Mobile Pascal TDP Tweaker Update and Feedback Thread

No, my default voltage under load is.....uh.....let me check.

@Coolane my default voltage is 0.8750v at 1886 mhz.

 

Oh well, I'm happy with the mods I've done. If only we could mod the voltage.

Sent from my SM-T560NU using Tapatalk

 

The thermal limits control is working as intended (at least on 8A Bios).

The "Throttling temp" controls when boost clocks start throttling. I set mine to 70C and once the temps reached 69C, the boost clocks (1886 mhz) started dropping by 13 mhz.
Apparently, the throttling temp for AUTO OVERCLOCKING cannot be changed. Desktop cards (afaik) have this issue as well. "GPU Boost 3.0" throttling cannot be removed. But the base boost clocks will not throttle until the temp reaches the Throttling Temp limit.

Were you guys having problems with something else? Or the rated temp and max temp limits?
Are your basic boost clocks (base clocks +200 mhz) throttling even with a high throttle temp limit?

 

since I can't get the 8A vBIOS to work I wonder what the deference is on a hex code level.

 

@thegh0sts
what is your "temperature throttling target" that you set in @Coolane 's flasher?
You said you're getting throttling.
What clocks are throttling? The boost clocks or the auto overclocks?
If you're not getting "PERFCAP: PWR LIMIT" flags, your boost clocks should not be going any lower than core clock + 200 mhz.
Your boost clocks should always remain at or above core clock+200+(custom overclock).

(e.g. if you set +200 mhz in afterburner, 1443+200, that's 1443+200+200= 1843 mhz (actaully it's more like 1860...I remember I changed core clock by ONE mhz (unstable) when going from 265 to 266 and clocks went from 1898 to 1911...I guess they are in 13 mhz steps (exactly the same as the throttling step drop).

 

I've set the temp to 60 (same as the desktop). will check the perfcap reason when I get home.

EDIT: I wonder if I can display the perfcap reason in my OSD?

 

Is anyone here using MSI GT73VR or MSI Barebones systems?

 

I think it's just you so far. Me, @Coolane, and @Mr. Fox are all on Clevo systems.

This thread is in the wrong place TBH.

 

I just tried lowering the voltage to 0.875v, no more thermal throttling until it reached 65C.
I am going to check out the default 65C value in the vBIOS, to see if raising that will help.

 

is the 8A vBIOS working for you?

Wait, I thought setting the throttle temp wasn't working either. Though I did change it to see if there was any affect.

 

hmm....perfcap reason varies between idle, VRel, and Pwr mostly as per GPU-Z.

Looks like it's mostly VRel related caps. Using vBIOS 86.04.56.00.3A. peak voltage is 1.063v.

 

If you raise your power limit further you'll probably end up just seeing idle & VRel caps. VRel cap is just showing that GPU boost is doing it's thing and maximising your clock for any given temperature within the scope of how much voltage has been allowed. I don't think you can change the voltage slider on mobile Pascal, but if you could increase it to 100% voltage for instance then you would see higher clocks at any given temperature during Boost 3.0. Of course it will end up running hotter with increased voltage, so it kind of negates the previously described effect a little. To me, I think the best way to tweak it is to max out your power sliders and then run the highest core offset overclock that you can (just under the line of stability), and then your challenge is to keep temperatures as low as possible to ensure maximum possible frequency as Boost 3.0 does it's thing.

 

i've only bumped the power slider to 112% (170w) and that's all i can do with regards to power. I'm trying a +150/+300 and seeing how the temps are.

 

You've got a modded vBIOS though right (purpose of this thread), so you can just increase that max power limit to whatever you want right? I'd just set the maximum to whatever you know is a safe maximum in terms of wattage draw.

 

I'm happy with 170W. I could allow up to 180W but I need more thermal paste to do testing.

 

I think running at whatever is your max stable core offset overclock is the best way to increase efficiency in terms of performance per watt. Increasing the core offset frequency doesn't allow more voltage to be used, it's just increasing the frequency for any given voltage/temperature point, so it's actually more power performance efficient to run a max stable overclock as long as you're not adding extra voltage. (which helps if you're limiting yourself to 170W max)

 

I am just wondering why I am getting random hitching in games like GR: Wildlands and I thought it would it was because of the OC.

 

I don't ever remember having hitching in games due to an overclock, either an overclock is stable or it crashes or produces visual artifacts, I've not seen hitching - but that's for you to test OC vs no OC for instance. (probably not that related to this thread).

 

I haven't tried it yet. We don't really need to use it because the voltage is low and thus limits overclocking.
The default 54C that you see in the tool is not working right now. There is another temperature value in the vBIOS that is not showed in the tool might affect the throttling temp.
I will try that out later, if it works I will put that value in the tool.

 

@Coolane IF you're here I sorta need help
You said I can probably run at 2000 mhz at 1.00v, or even 0.950v, but I dont know how to increase the voltage!
There's no place in afterburner to do that, and pressing control F brings up some weird curve thing which doesn't work...even if I move the points around it still says "GPU voltage 0.875v".

I really need help
@thegh0sts @bloodhawk

If any of you guys can help me increase the voltage so I can run faster than 1886 mhz it would be great.

 

If you could raise the voltage, could all other also. Maybe your voltage limits is for your vBios version(Boost 3.0 disabled). Aka lower limits than what the others use.

 

Hi @Papusan
It wasn't working even on other (original) Vbios. (86.04.56.00.3A).
I tried it long ago and it didn't work. All it did was give the green "VREL" throttle limit and ran slower than stock, and I just gave up.
Or I didnt know how to make it work.
(Yes I don't know how to swim).

I dont understand the graph.
there is no voltage slider. There's just a graph that doesn't seem to make any sense.
I somehow on accident got it to use 1.0v but I can't seem to do that again.
And this caused another problem (i can tell you about in PM). Now i can't get it to 1.0v again.

And all of the forum sites that talk about the graph don't make any sense...

The whole problem is I dont know how to use the curve nor do I know how to set the clocks I want on the curve.

 

The ctrl-F graph that comes up is just another way of overclocking within MSI Afterburner - it's an alternative way of overclocking in contrast to the normal core frequency offset slider that we all know & love. Some people say they can get higher overclocks by changing specific points on the ctrl-F graph, but in my own experience it doesn't help at all - I just use the good old core frequency offset slider. I'd forget about the ctrl-F curve/graph thing. The ctrl-F curve/graph doesn't let you add voltage, that's what the voltage slider is for - which is not usable on mobile Pascal apparently.

 

Calm down first, .
You said somehow you made it to 1.0V with the graph?
Can you try to do it again like the below picture, to see if you can lock it to 1.0V?

 

I managed to get 1.0v to appear four times (I know it was working because GPU power draw was 160W), then it kept going back to 0.875v.
I may open the laptop and flash another Bios and check.

 

@Coolane
I did that and a yellow dotted line appeared, and I hit apply after but nothing happened.
Clocks were 1886 mhz and 0.875v (This was after applying my fixed +250mv profile and then going into the curve editor and doing what you said).

 

Maybe try out the regular "3A" vBIOS and do undervoltaging like this:

 

I think I see what's going on here.
You seem to be using the boost 3 (the second boost clocks?), which I don't have access to, because you're at 1999 mhz.

I took two pictures to try to show you what's going on.
First picture: Used profile #1 (core =200, 300), pressed graph right away and took a picture
You can see 1445 mhz+200=1645 mhz (base clock).



Second picture: Hit control L on the 950 mV point, pressed control L as you said
The GPU instantly went to 1835 mhz on the windows desktop. But as you can see, on the chart, 1835 mhz is using 0.875v (it's right on the chart itself).
and when I ran valley, it showed 1835 mhz and 0.875v.
This is the main boost clock.



Ahem....
I guess I DO have to flash the other Bios and try the same thing...
There doesn't seem to be an easy way to make 1835 mhz use 0.950v.
Because if you 'drop' the graph to 'bring down' the 950mV point onto 1835 mhz, the GPU is only at 1635 mhz....and using the lower voltage.
I'll disassemble and reflash later.

 

I'll reflash later.
Found another problem.
This might be related to the "MSI Bug" that Prema was trying to fix at 4.9 ghz on the Evoc systems.

When total power draw came close to or exceeded 230W after awhile, "Power Limit 2" throttling started happening. The CPU then became TDP throttled to 45W. I read about this in one of Mr Fox's comments, but that happened only at 4.8 ghz or higher on an Evoc MSI Barebones system.

@Prema @Mr. Fox any comments? @Coolane

 

How are you monitoring the voltage? I dont see it in afterburner.

 

Sad to say so, but this unacceptable behavior is becoming common with notebooks. It has been a problem on Clevo P7 and P8 with the original cancer firmware for a long time. It also affects other brands, not just MSI. That was one of the serious flaws of the Alienware 18 as well. @Prema worked very hard to fix this problem, but MSI cancer is fully metastasized and there is a limit to what can be fixed with firmware. There are hardware and power gimping limitations in place, and it appears MSI pulled an Alienware on us in that respect. @Prema mentioned some of this in the his last post in the 16L13 Owner's Lounge.

This can also happen if you do not use the right BIOS settings. If your power limits are set too low it will cause this. Max them out and it may stop the problem. It may also cause the machine to exhaust its available power limits and shut itself off.

 

What is your CPU settings in the BIOS or XTU? Maybe take some screenshots.
I could draw 100W on the 6820HK along with a 160W 980M at the same time when I had my GT72S(official BIOS, 230W adapter, and battery drained). I am doubting you set the power/current limit too low.

 

When I run benchmarks like 3DMark and unigine Heaven they all seem to flucuate between 1999.5MHz and 1885Mhz....even with a TDP bump.

 

@Mr. Fox I love you!
If I could donate for all the truth you've given about how companies keep preventing enthusiasts from properly tweaking their hardware I'd love to donate to both you and Prema. But i'm on SSI so it's hard for me to throw money around

@Prema we need you! Please.

@Coolane the Power Limit 2/EDP throttling didn't happen right away.
It never happens if my CPU is loaded to, let's say, 70W (8 core chess engine or prime 95). It also never happened with the 1070 at 115W and the CPU fully loaded.

What triggered this is the following conditions: (This is one reason I've been here all afternoon; I first noticed this last night and thought "Ok, Hybrid Power Shenanigans again" so I unplugged the battery (check the MSI GT73VR thread) and then was getting power throttling that was NOT detected at all on Throttlestop. This throttling seemed to hard cap the total CPU and GPU power to 130W.

1) CPU and GPU power (not sure about "combined system power in HWinfo: this is only reporting like 20W) exceeds 200W for a certain period of time.
2) Battery is draining slowly while CPU and GPU power are exceeding 200W.
I can only do this by overclocking the CPU to 4.5 ghz, loading a high load program, then loading the GPU past 130W. CPU was at 4.5 ghz, RING was at 4.2 ghz (I did not test ring at 3600 mhz. Of course I might later....but seriously not right now).

Under these conditions, the GPU will keep working happily, but "TDP Throttle" checkbox appears in throttlestop, with "Power Limit 2" and "EDP Other" flashing in red. It keeps doing this even when Valley or the video card benchmarking programs are closed and only the CPU is in use (if the CPU has remained in use).

Even if I stop running the chess engine it will start throttling again if I activate the engine too quickly. I have to wait over 30 seconds, with no load, then start the engine and the throttling is gone.

Now I have another question (and I think only @Prema can answer this).
My GT73VR shows in the Bios as "Model 7RE".
However 7RE and 7RF use the exact same EC firmware and Bios version.
So how is it detecting 7RE? Is this hard coded somewhere, and intentionally preventing any power draw above 230W from happening?
Or is it actaully detecting the MSI 1070 video card and then showing up as 7RE? I don't have a 1080 to insert and test.

Since 1080 users (who have 7RF in their Bios) are not getting TDP Power Limit 2 throttling at 230W of usage, will a 7RE change into a 7RF if a 1080 card is inserted ? or will a 1080 be completely unusable because the system is somehow hard limited to 230W power draw and will throttle the CPU? How would this be hard coded (which is bad) rather than detecting the GPU (which means it would allow up to 330W if a 1080 is inserted)? This is a rather.....VERY Important question, because if there is some hard unavoidable power cap in 7RE, a 1080 would be unusable in the laptop.

I'll take pictures of the Bios screen shortly. I'm typing this on my desktop.

 

@Coolane here are the bios settings. If you need others tell me what field to enter (or change): I changed some of these settings just now.
Originally I had all the power limit options set to disabled, but that didn't help this "bug" with power limit 2 and power draw.







@Mr. Fox If you have anything you want me to change, let me know.

 

@Falkentyne - Your Power Limit 1 is disabled. That needs to be enabled and set to a higher power limit than the laptop is capable of achieving. Your other power limits are fine at 300000 for the PL2/3/4. See if you can enabled PL1 and set it for the same value.

Your TDP max looks like it is capped at 45W. You need more than 45W. You need at least double that to realize good performance.

Edit: the BIOS is set to use cTDP. It would be better to not use cTDP and use adjustable core ratios if your BIOS will allow you to do that.

What laptop is this, and what CPU does it have? The 45W TDP limit thing might not be fixable depending on the circumstances.

Good chance that is temperature related. It doesn't matter how high you clock a Pascal GPU, it's going to pull back to stock boost or less before the GPU reaches normal operating temperature unless you keep it abnormally cold or find a way to remove the cancer code that causes that behavior from the vBIOS.

 

You have unchecked BD PROCHOT in TS main menu?

 

@Mr. Fox thank you for the quick reply.

This is a GT73VR 7RE Titan (GTX 1070 model); ships with 230W PSU.
I am using the *330* psu with it however.
But it doesn't seem to care, haha.

How do I change the TDP? There is no option anywhere for that directly . I changed "ICCMax" which seems to raise the TDP (if i don't change ICCMax, I will get "EDP Other" (not power limit) TDP throttling down to 45W every time. ICCMax (VR current Limit) is the same as the ICCMax value in XTU. I sets that at 400A (100W).

 

Yes and disabled BD prochot in the Bios also.

 

Try disable C-states and run override voltage in bios. Then test with different settings in TS. Post your settings in TS as well. Decrease Ring to stock. You don't loose much performance if you decrease the ring ratio.

 

Since you have a BGA CPU there are going to be some hard limitations that cannot be fixed. As I don't have access to such a system to experiment, I can only speak in general terms that may not be applicable to your hardware configuration.

It looks like you may have conflicting BIOS settings. The BIOS is set to use cTDP, which is not ideal. If that is your only option, enable cTDP Up for the Boot Mode if you can (it is currently set to use Nominal instead of TDP up) and try tweaking those settings. Better yet, see if you can change Nominal to Disabled and cTDP BIOS Control to Disabled in an effort to force it to use core ratio clock speeds. It would be better to not use cTDP and use adjustable core ratios if your BIOS will allow you to do that. It looks like you have that part configured correctly to use core ratios, but if the core ratios are not holding you can try setting cTDP BIOS Control to Disabled and find out if the BIOS allows manually setting the clock speeds that way. It may not work, but it's worth trying to find out. I am referring to this photo. Notice in the photo below that the Boot Mode is Nominal and the PL1 and PL2 are both showing 0.0W instead of 300. Power Limit 1 is also showing to be Disabled on this screen.

 

Are you having the 7820HK or 7700HQ?
The ICCMax is current, 400 is equal to 100A, setting it to 400(100A) will help.
Like Mr.Fox said, see if you can enable the Platform PL1 setting. Maybe you will see a power value after you enable it.

Also, if you really cannot find the 45W, maybe use XTU and see if you can adjust it there.

 

@Mr. Fox @Coolane Platform PL1 is already re-enabled and set to 500000.
This is a 7820HK (the 7700HQ is what I returned for this. What a joke of a BGAbook!).

ICCMax is what controls EDP throttling (EDP Other). The same setting in XTU (ICCMax) is in the Bios (in two places; VR Current Limit and ICCMax). I've had it at 400 the whole time, except just now when I set it to 800, and the below happened (which might have not have anything to do with ICCMax, might have been an Afterburner unstable setting).

Thank you. I'll look at the cTDP up setting and change that later. I tried changing that several weeks ago and I gave up with that setting because it didn't seem to do anything.
I need to take a break for awhile though.
I had a scare that I bricked my laptop.

I was trying to change all the power limit settings to the maximum, and i set ICCMax to 800 (200A)--it was previously at 400 (100A), power limts to 500 (as @Mr. Fox said, I didn't change the cTDP settings yet though), and rebooted into windows. I dont know if it was a Bios setting or if MSI Afterburner was loading up an unstable profile (nor do I know if it activated the curve profile first or my 100% stable profile), but when I activated the 250/300 AB profile (i dont know what it loaded up automatically, might have been the 0.967v curve i was messing with), then clicked to run HWinfo, the video driver crashed instantly. Then when I tried to simply reboot the PC, it instantly locked up on the shut down screen, with sparkles(?) on the screen, crashed again then shut down.

So I powered it off, rebooted, and the MSI Logo showed up, and then a complete black screen. :/
I powered it off, powered on and nothing happened. Uh......

Powered off, and back on and it rebooted and I'm looping Valley again on tested stable settings.
So far it's reached 190W total on the stable profile, with CPU at 60W (Chess engine), GPU at my 1886 mhz overclock (130W max) without messing with the curves to increase voltage, and it hasn't PL2 throttled yet.

I need to break for a little.

I guess the afterburner curve profile i was messing around with to get 0.963v was making things unstable and I don't want to see that again right now.

I'm going to have a LOT of fun when I get my X299 board someday and 7820X 8 core CPU or 7900X 10 core, and 1080 Ti...:/

 

You still have power throttling limit with 170W. In order to stabilize the clocks with default voltage you will need over 220W for it. Or you can lower your voltage to <=1.0V, that will help your clocks fluctuate less.

 

The GT72S that I had, if I set the ICCMax to over 400(100A), it can cause my laptop instantly shutdown under some certain loads. If you experienced instant black screen/shutdown, maybe try to keep that value under or equal to 400.
Since we don't have your laptop model, see if users in GT73VR thread can help you resolving the issue.

 

Thank you.
I'm going to call it a day for now. I looped Valley+ Chess engine for over an hour, 130W from videocard and 60W from chess (only 4 threads; if I did 8 threads it's too hot in summer heat to keep temps under 90C) and there was no "Power Limit 2" throttling. PL1 was set to 500,000.

Again I have no way right now to test if higher power limit settings in the Bios did anything, or if the "PL2" bug only happens if total power draw exceeds 230 watts. But that's something I do need to find out.

I honestly think that it was the 230W draw that triggered the PL2 limit bug.
But that's a guess.

That's why i tried to page @Prema

I also need to know (as this can cost me $1,000) if the GT73VR 7RE has some strange power cap to 230W if a GTX 1070 is installed (detecting and setting it as a 7RE) and thus causing that power draw bug with PL2, or if a 1080 is installed, if it raises the power ceiling automatically.. (if someone who knew something about this could find out, that could be life saving, because I do plan on upgrading to a 1080 MXM card, but that's assuming the GT73VR 7RE is not somehow capped to 230W power draw, even though the firmware and Bioses are identical between 7RE and 7RF (this also goes for 6RE vs 6RF).

 

See the image below and be sure "Startup" is not enabled. If it is enabled, Afterburner will try to apply your last GPU overclock settings at Windows logon regardless of whether or not those were stable values. This is trouble looking for a place to happen.

 

@Mr. Fox
I never had it loading on startup!
That box was unchecked!

But yet it was still loading "profile 3" (the buggy one where I was trying to force 0.967v) on startup EVERY time.
Even though startup was unchecked.
I have no idea why or if this is a bug in AB or if this some good old BGABOOK STYLE windows 10 Shenanigans! Managed to delete the profile. It was automatically loading every time. Maybe there was a setting for it i made on accident. I was using the eVGA Skin and may have done some button combination even though "startup" was NOT checked. I deleted the profile though.
I think this has something to do with windows 10 "fast shutdown" and trash BGAbooks. btw if @Prema is not too busy, can you ask him about this? This is rather important because it could make me waste $1,000 on a GTX 1080 that MIGHT not work if the GT73VR 7RE is hard capped at 230W even though the EC firmware and Bioses are identical to the 7RF (Exactly). If the MSI board detects the 1070 and then hard caps the power draw to 230W, but if a 1080 is installed, the ID changes to 7RF and caps it at 330W, that's at least a good thing. But only @Prema would know!

Can you do me a favor and ask him? Because a mistake could cost me $1,000! I'd rather risk @Prema being annoyed than me buying a 1080 I can't use !

 

You could ask Svet. As he have worked with MSI firmware - laptops long time. Maybe @hmscott know it?

 

Thank you. I'll go PM him.
I suspect something is up because the Bios does detect that a "GP104" board is installed and the Bios even lists the GPU vbios version. That's why I 'suspect' it labels itself as a 7RE, and then the EC firmware somehow limits the power draw to 230W. I'll ask svet if he knows anything about this.