(Part 2) If U Can Solve This U R King

*UPDATE* Solutions that worked are posted here.

Specs: MSI GT62VR, i7-6700HQ, GTX1070, nv 385.69, Windows 10

Dear Friends,

Thank you for all your support on my first thread. I spent some time to narrow down the root causes of my underperforming MSI GT62VR problem. You are my only hope in fixing this. Here are my 3 observations:

1. Constant fluctuations in GPU clock while gaming, even when staring at a wall

2. GPU load constantly fluctuates between 60% and 90% and never reaches and stabilizes at 99% during graphically intensive gameplay like other owners of my laptop

3. Underperforming CPU: i7 6700HQ was benchmarked 106 times by notebookcheck.net and it scored an average of 672 on Cinebench R15. The same laptop as mine scored 678 on Techradar. My CPU scores 644-655 after a cold boot in a bloat-free environment. Similarly, below average CPU scores are observed in 3DMark

Owners of MSI GT62VR, or i7-6700HQ, or GTX1070, may I know if you experience constant fluctuations in GPU clock speed while gaming? And is your GPU load stable at 99% when running graphically intensive games at high performance power plan?

------------------------------------------------------------------------------------------------------------------------------

Observations 1& 2: GPU Clock & Load Fluctuations

My gtx1070 keeps fluctuating anywhere between 1640Mhz and 1840Mhz every 2 seconds even when not changing scenes or no action is happening. Framerates keep fluctuating a lot as well.

Other users reported steady 99% usage and steady 1840Mhz or 1911Mhz clockspeed while gaming.

Do your games behave the same?

Example: battlefield 3 while staring at wall



Observation 3: Underperforming CPU

Also my CPU is underperforming. My cinebench R15 multicore CPU score is 644, lower than average which is 671 as benchmarked 106 different times by notebookcheck.net. Techradar benchmark of this unit scored 678.



Gaming performance:

My framerates are low compared to all other owners, but still smooth. Example: in Metro map of Battlefield 3, MSI GT62VR owners get 130-150+fps where i get 80-112fps as u can see in this video. Can u kindly confirm your framerates if you have my specs?



My BIOS CONFIGURATION:

Spoiler: BIOS

Troubleshooting Covered so Far:

As some of you know, extensive troubleshooting has been done in the first thread on this forum. All software troubleshooting has been covered and it fixed nothing, example: multiple nvidia drivers where installed using DDU, High performance power plans are set, Vsync Gysnc disabled, xbox dvr disabled, clean windows install...All my system drivers are installed from the MSI product page.

Please help me understand what is going on. I believe I need to have my motherboard and/or CPU and/or GPU replaced but I am having difficulty convincing MSI RMA of this problem. They see it as a non-issue.

 

1.) On pascal that's completely normal
2.) This is the pascal problem u're facing
3.) 655 is a completely normal score, I've seen it millions of times already.

 

I disagree Danish, this is more of My Problem not Pascal Problem. See below and countless other users with GTX 1070 having:

1. Steady GPU clock while gaming battlefield (opposite to me)
2. High framerate range in multiplayer 140fps to 190fps (opposite to me, 80fps to 112fps)

Again, my observation of poor performance is referenced against other Pascal users of similar specs, so it is more of my problem than pascal problem.

Battlefield 3-GTX1070: Extremely stable GPU Clock


Battlefield 3-GTX1070: Extremely stable GPU Clock


Battlefield 3-GTX1070: Much higher framerates than mine, AMAZINGLY Stable GPU Clock


Countless others with GTX1070 and i7-6700HQ or 7700k are getting up to 60fps higher than me in multiplayer plus a stable and locked GPU clock, unlike my fluctuating one. Even when I am staring at a blank wall in a dark room my GPU clock keeps playing and the framerates lower than other users.

 

You cannot compare desktop GPU's to Pascal notebook GPUs.
This is the pascal problem, this will eventually happen for any notebook at some point, the SMC chip inside the die is doing powersaving while gaming hence your GPU usage and low performance. Something is triggering the card to do some kind of powersaving hence doing your performance issue, this is also known as the "Pascal problem", due to the fact that the chip is crypted which prevents people from writing custom vBios and forcing the chip to perform at best at all times you're basicially stuck with this until you figure out how to stop this.

Wether this is hardware or software based is still unknown, considering the fact that you still have full performance on benchmarks like 3Dmark I take it that it's software based.

 

Try copying the EVGA Precision X 16 default skin into your Afterburner skins folder and turning on K-Boost in the Precision skin. See if that fixes it.

 

To help limit further assumptions and speculations in this thread, can the owners of any GTX1070 notebook with windows 10 (MSI GT, Asus, Sager, Clevo...) please report the behavior of their GPU Clock during heavy gaming, and the degree of fluctuation of framerates in one static scene? Reporting on Battlefield 3 would be ideal and really appreciated.

CPU Bottleneck Investigation: Could it be that my GPU load is not stabilizing at 99% while gaming because the CPU is not pushing frames fast enough to the GPU and hence keeping the GPU more relaxed? i.e. CPU bottleneck? My CPU scores in 3DMark and Cinebench R15 are lower than average and might confirm this theory. However, I am not saying this is an i7-6700HQ problem because it performs great in old games with the GTX1070. I believe something might be wrong with how MY CPU is behaving. How can we further confirm this? Test or tool? Please no speculations. Help me generate numbers.

 

Does kBoost work with pascal?

 

Updated Screenshots of my BIOS configuration in original post.

 

Another owner of my laptop on this forum just reported a score of 687cb on Cinebench R15 without any overclocking. I only get 644-655max. He has GTX1060, I have GTX1070. We both have the MSI GT62VR 6RE with exact same processor and motherboard. Even on the 3rd run he gets a much better score than my cold-boot score. All his scores are equal to or above the average of the i7-6700HQ based on 106 different trials done by noteboocheck.net

 

Have you tried the settings what I sent to you via PM? Any change you noticed after applying it?

 

UPDATE: Running BF3 at 4k resolution on TV achieved 3 things:

1. Brought GPU load up to 100%
2. Stabilized GPU clock- much less erratic than running at 1080p
3. Stabilized framerates -but still lower than benchmarks



Why cant I achieve the same stability and GPU load in 1080p? Since framerates are low in 1080p anyway, so If the GPU can work harder, why is it not at lower resolutions?

 

You wouldn't ask these questions if you'd bother to read the answers presented to you.

 

Maybe I missed it, which post?

 

4# post.

 

1. The so called "Pascal problem" is irrlevant to my case. All instances of higher and more stable performance carry a Pascal GTX1070 card, even more so, these instances are generated by the same laptop model as mine owned by other users. There is no statistically meaningful number of Pascal users complaining about my exact same underperformance for us to label it as "Pascal Problem" like you are trying to push everytime. Is a software configuration messing with the power management of my GPU? Possibly yes but this remains in the realm of speculation. I am trying to quantify my analysis as much as possible. Also possible is that my CPU is underperforming as showed by the empirical evidence: low cpu benchmark scores.

2. No, I do not have full performance on 3DMark scores. CPU side of the scores is always lower than average and all other results I've seen coming from the same combo (6700HQ, GTX 1070, 16GB RAM, Win10) are higher than mine. Cinebench R15 further confirms this direction after the 106 benchmarks performed on the i7-6700HQ, giving an average score of 672.

 

Well, try a driver update to nvidia 388 WHQL.

 

Looking at your BIOS configuration your RAM is configured in single channel (you have 16gb in one slot). My laptop is configured in dual channel mode, 8gb + 8gb..i'm not sure if that could cause a perfomance drop in games, maybe someone more expert than me knows about it.

 

For crying out loud NO. Your CPU is operating completely fine and when under load like benchmarks your GPU as well.

Here is your score (quote from first thread):
https://www.3dmark.com/3dm/23367665
16406 Graphics, 9810 CPU, 5860 Combined, Total: 12808





By all means, go blame your CPU, If you don't want to fix your issue then live with it.

 

Makes hell of a difference in CPU bound games and benchmarks. Even I'm running 6700HQ on 2133MHz dual channel.

 

@macmyc , another owner of my laptop on this forum, kindly volunteered his Cinebench R15 scores and Intel XTU readings. At stock settings his unit is similar to most other units out there, and scores much higher than my unit. Here's how we compare to one another:

MSI GT62VR (1) - MacMyc's Unit Cinebench R15 Scores: (He has GTX1060 instead of GTX1070)



MSI GT62VR (2) - My Underperforming Unit Cinebench R15 Scores:



MSI GT62VR (1) - MacMyc's Unit XTU Readings



MSI GT62VR (2) -My Underperforming Unit XTU Readings



Help.

 

You might ask @macmyc for default xtu settings and see if applying that fixes your issue. I can see incorrect settings in turbo boost power in your XTU profile when 45W and 56W should be ideal settings since 6700HQ can't go beyond that. When MSI EC sees your settings it is flagged to have power throttled or even worse.
See mine and yours cloud gate benchmark https://www.3dmark.com/compare/cg/3983055/cg/4152755#

 

Thank you @Vasudev and @macmyc for those witty observations. We are definitely narrowing in on the problem despite those disappointing scores you all are sharing with me You have a GTX980M and still scoring better than me in 3DMark Vasudev

If the lack of dual-channel memory is a problem, then I will buy another 16GB stick, but first i want to make sure there are no hardware glitches in the motherboard and CPU and GPU.

Please tell me more about the comparison between the 2 XTU readings above. I did not understand what you mean.

 

Set the following in XTU:

1. reference clock to 100MHz
2. Turbo boost power max to 45W
3. Turbo boost short power max to 56.250W or 56W.
4. ICC max to 70A and System Agent ICCmax to 15A {Alienware has ICCmax to 100A and System agent Icc max to 12A}

 

I know those are the right values, but I am scared to apply them since it is such a big change for my current CPU settings. Is it safe to change from 200W all the way to 45W and 56W etc...? I don't even know how my current values were set so horribly in the first place. Can I have a second opinion please?

 

Then how did you change it to 200W before? There will be no issues you just need to reboot after applying. Apply the profile at once what @macmyc sent you.

 

Yeah it is, our CPU won't use more than 56W and it will stabilize on 45W mostly, this happens because it's a locked chip.
The 200W limit makes sense in cpus like the 6820HK where overclock is possible and if done it's essential to keep the frequencies up without downclocking on load.

 

I never downloaded and installed XTU before, let alone change any CPU settings. I only downloaded it because fellow forum members asked about its readings. i have no idea how my max turbo power got set to 200W and how my Turbo Power window is set to 0.25 seconds etc...

 

Optimal Turbo boost time window is 8 seconds on 56W and 12 secs on 45W. Me and @macmyc have set it to 28 secs.
XTU is confusing at first and too much bloated. I prefer TS. You can just unzip TS i sent you earlier and see the magic. Just exit XTU before running TS and set the XTU service to manual.

 

Macmyc, this is my system info in XTU. Can I please see yours as well to make sure we have same hardware? I tried loading your default XTU profile into mine but it said "profile invalid".

 

I manually adjusted 3 settings in XTU (i7 6700HQ) to bring them to the default settings other users have:

Turbo Boost Power Max set to 45W (was 200W)
Turbo Boost Short Power Max set to 56.25W (was 200W)
Turbo Boost Power Time Window set to 8seconds (was 0.25s)

I hit apply, restarted laptop, and ran Cinebench R15. I got even lower score than before. Did i miss anything?

 

Set that reference clock to 100MHz as well and exit XTU and use the unzipped TS folder I sent you.

 

Get another stick of RAM for dual channel. That'll solve your low CPU performance. Dual channel also makes a difference in Battlefield games.

Also you should probably try to improve your GPU temps. Getting over 80C at 50% GPU usage in BF3 1080p is not great.

 

Thanks. I will eventually buy a second stick of RAM.

Repasting and cleaning was done twice this November. MSI did it for me 1st of november then the computer shop last week. They said nothing is wrong with cooling. But i still reach 87c gpu and 88c cpu cores while gaming. they dont see it as a problem because temp stabilizes at 86-88C which is under critical temps. But then my clocks throttle during gaming, frameates are low and fluctuating, and as u said gpu load is not even high. How can i convince them that HELLO i have a serious problem here?! It is still under warranty. All they do is run furmark at MSI repair center (or at least ask me to do that whenever i call RMA), and just because it does not crash during furmark, they say nothing is wrong and ship it back....

What people at MSI repair center also dont know is that Nvidia drivers have capped the clock speed of the GTX1070 to 1440Mhz when running furmark therefore it never overheats, fluctuates, nor does it yield poor performance as it does in games. Furmark is no longer a valid stress test in 2017. In games, the gpu tries to reach its maximum factory clock around 1848MHz (or 1911mhz) but it cant. They simply dont know that and cant analyze deeper than basic testing by the book.

Let alone explaining to them my CPU has below avg results in all benchmarks. It is a huge headache trying to communicate such complicated problems to them. Feeling helpless and stuck with a lemon.

 

Yes. All it does is change some PowerMizer related registry keys.

 

I'll try this out, if it works as it should on pascal then you fullfilled one of my wet dreams.

 

For your GPU issues you can try setting a custom Voltage/Frequency curve in MSi Afterburner.

 

*UPDATE*

The Witcher 3 behaves differently and is much more stable than previous games. At ultra settings 1920x1080p and hairworks OFF, here are the results:

- GPU Clock relatively stable
- GPU load stable at 99%
- Framerates not fluctuating erratically. Stable between 70-80fps inside first room, drops to 60s outdoors.
- GPU temp stable at 87C first 30mins. Still High. Fans RPM is maxed (Cooler Boost ON). Below video is taken in first 5 minutes so temp went up to 87C afterwards.

Except for the high temp, this is what I want in Battlefield 3, Ryse Son of Rome, and all older games. In those older games, GPU load keeps fluctuating bw 50% and 70% and so does the clock speed and I get lower framerates than peers.

 

You should really get that second stick of RAM ASAP.

 

The GPU load on older games does fluctuate because they can't put 99% load on the GPU or they are more CPU bound so if CPU isn't fast enough you may not get those extra frames that can ramp up the load on the video card. For example if i run GTA IV on stock graphics i will get as maximum GPU load 50% while with a high end graphic mod like ENB installed i can easily reach 80-99% because of the extra load.

The witcher 3 is much more demanding than Battlefield 3, please remember it's a 2011 game and that GPUs have evolved, they're more powerful and they shouldn't struggle in older games (there is always the exception of games that are more CPU bound like battlefield) so you will always see a lower usage on them, especially in your case where the cpu is already weak and kinda struggling to give you extra power for more frames.

 

Gee I wonder why....

Better try to fix CPU then. Gotta be the cause.

@macmyc
Ryse Son of Rome is pretty demanding, not as demanding as witcher 3, but it's up there.

 

Macmycs, Good Day. Please remember that I am comparing my performance to other notebooks with GTX1070. In Battlefield 3, many GTX1070 users have a much more stable GPU clock than mine, much higher framerate range, and a higher and more stable GPU load. So my issue with BF3 and other games from 2012 and 2013 still hold. My second issue is the below average CPU performance in benchmarks if you compare it to scores generated by this same laptop model. In a nutshell, those are my 2 remaining areas of improvement surrounded by question marks.

 

Yeah i understand that, those users having a stable clock on battlefield got your same CPU aswell ?
Can you share the graphic settings of Battlefield 3? So we can at least try to put more load on the GPU by tweaking some and see if that changes anything.

 

Some of the comparable systems running Battlefield 3 have the same CPU as mine and some have a better one, but in all cases, the i7-6700HQ should not struggle at all with BF3 and should never act as a bottleneck in this old game.

My graphic settings of BF3: I just choose the ultra preset in game, and 1920x1080 resolution. No customization.

I tried running 3 youtube videos in the background while playing BF3 but that did not affect the GPU load at all, framerates dropped by 5-10fps. So not sure.

GTX1070 users (where are you guys show yourselves please) get 130-160fps in multiplayer BF3 while I get 80-130fps, depending on map and location in map.

 

Multiplayer games can be really demanding even if it's an old game and one core is stressed much more than multiple cores our 6700HQ, i'm afaik, will struggle. I've tested it in rainbowsix and some other games. Also, if you look again at your BF3 video, CPU2 is getting around 70% usage while in The Witcher 3 it doesn't (and TW3 is also demanding on CPUs in some areas) because it stays around 30-50% like the rest of the cores. I hope you are understanding what i'm trying to explain.

You should check every graphic settings instead of choosing the ultra preset, usually there are guides that explain which setting impacts more the CPU or the GPU and that can really help you out. I've managed to get my fps stable between 90-120 on rainbowsix by adjusting the CPU intensive settings. Presets can help you half way but you should always hop in and get the best configuration for your hardware manually if maximum perfomance is what you're aiming at.

edit:

this is outdated but it may help a bit

https://www.geforce.com/whats-new/guides/battlefield-3-tweak-guide#9

 

I almost guarantee your remaining issues will be fixed with dual channel RAM.

 

Including the 87C GPU temperature under load? this dual-channel thing must be magic.

By the way can anyone tell me whether dual-channel RAM has any effect on CPU benhcmarks such as Cinebench R15 multicore? (as opposed to no dual-channel). My Single Core benchmark is spot on like everybody else but multicore is below-average. Would activating dual-channel RAM improve that?

I wonder if I should buy a second RAM stick that has the same CL (CAS Latency) as my existing stick or not. My current 16GB stick is Kingston with timings 17-17-17-39. The one below is 14-14-14. Would this be a problem if I add them together?

Kingston Technology HyperX Impact 16GB 2400MHz DDR4 CL14 260-Pin SODIMM Laptop Memory HX424S14IB/16

 

Well except the GPU temps. That you gotta fix on your own since the MSI techs seem to be incompetent. As long as it doesn't void your warranty to do the repaste yourself with a good quality TIM.

Dual-channel RAM always has an effect on CPU benchmarks and CPU-intensive games, such as Frostbite and CryEngine titles like BF3 and Ryse.

Just get another RAM stick of the same capacity as your current stick, if it's higher frequency or lower latency it'll be limited by the slower stick and match settings automatically.

 

BTW did you just start The Witcher 3? Play until you get inside Novigrad City, then you'll see the same problems with single channel RAM.

 

I think if I was going to choose a game to prove CPU as definite problem I would have run and then compared my system to similar/same systems in GTA 5. If you look at scores for it across different GPU's they are very close due to CPU bottlenecks especially once you get up into higher graphics settings and resolutions.

 

On single channel mode the max RAM freq you will see 2132/2133MHz. So dual channel would give extra boost in frequency upto 4266MHz so your CPU can maintain upto 4.2GHz w/o RAM bottlenecks(provided you have an HK chip). 2400MHz RAM stick is a waste of money since 6700HQ doesn't support more than 2133MHz. So, buy a kingston with 2133MHz.