I.nfraR.ed

All old dropbox share links are not working anymore. I've shared the folder again (no individual links): https://www.dropbox.com/sh/b702d3y6guzdzyy/AAAq_b0WcaF1fIiQNrUdTjB5a

He has a better one. I literally have no idea about tweaks yet.

Yes, that's because of the hw prefetchers, but on XP and Zen3 haven't experienced the same. Granted my bios lacks those options. Most of the times 3rd run is very slow already and have to restart the system.

I've tried all the tweaks from the old Bulldozer Conditioner (yes, I've disassembled it some time ago and know exactly what it does), also some of my own test tweaks I was working on for the previous zen generations. If anything, one of the registers makes it extremely slow and the whole system gets laggy. Most of the registers are undocumented, the BKDG for 15h has more information than the document for 17h, but there's probably no parity between all the features and a specific bit in a register on 15h might be something else on 17h. Also Zen doesn't support Trailing Bit Manipulation (TB

@TerraRaptorI don't think so. I will take another look on Zen3, but I could not improve performance on Zen2 much. It had some effect on Zen1 though. The Conditioner uses some MSRs which are also used by the Performance Bias on Ryzen platform, which has no significant effect on the newest gen. Zen doesn't support Trailing Bit Manipulation anyway. I will try to boost performance in SuperPi, but I doubt there's anything significant to gain.

First tries with 5600X, currently on Win 10. Not bad for this board, I think. I can't boot XP anymore with this gen and bios.

Great one. My stick died at similar voltage IIRC.

True, you can't get the real CPU name, it's not fused on the CPU, but is matched to a table depending on the detected features, core, FSB and multiplier. That's why if cache is damaged (or unlocked) it gets detected as something else. Anything running out of stock invalidates the detected model and "stock" sometimes means completely different on different motherboards. It's really easy to cheat with model numbers, but I hope people benching those CPUs are honest. You can basically submit all XP-M Bartons with just one CPU. Even a photo of the CPU doesn't prove anything, exc

B350 motherboard running with B450 bios. Wraith Prism cooler.

Oh, you finally did it! Missed that submission. Congrats I will come back for it some day, promise!

Wait, 2033 is much faster than 2100? Is that the slow down some people talk about? I guess you don't want to disclose your timings, but if you run latest beta of ZenTimings, could you spot some differences in non user-configurable timings? I don't have 5000 series to check myself. Is it strap related? For example, is it the same if you use the lower strap and clock the bclk up or is it just some internal strap that makes things waaay slacker?

Possibly yes, I don't recommend it. It was a suicide and I was aware of potential loss of hardware. Especially on non-ceramic Athlons. Just for reference - I don't have it noted, but if I have to assume, knowing how they scale on cold, probably 2600-2700 valid with ~2V.

I benched it before it was cool :p

The "official" 1.2.1 is available on the site and has Vermeer support fixed.

New beta with Vermeer support somewhat fixed. You might have some readings missing though. Don't have a CPU to properly debug. ZenTimings v1.2.1.65 beta2 https://drive.google.com/file/d/1dC12pYM2qyY8D9aoEd192NtAeVh7m9uJ/view?usp=sharing