pastrychef
Moderator
- Joined
- May 29, 2013
- Messages
- 19,460
- Motherboard
- Mac Studio - Mac13,1
- CPU
- M1 Max
- Graphics
- 32 Core
- Mac
- Classic Mac
- Mobile Phone
Hi, there @pastrychef.....hope you are doing well......I have a question about a Vega 56 and liquid metal, and I thought you might have some insight.
So, as part of my upgrade to the Radeon VII, I sold my two Sapphire Nitro Vega 64s and purchased a used Vega 56. I didn't pay much, and in testing it, I can tell it's been modded with the Vega 64 bios. That's fine with me as I planned to do that anyway. I can always switch it back to the stock bios if I want (that wonderful little switch in the back of it).
The performance is really decent (a very pleasant surprise). However, even at idle, it runs a bit hot. Not awful, but just hotter than I would think given it's just sitting there (45 degrees, when it's doing nothing). So, I'm wondering about applying liquid metal to it.
Which leads me to two questions:
1. Did you see a large temperature drop after the liquid metal application on your Vega 56?
2. In applying LM, do you apply it to both the GPU and the heatsink/ihs? Because in the past, when I did CPU paste, I always put it only on the CPU. But I'm under the impression that to properly apply liquid metal, applying to both the processor unit and the heatsink covering it is the proper way to do it. Is that correct?
For idle temps, I don't think the liquid metal does much. I think that how your fan is set and how the core and HBM are clocked matter more. I currently have my fan set to be ~800rpm during idle and my temps sit around 42C which equates to ~108F. I don't think that's too bad considering my CPU sits around ~35-36C which equates to 95-97C and it has a much larger heatsink and two big fans.
To me, the important thing is how temperatures are under load. To really maximize performance on the Vega cards, it's very important to prevent them from throttling. I did lots of trial and error testing to try to keep temps under 80C when running Heaven and/or Valley benchmarks. For this, the liquid metal seemed to help quite a bit. I noticed that fans didn't have to work as hard to achieve the same temps and the temps dropped back down faster after applying the liquid metal.
Early on, users used Vega 64 firmware to try and improve Vega 56 performance. The reason it helped was that it allowed users to increase the power limits beyond what was allowed on stock Vega 56 firmware. However, as time progressed, people began to realize that undervolting resulting in better performance and thermals. On my card, I went from a stock 1200mV down to 1010mV. Then, I overclocked my HBM to 920MHz. This resulted in really, really good performance. Better than most stock Vega 64s.
Please try my VGTab kext to see how it performs and how temps are...