

 

I had a question for you folks. The 2060 is pretty much just power limited in my opinion. It is the sole reason why the 1660Ti is within catching range despite the lower count in Cuda cores. It's Cuda cores are simply juiced a lot more than that of the 2060s. However, in the GS65, the 2060 runs incredibly cool. For me, it averages at around 72-73 at full load with peaks at 75-76. I know @xLima has tried the 90W version, but has anyone tried to put in a higher modded vbios for their 2060? a 115W or something? Or a desktop 2060 vbios with power limit to 100+ watts?

 

Is that 2060 with undervolt or stock? My 1660Ti will go up to 77C stock fully loaded stock but ~72C when I undervolt it running on max stock boost MHz (1590 MHz and no cooler boost)
Both are 2060 / 1660Ti are throttled to 80W so I'd think the heat dissipation would be similar.

I bought my 1660Ti version with 16GB RAM from Best Buy when they had the sale at $1399 but recently they reduced the pricing on the 2060 version (with 32GB RAM too) so it is only $200 more than the 1660Ti/16GB regular price. I think $200 more is definitely worth it for 2060 / 32GB combo. Too bad they do this right after my returning period ended.

By the way, Best Buy is currently running sale on 2060 RTX/32GB version for $1699 right now if anyone is interested. ($200 off, was $1,899) Without any discount, it is the same price as 1660Ti/16GB so no brain-er to go for 2060!

 

72-73C with undervolt along with fans at 100%. The clocks are dependent on games. For games like DotA, I get clocks around 1700-1800MHz, and for stuff like BFV, I get 1450-1550MHz. The card is definitely power limited. But the other problem of flashing the chip with a higher TDP is that the motherboard power supply lines will have to deal with more power than its designed for. While the GPU may be able to handle the extra power, I don't quite know if the motherboard will.

 

Vrms are typically overspecced to reduce their heat emission. Say, a 4 phase 50A vrm may be able to handle 200A but typically people run parts that use 80-100A on them because its more efficient at non-peak loads.

I'd say 10-20w extra would be no sweat

 

Currently the only way to flash a vbios that isn't signed by NVIDIA for the exact gpu (in your case only other notebook 2060s) is by manually flashing the bios chip on the board. I think I am in the company of many others in that I am willing to get the tools to try this, as long as someone else takes the plunge to try it first and document the results.

There are other factors as well, my old desktop had a CPU with a motherboard vrm setup that should have supported it but since I was water cooling and thus didn't have much airflow on the vrms they started to overheat, under preform, and cause power limit throttling.

I wonder why MSI took a conscious decision to not go beyond 80W for any of the offered GPU configurations; the 1660 ti, 2060, 2070 maxQ, and 2080 maxQ are all 80W.

 

Just got my GS65 w/ RTX 2060 last week. Have had it for over a month but I had my bro send it to me while I was away overseas. Still reading from page 25 onward. Can anyone speed me up lol?

So just weekend I replaced all the thermal pads from 0.5-2.0mm sizes with Thermal Grizzly Minus Pad 8, some of the ones preinstalled weren't even touching despite "newer" design with the added heatpipe. Replaced the stock tim with Thermal Grizzly Kryonaut. Didn't go to the LM route atm since I move my laptop a lot. Did a TS -0.150-0.160~ and so a few C shaved off. With the iGPU (had to uninstall Nahimic) I'm seeing a good 5-6HR battery estimate with tweaks and lower brightness you can probably get away with 7-8 YMMV.

Purpose of the laptop? Well now that I have a desktop then most of my gaming has been moved to there so the laptop is more for light/casual gaming but more for productivity, internet, browsing.. media on the go. Might be putting up some VMs on an external to do software testing on and be able to plug in back and forth from laptop and desktop.

Still looking at sleeves or bag that will fit the laptop and possibly the PSU and wireless mouse. Might be looking into skin or cover for it but they seem to be very limited. Might be adding another NVMe in the future as well as putting in more ram, but for now maybe tweaking up the ram, overclock? tighter timings? rewrite xmp? anyways back to reading the rest of the thread lol. Thanks!

 

No need to rewrite XMP, just hold left alt, right control and right shift then press F2 when you're in bios to get into advanced bios.

Once you're in go to Advanced and down to overclocking. Ram overclocking setting can be changed there. Increase uncore offset to +125mv for better stability and you can start overclocking.

You can also remove the cpu power limit allowing it to boost to 4.0Ghz all core.

Go to Advanced > Power & Performance > CPU - Power Management Control > CPU VR Settings > Core/IA VR Settings

Set IMON Slope to 50
Set IMON Offset to 31999
Set IMON Prefix to "-",

And you're all set.

 

Quick question about Optimus.

Since I have many dGPU monitoring apps running, it seems like I have to close all of these to reliability get optimus to work.
-MSI Afterburner
-HWInfo (yes, even this)
-Steam
-ThrottleStop (only if NVIDIA GPU being monitored)
-Dragon Center

If I quit all of these then the power light will turn white. What is really confusing is sometimes the light will turn white w/o closing some of these apps (especially ThrottleStop and HWInfo) but that is not the always the case. Sometimes I have to put the laptop to sleep after closing all of the apps above then bring back for Otimus to work.

I had Razer Blade before, and all I needed to close was MSI Afterburner for the Optimus to kick in. I just love seeing that battery estimated 7+ hours remaining as soon as power is unplugged. On my MSI GS65, it is usually 3 hours because Optimus is not kicking in automatically until I close other apps as well. Then the battery estimation updates to 6+ hours.

I have configured NVIDIA control panel to force using iGPU for all of the apps listed above (and also all other apps that I've seen that tried to use dGPU, even web browsers and Microsoft Photos for crying out loud!) but unless I quit apps listed above which monitor dGPU, optimus is not guaranteed.

I do have NVIDIA notifications panel listed there is nothing listed there. For some reason Dragon Center is still occasionally there even though I set it to use iGPU only in the control panel. Closing the dragon center fixes this, but I have to do this manually. I do have Silent Option running still so fan control is not a problem. What is really perplexing is, I sometimes have to open Dragon Center THEN close it for it to go into Optimus - otherwise it won't even w/o Dragon Center running in the background.

Anyone else have this issue? I can close everything whenever I'm on battery but I'm getting tired of it.

Edited to add - I'm using whatever is the NVIDIA's latest DCH drivers, not the one listed under MSI support website. I will try that and report back...

 

Not sure if they fixed it or released the test version of Nahimic but I uninstalled mine to get Optimus to work.

 

If anyone is wondering, I went back to Nvidia driver 417.77 that was on MSI website and now as soon as I exit Afterburner the led turns white.
Running dragon center in the fore ground will turn it amber, but minimizing it turns it back to white, so now it is working as intended.

ETA - Nevermind. Worked for a while but after a couple of reboots I'm back where I was with newer driver.


ETA2 if anyone cares. The only thing that works consistently is reverting the profile to default on MSI Afterburner before exiting the app. As long as nothing is shown to be actually running in the NVIDIA GPU activity notification, if I reset the MSI afterburner profile settings before closing it then it reverts to iGPU. Sometimes it switches to iGPU even when power is plugged, sometimes it will only do so when I unplug it.

I noticed this happening when I switch Dragon Center profile modes because that seemed to revert dGPU settings to default (and subsequently running hotter when playing games w/o manually applying undervolt again) while at the same time allowing the laptop to use iGPU.

 

I do have a similar list on my old GT70 - had to set those executables to run on Intel to get white light. And some more, too - MailWasher, Windows pics, whatever. It was quite a task to hunt down those that tried to use Nvidia by default even when the preferred GPU was Intel.

 

I had read some of the other threads about bios editing and saw occasional posts of people messing up their computers/too risky for me. This seems pretty low risk/easy to try?--what are the main risks of this?

Thanks,
N123