Try running with your case open and a box fan blowing at it. Several reviews said water-cooling ryzen can overheat components on the motherboard because there’s no fan blowing air over them.
1 Like
I will check that out after my re-paste. I don’t think that would be an issue, as I have 9 120mm fan cycling air in this case, but none of them are blowing directly on the socket area.
This seems to have helped a little. I am now able to hold a steady just barely shy of 4.3 single thread, where before I was at 4.2-4.25 and erratic. Before results were 80c max CB20 4848 at 4.05GHz and erractic, after 76c max CB20 4926 at a very steady 4.1GHz, single core 494 4.3.
I am still disappointed that the single core boost never really passed 4.3, but the clocks are looking much more consistent now. I may test it again with PBO off, but last time I did that I lost performance.
Thanks for the help!
for me, it is the opposite: pbo on = loss of performance. pbo off= on the box speeds.
You could experiment with cooler rotation; that does seem to make a difference. As does paste application. It seems to be super picky.
1 Like
Yeah, I noticed when I pulled the block off that the paste was spread out unevenly. This time when I put the paste on I knew where the cores were underneath, so I put a dot right on top the core. I am guessing my IHS isn’t very flat. May have to break out the sandpaper one of these days.
I’m playing around with undervolting myself currently and seeing some interesting results. For reference I’m now running a Asus B450-I Gaming and EVGA CLC120 (albeit with two Noctua fans) in a Silverstone SG13.
I’ve been Cinebenching (R15) various vcore voltages and testing with PBO enabled and disabled. Firstly, I’ve seen my highest overall clocks with default settings but also my highest temps and voltages (upwards of 74C and 1.5v). PBO on vs off are within 10 points of each other. The top three runs are both sans undervolting:
However, the next nearest is at 1.0v (1810 CB score). So there is a pretty big disparity there. However, I was trying this last night initially unaware of the Optimum Tech video and ran through a bunch of games and couldn’t discern a difference stock versus undervolted.
Another odd behavior is that I noticed a much higher sustained clock speed under Prime 95s Large FFT with lower voltages…
Needs more investigation but I’m inclined at the moment to leave it at 1.0volts… didn’t screen cap it but pretty sure my temps running CB were under 45C (and my fans are running around 800 RPM at anything under 55C).
1 Like
When I was playing with undervolting my temps would improve, clock speeds would still read good, but my CB20 score would go down directly in conjunction with the voltage, it was weird.
I’ve done a whole heap more testing and basically, undervolting to 1.0v (while leaving the boost settings to auto) essentially puts performance on par with manually setting the all core frequency to around 3400-3500.
So far, my best results have come from an offset adjustment of -.05v which was essentially margin of error change from stock. Granted that’s only a mild drop in temps (maybe 2-3C load)