Soldato
I did notice that in the BIOS it is sitting at 1.48 volts on "Auto"!?
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Zen 2 Owners Thread!
- Thread starter SiDeards73
- Start date
I did notice that in the BIOS it is sitting at 1.48 volts on "Auto"!?
Caporegime
That's high!. Mines on 1.375. I think others have gone lower than that too.
Permabanned
Is everyone manually setting volts or leaving it on auto?
Caporegime
Mines on auto. It’s a 570 Gb Ultra.
A lot ot bios show false reading in the bios though and it could be showing 1.48v coz only 1 core is active at boost speed etc exp on these 3000 cpu ignore the bios readings who knows if these updates bios even remover there is no 20 degrees offset etc anymore use Ryzen master in windows is best way to gauge the temps although at idle expect them to be a lot higher then the intel cpu
Soldato
My 3900x idles about 34c custom water loop, fixed 1.325v 4.35ghz
Can owners please cast an eye over this:
https://www.reddit.com/r/Amd/comments/ch1kkl/3900x_memory_speeds_3600_or_above_limit_boost/
Relates to Memory and Infinity Fabric speeds affecting boost clocks and idle states.
https://www.reddit.com/r/Amd/comments/ch1kkl/3900x_memory_speeds_3600_or_above_limit_boost/
Relates to Memory and Infinity Fabric speeds affecting boost clocks and idle states.
Soldato
When you say "normal" for voltage. Do you mean auto or static? If so, this could be boosting higher at a lower FCLK due to lower power usage and being further under the FIT mark to allow for further boosting.
Neither.
BIOS has 3 options. Auto, Normal and Manual. Normal allows an offset but I have +/-0v.
My SOC is manually set to 1.1v so I'm not sure how the available power is being eaten by Infinity Fabric usage to such an extent as to rob the cores of 275Mhz of boost.
Soldato
The 3000 series FIT seems to be quite sharp/strict, it really wouldn't surprise me that boost clocks would fall off a cliff at one given point.
Though equally I wouldn't be surprised if this was a bugged BIOS/AGESA issue.
Worth @-ing The Stilt in your post over at OCNet
Edit: Maybe @gupsterg could also test and see if he can recreate similar results.
Last edited:
Ive been data logging R20 single core runs at 3533 / 3600 / 3773 and will graph them out.
One thing I noticed was a significant jump in CPU Core Power from 3533 to 3600 but not again from 3600 to 3733. SoC power saw a marginal uplift but almost insignificant.
It's odd behaviour that the Core Power raises when the average boost clock is lower.
Having lowered my VDDG to 1000mv I've noticed that the 3600 runs will exceed 4250Mhz now (But still won't idle properly) but I've not got evidence to support a correlation yet. Next on my list.
So I've been graphing out some power usage plots between memory speeds and also the impact of adjusting the VDDG. I've noted that with a manually set VDDG of 1000mv my 3600 and 3733 boost clocks improve beyond 4250 max up to 4300-4400ish. As a result i decided to plot power differences between VDDG of 1000mv and 950mv also. All readings were logged from a single core Cinebench R20 run.
Memory Speed Comparisons - Power Consumption
The take away from this plot is that there is a noticeable jump in Core Power usage from 3533 to 3600 but almost no jump at all from 3600 to 3733. The SoC power consumption has a marginal bump for each increase in memory speed but overall nothing significant.
VDDG Comparisons at 3600 - Power Consumption
Reducing VDDG universally reduces power consumptions as expected. It's also seemingly linear across all the readings.
CPU Vcore Comparisons
Having changed nothing else increasing memory speed lowers CPU Vcore
Memory Speed Comparisons - Power Consumption
The take away from this plot is that there is a noticeable jump in Core Power usage from 3533 to 3600 but almost no jump at all from 3600 to 3733. The SoC power consumption has a marginal bump for each increase in memory speed but overall nothing significant.
VDDG Comparisons at 3600 - Power Consumption
Reducing VDDG universally reduces power consumptions as expected. It's also seemingly linear across all the readings.
CPU Vcore Comparisons
Having changed nothing else increasing memory speed lowers CPU Vcore
Last edited:
Permabanned
its called beta bioses
Higher memory speed appears to be drawing more power from the package as a whole and is resulting in lower boost clocks hence lower vcore. Quite why the SoC is sapping power from the cores to such an extent when there is plenty of PPT headroom is beyond me.
Caporegime
They need to fix that
Soldato
The gigabyte guy at reddit said AMD and board partners know about it and is getting fixed.
The SoC Power usage [W] between 3533/3600/3733 is largely unchanged (<0.5w variance). There is a marginal uplift of Amperage for each speed bump (<0.5A) and SOC voltage is largely unchanged.
A Single Core Cinebench R20 is not power limited though. Max CPU Package Power of around 77w.....PLENTY left out of the stock 142W on a 3900X. So why the throttling of the cores?
Soldato
I cannot find what VDDG is on the Taichi to help you more, and I believe LLC = 3? (there is no standard in Asrock ever)
However when I had the CPU on Game mode (6 core) and tried the 3533Mhz ram with 1767 IF, I saw the cores clocking 4540, otherwise gets to 4270ish.