Modifying vBIOS for minimum fan speed

[Smitty2k1]

Hey everyone, I recently added a 1050 ti low profile card to my tiny gaming PC (this one: https://www.gigabyte.com/Graphics-Card/GV-N105TOC-4GL#kf). There was enough space that I removed the small fan and hooked up a pair of Noiseblocker 80x15mm fans to a splitter off the fan header on the graphics card. Strangely enough, the two Noiseblockers spin fast enough to make some serious noise even at idle.

I determined through MSI Afterburner that the minimum fan speed for this card is 30%, which the fans spin at even when the graphics card is idle. This 30% is enough to cause too much noise.

I then hooked up a Noctua low-noise adapter. With this attached the fans still spin at idle, but slow enough as to not cause significant noise. However, when stress testing the card with Furmark, the fans ramp up to 100% but the card still thermal throttles. It takes about 10 minutes for the card to thermal throttle, and it only has to downclock a little bit, so I'm pretty sure the fans are almost cooling the card enough with this low noise adapter.

So at this point, I think it would be best if I could modify the vBIOS to allow a lower minimum fan speed, but I also see a few other options:

1) try some more real world tests with temperatures in games and see if the card throttles, if it doesn't there is nothing to worry about - keep using the Noctua low-noise adapter
2) try to find a more 'mild' low-noise adapter and see if the fans can achieve a higher maximum RPM under load
3) try to determine a better heat sink for this card (unlikely and a lot of work), or try to reseat the heat sink with different thermal paste
4) modify vBIOS or find some other way to lower the minimum fan speed for this card and use it without the low-noise adapter

Any help you all have would be great, here is a (very long) build post of my rig for more info, see the latest edits for most up to date build: (https://smallformfactor.net/forum/threads/relan-e-i7-htpc-build-and-mods.8360/#post-107174)

 

[Advil]

I'm not understanding the fan problem. If those fans are too noisy at 30% just get some different fans that are quieter and slower. That's the cheapest easiest thing to do. $10 for a couple of basic fans.

I find it hard to believe it won't keep it "cool enough to work."

If it's melting in there set up like that, you'll need to mount a fan in the case over the card.

 

[doubletake]

It's probably because you have the two fans next to the GPU set to exhaust (at least, it looks that way from your imgur album). That paltry heatsink probaby needs plenty of air forced through the fins to be able to cool even half-decently, so having only exhaust airflow is going to leave you with mediocre thermal performance.

Also, stop using Furmark.

 

[Smitty2k1]

I'm not understanding the fan problem. If those fans are too noisy at 30% just get some different fans that are quieter and slower. That's the cheapest easiest thing to do. $10 for a couple of basic fans.

I find it hard to believe it won't keep it "cool enough to work."

If it's melting in there set up like that, you'll need to mount a fan in the case over the card.

The fans need to be 80mm slim fans, and the Noiseblockers are considered very high quality and well performing fans for their size. The fan header on the graphics card is only 3 pin, so I assume "30%" correlates to a voltage and that voltage is spinning the fans at closer to 60-70% of their max RPM.

It's probably because you have the two fans next to the GPU set to exhaust (at least, it looks that way from your imgur album). That paltry heatsink probaby needs plenty of air forced through the fins to be able to cool even half-decently, so having only exhaust airflow is going to leave you with mediocre thermal performance.

Also, stop using Furmark.

Good point about the exhaust, I was replicating another user's setup that I found and he did some quick thermal testing and found exhaust to work better, but he had a slightly different version of the 1050 ti. Easy thing I can test this weekend.

Also I'm interested in your comment about Furmark. What other stress testing/stability software do you recommend, and why is Furmark not recommended?

 

[cybereality]

Furmark is good to test thermals for a few minutes, but it's not a realistic test for what you see in a game (but will show you what a worst case scenario would be).

I usually try Unigine Heaven or Superposition stress test.

 

[Smitty2k1]

Furmark is good to test thermals for a few minutes, but it's not a realistic test for what you see in a game (but will show you what a worst case scenario would be).

I usually try Unigine Heaven or Superposition stress test.

Thanks! My low-noise adapter from Noctua allowed Furmark to run for 10 minutes before it hit the thermal limit so maybe Heaven or Superposition will be acceptable. I'll try out some tests, then try swapping fan airflow direction, then try more drastic measures.