b-die tweaking, timings ok?

[Waldorf]

finally found the cause for my "instability" (board added 0.01-0.08 Mhz to Bclk after cold boot scramble bios),
im finally able to tweak ram.
i got a corsair 3600c14/1T/1.45v LPX kit (2x8gb), and wanted to double check the timings.
im able to pass 100% of HCI (dos) with 14 flat, but want to tweak all other timings, before running 1000% pass,
and i still have trouble figuring what is interconnected/affected, or what makes (more) sense, vs just going lowest.
trying to go as fast as possible (within reason), but dont need to get the last 0.001%

5800x in eco mode, Gb Ultra x570, soc 1.05, vddp 0.95, vddg 1.0, proct 40, GD and PD are off.

(14-14-14-14)-30-46/1T
CWL ?
FAW 16
WR 8 (10/12)
RTP 5 (8)
RFC 288/280/272/264/256
RRD S 4, L 4
WTR S 4, L 8
WRRD 1
RDWR 8
cKe 1
RD SCL 4 (3)
WR SCL 4 (3)
WR SC 1, SD 7, DD 7
RD SC 1, SD 5, DD 5

 

[freeagent]

soc 1.05, vddp 0.95, vddg 1.0

For 3600?

I would do 1.1 SOC, and for just 1800 I have my VDDG's @ stock as well as CLDO. But for over 1833 I will bump them to .999 IOD @ 1.055, .999 and it runs fine. SOC gets a bump to 1.125 and is good for 1900. I am using the boards default subs for 2133 15-15-15..

 

[Waldorf]

yeah, thought that was enough (?!)
have fluctuations, so i tried not to have it go too high, 1.05 is 1.06-1.07 eff,
so i lowered it a bit, as it seemed to mess with usb.

vddg got set to "amd oc" to 1.05 after i set everything else, so i manually set it to 1v, to match soc.

just trying to see if timings are off, e.g. i set faw to 16, but you and others have higher numbers,
cke to 1 (stable with cl16), but see others running 3-7, and dont know much about things like proc/drv str.
all connected to the higher IF, or just not able to run lower (or even if it makes sense)?

 

[Ibizadr]

I have my gskill kit 3600 cl16 to 3800cl14 @1.5v. It's a b die kit so give it a try. I think it's the same chip gskill advertise 3600 cl14 @ 1.45.

 

[tabascosauz]

procODT usually no reason to touch if nothing wrong or obscenely off, not a performance enhancer. Same with drive strength unless you want 1T. Same with tCKE, varies with board/kit. None of them exist on a sliding scale where bigger or smaller number is "faster". Also don't copy others' procODT and tCKE especially if not single rank like yours.

3600 should be pretty chill but if you want 3800 then VDDG_IOD and VSOC gonna have to come up. At 3600 I think I was at 0.95V for both VDDGs and something like 1.03V VSOC.

For me 3600 14-14-14 and 3800 15-15-15 run the exact same VDIMM stable. Once I went to a loop I no longer had the ability to cool more than 1.45V so 15s is fine.

Generally no need to have VDDP any higher than 0.95V, doesn't seem to help anything, it's usually on VSOC and IOD to make a difference for IF if you struggle with IF stability.

 

[Waldorf]

lol, had hoped the usual suspects could comment

forgot to mention i wont touch if or cpu for now, as even in eco mode its fast enough.
im done buying fast kits (e.g. 4000@19) and hope lower timings at lower speeds will work,
so went with the 3600c14/1T, to have an easier starting point for timings.

didnt see a drop in latency with rfc264, so i upped to 280.
or should i go 1 more step up/down?
besides stability, anything on 2nd/3rd timings that can lower perf if too low?
as in, it might make sense to go with say 2/3/4/5/6 vs my 1 (cke/RD/WR stuff)?

thanks for all answers btw.

 

[Waldorf]

what volt is usually causing the usb trouble (sticks switch between off/online, kb/mouse acting up)...
still have it now and then.

 

[freeagent]

what volt is usually causing the usb trouble

RMA your CPU. I have two Ryzen 5000 series chips and none have usb issues.

Ram looks ok.. way tighter than I run mine

 

[Cowboystrekk]

what volt is usually causing the usb trouble (sticks switch between off/online, kb/mouse acting up)...
still have it now and then.

What agesa you on? Try a later agesa and see if it helps. I had those issues with agesa 1.1.0.0, got fixed with agesa 1.2.0.0. I would try running ram at 3800cl15, it will improve speed at bit and doesn`t require that much more voltage. Mine run 3800cl15 1t 100% stable at 1.07v soc, 0.98v iod and 0.82v ccd and vddp. Running in eco mode I would consider curve optimizer, gives you 5-10% extra performance at same consumption\temp etc.

 

[Waldorf]

its not the cpu, as a 3800xt on different board does the same (with identical settings),
nor do i have it with stock use.

1.2.0.4A

 

[tabascosauz]

what volt is usually causing the usb trouble (sticks switch between off/online, kb/mouse acting up)...
still have it now and then.

its not the cpu, as a 3800xt on different board does the same (with identical settings),
nor do i have it with stock use.

1.2.0.4A

USB issue really isn't something you can just easily tweak away......you could try changing VSOC either direction or IOD.

AGESA 1200 and 1201 were supposed to fix the USB shenanigans and they did the most pressing concerns but I've never had a B550 board that didn't have the occasional disconnect sound. B550M TUF, B550I Aorus AX, B550 Unify-X. My current X570 Impact is the only board that's completely problem-free on all BIOSes I've tried.

 

[Waldorf]

not the known usb issue, just minor things like num lock turning off etc.
not happing outside the oc, just didnt want to apply randomly, but expected it to be those.
i had them on the lower end, adding .01 fixed it.

 

[Chomiq]

The a patch of agesa 1.2.0.4 has a VDDG CCD and IOD bug that puts them fixed 1V no matter what u have selected in BIOS

https://www.reddit.com/r/Amd/comments/rh6s7q

Can't confirm that as I opted for one bios before 1.2.0.4a.

Aren't USB disconnects related to buggy AGESA?

 

[Ibizadr]

What agesa you on? Try a later agesa and see if it helps. I had those issues with agesa 1.1.0.0, got fixed with agesa 1.2.0.0. I would try running ram at 3800cl15, it will improve speed at bit and doesn`t require that much more voltage. Mine run 3800cl15 1t 100% stable at 1.07v soc, 0.98v iod and 0.82v ccd and vddp. Running in eco mode I would consider curve optimizer, gives you 5-10% extra performance at same consumption\temp etc.

Its any benefit in try to lower VSOC vddg vddp?

 

[Cowboystrekk]

Uses less power on IO-die so CPU clocks boost higher. On dual ccds you usually need higher ccd than on single.

 

[Waldorf]

AGAIN, never had any issues with disconnects.
only with KB/mouse not working as expected on manual config.

@Chomiqcve
you need to tell that to my crappy Gb board, can set all (needed) volts manually, and able to see the difference
guess my bug collection cancels each other out
should i still go back to pre 1.2.04?

 

[Ibizadr]

Uses less power on IO-die so CPU clocks boost higher. On dual ccds you usually need higher ccd than on single.

So by your point maybe I have to much power on ccd and vddp? I think I can't lower my iod what you say about?

 

[Cowboystrekk]

So by your point maybe I have to much power on ccd and vddp? I think I can't lower my iod what you say about?

It may be needed, but if you can run them lower you can get a bit more performance since IO-die uses less. Try aida, if latency increases atleast one volt is to low. Suggestion: try iod 1000mv, ccd 950mv and vddp 900mv, it may be to low for you, but will probably work, maybe you cab run as low as me.

 

[Waldorf]

Still had blck fluctuating, and while it now stays below 101 (101-109Mhz before),
prev release seems to need more soc (1.031xxV) to prevent cold boot trouble.

gonna keep it set like this for a few days before redoing ram tweak.

 

[freeagent]

You can turn spread spectrum off and that should keep your clocks steady.

 

[Cowboystrekk]

You can turn spread spectrum off and that should keep your clocks steady.

And it might stabilize ram oc, did that for me

 

[Waldorf]

lol, not sure why anyone would assume i have it enabled, even without looking at bdie or tweaking secondary/tertiary timings,
as this should never be used enabled for 99% of users.


the bsclk (only one affected) is set to 100.00, changes happen after a cold boot messes up bios settings,
short of reflashing bios and reapplying (prior) profile it keeps happening with stock/auto SOC V..

 

[freeagent]

Did you try a touch more vdimm?

 

[Waldorf]

board already boosts V to 1.46x
rn i just wanna observe soc/cold boot for a few days, just to make sure i dont have reflash after starting tweaks.

 

[GorbazTheDragon]

Probably just crappy 4dimm board+GDM off