SOME RYZEN 7 1700 OVERCLOCKING!

steve258 · 27 Mar 2017 at 09:51

BCLK clock adjustment on B350 is very interesting indeed. Does anyone with the Tomahawk have any actual success with BCLK OC?

Scougar · 27 Mar 2017 at 14:57

steve258 said:
BCLK clock adjustment on B350 is very interesting indeed. Does anyone with the Tomahawk have any actual success with BCLK OC?

This I really want to know.

The only downside is the power phases of these boards are much worse than the X370's. The best board for power phases is the ASROCK Pro 4 (B350) I believe. 9+2 or something.

djbully · 27 Mar 2017 at 16:01

Interesting, on IBT, I have used it in the past, but it always passed on my Intel that I setup with P95. I did notice that it gets the temps 1-2 degrees hotter though!

I will test my ryzen tonight with it, just to make sure of stability.

FYI, I'm @ 3.9 1.42-1.48V anything less is not prime stable.
I have put close to 1.6V into this thing but no luck getting 4.0 stable.....

gavinh87 · 27 Mar 2017 at 16:12

djbully said:
Interesting, on IBT, I have used it in the past, but it always passed on my Intel that I setup with P95. I did notice that it gets the temps 1-2 degrees hotter though!

I will test my ryzen tonight with it, just to make sure of stability.

FYI, I'm @ 3.9 1.42-1.48V anything less is not prime stable.
I have put close to 1.6V into this thing but no luck getting 4.0 stable.....

I think a lot of people have been posting results with low voltage at 4.0 running cinebench and calling it stable. I think realistically 3.9 at around 1.4 is more likely. Especially when you look at silicon lottery results where they have the same.

Awahwah · 27 Mar 2017 at 16:50

I think you guys are on to something here. I've dropped my core down a little and the jump downward in volts is big. 3.9 needed 1.36v to be stable, 3.85 is 1.30v. 50mhz for all that voltage, will I notice the 50mhz - no sir. What I've found, we the chip is on the edge, Cinebench produces a lower score than you'd expect - eg 3.9 was 10 points higher than 3.8.

I now run multiple tests to ensure stability and gaming for a couple of hours. One thing I have noticed, I have greater stability at a lover core voltage now I only have one Graphic Card.

flea-rider · 27 Mar 2017 at 16:55

ok so I've got it at 4ghz and 1.37 volts
so what tests do I need to run ? just to make sure she's ok ?

gavinh87 · 27 Mar 2017 at 16:57

flea-rider said:
ok so I've got it at 4ghz and 1.37 volts
so what tests do I need to run ? just to make sure she's ok ?

Realbench stress test at max ram for at least 1 hour. I'd also give intel burn test at maximum settings a go to, this found my instabilities where others failed. (10 runs of IBT should take around 90mins)

Scougar · 27 Mar 2017 at 17:16

Awahwah said:
I think you guys are on to something here. I've dropped my core down a little and the jump downward in volts is big. 3.9 needed 1.36v to be stable, 3.85 is 1.30v. 50mhz for all that voltage, will I notice the 50mhz - no sir. What I've found, we the chip is on the edge, Cinebench produces a lower score than you'd expect - eg 3.9 was 10 points higher than 3.8.

I would suggest that there is internal throttling going on, or error control is occuring where it identifies errors corrects it and that impacts performance. So, either throttling is not shown, or errors are occuring.

I don't have Ryzen yet, but on my overclocked 7850K, in order for the chip to not throttle on the limit, I had to manually edit the P-states to stop it dropping the multipliers. I really hate the internal protection system, that they make it awkward to get around.

FredFlint · 27 Mar 2017 at 18:01

Just installed 2*8GB G.Skill 3200 CL14 kit, went into BIOS and changed it to the 3200 profile, saved and booted into Windows. Get 1656 on Cinebench, on 2133 CL15 it got 1596. Will need to test it more but good start.

cliffy · 27 Mar 2017 at 18:59

steve258 said:
BCLK clock adjustment on B350 is very interesting indeed. Does anyone with the Tomahawk have any actual success with BCLK OC?

I tested it ibn mine. Max value I was allowed to enter was 103 anything more could not be entered. System was fine for normal day to day use (email, watching video, web browsing) but crashed quite quickly when using handbrake. Haven't had a chance to do any further testing yet however.

flea-rider · 27 Mar 2017 at 19:32

if you use hwmonitor .. can you tell me what your cpu vid is for ryzen plz ..

gavinh87 · 27 Mar 2017 at 19:35

I think it's 1.550 but it's incorrect

Scougar · 27 Mar 2017 at 19:45

cliffy said:
I tested it ibn mine. Max value I was allowed to enter was 103 anything more could not be entered. System was fine for normal day to day use (email, watching video, web browsing) but crashed quite quickly when using handbrake. Haven't had a chance to do any further testing yet however.

Only very minimal adjustments? That sucks. It is the ONLY reason to really buy an X370 unless you somehow need all the features, and that is just silly to buy a really pricey board if you are buying one of the budget options (1700 or lower, especially the R5 stuff)

Awahwah · 27 Mar 2017 at 20:21

Scougar said:
I would suggest that there is internal throttling going on, or error control is occuring where it identifies errors corrects it and that impacts performance. So, either throttling is not shown, or errors are occuring.

I don't have Ryzen yet, but on my overclocked 7850K, in order for the chip to not throttle on the limit, I had to manually edit the P-states to stop it dropping the multipliers. I really hate the internal protection system, that they make it awkward to get around.

You are probably correct there.

KrisWragg · 27 Mar 2017 at 20:53

Thanks to the comments on here about IBT, I had my 3.9Ghz OC RealBench stable for 4hrs at +7500 offset, IBT failed in about 1 minute... pushed up load line calibration from level 5 to level 3 and its now stable with IBT.

Seems IBT is a much quicker way of finding instability!

I tried quite a bit for 3.95Ghz, a mere 50mhz extra but the voltage jump was getting silly and still wasn't stable. In the end I stopped at 1.437v (under load), compared with 1.381v under load for 3.9Ghz.

BTW what exactly does load line calibration do, other than control vdroop? Is there any reason not to just set it to level 1 where I get absolutely zero vdroop? Looking at HWiNFO my VRM temps are not exceeding 51*C under full load.

flea-rider · 27 Mar 2017 at 21:02

KrisWragg said:
Thanks to the comments on here about IBT, I had my 3.9Ghz OC RealBench stable for 4hrs at +7500 offset, IBT failed in about 1 minute... pushed up load line calibration from level 5 to level 3 and its now stable with IBT.

Seems IBT is a much quicker way of finding instability!

I tried quite a bit for 3.95Ghz, a mere 50mhz extra but the voltage jump was getting silly and still wasn't stable. In the end I stopped at 1.437v (under load), compared with 1.381v under load for 3.9Ghz.

BTW what exactly does load line calibration do, other than control vdroop? Is there any reason not to just set it to level 1 where I get absolutely zero vdroop? Looking at HWiNFO my VRM temps are not exceeding 51*C under full load.

got my load line calibration set at lvl 1 ... if your pushing it set it high

KrisWragg · 27 Mar 2017 at 21:47

flea-rider said:
got my load line calibration set at lvl 1 ... if your pushing it set it high

Given that a try, can drop the offset a little with LLC level 1. I'm finding the temps on Ryzen a bit weird, with a Noctua D15 I am getting temps of about 76*C Tdie (96*C Tctl due to 1700X offset), but the cooler is literally luke warm, I can put my hand on it and it's barely noticeable. Might try reapplying the Noctua TIM and reseating the heatsink?

BigDaddyKong · 27 Mar 2017 at 23:03

Level 5 is the highest LLC setting isn't it? I seen one chart that showed it backward. If I put mine on level 1, it does almost nothing.

gavinh87 · 27 Mar 2017 at 23:09

BigDaddyKong said:
Level 5 is the highest LLC setting isn't it? I seen one chart that showed it backward. If I put mine on level 1, it does almost nothing.

You are correct 5 would be the highest.

BigDaddyKong · 28 Mar 2017 at 03:24

How do you get rid of the power management error in the event viewer after you overclock?