Adventures in 1700 Overclocking. Somewhat success

parsec wrote: CocoaThumper wrote: [URL=][/URL] jj22ee wrote: I have the same problem.? If I change the voltage in pstate0 to anything besides stock.? It defaults to pstate1 freq.? So I am limited in my overclocking to stock voltage and 3.7 ghz.? If I overclock the non pstate method it works.? But it will not throttle down. 3.7 is plenty fast right now.? Overall I am loving ryzen and my asrock board.? Just ready for some of these bugs to be worked out. Hopefully we get a new BIOS update soon. I'm enjoying Ryzen and my Asrock board despite the kinks. nangu wrote: @CocoaThumper: Thank you, very informative post. I managed to pstate0 overclock my R7 1700 to 3.8 Ghz this way. It only needed one "click" to the CPU offset voltage to reach it, really easy to do. I have to do more testing, and I hope these bugs will be ironed out so we can set P1 and P2 with custom voltages too.? Which voltage value did you set for your offset? Unfortunately for me I can only offset voltage for a 3.7ghz overclock. If I use an offset higher than +130mV, I cant load Windows, even though the motherboard Posts. Hopefully sometime soon we get a new BIOS so I can use the higher offsets properly so I can go higher than 3.7ghz I know you are using PStates, but in the OC Tweaker screen, is the CPU Frequency and Voltage Change option still set to Auto? If you set it to Manual, you can then enter a VCore value, which will be reflected in PState 0. I use Offset voltage greater than +130mv (+300mv), and Windows starts fine. That you cannot does not make sense, not sure what is happening.

I still have my OC Tweaker CPU values set to AUTO. If I set it to manual, then both Pstates P0 and P1 will change to whatever is in OC tweaker. And then I lose my 2.7ghz middle downclock.

Right now I have OC Tweaker on Auto, and set my overclock using P0 manually, with P1 and P2 on Auto. This lets me downclock properly to P1 @ 2.7ghz and P2 @ 1.55Ghz

My goal is not just simply getting a higher overclock. I can get to 3.8ghz by using OC tweaker, or further messing around in the Pstates. However, my overclock ends up running 24/7 and never downclocks. Plus there's the fact that my system will not load Windows if I use a VCore offset higher than +130mV

The problems I highlighted in this post and in my OP are something Asrock need to address so we can all properly use several manual Pstates, proper voltages, and have custom downclocks work.

PS - Based on a couple responses in this thread, and from talking to folks on reddit, these problems are pretty common.

Edited by CocoaThumper - 07 Jul 2017 at 9:50pm

Watching this thread....haven't attempted offset overclocking yet.

I don't get any downclocking issues when OCing with OC Tweaker so haven't tried pstates yet, thats with a Fatal1ty Gaming K4 Asrock sent me. I wouldn't necessarily hold out hope that newer AGESA code will improve overclocks all that much as the biggest limitation is the lithography used with Zen.

Long story short, the lithography has 2 "critical" points (when things start to stop scaling linearly) The first critical point is right at 3.8GHz, the second at 4GHz. Newer CPU revisions might extend these critical points as new CPU revisions / steppings often use tweaked lithography techniques but don't be surprised if you have to wait until Zen 2 / 3 before overclocks of 4.5GHz or more become common place.

One area you can hold some hope for though is regards SMT. With some testing I did the 1700 I have would run and be semi-stable @ 4GHz with only 1.3v but that was with SMT off. With SMT on 1.42v+ was need for 4GHz to be about as stable as it was with SMT off. Heres a copy / paste of part of the results I noted down post K4 review when I had more time for some testing;

3.7GHz - Stable @ 1.25v SMT on (might of been stable on lower voltage I was aiming for maximum frequency without crazy voltage)
3.82GHz - Stable @ 1.32v SMT on (you'll notice the non-linear voltage bump for 3.8GHz as "critical 1" is hit)
3.9GHz - Stable @ 1.38v SMT on (done post review of the K4)
4.016GHz - Semi-stable @ 1.45v+ SMT on

Needless to say, the 14nm lithography really doesn't scale well once you hit 3.8GHz+ and even then it's debatable if the voltage increase is worth it for a extra 50-100MHz  but even at that speed by comparison the Zen 1700 stomped all over the old i5 3570k @ 4.5GHz with 2133MHz RAM test system in every scenario, gaming included.

At the end of all this testing one stark thing that can be taken away is that while Zen may not have huge scope for overclocking ("huge" being relative of course, in my case the 1700 when left to its own devices will boost to 3.2GHz, where a manual OC can take that all the way up to 3.9GHz.. so a 700MHz OC certainly isn't what I would call bad by any means) memory speed for 1080p gaming at least can have a large impact on performance. I gained more FPS with memory speed increases than I did overclocking the CPU itself.

You can take a gander at the review I did for more details HERE