If I were a millionaire... Infinite Fabric would be perfect.
If I were a millionaire... Infinite Fabric would be perfect.
I did what I planned, that's how the mobile device was configured. I just adjusted it to 1600 so everything balances at a 1:1 ratio. The single-core performance in GPUZ dropped slightly—from 481 to 485—and multi-core stayed roughly unchanged. However, my valley benchmarks improved from 6400 to 6600, which is a noticeable change. Valley doesn’t rely heavily on the CPU, but it still uses it (about 14% on average).
It matters when you run 10 to 20 times and calculate the mean. One session by itself offers little insight.
Sure, I followed four runs each time and observed variations, though the outcomes rarely surpassed the top scores—typically around 485, 486, 481, 485. When I adjusted the IF to match certain values, a pattern emerged, suggesting some consistency. However, I’m unsure how trustworthy this method is, especially with the rapid results it produces. The valley runs also came up twice, but that might not reflect true performance either. Overall, I’m more concerned about those low-output segments than the overall test reliability.
For dependable evaluation, consider this alternative. https://benchmate.org/
There are quite a few benchmarks listed... it seems they’re all executed sequentially, forming a kind of benchmark collection. That made me consider reverting to my initial configurations for more accurate comparisons. I reset the FCLK and PBO settings to "auto" and did the same. Interestingly, the FCLK remains at 1600 instead of the previous 1800 when auto was active—seems like the motherboard adjusted its behavior. Also, my Prime95 temperatures have dropped to around 76°C compared to about 86°C with PBO enabled... that’s definitely a change. Plus, CPUZ scores are performing better than ever, so having FCLK and UCLK aligned at 1:1 might be beneficial. Multi cores didn’t vary much, but single-core improved by 8 points over when PBO was active—still a noticeable difference. Cold as ice! I’d love it more, but the 10°C gap with PBO on auto still stands out, especially considering how Ryzen boost works and how cooler components help.
It seems like you're noticing something unusual about the CPUZ benchmark. You think it might be related to a power-saving mode or something similar, but it's not always clear. Right now, you're trying to understand why the results look better when the CPU is off. It sounds like you're considering 3DMark as a possible test for that scenario—quick, practical, and reliable.
Yeah, I will at some point, for comparison sake, but I'd prefer something with a more realistic workload. I mean for stability P95 should be fine (and it's quick) Okay, but now I tried something and I don't get it at all... So I set PBO to "advanced" Clock "override" to +50mhz (25 is smallest step) and voltage "offset" - 250mv ... And it boots, but now it's back to 1800 for FCLK ! I didn't change that at all, just looks like the mobo really likes it at 1800, I did notice it did feel a bit sluggish with IF at 1:1, so opposite of what it should be, weirdly? Anyways I'm not touching that again for now, now i have to check my "undervolt/ overclock/PBO" thing for any real world gains and stability, if it's stable I probably should push it a bit more, 50mhz isn't exactly much (though of course I need to see how it actually behaves first to judge this)
For optimal results, try a clock tuner tool like this one: https://www.guru3d.com/files-details/clo...nload.html