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

What.

That's not how this works. That's not how any of this works.

Here is a GT62VR 6700HQ running 2400MHz DDR4 just fine:

Spoiler

 

Absolutely not, double channel got 50% more bandwidth (verify by aida64 bench or MAxxMem bench or IBT same size, or by games on anandtech article) on every cpu/ram intensive task Vs. single channel ram mode (same on triple and four-channel ram mode by different cpus).
Bandwidth, not speed.



Can you list the gpu driver that you also try?

Edited for wrong quote, plus found driver tested:
"1. Nvidia Drivers: I used DDU to try various nvidia drivers but that had no effect whatsoever on the type of problem I am facing. 381.89 were working great for 7 months up until the problem happened in October. Then I tried 382.05, then 388, and now settled on 385.69. Again no change. I am aware of the need to restart after setting power plan to high performance in nvidia cp, thanks.


So dear OP, other than mentioned thermal issue, because 86°c isn't very good, could you install a debloated version of Win10 (LTSB 1607 here) or better Window7x64 (hail to old king) and see if the issue still persist?
If yes, I vote for power adapter or gpu failing

 

Are you asking me? I don't have a GT62VR. My specs are in the sig. I'm just trying to help the OP.

 

well no.
You already fixed his issue to be honest. He just doesn't realize it yet because he's so fixated on fixing something that aint broke to begin with.

btw thanks, it worked like a charm, never seen FPS this stable before, downright frightning. You fulfilled my wet dream

 

This " is it a CPU or GPU problem" shouldn't be so confusing!

1) In your previous topic you mentioned BF framerates under 100 and also considerably lower framerates on all games (I presume that at least some titles are mostly the GPU intensive types at higher settings), this would indicate that the problem is GPU related.

Yes, an underperforming CPU would have a negative performance impact in some CPU bound titles, or at low graphics settings in most titles, but the performance impact at high graphics settings (in games that are heavy on the GPU), would be minimal.

Also, your slightly lower CB15 shouldn't translate in much lower FPS (double digit % lower) at high graphics settings, unless that game is CPU bottlenecked (this would mean that you couldn't get significantly higher FPS in that game with the same CPU, even if you had a GTX 1080Ti, and shouldn't have considerably lower FPS even with a gtx 1060).

You can clearly see in NBC reviews that most GPU intensive titles aren't highly affected by the CPU performance (unless the clocks drop too much). There are laptops that can get away with a CPU that doesn't even hold any turbo boost (or have even lower clocks than the base) because the cooling system is inadequate (or they are limited by the firmware, probably because of the same reason... the cooling system), and perform similarly (in quite a lot of games) to those that can hold high CPU clocks.

The same is true for CPU OC-ing in games, there are some titles that would benefit from OC-ing, but you're mostly GPU limited (unless you lower the graphics settings, enough to be CPU limited) in the rest of the titles.

2) As you probably noticed, the GPU is underperforming (hence the big improvement in FPS when you applied an OC in MSI AB).

What I don't understand is... did you already found a fix for the GPU issue (other than OC-ing with AB), because it appears that you are mostly focused on the CPU (that is slightly underperforming in CB15 which should have a lesser performance impact in gaming... even higher FS scores don't always translate to a similar % performance gain in gaming)?

You can test how much does the CPU clock affect those games by testing with turbo boost on (if the CPU TB does hold in gaming) and compare the results with TB off.

The GPU issue seems to be similar/related to the "Pascal problem" (where the performance is lower at higher clocks when not enough voltage is supplied for those clocks).

3) Why would adding more ram solve your problem, since your laptop performed at the expected levels for some months and I presume that you didn't remove a stick of ram when the problem began to occur?

4) You previously mentioned that, before you sent your laptop in service, you could regain the full performance with a shift+power off and leaving it unplugged for some hours.

The EC is responsible for some some power management stuff and your problem seems to be related to something that triggers a power saving feature, so this could be something triggered by the EC (though I don't know in what way it would affect the GPU).

You also mentioned that you updated the BIOS, but did you update de EC firmware?

Did you try resetting the EC?

There are previous records of msi laptops that behave oddly after a BIOS update and is recommended (by MSI) that you do an EC reset after a BIOS update (or when you encounter odd behaviors that could be related to the BIOS and EC).

 

1.) You forgot to mention that his 3dmark CPU score is above average, which translates to that his paranoia towards his CPU is even more unjustified.
2.) Correct.
3.) I don't know either, the guy seems to not want to accept the actual answer for reality, why? I don't know.
4.) I doubt it's the EC considering that other users with GTX 1070s have the same issue (even desktop ones), but your hunch about the powersave is correct. Sadly this is caused by the SMC Chip inside the Pascal Die, which can't be modded since the card is crypted. EC reset might be a temporary solution tho, since it could trigger the pascal cards SMC chip to behave normally again.

 

I know that those numbers what I gave was incorrect. But, the gist was dual channel memory really works wonders. When your task involves heavy use of RAM for example data mining or even AIDA64 memory test/Linpack single channel simply can't transmit data to and fro to the CPU all the way like Dual channel memory.
@yrekabakery check cpuz memory tab while running another instance of CPUz test bench using 2 threads. You will see 3.3GHz NB freq and 3.4GHz on single core/thread.

@aaronne I had similar experience like yours. For me, 382.19 hotfix driver had best performance followed 385.41, 387.92 and 388.00
Although I do get momentary stutter on GTA V but it works.

 

These 2 fresh videos should clear things up and help in the analysis. Lowering video settings to the lowest preset as well as the resolution to the lowest had absolutely no effect on framerates in battlefield 3.

Lowest Settings and very low resolution: (130-135fps, GPU usage 15%, CPU cores usage up to 64%)


Ultra Settings at FullHD 1080p resolution: (130-135fps, GPU usage 50%-60%, CPU cores usage up to 77%)


What does that tell you?

 

The same everything else you posted did.

I did a little test on my system which has a GTX 1060 and a 4940MX @ 4ghz and got this:
720P @ Low settings


1080p @ Ultra Settings


I assume my bottleneck is around 180FPS due to CPU. Since your CPU is a great deal weaker than mine I assume your bottleneck is around 140s - 150s.

 

Your GPU is hitting the power limit aka TDP limit. Use GPU caps viewer and post a screenie. I think 150W should be the TDP. Ask @Falkentyne how to flash that power mod vBIOS.

 

GTX 1060 MXM 3.0b?

 

yeah, it's quite the performer.

https://www.3dmark.com/3dm11/12483255

 

Here are some insightful graphs while benchmarking on Heaven. GPU usage is 99% during Heaven benchmark with few random dips. I got a score of 84.5fps on my GTX1070 notebook, is this healthy? Unfortunately, there is not one GTX1070 user on this thread. Do you see a problem with CPU or GPU from below graphs?

 

Yeah its better than BGA gtx1060.

 

Its that dreaded 800MHz bug on 6700HQ due to throttling. @VICKYGAMEBOY @Papusan

 

You do realize that the CPU ran the whole time over 3ghz right? The only time it went down in frequency was when he didn't benchmark or ended it. You can clearly see from the graph when he started and ended the benchmark.

I will do another test since I changed a couple of things and it now runs even better.

Seriously, stop with the CPU underperforming crap already. You can clearly see that it operates completely normal, his scores are normal in all benchmarks, the moment he plays a game like Bf3 that doesn't utilize his GPU his performance goes down the drain, it takes no rocket scientist to conclude that his graphiscard is power saving because pascal is crap.

 

My 4 observations from above Heaven Test:

1) GPU Core Clock was running at the lower end between 1440Mhz and 1530MHz during the benchmark and did not try to boost to 1911MHz like it tries to do in games> Hence GPU temp in this benchmark is relatively lower than in Games, and less fluctuations in FPS. Why did the GPU Clock not bother going to max possible clock speed? Driver limitations maybe just like nvidia did to furmark? I saw the 1440MHz-1530MHz readings in benchmark window if you are wondering. This test remains inconclusive because as you can see in my previous battlefield 3 videos, GPU clock is fluctuating and framerates are lower than other owners of this GPU and lowering all graphic settings did not change framerates at all.

2) CPU Cores speed: CPU1 to CPU4 running at 3.2GHz, CPU5 to CPU8 running at 3.4GHz. Is this normal?

3) Power Limit flag in last screenshot. It is very recurrent. Still not sure if that is normal to Notebook GTX1070 or not.

4) Notice when the GPU clock increased to 1911MHz at the very end, I got a Voltage limit flag.

 

1.) yes it's normal, the more taxing something is to your GPU the lower the clockspeed gets regulated because the graphicscard only has so much Watts to play with. See my Bf3 screens, on low settings where the CPU becomes the bottleneck my 1060 went to stable 2088mhz core clock not moving an inch, while on load it would go from 1750 - 1800.

2.) It's normal, if you run stresstest like prime95, intel burn or do a benchmark like cinebench, all cores will run at 3.1ghz, if you don't stress them, cores are clocked like this:
1. core: 3.5ghz (1 core active)
2. core 3.3ghz (2 core active)
3. core 3.2ghz (3 core active)
4. core 3.1ghz (4 core active)

3.) welcome to pascal problem
4.) This is caused by the GPUBoost 3.0 crap by NVIDIA.

The observation you make is always the same over and over and over again.
On stresstest = normal performance
On something that doesn't drain as much performance = power saving crap from pascal limits itself causing underperformance.

 

OP. Run and buy your missing ram stick. And why not change Win power plan to high performance? Tested with C states disabled? Tested with TS ?

 

Dual Channel will make minor improvement, nothing significant to my case, please see below tests.







Windows is on high performance power plan as mentioned in original post.

No I did not disable C-State in BIOS, there are also other CPU Stepping readings that i would like you to see on my BIOS screenshots. I can disable C-State and see what happens.

My BIOS CONFIGURATION:

Spoiler: BIOS

No I never used throttlestop.

 

Friends,

At this point all I need to know, with as much scientific evidence and empirical data as possible, is if what you all see in Post #58 and Post #20 is normal or not for my MSI GT62VR laptop.

That's all.

P.S: Friend DanishBlunt, I well received your point of view and there is no need to repeat it again after this request. Appreciate it.

 

I benched my cpu and it hit 696cb instead of 687. Use my custom TS folder I sent you and see the difference. http://hwbot.org/submission/3721811_v56m_cinebench___r15_core_i7_6700hq_696_cb

 

Sorry if this is already stated but. Make sure you have gpu performance set to "prefer maximum performance" in nvidia control pannel. Your cpu looks stable but in my experience the cpus on these laptops need some love esp the locked versions. Try undervolting it to where it is stable. This will reduce your power draw and your chances of minor power level throttling, and thermal throttling should go down. Set fans on high and elevate your laptop or put it on a laptop cooler. This happening after a short peroid of the laptop working point to it being some kind of thermal issue. You may want to consider repasting and repadding your laptop with good thermal paste and thermal pads if you are able and can do the work.

You can also try locking your cpu at 3.1ghz using throttlestop speed shift settings. Fluctuations in cpu performance can change the load on the gpu. And the gpu can play a game where it clocks up then uses too much power. So it clocks down. Then it has more power that it can use again so it clocks back up. This should normally be minor but if the cpu is fluctuating as well it can exaggerate the fluctuations of the gpu.

Basicly I would play around with anything that lowers power consumtion, reduces heat, or makes the cpu more stable, and also try different bios versions and performance settings in the bios if there are any.

 

You should also read the final words in that article you get the pict... "But please keep in mind that those results apply only for the hardware and software configuration we used. With a different setup, results may be different." And 6700hq ain't exactly the fastest cpu on earth.

Spoiler: 6700hq vs. 6700K

 

I did read them and I do acknowledge the performance boost while using dual channel mode but it is still insignificant to my case where my fps loss compared to benchmarks is way larger than the potential framerates I might gain when plugging in another stick of RAM. And that is why I also showed you tests from other articles with different hardware setups as well. It does not explain why I only get 80fps-130fps in BF3 metro map, even when i turn all settings and the resolution to the lowest possible with no change in frames. Is single channel RAM bottlnecking the CPU in that game?

 

What's TDP of ur card 1070?

 

115W on paper, reaches 123W while gaming.

 

Different setup, can give different results. Have you checked if you run same firmware as working machines? Tested with different drivers?(I haven't seen, or can't remember all your and others posts). Tested with previous Windoze X Build?

 

123W is max load or current load?

 

How do you come to the conclusion that TDP is his issue when he has normal scores on all benchmarks he did?

Also just to showcase how worthless the cinebench score is:


Oh nooo....my CPU is garbage, it's underperforming so hard.... Nooooooo....

Spoiler

Syke!

Same system, same CPU etc, didn't change a single setting on the CPU.

 

max

 

I get 644-655 consistently everyday all the time on cb r15.

I am generating the same graphs i did for heaven benchmark now for battlefield 3. 2 sets, one at ultra and one at low. this should definitely tell us what is wrong exactly with my BF3.

 

And you continue say @Unhappy User's 640-650cb score is perfectly normal for 6700hq. Maybe below 640 is the normal as well? I don't need to say more.

 

Again, for the last time. Single channel RAM is your problem.

You can keep chasing your tail around and around, and pretty soon this thread will go off the rails like the last one did. Or you can nip it in the bud, right now.

 

it is not fair to use benchnark gpu readings to assess the problem danishblunt. because in heaven benchmark gpu clock is stable at the lower end 1440mhz-1530mhz unlike games wherw it struggles to stay stable bw 1600 and 1900mhz. so power consumption question is valid.

 

Gentlemen,

Kindly postpone your personal arguments as very important graphs will be posted shortly. They require the reader's attention.

Thank you.

 

What are you talking about. Benchmarks stresstest your GPU which forces it to go to the limit unlike games do. If a card runs as expected in benchmarks then TDP is NOT an issue, since the card draws less Watts from games than benchmarks, hence TDP limit set by the vBIOS is not the issue.

 

@Danishblunt Watch his video and you see RTSS firing PWR limit for very short duration.

 

Please dont argue about power limit with Danish this will derail the thread towards the "Pascal Problem" and it will never stop. I have something important to share very soon.. Please guys.

 

yeah I'm aware. It has nothing to do with his TDP limit set by the vBIOS. If you insist that this is the TDP limit set by the vBIOS then tell me how he scores perfecly fine scores in 3Dmark and Valley if his TDP limit is indeed limiting the performance.

Or are you implying Battlefield 3 is more demanding than firestrike?

Here are his benchmarks quoted from his "part1" post:

Also note that his notebook ran flawless for a long time without any issues until a certain point, which again, completely crushes your theory on the TDP part.

 

This is the ultimate post.

After sharing all GPU and CPU readings while running Heaven benchmark (GPU usage 99%, score: 84.5fps), I will now share the same readings for the problematic game of Battlefield 3, at ultra settings 1080p, and at lowest settings 720p. Reminder: Framerates surprisingly remain the same in both cases, so let us analyze together what is happening at the CPU and GPU level.


Battlefield 3 Ultra Settings 1920x1080

Spoiler: BF3 Ultra Settings: Please click to see graphs

Battlefield 3 Lowest Settings 1078x720

Spoiler: BF3 Lowest Settings: Please click to see graphs

 

Wait they're not using dual channel memory config? That could explain some benchmark differences.

 

You have really odd issue. For monitoring use HWINFO next time. I hope you have tweaked services.msc and your AV.
Maybe your card is unstable at higher clocks so you need to undervolt it 1800MHz in MSI AB. Check Latency of ur system using this link http://www.thesycon.de/eng/latency_check.shtml

 

Yes, OP has been running single channel this whole time. The lower CPU benchmark scores, lower/inconsistent performance in CPU-intensive games like BF3 multiplayer, etc, it's all root caused by the RAM. We've tried telling him a bunch of times but he won't listen, and keeps spamming these graphs over and over. That's why this thread has dragged on so long with no headway, and why the one before this turned into a flame war before it was locked.

 

The OP has been running that same single channel ram before the problem started (~1 month ago if I remember correctly) and according to the OP, the performance was considerably higher (and stable) before (for ~7 months) with that exact single channel stick!

 

so the another thing that maybe is changed is the ambient temperature and so the way how pascal turbo boost act (lower)

meantime I'm downloading BF3

 

You say ~1 month ago. Have the machine ever worked as it should with Win Fall Update? Have @Unhappy User upgraded or done clean install? Used exactly same drivers. Same firmware etc.

 

I would appreciate it if we keep some room for readers to focus on Post #92 before we move on to other issues. Carefully analyzing the data in that post will help us understand what is going on once and for all.

 

If you know how pascal works you'd know that your graphs are completely utterly normal.

There is nothing to say. It would be beyond amazing if your pascal card wouldn't power limit without doing a TDP mod.