Mobile Pascal TDP Tweaker Update and Feedback Thread

Did you use just a small diameter cooper wire, or did you add more shunt resistors in parallel to achieve a lower total shunt?

 

Just solder. Its the same thing as a liquid metal mod, just solid.

 

You have a lot more confidence in the engineering of the motherboards than I do
And even with a solder blob bypassing the shunt resistors there is no safe mode, I read around that on desktop GPUs if you just bypassed the shunts the card would be stuck at 135Mhz.

 

I've yet to encounter that but yes I've read about it. Just don't overdo it and youll be fine.

 

and your card isn't downclocked ? usually when the shunt resistance is too low, it's locking up at 2D clock

 

No its not running into any safe mode. I tried soldering a piece of wire across the shunt and that resulted in a 3-4w power reading under full load and that resulted in the power throttling indicator coming up intermittently and a negligible performance increase. So i removed the wire entirely and just made a small bridge just using solder. That resulted in a 10w power reading and no power throttling indicator in afterburner.

 

Do you have any photos of your work as I am thinking of shunting this way one laptop.

By the way I didn't knew that that my gtx1060 and 7700hq is almost in pair in resaults with this zepherus.
In FULL stable oc condition my gtx1060 scores just around 300 points less then this 1660ti and cpu just 3 fps less then this ryzen. I think this Asus laptop is underpowered any way

 

Well it Is the slowest zephyrus. The 3750h is also a really slow processor, slower than an i5 8300h. Your gpu score is also really good regardless.

I didn't take a pic of the shunt mod unfortunately. But ideally it should look like this.



Melt solder onto the pads on either side of the shunt resistors (marked as R005) and then when it's slightly taller than the shunt, bridge the two sides with one swift stroke.

As I understand on some thicker laptops the shunt resistor will look like desktop shunt resistors so keep than in mind when locating the shunt resistors on your motherboard. The shunt resistors should also be located close to the vrm of the gpu and will come in a pair.

 

Tweaked the overclock to fit the shunt mod and got a new high scores

https://www.3dmark.com/fs/22961346
https://www.3dmark.com/spy/12671011

What are your scores?

 

@seanwee, I'm planning on shunt-modding my Quadro RTX 5000.
Below is the front and back PCB of my card; how should I go about soldering across the R005 resistors?



Thanks, and great job on the 1660 Ti records, that GPU is nearly as fast as the 2060/RTX 3000!

 

I like my desktop 2070 super equivalent 2080 Max-q better

Since your mxm quadro uses full size shunts, I'd recommend you buy two R003 resistors to replace the R005 resistors instead. Getting a thin enough solder line across without tripping nvidia's built in protections will be nearly impossible. Wires are out of the question as they will drop the resistance too much.

 

Thanks for that. What would be the actual power draw if those resistors were used instead?

 

Took a closer look at the pcb and I don't think it's a wise choice to do a shunt mod on your quadro 5000. The shunt mod will allow it to draw about double the power but realistically it will top out at 140-150%.

Its using 5 NCP 303150 mosfets which is the exactly the same configuration found on my GS75. Even at stock (150w) its going to produce a ton of heat.

For reference other 150w 2080 laptops like the GE75 and Hp Omen 17 have 6 NCP 303150 mosfets.

Keep it at stock. Don't mod it. It'll likely burn out if shunt modded.

 

Ah, snap. The 80 W power limit is really annoying.

How about a shunt mod to 120 W?

 

Wait it's a Max-q?

In that case you can go ahead with the shunt mod.

 

Awesome. So replace both of the R005 resistors with R003? That's a ~40% drop in voltage/power across the resistors, so I presume the shunt mod would bring it up to ~110-120 W?

 

Yeah more or less.

 

@Ionising_Radiation just solder another 5mohms resistor on top on each resistor, it would double power limit, you can also solder 10mohms resistor to get +50% power limit or 20mohms for +25% etc

Or follow this formula to find the resistor you need to solder on top of the other

 

Being a laptop gpu stacking resistors doesn't make sense as it'll likely have space constraints. That's why I suggested replacing the R005 resistors with R003 resistors outright.

R002 would be a bit too much.

 

Hello, but how to shunt here such a resistor that has 4 contacts ?? There is no way to solder on top, there are no contacts. And is it possible to replace the usual one with 2 contacts? Sorry for my English)I have clevo n950ts laptop with 1660ti.Gpu limited to 80 watts in tdp. According to the service manual, 2 resistors in the Gpu power control circuit RL1632T4F-R005-FNH.

 

Watch these videos

 

You can use a normal resistor, the smaller two pads are used to route the small traces that go into the current sensing circuit, but its only one big resistor, its to reduce thermal noise into the current sense circuitry, but a normal resistor will work just as fine.

 

does this tweaker work with the gs63 8re gtx 1060 laptop and what exactly will it do to my scores number wise

 

@Krzyslaw has a tdp modded 200w 1060. You can ask him for help.

At 15k+ firestrike graphics score it beats even overclocked desktop 1060s

 

I have modded gtx1060 to 200W but the max that I have ever recorded was around 170W and it was under unreal conditions.
But this way I am not Power Limited.

Max score was 15.4K but this was with some old driver (wddm 2.4 I think) that was giving me best bench scores and also I have a bit higher mem oc+ 808mhz but was sometimes unstable.
Now with most recent driver with the same max core OC +190 and mem oc +704 which is stable in all gams and tests I am scoring in FS 15,2k -15.3k graphics score.
Very happy with the resaults as it is more than desktop 1060S even after OC.
I can provide stock 78W bios, stock 88W bios for bga gtx1060 ( both can be flashed with nvflash) and also modded vbioses. But to deal with +110W and more in laps that have gtx1060 user must work much on cooling core ram and power section.
Any way best way is to flash 78W if you have even lower tdp and check the termals and if you still good flash 88W stock bios and check again and than if you still have room I would play with tdp tweaker and programer.

BR

 

Hi @seanwee. Your sharing on your shunt resistor mod and benchmarks is pretty cool! Your timespy CPU score is also pretty high. Did you buy 3200MHz RAM or was that from manual overclocking? My best cpu score was ~7300 but typically caps at ~6900 in timespy. Clocks run at 3988MHz instead of full 4GHz too, not sure why. It takes ~56W when running timespy. I know my laptop can do about 86W for about a minute in blender before vr throttling and it can sustain 76-78w for extended periods without power/vrm/thermal throttling.

For mobile rtx2070, do you know if the shunt resistor is typically on the underside of the mainboard?

I have searched around for vbios mods with limited luck - have you come across anyone successfully modifying power limit in vbios and applying with nvflash instead of hardware programmer?

I'm currently using Razer blade pro rtx 2070 max-q 100w vbios on my clevo p960rf and its working well. Laptop still has headroom so i'm interested to try 110/115W but haven't found any working vbios from other manufacturer. There are many 115W vbios i can apply but my GPU would be pinned @ 110-115W power consumption with only 1300-1500MHz clocks and fans ramp up to max eventhough temperature is only 73c. On the working RBP 100W vbios i can get 1700MHz.

Also - i7-9750h, know of any way to change multiplier to have higher clocks when multiple cores are active? I've got all hidden options in my bios exposed, among which are the cpu turbo multipliers. it would let me set the 5-6 core multiplier (1-4 core locked @ 45->42) but it doesn't do anything even with throttlestop. 6 core active still capped at 4ghz.

 

Yes, I overclocked the stock ram that came with my GS75. I've also bypassed the cpu power limit by changing the imon slope in the bios so that the cpu can draw as much power as it wants. The cpu running at 3988mhz is due to the bclk being set to ~99.8mhz instead of 100mhz. I plan to try my hand at hex editing to unlock bclk overclocking on my GS75 as well.

The location of the shunt resistors changes from laptop to laptop so there is no fixed location for the shunt resistors. Just look for two R005 resistors located close to each other, those will be the shunt resistors.

As for vbios flashing, no, nothing works. I've tried about 10 different vbioses for the rtx 2080 and none worked, what you're seeing is a fake power reading and the gpu is set to safe mode which will result in lower performance than stock. That's why I went all in on the shunt mod, no software options worked.

And no, the i7 9750h is a locked processor so you can't overclock it. The default 45,44,43,42,41,40 multiplier is fixed.

 

Thanks for that. I'm asking on the win-raid forums about vbios power limit changes as well, will see what they come back with but not expecting much to be possible without a hardware programmer.

I've looked at the top of my mainboard and did not see anything that could have been the shunts. I did see 2 sets of 2 very small resistors (the 2 small resistors are on adjacent pads) from my laptop service manual PDF (has a very grainy image). I think those would probably be it but they're physically much smaller than the r005 resistors in the youtube vids i've seen and the images posted in this thread. I'll post some pics if i get around to that next level of disassembly.

any reason for doing the imon slope changes in bios rather than using throttlestop? Seems to achieve the same outcome to me but haven't researched it in detail.

What was your stock RAM speed? I haven't tried changing settings on mine but am not all that optimistic considering i've got sticks from 2 diff brands (1x teamgroup 1x crucial). Neither has xmp or any jedec profiles above 1333MHz for an effective 2667MHz.

 

Throttlestop can't change cpu tdps on msi laptops as its locked.

And my stock ram speed was 2666 cl19. I changed it to 3200 cl18 with optimised subtimings.

 

Wow. Big gains on the RAM.

On my laptop throttlestop allows adjustment of pl1 & pl2. I did look up the imon and Ac/dc loadline changes on some other razed and msi forum threads and they find it more optimal that purely using cfu/throttlestop. Will give it a go.

 

My 2x16GB e-die did stable 3200 cl17 at 1.2V. Unfortunatly on hm175 Chipset you cannot play with ram voltage. Modules works only at the voltage that are programed in spd or xmp.

 

What if the XMP specifies 1.35v? Would it run at 1.35v then?

If yes then you can try editing the xmp with Rweverything to get more voltage.

 

Yeah xmp is 3200 cl16 @1.35V and if I load it they run 1.35V But with bclk oc it gives me 3300 and they are not stable so I assume that need a bit more voltage.

If you could write me how I can edit ddr4 xmp profile with RW plus some photos it will be easier to me to understand and I will definitely try to mod it.
You can write it here or in pm
I was looking for that info many times but unfortunately nobody anwsered, so if you have time please guide me.

I can try even higher voltage than 1.35V for testing pourposes.

 

Has anybody managed to get this thing working with RTX VBIOS? I tried flashing a file edited with it and get boot loops and Error 43s in Device Manager. :S

 

All guys ends with error 43, For now I think only way to go with RTX is shunt mod.

 

Does a RTX 2060 Mobile even have shunts to mod?

 

ALL nvidia gpus use shunt resistors for power monitoring.

 

Follow this guide

http://forum.notebookreview.com/thr...e-timings-for-any-ram-on-most-laptops.805589/

 

@seanwee i tried out the ac/dc loadline and imon method of managing power limits. Did not get any real appreciable gain in my case over using throttlestop. Interesting that it works well for others with some quoting +400 higher cb20 scores. In my case, throttlestop vs loadline+imon both get me ~3150 on average. My laptop does let me manage PL1/PL2 within throttlestop and i did spend a lot of time testing my undervolt to make sure its stable. I'll mess around with this for a couple of days but i think i will eventually switch back to throttlestop since i like being able to look at the power reading in software.

 

If it works for you then all the better. The only way msi users can bypass the power limit is by using the imon slope tweak.

 

Thanks but this is for ddr3 and I thought that this is not applicable for ddr4

 

It should be the same process, sure the timings and clocks aren't the same but you're only interested in changing the xmp voltage right?

 

Yeah but he main problem with this guide for me is that the tools can't read ddr4 SPD 2.0. Second thing is that I can't create xmp profile with higher voltage with taiphoon in free version so I am not able to save it.

For me the best scenario would be if someone could point which adress should I edit in RW to program XMP voltage higher, as all the utilities and software are lets call it gui for RW, so everything can be done in RW, but I don't know which values should I change to program xmp to lets say 1.4V from 1.35V

 

Do you have the inside scoop on how to do this?

 

Inside scoop? No

Watched/read lots of shunt mod videos/posts? Yes

There's not much to it really. Just find the shunt resistors, short them so that the resistance is low enough to trick the power reading but not too low until it trips nvidia's built in protection and you're done.

 

You got a link on a vid doing this on a RTX 20xx mobile chip?

 

No, as far as I know I'm the only one stupid enough to try this.

I just transferred my knowledge of desktop shunt modding to laptops.

 

Maybe a bit crazy for me. I'm sure there's a way to hand edit these VBIOS files and flash them back on with the programmer. I'll see what the guys on win-raid say.

 

So far no one has been able to crack the latest VBIOS signing of Turing GPUs, you can edit the VBIOS all you want, because the places of where the power limits are known, but you wont be able to get the card in working order with the modded VBIOS.
So, a soldering iron and two resistors is way easier to do, and in fact, the only way to do it with Turing GPUs..

 

Edited bios for RTX series is not working - will give you and Error 43 in the device manager.
However there is available 115W bios from MSI (and also newer max-q bios from ASUS) that is original, signed but require hard flash via programmer (different chip ID, so nvflash will refuse to flash it).
Anyone willing to try?