The ThrottleStop Guide

Dear fellow users!
Since upgrade from Windows 1803 to 1809 and TS to 8.70, it now can't detect my base clock. So I'm always 0 MHz in the tray. Please suggest.

 

On linux use nvidia gpu because Intel drains battery like hell. If you update nvidia using intel gpu then nvidia driver won't work at all. Had to restore my Clean install backup via Macrium.

Disable any virtualization features such as core isolation mode or memory integrity checks in Win defender.

 

Thank you so much! Disabled Windows Security->Device security->Core isolation details->Memory integrity. Going full throttle again!

 

You mean full speed in the computer world the word [ Throttle] means a bad word even the name of the program itself is named [ThrottleStop]

 

Yeah, but in a racing it means the opposite Anyways, going optimally undervolted and cooled. TS is great!

 

So I got my hands on a MSI GE63 8RF (i7 8750h + 1070). Updated BIOS and EC. Loaded optimal settings in BIOS. Checked a bit the 120hz display, the rgb configuration,etc on the default windows installation. Did a clean Windows 10 LTSC 2019 install. Installed updated drivers, and proceeded to bench.

It throttles when using CPU intensive tasks, the CPU wont use more than 17w, despite temperatures barely reaching 60º.
Short benchmark like Cinebench gives good results: ~1100 and +1250 with undervolt.

In Prime96, OCCT, the CPU frequency fastly drops to around 2,9Ghz. Same in Overwatch, it runs at ~90fps with everything at low due to the frequency. My 7700HQ +1050ti keeps +130fps (thought it has dual channel which is a big deal for the game)

What could it be? Its does that without and with throttlestop, and its not a temperature problem.

EDIT:
With Cinebench (thought really short test) and TS 1024MB test, it goes for over 2 minutes at 3,9 GHz on all 6 cores, with 45W, yet the moment I run OCCT or Prime95, the frequency go down to 2,9Ghz and the W its betwwen 15-20. Same with the 2 games I tested: Overwatch and Quake Champions, I actually get less fps than on my 7700HQ. The 7700HQ actually runs at 3,4Ghz the whole time. They're set at lowest settings so they're CPU bound, I guess if I tried any GPU bound game I would get a lot more fps, as I did with Uningine benchmark, but I actually want to play those 2 at 144fps...

I just dont get why it uses 45W sometimes and sometimes doesnt get past 20W (max temperature was 74º on the TS 1024MB test at 3,9Ghz). Is my unit faulty or is it how the 8750H actually works?

 

@Temp1234453 - Are you using the FIVR - Disable and Lock Turbo Power Limits feature? Once that is checked off, how about post some TS pics and run a TS Log File while playing your games. Exit TS when you are finished playing so the log file is finalized. Some people have had trouble with running too many monitoring programs so just use TS. The data in its log file is usually pretty good.

@PascalD - Thanks for posting how to fix the 0 MHz problem.

 

I think AVX2 based testing severely underclocks your CPU and increase the CPU watts. You need sweet spot for that AVX offset ratio.
@Falkentyne Any way to work around that underclocking issue in MSI via RW everything?

 

Has anyone found TS Bench inside ThrottleStop to be inconsistent? (I am undervolting with XTU but using TS Bench for benchmarks. I plan to eventually move over to ThrottleStop.)

I thought I had undervolted too much because at 12 threads / 1024M it was giving me errors. So I backed off to -145 mV no errors. Then I decided to try again and errors showed up at -150 mV. I went back to -145 mV and the errors stayed. Then I restarted ThrottleStop and tried again and the errors disappeared at -145 mV, then tried -150 mV again and no errors.

I read somewhere else to let Prime95 run with 12 threads for 15 minutes and if it gets a mathematical error it will terminate the tests. I have no issues at even -155 mV to -160 mV running Prime95, CineBench, AIDA64 for even an hour. I'm reducing the amount of undervolt solely on TS Bench and now I'm beginning to wonder.

(System is a Thinkpad X1 Extreme w/ a 8750H CPU)

 

Interesting. I do not have this option.

 

I tried enabling it, but no difference.
Here are my settings:

Spoiler: Settings

I attached the logs, as you can see OCCTP instantly throttles (actuaylly it runs at 3,9 for 3 or 4 seconds), and Overwatch actually throttles before getting ingame. TS Bench keeps the turbo with no problems.

It actually doesnt increase CPU watts. I think it actually underclocks because the watts dont get past 15-20w.
The same happens with 2 games, dont think its because of AVX2.

 

Uncheck BDPROCHOT,C1E and Speedstep. In FIVR, change settings to save voltage immediately and in TPL make sure speedshift is started when TS starts. Hibernate must be disabled.

 

If you use 3rd party AV then those option might be missing.

 

I only use Windows security. Thanx for info @Vasudev. I will Google what is going on.

EDIT. I have virtualisation OFF in Bios !!!! Maybe this makes the options disappear.

 

A guy(DeepWeb from malwaretips) said the same thing, I said, it was new to me & he was saying the same thing you did.

 

Unfortunately, the behaviour is still the same

 

Already told you what to do in the MSI thread. IMON SLOPE/OFFSETS remember?
You didn't change VR Current Limit to 1000 yet?

 

Your cpu is working but somehow its going for short TDP burst and anything sustained load throttle down the cpu. Really odd.

 

Yes I tried those. The default VR Current limit was over 500, changed it to 800, still the same.
Also tried IMON/OFFSETS steps you provided in the old thread, but still throttles.
I retried reflashing the BIOS and not load the optimized default but nothing. I also cleared the CMOS with a clip in the hole under the base.

I already asked Amazon to swap it for another unit, it isnt normal that a clean install and just running OCTTP (not even TS running) it throttles for no reason,unless they ****ed up in the latest BIOS.

One weird thing I noticed (probably non related to my problems) when messing with some BIOS settings, is that if I set the core ratio to x41 for 6 core turbo, the laptop will power limit like crazy for no reason, just maximizing TS window would power limit it for some seconds.

 

That hole you pressed isn't the CMOS button at all
That' clears the EC RAM registers.
You have to take apart the laptop completely to remove the CMOS battery.

You didn't downgrade the bios to the one that was working fine?
Try clearing CMOS (power button pressed for 60 seconds straight) but this doesn't do a full cmos clear on the non titan laptops. It does on the Titan laptops.
But it doesn't hurt to try.

Could also be an EC firmware problem as well.
Invalid value stored in EC RAM Register E3 will enforce base TDP at all times. But IMON SLOPE and IMON OFFSET will override this.

The ICCMAX in throttlestop isn't helping?
This is beyond my ability to help.
I would downgrade the bios (and EC if you did that too) to the last working version.

 

Tbh the first thing I did was update the BIOS and EC firmware, so I dont know if old ones work fine. Ill try to find them.

 

What are you wondering about? If the TS Bench is reporting errors it is because your computer is not 100% stable. The TS Bench does the exact same calculation twice and compares the results. If the results are not equal, that is an error. Most of the time 1 + 1 = 2 but if sometimes 1 + 1 = 3; that's a problem. Your overclock or undervolt is not stable. Prime95 is a good test for software that heavily uses the AVX instructions. The TS Bench does not use any AVX instructions so it is a good non-AVX test of your CPU. It was specifically written for this purpose. A stable CPU should be able to pass both tests with 0 errors. Thanks for confirming that the TS Bench is a pretty good test when setting voltages. No need to run Prime95 for hours and hours of AVX torture.

@Temp1234453 - Your ThrottleStop screenshot shows that you have a check mark in the Speed Shift EPP box but you do not have Speed Shift enabled so that setting is not doing anything. Open up the TPL window, click on the Speed Shift box and hit Apply or OK. Now go back out to the main ThrottleStop screen and you should see SST in green. That is how to confirm that Speed Shift is enabled.

I think the default Turbo Ratio Limits for your CPU are 41, 41, 40, 40, 39, 39. The CPU will ignore any changes so setting these all to 41 does not do anything. Most people have the CPU Core and CPU Cache voltages set equally. I would be doing that until you get the problem you are having figured out. I would also maximize all of the IccMax adjustments. Can you post how you have the TPL window set up? I agree with @Falkentyne that this throttling problem seems to be Current related.

In the ThrottleStop Options window you can check off, Add Limit Reasons to Log File. This might give you a clue to what is holding you back.

Edit - Weird question. When you are gaming do you have monitoring software reporting your MHz, etc. on screen? Try turning that stuff off and just run TS with the log file option instead.

 

I tried many configurations, right now multipliers are set to default and SST is properly enabled. This is the current TPL window,althought I tried many configs:

Spoiler: TPL

I checked limit reasons to log file, but seems it doesnt give any (Limits reasons window also doesnt show anything)

As for the games, I dont use any external software, they have built-in fps counters, and could tell by the fps the CPU was throttling. So I put them in windowed mode, and confirmed it was throttling with TS.

I was thinking of installing Linux, and undervolting and checking there, but Im tired, and I think im gonna just wait for the new unit.

 

Stupid question: Is there a way to autostart TS with the icon in the Taskbar? I can autostart it but without the icon ... and because it "runs" i cannot start it again to get to the main menu ... any idea?

Thanks!

 

Use taskbar option in main window.

 

Sorry, does not work. I have "all icons on" ... and if I rund TS "normal" without the Autostart, I have the icon if I minimize it ...

 

This is an EC problem.
I've seen this before.

On the MSI GT73"EVR" (EC version 17A1EMS1.112), if the system total power draw exceeds 330W, the CPU will be instantly cut to 800 mhz with no throttling flags set in TS or HWinfo or anything.
This behavior does not occur on the CM238 EC (used by 17A1EMS1.108, .107 and .109 (.109 is embedded in the bios and has to be manually extracted or trimmed out with a hex editor...not sure if UEFItool can do it)

So are you telling me that if you set IMON SLOPE=50 and IMON OFFSET = -31999
that you still get throttling?

 

I found out it could be AVX extensions causing part of the throttle. Disabling it, Prime95 didnt throttle that fast and that bad, OCCTP gave an error and couldnt start, but Overwatch was still giving the same low fps. Didnt bother to test more,and already packed the laptop to return it and get a new unit.

The IMON tweaks didnt stop that weird throttle.

 

Check in HWINFO64 after triggering the throttling and see if IA:VR Thermal Alert was On.

 

Did you follow the guide in the second post of this thread? It sounds like you did not setup your Task correctly. Compare your Task to the guide or better yet, delete the Task you created and follow the guide exactly. The system tray icons work correctly for me.

How To Start ThrottleStop With Windows
http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/#post-6865107

 

So NOTHING shows up in Throttlestop Limit Reasons, and no flags get flagged in HWInfo64 and the CPU speed just gets cut ?
That's caused by the embedded controller.

You said you flashed the EC *AND* the bios?
You should be able to downgrade the EC (hell I've flashed to the Gt73"EVR" EC to test it and then flashed back after i saw that 330W power throttle limit).
BTW before anyone asks, I couldn't actually PULL 330W to the system (which would require more than 360W from the wall), even with a 230W TDP GTX 1070 and prime running. I simply set PSYS_PMAX in the bios to 2700 (= 337.5 watts).

Then ANY sort of load at all on the video card instantly throttled the CPU to 800 mhz. With no flags anywhere.

With the CM238 EC, this didn't happen (even went up to 8192 on PSYS_PMAX which is 1023 watts....)

 

Thank you very much! The mistake was on my site. I set it to "log in with everybody" ... I made it new and now that part works.

Last stupid question: If the Taskbar icon is red, everything is turned on and works? Is that correct?
Thanks in advance!

 

@Mythbuster - The Turn On - Turn Off, red icon, green icon really only applies to three features; Clock Modulation, Chipset Modulation and Set Multiplier. On many modern laptops, most people do not need to use any of these features so it really does not matter anymore. The voltage adjustments and most everything else works regardless of this setting. Personally, I go into the Options window and I turn off the Notification Area Icon so I do not have to look at it. After that I check off CPU Temp. That's about the only thing that I am somewhat interested in seeing 24/7. When it jumps up 20°C, that gives me a warning that some crap is running in the background that probably shouldn't be. Perhaps it is Microsoft trying to help me out again.

 

Yes I flashed the EC and the BIOS (in that order if I remember correctly). When I get the new unit ill try the tests again without updating them

 

Thank you for taking the time to reply. I'm undervolting and at -145 mV I am running TS Bench on Normal / 12 Threads / 1024M and it literally takes me having to run the test back to back 100x to generate an error and it usually just says "1 error." Am I testing this correctly? I don't know if I'm only supposed to run it just once and maybe running it over and over and over is not the right procedure? I assume if I was supposed to run it 100x in a row you would have provided that functionality? I'm sure your software is not at fault, I guess maybe I'm trying to rule out "human error" by my overthinking the situation and sitting there running the test 100x in a row. Is everyone else undervolting sitting there running the test over and over 100x? Since you are the author, I guess you are the right person to ask what the proper procedure is Thanks again for all your help.

 

Hi all,

New user of ThrottleStop here, a big thank you to @unclewebb. From a fellow developer, I admire your passion (especially while working without proper documentation)!

I started playing around with ThrottleStop a few days ago and read basically the whole thread to catch the most information I could before testing. I'm using a surface book 2 (13.5") that has the i7 8650u. I did perform a few round of tests on my laptop to understand a bit more how and if the power limits work on this laptop. The default power limits detected by TS were 30 and 35, with the clamp option turned on.

They do work, but are overridden by throttling, no matter the clamp option. For example, if I set the limit to 25W and 35W, it turbo boosts for 2 seconds at 28W and then goes to 25W. But later on, when throttling it goes down in steps to 15W, then 13W and at times even lower. Changing the power setting in windows also change the throttle limits, e.g. If the CPU is throttling under "best performance" the PC1 power limit goes down to 13-15, switching to "Recommended" lowers a bit the limit (my guess is that it has something to do with the speedshift affinity). Note that the limits are raised automatically after a while back to 25W, if the temperature is back down switching the power setting in windows resets the limits to 25W/35W.

Another interesting fact is that the power limits may be different from one boot to other, I guess depending from the load. Sometimes the power limits are set to 25W/35W, other times to 25W/28W. I can see consistently that the power limit eventually go back at 25W/35W.

So my guess is that there is a special "surface" only control for the power limits. Note that I didn't try yet the option to "disable and lock turbo limits", will try that next as I think that will prevent this behviour. About this, if I enable this option, do the aforementioned limits still apply? Or does it free the CPU to take as much current as it can?

Another question is about the GPU, I read in a previous post from @unclewebb that when undervolting the GPU also the system agent should be undervolted in pair, now I know that the naming changed for 8th gen processors and now there are system agent and iGPU unsliced, which one is which? I noticed that XTU pairs the system agent and the gpu voltage. Just to make sure, is the system agent in TS the same or should I refer to the iGPU unsliced? I will of course test with all the pairings but I wonder if someone has some consisent experience with undervolting the integrated GPU.

Thanks!

 

I have thought about adding a continuous option to the TS Bench test but have not got around to it yet. In my opinion, zero errors is best whether you run this test 1 time or 100 times. An error detected by the TS Bench is most likely being caused by one of three things. Either the CPU is not stable, or the memory or the on die cache memory is not stable.

If you are only seeing 1 error every 100 tests then that is pretty insignificant and will not likely cause you any problems during any real world use. Some CPUs degrade a little after initial break in so it is not a bad idea to do some more testing a month or two from now.

If this computer is going to be running the neighborhood nuclear reactor, I would play it safe and give it a hair more voltage. There is rarely a reason to reduce the core voltage to the absolute bare minimum. Intel sets the default CPU voltage a little on the high side so they can guarantee stability for the practical life of the CPU. If you require that level of stability then try using some more voltage.

You are welcome. I do not own a Surface Book and I have not done any hands on testing with any 8th Gen U CPUs. I believe the Surface Book uses a skin temperature sensor and it might be changing the turbo power limit values based on that data. I would definitely check off the Disable and Lock Turbo Power Limits option. That might throw a wrench in this throttling scheme. See if it makes any difference to what you have observed so far.

You could also try using your laptop in a cold environment (meat locker perhaps) or if you are not quite that crazy then just try pointing a high speed fan towards it to see if that makes any difference. I live in Canada so at this time of year, the great outdoors becomes a perfect testing environment.

I think in TS 8.70, when undervolting the Intel GPU, both the iGPU and the iGPU Unsliced need to be undervolted equally. Best to do your own testing to try and confirm this. Something simple like the Heaven benchmark, the render test in GPU-Z or something similar. Play around with iGPU voltages and see if there is any difference in package power consumption or the temps.

Don't be afraid to post lots of pics. Without hardware, I need feedback like this so I can see if TS is working as designed or could use another tweak.

 

I managed to fix the weird throttle. This time didnt even bother to try OCCTP from start, I was checking the results with Cinebench and Overwatch.
Today I got the new unit,and still had the same problems. Having my old laptop (7700hq+1050ti) running the test I do with OW, made me remember I had tweaked it (giving the .exe high priority,etc), and I kinda fixed it:

Spoiler: A bit more info about Overwatch

http://forum.notebookreview.com/thr...-raider-laptops.806824/page-171#post-10830147

The thing is, before trying the game again, besides the tweaks, I also installed the chipset drivers and messed with some BIOS settings. I restarted,and before testing the game, I tested Cinebench again, and it gave me a score of +800 down from +1260.
So I loaded the BIOS optimized settings again, and then it tested CB again ,it gave me a correct score this time. So I continued with the game, it actually runs at 3,9Ghz now and managed to increase the fps. Not as much as expected, but maybe is due to having single channel ram. Quake Still so-so, better fps,but not better than my old laptop.

But then I decided to try OCCTP just by curiosity, and it actually stopped throttling. Ran at 3,9Ghz for a couple minutes until I stopped it.
The only difference I cant think of, are the chipset drivers. Also, after windows install, I installed some unknown devices from device manager, from a folder with all the drivers folders. On past installations, I just let Windows install them from internet. but tbh Im not sure I checked all of them were installed. Besides sd card reader,radio switch,etc... I think it installed ME and couple of other system devices

Also found some weird entry on BIOS:

Spoiler: Test function 1

And so far I have a -.200mv undervolt on core and -.130mv on cache. Ofc havent tested enough, but hasnt crashed yet besides when I tried the cache at -.150mv.

Now Im not sure if I should update BIOS and firmware...

 

Unclewebb, Thank you again for your time and advice.

Do you have a ratio in mind since 1:100 errors is insignificant where you would stop undervolting? 1:50? 1:25? In real world use, when that 1:100 error pops up, what will happen? Does the whole system BSOD? I'm active on another forum and I feel that people are deciding on undervolt values based on just a few runs of TS Bench which is why people are able to undervolt to far lower voltages and think I'm crazy for running it 100 times in a row. I would like to impart your advice to them.

I also have Intel GPU, iGPU Unslice & System Agent set to -65 mV. I read these three values should be set to the same thing. Is there a way to run a test to properly determine the undervolt for these? Any recommendations?

Also, I find that all my benchmarks drop when I turn on "Speed Shift - EPP" which is hard coded to 128 SST(green) but a video I watched told me to turn it on (my CPU is the 8750H in the X1 Extreme).

Thank you again. I'll prepare some final screenshots for your approval so that I can share them with others.

 

Hi
you can avoid most of throttling in MSI laptops by enabling in bios Intel DPTF. After you do that you need to instal dptf drivers
Latest drivers(7212) here: http://download.windowsupdate.com/c..._9b9eca96252709eab823f61a646b3e331a69ff9d.cab
and apply reg tweaks made by @ 6.|THE|1|BOSS|.9 to be able to have full functionality of dptf.
He posted everything here http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-883#post-10797302.

After you aply reg tweaks open windows power plan and set everything as you want in intel dptf settings. Personaly I have set dptf to max power while plugged in.

This will help you with all strange throttling cpu slow downs. In oposite you will probably see an performance increase from your cpu, as it was in my MSI GE72VR 7RF. You will also notice much lower power consuption and lower temps on your cpu while having full power of your cpu.

Also check in throttlestop disable and lock power limits.
In throttle stop you can also enable speedshift insted of speedstep ( you can enable speedshift in msi bios also).
Apply your UV in Throttlestop and increase turbo boost active time, and also set under tpl 31 to your cpu and 0 for gpu

 

I will try it on this laptop. When I did in the old one, it would suspend or power off (dont remember) when reaching critical temperatures.
Btw ,thats a nice undervolt, my 7700HQ wont get past -130mv. What are the max temperatures of your CPU?

 

While gaming I have hitted max ~74 on cpu and ~ 70 gpu. That was mac after very Long gaming session. Normal 2-3 h of gameplay it was around 2-3C less for Both. running very silent. Fans are not even hit max rpm. Idle temps or just Office tasks like Word Excel Power Point or Simple web browsing 35-41C depending on tasks.
I had problem with suspend or power off with dptf but I have increased the temp values in BIOS in DPTF. I think there is some glich in DPTF that is trigering power off even if The temps are under the default values in DPTF section. Just increase them.
Also increase in throttle stop iccmax to Max values under cpu core cache and igpu.

As for my high UV, I manage to do that because I have disabled c states in bios and I requested c1 in throttlestop and I leave C1E in main Window of throttlestop. When you disable C states you will be able to UV much better.
Disabling c states will make your max cpu frequency hitting only the max turbo for all cores even on one core tasks but you will be granted with much better performance of your gpu. With c states disabled i have around 500 more points in graphics score in FireStrike on my gtx1060 and around 3-8fps more in games depend on situation.
Also consider DPTF because by enabling it my cpu scores went up because there was no throttle.
Also I lowered The tdp values in throttlestop from 45 Long and 60 short to 40 and 45, because apart from Prime I never hit max 45W tdp during gaming. When I was testing my ram oc with Prime it was hitting only max 39 W

Btw if you also disable C1E in throttlestop you will have another ~100points more in FireStrike but The temps will climb up quite much

Also it is worth trying to set in throttlestop options power saver to 0% and setting ac timer resolution to 1. when i did that my cinebebch 15 score went up from 765 to 772.
But you will be using around ~1.5W more in idle and load

 

You are absolutely right, the 13.5" version is fanless and has a sensor for the temperature on the back lid which acts independently. I actually tried disabling the BD PROCHOT but to no avail though. I will try the "disable and lock turbo limits" option soon together with some undervolting (so far I only went to -50mv which I think it's not enough to lower the temps).

We are neighbours Here in washington state is also quite cold, so good for testing

I will go on and do some testing later this week, I was also thinking to plot the logs of TS to better see patterns. It would be useful to include the CPU frequencies in the logs and maybe the current power limits and gpu voltage/power. So far I've been using HWInfo with its charts to monitor where I can see the current power limits, but I think it's not very optimized and may interfere with TS.

 

I also wanted to report a minor "inconsistency" (I wouldn't say it's a bug), not sure if this is a known behavior:

I saved my initial profile as default as to keep everything to stock values, therefore not adding any undervolting. I also created a testing profile so that I can test out various parameters, but without selecting the "OK save" option, just hitting apply to push the new voltage values.

Now if I switch between those two profiles the voltage offset (which is 0) in the default profile is not applied, I guess because I didn't unlock the voltage for that specific profile. I think it would be better when switching to a profile if any offset is set and set it to zero if not otherwise there may be a scenario where the voltage are not applied as expected (e.g. suppose that when plugged in I didn't want any undervolting).

 

Sorry, a bit lost here. If you don't mind me asking a few more questions:
- what exactly are these three sets of power limits, how they are named (the relevant MSRs and memory-mapped registers), any Intel docs on them?
- via WinRing0, ThrottleStop has access to only one set of limits? or two?
- via RwDrv, ThrottleStop has access to all three of them? or two?
- what about the Intel XTU, which power limits can it control, compared to the above?

I'm getting a weird issue with this feature. When it's enabled (i. e. "power limits disabled and locked"), cTDP setting does nothing. The TDP power limit just "disappears". Power usage stabilizes on ~27W, then thermal throttling kicks in, but never power throttling.

I have a feature request, if I may: is there any possibility of this setting getting a bit less coarse? Like can you split it in two (one for "disable" and second for "lock") or even turn it into a slider + lock checkbox, like everything else? And what about temperature limits? Can we have a configuration option to make temperature limits more conservative?
I'm kind of worried that the CPU runs so close to Tj_max. I'd really appreciate it if I could either set global power limits at 20W or disable power limits and set temperature limits to like 85-90 degrees centigrade.

 

This still won't stop the battery boost and AC power draw related throttling. EC Firmware controls that.

 

What do you want to achieve with this?

 

I got the laptop working properly now. The dual channel ram made a huge difference. Went from 16Gb 2666Mhz single to 2x8 2400Mhz and gained:

Overwatch: +50 to +100fps
CS GO: +40 to +70 fps
Quake: from 80-120fps to stable 144fps

Didnt expect such huge performance boost, considering the 7700HQ + 1050ti had more stable and higher "lower fps" in overwatch, better fps in Quake and similar performance in CS GO.
Ofc with settings on lowest, those games stress more the CPU than the GPU, but still one would think a 8750h+1070 would beat the old one easily, even with single channel ram.

Core undervolt currently at -250, (cache at -130, it crashes at -140) Is -250 actually working? HWMonitor also confirms the undervolt, but on HWInfo Im seeing the max voltage reached is 1.088, that would mean without undervovlt, it would be 1.338v, I dont think laptop CPUs go that high.
A run in Cinebech at 3,9GHz uses around ~1.010v

 

No.
It uses the higher of the two voltages.
So it uses -140.

When there is an AVX load detected however, it uses -250. However I am not sure if it matters if -250 is on the core with -130 on cache, or if -130 is on the core and -250 on the cache for -250 to be used in AVX. I did not test this because I think this is silly and just a bit of a "bug". I keep both linked.

The only way to get a reliable reading from this is to set Internal IA AC DC Loadline to "1" for both, but this requires an unlocked bios.

 

That makes sense. What if I didnt uv the cache at all?