Ryzen "single core" / "half-stable" OC

[Rebe1]

Greetings from a new guy!

My platofrm:
2600x + BQ Dark Rock 3 Pro
Asus ROG X470-F Strix
G.Skill FlareX 3200MHz CL14

I have been reading few interesting threads about some nice results of oldschool BCLK OC on Ryzen and decided to give a shot. Unfortunately, without a single positive result My mobo not only limits "BCLK OC" possibilities to integer numbers (so no chances for sweet 103.53) but also remains completely unstable when applying anything higher than stock 100 MHz... I was able to reach windows and open HWmonitor for few times, but usually system crashes before OS is loaded. I though the reason behind that was my RAM (I OC'd my FlareX to 3400 CL14 - rock stable after 50 loops of memtest) so I reverted any OC-related changes and keep entire mobo at factory settings - still no go

To be honest, I think most of the "BCLK OC" successful attempts I saw was in fact made on Asus Crosshair (VI and VII) or AsRock Taichi mobos. Is that the case here or am I doing sth wrong? Do you have any experience with Base Clock OC on Ryzen?

 

[Rebe1]

I forgot to mention: the purpose of the BCLK tunning is to see the highest stable clocks in e.g. Cinebench R15 single core test or a CSGO session - it do not have to be stable in normal, all-core stress test.

 

[Johan45]

Welcome to OCF Rebe1
BCLK OC affects all the other busses like memory/Fclk/PCIe etc.. Unless you have the synchronous setting in BIOS the you need to drop ram speed, PCIe gen from 3 to 2 and try to get your HDD on a connection from the chipset and not the CPU
I don't bother with BCLK OC, not that much to gain really and it makes so many other things unstable

 

[Rebe1]

I don't mind lowering RAM speed and setting PCIe to gen 2 mode if that will "unlock" the BCLK OC (even purely for the tests, not for 24/7 overclock) - rly thanks for suggestion. I started to read about synchronous/asynchronous base clock generators after I ordered motherboard. If I knew the fact, that my X470-F will not have this feature accessible (I believe each X370/X470 might have asynchronous BCLK generator but only in those top-end models user can make modifications to it) I would double think about Crosshair VI Hero (X370) as it cost the same money now.

My 2600x is running at 4.25 GHz all core (42,5 x 100) and the "sweet dream" was to push it a bit higher with previously mentioned 103.53 BCLK (4.4 GHz).

 

[Rebe1]

Thx, @Johan - I passed few CB R15 tests with new setup: 102 x 42.5 (~4335 MHz)! What I did was switching OCCT profile to MANUAL for RAM and setting its speed "1 lvl lower" than stock (3166?). With BCLK = 102 my ram was 3195 MHz. I passed first few "single core" tests (183 pts). Later I tried to tune my platform to pass 6-core test (SMT off) - to achieve that I had to pump VCore to 1.538 with LLC lvl 4. Max requested voltage was 1.58 (similar to values I see when XFR and PB are running with Performance Enhancer lvl 3 and LLC lvl4 @ voltage and cpu speed set auto. Result = 1048 PTS in CB R16.

 

[Johan45]

That voltage is way too high for ambient cooling.Max should be 1.4-1.45V I realize you're just testing but be wary of high voltage it can cause damage to the CPU

 

[Mandrake4565]

That voltage is way too high for ambient cooling.Max should be 1.4-1.45V I realize you're just testing but be wary of high voltage it can cause damage to the CPU

I agree with Johan....You may want to start considering some better cooling if you're going to be pushing the voltages that high. Being in Sweden you can always take advantage of the ambient outside air when the temps drop.

 

[Scu84St3v3420]

I was going to say that is WAY to high of voltage on ambient cooling and the gains aren't substantial enough to justify either. I have my 2600X @ 4.2 GHz 1.45V LLC 4, nearly 1.55V isn't justifiable for a gain of 135 MHz and 6 points in CB15 Single Core... Just my opinion, now if you were to run LN2 and pump way more juice into it for substantially higher clocks for competitive benching, that would at least make sense.

 

[Rebe1]

THX for feedback, guys. OFC, this was just purely for testing and to see how far can single core go on Ryzen 5 2nd gen. I also tried to play a bit with multiplyer only and have much better results (no pic though since It was during late night hours ).

I was able to get 4 successful runs of CB R15 (again, single core only) with 44.25 x 100 which ended with 190+ pts (fixed VCore of 1.45V, LLC4). I do not know if I am lucky or sth but my CPU had no issues booting up to 4.5 GHz (didn't try higher) with fixed x45 multiplyer. However with this setup only few times I managed to see the end score, never 2 in a row (result: 198-200 pts) but it gave me some idea what to expect from upcoming Ryzen 3 made in 7 nm. I'm pretty sure that the new CPUs based on Zen 2 chips will have a boost clock not lower than 4.6 - 4.7 GHz.

Since I still have a "time window" to send back my 2600x I bought a new CPU: R5 2600. Someone has returned it to my local store with damaged fan and scratched radiator (also, no thermal compound on it) so the end price for me was 147 USD. If it can run 4.2GHz @1.42-1.45V with LLC 4 I will save 100 bucks