[Video] Der8auer Checks his GTX 1080 Ti Liquid Metal Power-Mod, 5 Months Later

Around five months ago Germany’s renowned overclocking guru der8auer published a video in which he showed us noobs how to do fairly straight forward power mod on a GeForce GTX 1080 Ti card. The mod basically involved applying liquid metal to the three shunt resistors (PCIe slot, 6-pin and 8-pin) that measure the power going to your card. This shorts them, basically fooling your card’s measurement capabilities which allows you pump more power through the GPU. A nice quick but effective mod. A few days ago however, he got an email from a YouTube subscriber called Chris who had applied the power-mod demoed by der8auer, but in doing so had damaged his card.

Chris in fact reported that after five or six months his rig was failing. After debugging the system he found that in fact the shunt resistors had literally fallen off the PCB. The only real explanation is that the gallium is reacting to the solder or tin used to attach the resistors to PCB. In fact, in certain communities this has almost become a known issue with some overclockers recently starting to insulate the solder.

In his most recent video der8auer goes back to his original GTX 1080 card to see how it’s faring. The card in question has been in use on his regular day-to-day use rig and has never shown any signs of instability or failure. In fact as far as he can tell, the card still manages the same level of performance that it was capable of when it was new. Der8auer removes the heatsink and strips it down the PCB and takes a good long look at those shunt resistors. What did he find? You’ll have to watch the video to find out.

Catch the video here on the der8auer YouTube channel.


14

Taiwan sdougal says:

Anyone finding issues with shunt resistor modding with liquid metal?

H2o vs. Ln2 says:

The only issue... will not improve performance on yours videocard at all.

Croatia dabar_Solta says:

why not simply solder 0r resistors of same package?

United States The Undertaker says:

Liquid metal is easy to apply and remove, that's probably why. :)

Germany der8auer says:

dabar_Solta said: why not simply solder 0r resistors of same package?


0 resistance will cause the card to go into an error state so it will always clock at 137 MHz


H2o vs. Ln2 said: The only issue... will not improve performance on yours videocard at all.


Of course it does if you do heavy OC you will always run into the power target.

Croatia dabar_Solta says:

der8auer said: 0 resistance will cause the card to go into an error state so it will always clock at 137 MHz




Of course it does if you do heavy OC you will always run into the power target.


so this liquid metal has some resistance? what is original rsense value? could you simply put other smaller value ?(like this WSL2010R0500FTA Vishay / Dale | Mouser Europe)

H2o vs. Ln2 says:

der8auer said: Of course it does if you do heavy OC you will always run into the power target.


Run GPUPI in your's Liquid Metal mode to find out how much you can increase the speed. ZERO MHz for me under chiller at 0Celsius with GTX 1070, 1080, 1080 Ti, Titan X Pascal or XP. I don't believe any subzero Ln2 users will mode the card as you do, but thanks to give a try.. and make me busy :)

United States bigblock990 says:

H2o vs. Ln2 said: Run GPUPI in your's Liquid Metal mode to find out how much you can increase the speed. ZERO MHz for me under chiller at 0Celsius with GTX 1070, 1080, 1080 Ti, Titan X Pascal or XP. I don't believe any subzero Ln2 users will mode the card as you do, but thanks to give a try.. and make me busy :)


Have you tested with real 3d like firestrike, catzilla ect? gpupi is only compute

Germany der8auer says:

H2o vs. Ln2 said: Run GPUPI in your's Liquid Metal mode to find out how much you can increase the speed. ZERO MHz for me under chiller at 0Celsius with GTX 1070, 1080, 1080 Ti, Titan X Pascal or XP. I don't believe any subzero Ln2 users will mode the card as you do, but thanks to give a try.. and make me busy :)


We had to use this on our Titan Xp to run Vantage at 2400 MHz+....

United States The Undertaker says:

Not gonna lie... chilled water is slightly different to LN2 and what sort of power limits you can hit... Roman definitely didn't do it for no reason, and neither do the other big players. Come on... :D

H2o vs. Ln2 says:

bigblock990 said: Have you tested with real 3d like firestrike, catzilla ect? gpupi is only compute

Catzilla always first try because my setup but the performance decreased. On Time Spy don't see any improvement then fires trike never try. Regarding "power target" with GPUPI, I was not able to increase the speed by 13Mhz before benchmark crash.

H2o vs. Ln2 says:

der8auer said: We had to use this on our Titan Xp to run Vantage at 2400 MHz+....


Than work for LN2 only, I can't increase my GPU by 13 MHz at 0C.

United Kingdom mickulty says:

The power mod isn't to allow you to set higher clocks, it's to get the nvidia driver to run higher actual clocks at the same settings. The increased average clocks means it might even reduce what clocks you can set, but the performance at the same settings should increase.

Germany der8auer says:

Here are graphs I made of the initial power mod testing. It shows a run in Unigine Superposition:

Before modding:
[img]https://i.imgur.com/jYm9tzV.png[/img]

After modding:
[img]https://i.imgur.com/FJff4dv.png[/img]

You can clearly see how much more stable the GPU frequency is during load and also how the gpu voltage stays stable through the run. The final score was about 200 points higher without touching any GPU clocks. So saying it doesn't do anything is clearly not correct.

Please log in or register to comment.