Ryzen Pstate OCing - K17TK and similar software
Hi all, asking here because some people seems to have already tinkered enough with Ryzen's Pstate OCing.
Since my mobo doesn't support Pstate OCing in BIOS (AB350 Gaming), I was looking into the software side of things. Since I'm running Windows, 2 pieces of software came up:
- ASUS Zenstates (limited to ASUS motherboards)
Did anyone try the last one? Seems an user in another forum had some issues using it:
Originally Posted by Formula.350 @ [H]
Things to keep in mind... I do not suggest utilizing the CPB and C6 disabling options. The first time I played around with it, the program worked. All I did was change P0's multi to something higher. I tested in AIDA64 to make sure it really was working, and my performance increased quite a bit. Then I started to tinker with more settings, and somewhere in doing so, I've stumbled across some kind of bug.
Well, actually there seem to be 2 bugs. First one is that once you apply anything, Core Performance Boost (CPB) will cease to function. Meh, fine. However, the second bug is way more of an issue.
Let me start by saying that restarting the system will, naturally, revert all changes. Problem for me is, something occurred (and seems to now happen any time I use the program to apply settings) where it now will degrade performance and cause things to behave as if it's not actually running at the clock speed that everything reports it's running at. For example, running normal, before applying settings, my scores in AIDA are as normal. One I apply a setting, even if it's changing the frequency +25MHz, performance drops. I don't feel this is due to loss of Boost speed, either. I can crank the clock up to 3700 and performance isn't what it was the first time I used it to increase multi to 36.25, as it falls quite a bit short actually. It's all the AIDA tests, too, not just one in particular.
Oh, also, the DID (Divisors) don't work 100% correctly. For example, you'd expect that a Multi of 37.5 with a Divisor of 1.125 would equal 3333MHz, and it's little output does indeed state that. However, it actually just seems to select the closest actual Multiplier that it would come close to. In other words, despite setting 37.5 FID and 1.125 DID, the resulting frequency is 3325MHz, and so really all it has done is set the multi to 33.25. Not a HUGE deal, but still a bummer heh
Also note that on Windows 10 and likely Windows 8, changes to FID or DID using this utility will cause performance to be skew, at least in AIDA64 tests for certain (for example, any CPU/FPU or Cache tests. As such, if you're looking to compare overclocks to your own previous results, this will likely invalidate your data. I suggest using this tool only to find max table clocks with voltages OR for finding undervolting levels OR to simply use it to create your own PStates from Windows.
In any case, thanks in advance! '); }
Read responses in overclock.net