Can I undervolt Palit GTX 1050Ti Kalmx?

[Marek Ogarek]

CPU undervolting worked like a charm (15C lower stress temperatures). Now I want to try the same with GPU, especially since it gets very hot, up to 87C (not faulty, it's to be expected from a fully fanless GPU in a case with no fans).
I'm not going to change heatsink or paste because one of heatsink screws has a warranty sticker and I don't wan to void my warranty). I can't change fan curve because my system is fanless. I don't want to sacrifice performance by underclocking. Please don't recommend the above.
I've tried searching the Internet about GPU undervolting but I only found mentions about undervolting AMD GPUs and about MSI Afterburner, which presumably only works for MSI cards.

 

[frostedtim]

Dropping below the rated voltage requirements on you video card will hinder performance, that is if it doesnt cause the card to fail outright.

Your issue is cooling in a fanless case. You need to add fans or figure out better cooling and air flow for your rig.

Can you post your rig's specs so we can try to come up with a proper solution, please? If its a pre-built rig, the make and model info will help so we can look it up.

 

[Marek Ogarek]

Do you mean it as in you don't know what undervolting is and that it can work without changing performance (that's exactly why people do it instead of underclocking) and doesn't cause stability issues if you test it accordingly? If this, then just google it (and maybe try it yourself - it's really great).
Or as in my card is volted so tightly that you can't decrease it by even 0.005v? If this, then I admire manufacturer's ability to put a stop to silicon lottery.

 

[frostedtim]

Do you mean it as in you don't know what undervolting is and that it can work without changing performance (that's exactly why people do it instead of underclocking) and doesn't cause stability issues if you test it accordingly? If this, then just google it (and maybe try it yourself - it's really great).
Or as in my card is volted so tightly that you can't decrease it by even 0.005v? If this, then I admire manufacturer's ability to put a stop to silicon lottery.

The silicon lottery is exactly why this is a risky thing to do, and you do not know what you are doing. You are looking to undervolt a video card without loosing performance and still lower the overall heat output of your rig.

You need more cooling. Your current rig has none. You need to find a way to get that heat ventilated or mitigated. Undervolting your GPU will only do so much, and you risk performance loss as well as outright failure if you go to far. You have a system that is not staying cool enough because you opted for "fanless," which, at a guess is because you wanted the system to be quiet... maybe?

People undervolt video cards in small increments to see if they can keep as much performance as possible. But the key words here are "SMALL INCREMENTS." If you are looking to drop a significant ammount in thermal output, you are going to need to cut a lot more power than that. Your best option here is to figure out a better cooling solution.

All that being said, I did ask for you to post your rig's specs so we can help you come up with a proper solution to your issue. If you dont want to do that, that is fine, but please understand that messing with voltage, up or down on any electronic components pose risks, and if you dont know what you are doing, you could quite literally be playing with fire. If you want your PC to run at optimal performance, you have to make sure it has optimal conditions for it to do so. Your rig is running hot. You need better cooling options. Fanless is not working for you, and undervolting is not a great way to deal with it either.

 

[Marek Ogarek]

1) It's not risky if you don't go too fast, at most you will get problems during stress testing, then reboot and change settings. Exactly 0% chance of destroying anything. Unlike overvolting or overclocking it can't fry any components.
2) Undervolting CAN drop temps by a significant amount, as exemplified by 15C drop on my CPU.
3) I just found out that the Palit's target temp is 83C and that according to NVIDIA's specs "Maximum GPU Temperature (in C): 97" https://www.nvidia.com/en-us/geforce/products/10series/geforce-gtx-1050/

I guess it means I don't have to worry too much if an hour of FurMark only gets it to 88C. I'm just not used to such high thermal resilience. I guess technology does move forward.

Still, would be nice to drop it even further by undervolting, but we still don't have Pascal BIOS editor. Afterburner's Voltage/Frequency curve only goes down to 0.8V and my card doesn't even go that far regardless of temp (on second thought maybe it means it's factory undervolted? seems way lower that what most people are getting).

My rig is Seasonic 520W fanless PSU (with thermal failsafes, so I'm not worried about it overheating and anyway it is at the bottom so it gets cool air, my power draw is low and the efficiency is through the roof, so it's not like it has a lot of heat to dissipate), FX6300 (@ stock clock) with NoFan CR95C cooler and good paste (thanks to undervolting CPU only gets to 65C in one hour synthetic stress test), Palit 1050Ti Kalmx, MOBO is infamously hot GA-970A-UD3 (but thanks to undervolting NB and other parts I actually have lower MOBO temps than when I still used fans in my case), RAM is low speed so it shouldn't overheat.

Setup works by convection, so a very important part is having case with a lot of fan grills, most importantly two top fan grills.

EDIT:
Only real danger of undervolting is being unable to boot. For CPU/MOBO undervolting of MOBO without dual BIOS you just have to remove battery for a minute. Where it gets more dangerous is GPU BIOS flashing. (Not sure how to undo flash that doesn't let system turn on.) Still if you go really slow it will start crashing synthetic tests on higher voltages than voltages needed to fail a boot. Anyway what I wanted to do was find some software way to do it, because those are even safer.
After hours of googling since my original question I finally found it - Afterburner is totally safe, even if you go down too quickly, because you don't have to run it at startup for testing, so when something goes wrong it's easy to undo everything. Real shame they didn't fathom somebody wanting to undervolt it even below 0.8V.

 

[frostedtim]

Ok, Im not sure Im getting through here. I understand from your original post, that you are looking to reduce voltage on your GPU in order to lower the heat output, and you have done that with your CPU as well. But you still have a high temp issue, though you are running within acceptable levels so far.

People who mess with their clock speeds or voltage are trying to either squeak out extra performance, or keep the rated performance levels on reduced resources. You are looking to manage your cooling issue... and you dont have any. You found people on the internet that tweek their systems using these clocking and voltage tricks, but these are not common things to do for most enthusiasts, and the ones that do, usually have some sort of cooling solution to deal with the added voltage and clock speeds when adding to them. Undervolting your CPU is going to hinder its overall performance, even if you can set it to stock speeds. When you truly put it under load, you will either get a performance hit, or the system will reboot.

Your PSU is nice, but understand that when you say "thermal failsafes" it means that the device will shut down if it gets too hot. Same with CPUs and GPUs. Having a NoFan cooler is fine, but again, you have to be able to vent the heat that comes off of it. Having good thermal past helps with the transfer of heat to the heatsinc, but that heat still needs to be dealt with.

This is the heart of our problem. You are trying to solve it with clocking and voltage, and anybody who tells you this is a good idea based on how you built your rig is not doing you any favors. You need some sort of fan to blow the hot air out of your case (which you forgot to list). You need to find some way to get the air flow up in that case. If you need to add a fan or three, then do so. Otherwise, you need to find a different case that allows for better air flow. Maybe both.

You may have overthought your situation with this fanless build, and going to the clocking/voltage options to deal with the heating issue is sub-optimal at best (please read that as not good), and getting things to just under the red line for rated thermal tolerances is just bad for long term operation.

I get the point of going fanless: less noise and fewer mechanical parts that can wear out. But the fanless setup will always need some sort of air flow otherwise the heat will build up and your system will fail (shutdown and wont restart untill it is cool enough is best case scenario). And if you are relying only on convection (heat transfer to a less dense material), you really need that air movement.

 

[Attenbach]

Hi,

i love silent PCs and that's what i was aiming for too, when i built mine some months ago.
I stronlgy recommend getting two BeQuiet fans and strapping one to the GPU and (at least) one for the case using Zip ties. Mine is inaudible till you get 10 inches/25cm close, and i have very good ears.
This is the most reasonable solution that doesn't compromise on computing power and also increases your hardware's lifespan!
Go for it, it's definitely worth it. Expect -40 on your GPU.

Atten