Mobile Pascal TDP Tweaker Update and Feedback Thread

And you still didn't say what is your GPU model, is it a Maxwell GPU, Pascal, something older?
What laptop model as well?

 

Hi, I will answer what I can based on the information available to me.

I do not use TDP Tweaker, so I cannot tell you. Maybe someone that does can answer.

I can only guess, but I have the same problem using different vBIOS on a GPU it was not designed for. The video output is probably being routed to a port location that does not exist. For example, if I use a Clevo vBIOS on my laptop 2060 the external mini-DP does not work. The laptop screen does, but there is no signal going to the mini-DP connections. I have also found that using a foreign vBIOS on desktop GPUs can cause some of the ports to no longer function. For example, using an ASUS Strix XOC vBIOS on my EVGA 1080 Ti caused 2 of the 3 DisplayPort connections to no longer have a video output signal. Only one of them worked.

Absolutely. Even more. The original spec is ancient and it wasn't designed for that in the beginning, but many systems pull a lot more than what the original spec contemplated and it hasn't been a problem in any way.

 

Thank you Mr. Fox,

I got two issues one is the Black Screen after driver install (MSI Vbios).

The other one is the TDP Power Limit that lets my Card throtheling (max-q Vbios). It’s not the Temps they are around 70*c while Gaming.

I got a P5200 Quadro Card currently running with a 130w Bios.

It’s a max-q Bios. Hope the 140w Bios mod Stops the Trotheling and holds Clocks more stable.

I’d like to use the MSI Vbios because of its higher Clocks.

are The Max-q Cards physically different compare to non max q ?

I taught it’s just a Vbios difference seems it’s not ?

so could anybody please guide me threw the settings of TDP tweaker to set 140w and get the best Performance out of my Card.

any Suggestions how to get the MSI Vbios Running ?
I got it running once but with a oem 27id. Out of desperation I used pre modded inf files replaced all in the Nvidia driver folder. It worked with 3D Mark but fail on AAA games, it crashed. Just modding the MSI Inf Leads to mentioned Black Screen after successfull driver install.

Are Pascal Cards abble to OC ?
MSI after Burner Sliders Revert back after I hit apply. Is this a Vbios issue ?

Thank you all for your suggestions and help.

 

Max-q cards aren't any different physically speaking. Maybe they have weaker Vrms but aside from that it's just a lower tdp part with undervolted and underclocked memory.

 

Thanks seanwee,

i expected max-q and Max-p are physiclally the same. But was not sure about that. Thanks for confirming.

Could anybody please help me dail in correct values in tdp Tweaker for 140w ?

Do anyone know why the max-p MSI vbios don’t let the Nvidia driver install finish properly with modded MSI inf ?(black screen it’s the last Setup step) usually after few seconds the desktop reappear but not in my case.
Any suggestions on that ?
Is that what tdp Tweaker call black screen issue fix ?

Beside Powerlimit throtheling of my max-q 130w Vbios Nvidia driver install passes with no issues and no system Crashes with modded Clevo inf.

the 130w Max-q is a Clevo Vbios
The Max-p is a MSI Bios
My Laptop is a Clevo Laptop (mxm 3.0b)

Thank you all for your suggestions.

 

I apologize, I'm not going to be any help with Max-Q stuff. I have no exposure to it and no interest in products that incorporate Max-Q design. I'd hate to tell you something wrong and cause you to mess something up.

 

Thank you Mr Fox,

couldn’t anybody help me with TDP Tweaker ?

What Values do I need to Change to rise the TDP Limit from Currently 130w to 140w ?

Thanks for any help

 

Its a Quadro card, its locked, so far only Prema was able to mod one, if you want more TDP just use the card original VBIOS and shunt mod it, using other VBIOS will lead to either non working outputs, or bricked card due to mismatched memory timings.

 

I know that modified pascal Vbios can’t be flashed with nvflash. I use a Programmer to write Modified Vbios directly onto the eprom Chip.

Please anyone help me setting up 140w in tdp Tweaker and if possible disable or rise throtheling limit to 85* (no throtheling below 85*)

Thank you.

 

Quadro cards wont let you do the mod, even if they are Pascal based you will end up with a black screen, there are extra checks on the Quadros to prevent TDP modifications.

 

Is this what i think it is?

 

That might be something related to Dragon Center profiles, given that its under software.

 

I don't have dragon center installed or any msi software installed for that matter. I tried playing around with it, no values worked and adding a DisableHPpower line did nothing.

 

I run a 1070 MXM 3.0B and the VRM zone gets really hot at 90W TDP. If there is no active venting to the VRM zone then some MOSFET could die and kill the card.

 

on laptop you NEED to cool the mosfet with an heatsink, airflow is not enough to cool them unless it's not in a laptop.

 

I agree with you there, GTX 1060 140w TDP modded (adjustable) and not getting hot VRM's issues which means @IllusiveMan might want to check the thermal pad / thermal paste (as some now use thermal paste) contact to ensure they are making contact with the heatsink.

Like @TheQuentincc stated Laptops based system VRM's need proper contact with the Heatsinks else they are going to get toasty and depending on the VRM will depend if they save themselves or thermal run away latter being what you don't want.

 

so, let me get this straight, can someone make video of how to do it on rtx2060 on msi laptop please? i am so confused right now, as i am trying to raise my tdp to around 95-100w from 80(msi afterburner showed max 91w but gaming it only draws 80)
(i could go higher also if someone has had success before, as i am on liquid metal and running modified cooling solutions and getting 2085mhz stable gaming @Around 60c on bf5)

 

Not sure coolanes tools works on RTX cards.

Sent from my SNE-LX1 using Tapatalk

 

so can someone walk me through what should i change here to make it 95-100w? seems like the power limit is 80w altho msi afterburner clocked in 91w highest draw for gpu?

 

I meant to cool the area of the heatsink which covers the VRM zone. The P750DM does not have vent holes there, the bottom cover gets hot even during normal workload.

I modded the 980M heatsink in the P750DM to fit with any 3.0B MXM, in my case the gecube/aetyna model (black color PCB):


All VRMs are covered with thermal pads, but since there is no heatpipe on the area, cooling is not doing too well. I am still thinking to add maybe some slim heatpipe on the zone.

 

Are you able to modify the heatsink to add a heat pipe to it, as this will be the only way around your issue.

Sent from my SNE-LX1 using Tapatalk

 

Yes, though I'd need a very thin heatpipe or several.

The heatsink was already ripped apart in the VRM zone, so it's worth trying.

Anyway, stock heatsink for maxwell cards is really poor in the VRM zone.

 

I take it there is no pascal heatsink for your model? anyways lets keep to topic feel free to DM me if needed.

 

Well, I did this but it seems like it was mostly a waste of time. Overall pulls about 260W (maybe 70-80W more than before) from the wall running Time Spy, but the performance increase is largely unremarkable. The BGA 2060 GPU is so firmware cancer crippled and voltage-limited that the shunt mod seems futile... meh. I can't overclock the GPU core any more than before due to being voltage-starved. Basically the same max core overclock due to a wimpy wuss cap of around 1.065V, which it can't hold indefinitely.

Oh well, at least I got a chance to practice my SMD soldering. Nothing ventured, nothing gained. Que sera sera.

 

Damn, 260w on a 2060 system. What are the scores before and after though? I cant imagine that it's not within 5% of desktop 2060 performance at this point.

Besides, desktop cards also cap out at around the 1.1v mark so there not much to gain in any case. Maybe 2050mhz instead of 2125mhz. A more noticeable decrease performance would be from the higher temps on laptops causing the boost clocks to be lower.

And overclocking limits will stay the same regardless of power level. That depends on the silicon quality.

 

Hey guys, What happens when I flash a GTX 1080 Mobile bios on a GTX 1080 Max-Q laptop?

 

@Mr. Fox
you need to play with RTX2070 instead. THAT thing is really power starved, nowhere near any limits but power.

 

Yeah 2060 is a low budget gamer-only GPU. I ran a bunch of Time Spy and Fire Strike passes on my portable AC last night and it was a waste of time. The firmware is so castrated that the shunt mod makes no improvement. The low voltage throttling algorithms cancel out everything.

I kind of expected this. This was something I tried as a cheap thrill because it cost almost nothing (under $10 USD) and I had nothing to lose. Otherwise, I am no longer interested in spending money on laptops. They are all varying degrees of dead-end castrated garbage because all of the companies that manufacture them are losers that half-ass shortcut everything they are involved with. They are focused only on things that are either totally unimportant to me or way low on my list of priorities, and there are far too many intentional restrictions in place to interfere with the successes of anyone with a goal of correcting their stupid mistakes. It used to be fun, but now everything is just a PITA to deal with.

 

I see. Point was, on 2070 specifically power limit is much stronger than on other two... stock 2060 and 2080 have nearly enough power to run at their capacity, while 2070 is basically 2080 with 2060's power limit.
Still not as good as desktop of course, and i agree their greedy limits are cancer. But, i've no idea where i'm getting, but i'll try power mod on mine, once warranty ends.

 

Oh, I got you. When you said " THAT thing is really power starved" I thought you meant the 2060 (which would be accurate) and were suggesting that I play with 2070 instead because it is less of a mangled piece of crap than 2060. You were saying more or less the opposite. It's unfortunate that all of them are screwed up in one or more ways. So were 1060, 1070 and 1080. Add Max-Q cancer to the mix and it's all just a huge cluster. I am almost positive the people that design laptops now have the intelligence and skill of chimpanzees and orangutans.

 

What if I were only modding the vBIOS to remove the UEFI portion? I change nothing else. Would it still give issues?

 

I can try multiple vbios on my P4000 if you want to try your idea

 

Do you have a M17xR3 by any chance?

Can you dump your vBIOS? I will have it modded same way it would be modded. But do understand the mod removes UEFI so you need to use CSM or legacy mode.

 

The extent of your complaining makes me think that your shunt mod failed. There is no way you are getting the same level of performance with 1.5x power draw on the gpu.

Either:
1. The laptop is thermal throttling (which is unlikely given a man if your expertise, you will find a way to cool it down)

2. The new shunts have too low of a resistance and has tripped Nvidia's built in protection. In which case you'll have to get higher resistance shunts.

3. Eluktronics's Eluktroboost is messing with an otherwise working shunt mod.

4. The vrm is throttling from the significantly higher power draw and heat causing poor performance.

There is no other "castration" other than the low power limits. I got massive gains on my machine from shunt kidding and I'm on the even more "castrated" Max-q platform. Don't fool yourself. Find the root cause and unlock its full potential.

 

It's right now in my M6700 but if needed I can try it in my M17x R4, I have no R3 :/

 

I think we can scratch 1, 2 and 4 off the list. If it wasn't doing something, then the system wouldn't be pulling more power. If NVIDIA protection were tripped, it would be locked at around 300MHz idle clocks (already been down that road on desktop shunt modding). And if 1, 2 or 4 were occurring my GPU performance would not be equal to what it was before. It would be reduced.

In both comparisons below the benchmark runs are with the laptop sitting on a portable AC unit and the "hottest" the GPU got was like about 50°C, so we can rule out temperatures. The core max overclock is the same as before. The voltage is crippled to the same maximum as before.

Fire Strike - Before and After: https://www.3dmark.com/compare/fs/20785001/fs/23301553

Time Spy - Before and After: https://www.3dmark.com/compare/spy/9018153/spy/13456337

Maybe 3 comes into play. Impossible to know since we have no idea exactly what "Eluktroboost" is.

More likely it is exactly what I already said... No way to increase core clock above it's prior max clock speed without a corresponding increase in voltage equals no increase in performance. My max core overclock is the same as before. There's no way the scores are going to increase without increasing the core clock. So, I still believe the shunt mod is providing me with zero benefit because I cannot increase my core clock due to no ability to increase voltage. This applies universally... desktop, laptop... doesn't matter. Clock is king.

If you were having power limit problems before your shunt mod, an increase in performance would be expected. A potentially large one. If Eluktroboost was allowing me to function without a power limit problem, the shunt mod would accomplish nothing. The only way the GPU would be able to take advantage of a higher power limit is by increasing core clock.

What I cannot answer is why the GPU is pulling more power than before. Maybe I didn't get an accurate read and I just think it is pulling more than before and it's not.

What I do know at this point is that I am not going to burn any more calories (or money) on it. Without unlocked firmware that allows me to increase the voltage, it's not going to reap any benefits for me.

What might be holding back the core voltage is the resistors near the battery connection. I did not shunt those two resistors. Maybe if I did the GPU voltage would increase. But, I am not sure exactly what they do. The shunt mod on my desktop 1080 Ti and 2080 Ti GPU allowed a substantial increase in core voltage using a different vBIOS. On the 1080 Ti it was the Strix XOC vBIOS. On the 2080 Ti it is the Galax HOF 2000W XOC vBIOS. Using the stock firmware the shunt mod on both of those GPUs only yielded modest performance increases. With both of those GPUs using the better firmware enabled major increases in core clock, which then induced power limit throttling without a shunt mod because the GPU could not provide enough power to handle the added voltage and corresponding core clock increase. In other words, it took both firmware with higher core voltage limits and a shunt mod to achieve the desired outcome. One without the other produced only modest and lackluster results.

To my knowledge, there is no 2060 notebook vBIOS with unlocked or elevated maximum core voltage limits. I wish there was, and maybe then we'd actually see something impressive happening. I think the reason we don't have that is because the largely artificial differences between 2060, 2070 and 2080 notebook GPU performance would instantly evaporate except for the limited circumstances where Cuda and Tensor/RTX core count matters. That is very deliberately manipulated and controlled by the Green Goblin as part of the "pay for performance" racket. Brother @B0B recently posted a video exploring some of this nonsense in notebookland.

TL;DR - the shunt mod didn't fail. It simply seems to be of no benefit on a system with Eluktroboost because there is no way to take advantage of it without a corresponding improvement in firmware with higher or unlocked core voltage limits. The firmware it has now is already optimized to function at its best in concert with Eluktoboost along with the currently stunted (very low) core voltage maximum. It is "optimized" but definitely not optimal in terms of reaching maximum overclock potential.

@Papusan

(Edited for clarity.)

 

Eluktronics's Eluktroboost is either going to be a hardware mod or something in the bios/ec to trick the system into think it's drawing less. Does the issue happen with Eluktronics's Eluktroboost is set to off?

Sent from my SNE-LX1 using Tapatalk

 

Eluktroboost is a hardware mod and may, in fact, be a shunt mod of sorts, which could explain the situation. The only people I know that know what it is cannot disclose what it is. I wish I knew.

As far as your question goes, I don't know. I never turn it off. I literally leave it on permanently and have since I bought it. There has never been a reason for me to turn it off. I will check later when I return to my computer.

And, just to be clear. Nothing is wrong. There is no issue other than the shunt mod did not have any effect on performance. Performance is exactly the same as before and better than most, if not all, comparable systems with the same specs.

I edited my previous post to add clarity after I had more time to think about it.

Think of it the same way you would a CPU. If your power limit it too low your CPU is going to power limit throttle running stock. In fact, we see that stupid nonsense all of the time with lots of the turdbooks out there. If you fix that problem it is going to run better and get higher benchmark scores... naturally. It's not overclocked, it is simply no longer malfunctioning. Raising the power limits to a value more than what the CPU needs is not going to change anything. It won't cause any harm, it simply won't provide any benefit. They'll just be higher limits and the additional power available won't be utilized.

 

Like night and day... massive difference. I just tested it and with Eluktroboost turned off there is about a 10 FPS decrease in Time Spy and the GPU core boost clock is 250-300MHz lower. The GPU core voltage is also about 0.100V (100mV) lower.

 

I see the problem now that you posted the firestrike and timespy scores. You were already at max performance before the shunt mod. 5% above stock desktop performance in fact. There is nothing left to be gained. I'm not sure whether to say congrats or my condolences.

So yeah the shunt mod wouldn't achieve anything more.

What's amazing is the fact that the Eluktroboosted 2060 at 115w can achieve performance parity with a desktop 2060 which draws 160w with an overclock.

Not sure about the power draw increase. Could be a misreading as you said or a dramatic drop in efficiency as you draw above the rated power for the AC adapter.

 

Yeah, I think you're right on the money now.

 

If the GPU is locked to max 2.1Ghz, and you are already hitting that or very near, dont expect more performance just because the GPU could use more power.
More power is not more better, leakage starts playing a bigger role the higher you go.

 

Okay could you dump your vBIOS? I will have it modified and at least we will know if the UEFI removal results in black screen or not, then it would allow further testing in a unsupported system like M17xR3, knowing the card/vBIOS mod work fine in a tested system.

 

I modified it myself, so I hope it's okay. Check your PM.

Spoiler

 

I am looking info about Pascal bios and memory voltage. Does anyone know if there is a way to modify mobile vbios with hex so memory could use for example 1.45V instead of 1.35V?

 

There is no way to do this, the best way to increase voltage is to hardware mod the voltage regulator

 

Ok, thank you.

 

Removing UEFI from Quadro P5200 resulted in black screen, so I guess you cannot modify Quadro 10 series vBIOS at all then?

Would this still work on those 1070s that are build on P5000 PCB?

 

Hey guys, If the TDP is modified using a mobile pascal tdp tweaker, the amount of power applied to the GPU of the laptop will be different from the amount of power intended by the manufacturer. Can you be sure that the laptop's capacitors will be able to withstand the power values changed by the user? This may be quite dangerous software. Am I wrong?