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

@Neothane @Ekami When you run BFV use msi afterburner and set it to view the temp of each individual CPU core. If even a single core hits 90+ the general CPU will be reading 90+. Generally your cores should all be within ~5 degrees of each other varying back and forth. I had one core that was always spiking above 90 while the rest maxed around 85ish, usually hoving around 82, so I had to get it sent back. Granted it probably got messed up in the first place because I chose the slow ass free shipping, so it sat sideways in a truck for a week.

FWIW they sent me a 2day shipping label, repaired it the day the received it, and sent it back 2 day, all free, so I only lost use for 5 days. I guess it is part of the risk with LM.

Also @JRey I've flashed to 90w but it's still being power limited to 80w. I'm thinking this is probably because I'm still using Dragon Center? EDIT: NVM it was dragon center, had it set to comfort mode instead of "sport."

 

I ordered a GS75 from them on February 16 and emailed with questions and haven't heard anything either. I know the 17" ones have come out later but I've been surprised to not get a response to my questions. Normally I wouldn't care but these laptops cost enough that the service matters to me. Fingers crossed that both of us will hear back soon and get our orders!

 

So I ran a number of tests, and what I have noticed is that the spikes are at the beginning of the test ramp up, and then it settles to some degree (in terms of throttling), however, I still get 92C temps.

I have included the stats and chart below for Auto fan. My spread between CPUs is 11C! (80 for the lowest and 91 for the highest). It topped out at 92C during the test. Look at the throttling...it maxed at 20%

So, I played with the fan curve and made it:

135 - 125 - 110 - 100 - 100

This basically turns the fan up at the first part and then settles down to 100% fan when test is running longer. This stopped the peak 92C and my max hit 87C, but there was still a minor amount of throttling.

 

Yea that gap is way too big IMO. Prob a bad LM paste or a long shipping time which can mess up the application if laying upright. Hopefully they do a quick and free return and fix.

 

You need to run on cooler boost 150% from the start
Let the fan get to 6k RPM and start test.

The reason it throttles at the start and then cools down is because the PL is hit and the CPU downclocks after sustaining a TDP>45W for longer than 28 seconds.

Assuming no BIOS tweaks or TS PL slider modification.

Sent from my BLA-L09 using Tapatalk

 

You need to run on cooler boost 150% from the start
Let the fan get to 6k RPM and start test.

The reason it throttles at the start and then cools down is because the PL is hit and the CPU downclocks after sustaining a TDP>45W for longer than 28 seconds.

Assuming no BIOS tweaks or TS PL slider modification.

Sent from my BLA-L09 using Tapatalk

 

I don't think it's the Power limit that's throttling for you, if a core goes over 90, it thermal throttles. You can check HWINFO on whether it's thermal throttling when you run your stress test of choice. After I did my own LM repaste, 1-2 cores still thermal throttled under load. I did the below BIOS changes to reduce temps and completely remove thermal throttling:

- I set IA loadline for AC/DC down to 1 (therefore I did not do any further undervolting with ThrottleStop, although JRey suggests a further -40mV is still possible)
- Set VR Current Limit to 800
- Raised the thermal throttle point (PROCHOT) from 90 to 95

After that your limiting factor should be the CPU cores power limit (45W) under load.

Core variance of 8degs or so is normal. You're looking at max max ctemps as well which can be misleading if you've only run it for a short while; 1-2 CPU cores can boost to 4.1Ghz on occasion for lighter threaded apps, which can make your max core temps look skewed. Let it run with a variety of loads and see if it flattens out over at least 10mins.

 

Thanks a lot. I believe we could probably open up the laptop and check if LM is there... but I'm very afraid to do that especially considering if there is LM in there. Donald from HIDEvolution contacted me back and told me in their tests the laptop never reached 90C... So I dunno what is happening here

 

Btw guys how did you redeem the game bundle which are offered with the laptop (Anthem/Metro Exodus/BFV), I registered my laptop on MSI but I couldn't find the bundle and MSI told me to contact HIDEvolution and ask them... Then HIDEvolution told me it was MSI that I should contact lol. Should I throw nvidia in the loop? Thanks

 

You had until the 28/02 to do the purchase I believe. There is a part of MSI website that has promotions

Sent from my BLA-L09 using Tapatalk

 

What's your ambient temp? Also depends what kind of load they put it under. Prime95 with small ffts will more than likely put you over 90 even with an undervolt.

 

Ambient temp is around 20C. Prime95 is indeed one of the worse bench I could experience lol. I used the aida64 benchmark, here is a capture with almost all options from the stress test:

 

Ah, so that thing is running a GPU stress test at the same time. I mean, yea that's probably equivalent to running prime-blend + prime small-ffts + furmark, it's going to push your thermals to the limit: That's not a bad result tbh, par for the course if you have not yet under-volted.

When I get home I'll run the same and I'll tell you what I get.

 

Thanks! I can rerun other kind of tests. The GPU test is actually stress testing the Intel GPU. Should I try to untick some options and rerun the test again? Do you also have LM repaste in your unit?

 

I just reran the tests but only with "Stress CPU" ticked for 6min. This time no throttle, the temp would also very rarely go above 82C. I feel fresh now. But I'm still curious in comparing my results with others, just to make sure.

 

Just as an update on mine ordered from GentechPC on 20th of Feb in case anyone is considering them: I spoke with Ken on the phone 3 days ago and he said the laptop would ship in 2 days, which it apparently hasn't...quite a disappointing experience overall, but still hoping that it does eventually arrive in good shape as I've configured it.

EDIT: Shipped

 

Hi Guys,

New owner to MSi GS65 w/RTX, and this is my first gaming laptop ever. I followed most of the tweaks on this thread so far, and it seems to be working pretty good. One thing I couldn't figure out though is the CPU boost clock speed.

I undervolt it and did the CPU stress test in XTU, and I do not see any thermal/power throttling. However, when I started to play games (Shadow of the Tomb Raider), the CPU will only sustain max 3.1GHz during gaming session. According to MSi's afterburner OSD, the CPU/GPU temperature is only in the mid-70C, so I doubt it is thermal throttling. My Cinebench R20 score is around 2500, is this comparable to your experience?

A quick search online, shows what I am experiencing is "Power Throttling." Some say this can't be avoided since i7-8750H is just that power hungry, and 45W TDP is just not enough. Is this true? How come some YouTube videos showing MSi GS65 (w/GTX) that has the same processor does not have any problem sustaining 3.9GHz while being benchmarked? Am I missing something here?

One of the BIOS tweak can help faking the total CPU power draw, is this one safe to do?

 

I did my own LM re-paste and I also have undervolted in the BIOS plus raised thermal throttle threshold to 95degC in the BIOS. The under-volting I reckon helps as much as the LM re-paste so I do recommend that if you haven't already.

Ok so I ran the AIDA64 tests. My ambient temp was 24 degC, laptop flat on desk:
1) Just "Stress CPU" -- After 7mins, my max core was 79 degC and max CPU package was 80 degC. Max temp variance between hottest and coolest core is 5 deg C.
2) Stress CPU, FPU, cache, system memory, GPU(s) -- After 7 mins, my max core was 93 degC and max CPU package was 95 degC. However, what's happening here though is that the cores all hit their power limits within 10 secs and then they all go back down after they get power limited and stay around 83-88 degC. It's weird coz I don't get that overshoot on Prime95; they just get power limited before they can shoot to 90+ degC.

Anyway, basically I think it's safe to say your LM re-paste is fine, IF my laptop and re-paste job can be considered the median standard of course. It is more or less I reckon, I've seen a few anecdotal examples incl. it being similar to what JRey was getting. Doing the under-volt should help you get closer.

Another thing to note here is that combined AIDA64 stress test is insane, it stresses everything, incl. BOTH the Intel iGPU and the Nvidia GPU!! (max 77 degC). I measured 195W draw at the socket, that's waay more than what Prime95 and Furmark can do together. However this might not be ideal if you want to test system stability as you're hitting all sorts of power limits so it might be worth testing your under-volt/overclock in isolation first. For reference HWINFO was showing other power limits being hit aside from the individual core power limits, namely: IA Package level PLs, IGPU PL ,IGPU Max VR voltage and fuses and, worryingly, power, thermal and "Reliability voltage" limits on the Nvidia GPU...

 

EDIT: @Mikhailovich Liovsky @Neothane Nevermind Mikhailovich is right I was looking at the max CPU temp spike which, while a little strange, doesn't mean it's a bad LM application. The average temps are well within range of each other. But also strange the general CPU temp was 89 while the cores were low 80s.

 

I could not find a single mention of "creaking" battery in this whole thread. Is it really so that MSI fixed this area in the 2019 models?

 

And one more thing: I am bit confused regarding the performance of 2060 vs 2070 MAX-Q.
It seem that the consensus is that the 2060 model is more bang for the buck. Sure. I could also find comparisons (notebookcheck.net) giving evidence that the 2060 can be in some scenarios even faster (up to 10%) than the 2070 MAX-Q, which is surprising. I was wondering, what kind of headroom does the 2060 model have given the smaller power brick and slightly narrower bus? Is there any?
Shouldn't the 2070 MAX-Q deliver more performance thanks to the bigger brick and more shaders, and even more so, if one would sufficiently undervolt it? Or is it like both thermal throttle long before they reach the 70/80W power limit?

 

For what it's worth, the stats on notebookcheck seem to be flawed - one of the systems (out of very few) has a desktop processor paired with the 2060 which would definitely affect the numbers. I could never tell whether the 2070MQ was actually faster or not.

 

So I have done some experimentation with my laptop. Currently, both my CPU and GPU are both undervolted. Also, I have configured my GPU to run at around 1450MHz at base boost clock voltage. With this setting, I am regularly able to avg 80W on my 2060 when I allow my CPU to turbo (I usually play my games with turbo disabled). Even if I allow for higher boost clocks through the curve settings, the GPU is usually not able to cross 1450 because it hits 80W (with a load of 95%+) at that clock speed. With these settings, my timespy graphics score comes out to be approximately 6050. Ideally, a properly cooled laptop (that can sustain default voltages, like the one in notebookcheck with the desktop CPU) should be able to achieve this score by default.

From my understanding, Max Q chips are essentially power limited chips compared to their their normal versions, thereby forcing the card to use much lesser core clock speeds. So, a 2070MQ has around 16% more cuda cores, but is essentially running at a slightly lesser speed. At the same time, since it has a larger number of cuda cores, the amount of GPU boost that a 2070MQ [80W config] does is lesser than that to a 2060 because it'll hit the TDP quicker for the 2070MQ [I might be wrong here. It'll be great if someone can verify this on their 2070MQ system]. Therefore, while the 2060 can reliably work at 1.1x of its design spec, the 2070 can probably run at maybe 1.07x its design spec. Taking all the ratios into account, a properly cooled 2070MQ is probably around 10-12% faster than a properly cooled 2060. In most games, thats a 6-10FPS difference.

 

Don't know about R20, but with R15, I get 1215 score. Others with LM get more.

About your clocks, there seems to be some setting issue there. Make sure that you are set to 100%fans/Coolerboost and run the stress test in XTU to confirm that your CPU cores hit 3.9GHz for at least a small period of time. If not, does XTU flag thermal limit/power limit? If I keep my CPU at stock config, I instantly hit 90C with cores at 3.9 for under 10 seconds before stabilizing to a lower value. With an undervolt of -155mV and a lower value of short period turbo max power(55W), I'm able to sustain 3.9GHz forever with temps jumping between 76 and 86. No LM by the way..

 

My CPU can sustain 3.9GHz in XTU's stress test (no thermal/power throttle) without any problem, but it won't while gaming. I don't have LM, but I did undervolt it by -150mV and fan setting to 100%. However, I will notice power throttling while gaming or during Cinebench R20 benchmark (XTU will flag power throttling). It will jump to 3.9GHz, but it will back down and settle at 3.1GHz (gaming) or 3.3GHz (Cinebench). The OSD temperature will fluctuate in the mid-70 to low-80.

When you say your CPU can sustain 3.9GHz, was it in the XTU or gaming? i will try Cinebench R15 and compare with your score when I get a chance.

 

My CPU can sustain 3.9GHz during XTU. While gaming, it jumps around 3.6GHz to 3.9GHz with package temps of 87C-91C. I'm sure my CPU both thermal throttles (because of the high temps) and power throttles (because of my 55W short max turbo power) at the same time. The thing here is that if I didn't have the 55W power limit, my CPU clock variance would have been much higher. Can you check if your clocks are higher in games if you limit short max turbo power to 55W?

I just ran a R20 and looks like I have a similar scores with my clocks settling at 34.GHz towards the end. My score came out to be 2597.

Edit 1: Also, what games are you playing?

Edit 2: Its also very interesting to note that you don't thermal throttle in XTU stress test. Can you check your temp logs to see the max Core package temp and the avg value as well for maybe a 10 min run?

 

How did you go about the GPU undervolt? I've seen people using MSI afterburner's frequency/voltage graph, but when I bring it up clicking ctl+f no line shows up and no dots are there to move around. It's just an empty chart. Any idea what I'm doing wrong?

 

I tried setting the PL2 to 55W. It will hover around that wattage for 28seconds at 3.5GHz, and lower down to 45W (PL1), which brings the CPU clock to 3.3GHz. Changing the PL2 didn't effect my gaming performance.

My R20 score is 2583. I can confirm no thermal throttling happened during benchmarking. Did you change anything to PL1 and/or short time boost duration?

I am playing Shadow of the Tomb Raider.

You said you undervolt with -155mV, which means you didn't do the IA AC / DC loadline, = 1 tweak correct?
I didn't do anything to the BIOS since I am new to this, so I am staying away from tinkering the BIOS for now.

Edit: Did you do a fresh install of windows? Will this help? I did uninstall Dragon Center, but could there by any hidden process/services left behind?

 

Anyone know if there is a discount code for gentechPC or HIDevolution? Looking to buy another GS65 but with 2080. Thanks!

 

Yes the fact that you have dragon centre installed means your AC/DC Loadline is 180, your TDC is enabled, your ICCMax is 628 (I forget the exact number).

I will link my post in the other thread with the differences between fresh windows reinstallation and stock recovery image.

Sent from my BLA-L09 using Tapatalk

 

Looking for some help, I repasted and I'm pretty sure I made the GPU worse, when running Valley (Extreme HD) three times at the end of the third run I'm getting about 62C on GPU with fans set to max. I think I was getting 60C on stock TIM, but I didn't make a note or screenshot. Can someone with 2060 repeat this and tell me what temp they're getting? Run Valley three times with fans set to max and make a note of GPU temp at the end (it takes that long for the temp to be stable). I just don't feel like opening this again but thinking that I screwed this up will bother me to no end.

CPU is fine, for this I did a screenshot and get 5C lower temps or so. 2865 score in R20.

 

Thanks a lot for these useful infos!! So here are my results after downvolting at a stable -0.90mv with XTU:


That's already better than my previous -20% with the same tests .

 

Post statistics for reference please

Sent from my BLA-L09 using Tapatalk

 

What kind of statistics? Btw I had to rollback to 0.070mv for undervolt as the TSBench was showing some errors below that value.

 

Statistics tabs in the AIDA, benchmark. No worries though.

Sent from my BLA-L09 using Tapatalk

 

OK new tests with the -0.070mv undervolt and the statistics. Ambient room temp is ~24C, dragon center shift is set to "Sport" and Fan Speed to "Auto".



Does this look decent to you for a LM repaste?

 

Your core differential looks perfect. I would run with the IMON slope to see if how temps go but looks pretty good.

Sent from my BLA-L09 using Tapatalk

 

Awesome, great to hear! For the IMON slope I don't even know what it is. I prefer not touching things that I have no clue about aha.

 

My BEST score in FS on a GS65-8RF with 32GB RAM (2*16)
https://www.3dmark.com/3dm/33698352 - OCed +210/+300
AVG FS were 14036 (OCed) - STOCK Settings FS 12792

My BEST score in FS on a GS65-8SE with 16GB RAM (1*16)
https://www.3dmark.com/3dm/34345569 - OCed +220/+250
AVG FS were 14332 (OCed) - STOCK Settings FS 13136

NOTE the difference in Physics score with Single vs Dual channel RAM (At least I think it is attributable to this)

 

@Ekami
GS65-8SE
Fans @ 150% (Both)
AirCon set at 20C
Aluminum X-Stand

Stock paste, pads. AC/DC Loadline 1, -74.6 UV, IMON Slope Tweaked.
16GB (1*16 stick)
TDC Disabled, ICCMAX 232,

Spoiler: 5 MIN AIDA64 - GS65-8SE

Prime 95
SMALL FFTs

• CPUSupportsAVX=0
• CpuSupportsFMA4=0
• CPUSupportsFMA3=0
• CpuSupportsAVX512F=0
• CpuSupportsAVX2=0

Spoiler: 6 min Prime95

 

Install the latest Nvidia drivers. Install MSI Afterburner 4.6.0 Beta 14.

Nope. I didn't touch the bios. No fresh windows install either. I think we both are doing something wrong here. @Thysanoptera seems to have gotten a much higher score. Probably need to replicate his settings.

What kind of pasting did you do? LM?

At what benchmark settings? Extreme with resolution to 900p? With these settings, my gpu goes to 66C within a minute (I'm running 1455MHz on core clock at 700mV). My ambient temps are 28C-30C.

Your CPU score seems to be really high. Could you kindly share the configuration? Did you do the Bios tweaks?

 

Just Kryonaout. Extreme HD in Valley, fans to full (cooler boost in Dragon center), but my ambient is ~22C - and that's a huge difference plus you have overclock. I just wanted to see if it is worth for me to bother repasting again - if 62C with full fans after couple of runs is acceptable and if my mind is playing tricks on me with 60C on stock paste.

I have the bios trick with power offset only, -140mV undervolt and fans set to max. But in this new Cinebench the CPU is thermal throttling - if your ambient is 30C and mine 22C then I will have much higher score. The computer has 2060, 2 x Kingston HyperX CL15s, Evo 970 2TB + Intel 660p 2TB.

 

CB R15: 1220
CB20: 2937 and recently 3000


Sent from my BLA-L09 using Tapatalk

 

You're not running this in a freezer by any chance? Your temps look awesome, I hit 90C and throttle, the best I can do is 2900 in R20, but in R15 I'm getting 1260.

 

AC was on 21C, but unsure of actual ambient temperature. Stock pads/paste.

150% Fan, AC/DC Loadline 1. Hence low UV. IMON Slope edit.
@Thysanoptera


Sent from my BLA-L09 using Tapatalk

 

Actually, the only difference is the ambient temps. From my testing, the GPU actually runs cooler with the overclock to 1455 at 700mV compared to the standard curve it shipped it. What amazes me is how your CPU is performing more than 10% better than mine. What are your TimeSpy/Firestrike scores?

My TimeSpy Graphics - 6050
FireStrike Graphics - 15600

What astounds me more is the graphics score @xLima is able to attain in FireStrike! 17000! Whoa!

 

Would you be willing to make a screenshot of HWinfo CPU voltages while the R20 is running? Like resetting the average/max/min at the beginning and making a screenshot before the benchmark ends. I'm wondering what's the end result of this Loadline and UV.

5629 and 14985 respectively, but that was bone stock, right after I bought the machine, no tweaks at all. And @xLima is another 4% better on CPU, with stock paste job. And I actually did drop good 5C on CPU after my own repaste.

 

Here you go.

This run was had a score of 2912, just turned AC on for a bit so room was not as cool as before.


I did PEG0 ASPM disabled. Which also helps the FS Graphic Score.


Sent from my BLA-L09 using Tapatalk

 

Thanks! But could you attach the screenshot in higher res, the text is not quite readable.
EDIT - actually I think the resolution is ok, but jpg compression very high

 

Hey guys,
I just measured my battery life with PCMark 8. This is very far from what I expected. Is this expected results? (Notice I went crazy and upgraded to 64gb of RAM so maybe that's what is sucking all the battery). I also checked while running the benchmark if the program was run under the nvidia GPU and it was not (although the light close to the power button was orange).