[nievz]

My timings are in red below. I'm using Hynix DJR, G.Skill Trident Z Neo F4-3600C16D-32GTZNC at 1.4v in BIOS.

I did another benchmark run with my startup programs closed and got a nice 59.3ns. I think F31k is solid! My FPS in Warzone is just insane. I've never seen my rtx 3090 get worked up this much until I got my 3800IF and RAM stabilized.

Voltages:

Yeah just the only thing i don't like is my 5800x's temp 🤣

 

[Nighthog]

@nievz about your Hynix DJR, can't your kit do lower tRFC?
I had a kit that could do 448 tRFC @ 3866/1933. Would lower your latency more.

 

[ManniX-ITA]

@nievz about your Hynix DJR, can't your kit do lower tRFC?
I had a kit that could do 448 tRFC @ 3866/1933. Would lower your latency more.

Yes I think at 1.4V with that kit I could go down to 512.
With a small VDIMM bump down to 496.

 

[iNeri]

Is 62.1ns good for 3800 CL16?

Thats weird guys, on f31 i got about 5 ns less.

Fixed OC

PBO+200

 

[panni]

Still can't boot (POST!) my Micron E-Die at 3800 combined with IF 1900 on F31k, no matter what I try. Damn.

 

[kundica]

Thats weird guys, on f31 i got about 5 ns less.

Your timings are significantly tighter.

 

[LionAlonso]

Yes I think at 1.4V with that kit I could go down to 512.
With a small VDIMM bump down to 496.

Ye, mine can do at 1.4 V the 497 that the dram calc indicates.
Definetely something he should try lowering.

 

[iNeri]

Your timings are significantly tighter.

I will try f31i and report back but on f31h I got worse result on Aida ns at same timings

Enviado desde mi ONEPLUS A6000 mediante Tapatalk

 

[kundica]

I will try f31i and report back but on f31h I got worse result on Aida ns at same timings

Enviado desde mi ONEPLUS A6000 mediante Tapatalk

I thought you were comparing your results to the other user. No? If you're comparing to your own results then I don't know, but if you're comparing to the other user, your timings are tighter than theirs.

 

[Netherwind]

I recently got my 5900X and am running F30 since so many have reported problems with F31. Frequencies (up to 4950MHz on CCD0 and up to 4800-4850MHz on CCD1) and benchmarks are mostly as expected but one thing is strange and that's my temps. In multicore stresstests like AIDA64, 3DMark TS or FS physics, CPU-Z and CB20, temps are quite low (64-67C) with frequencies hitting 4200-4250MHz and SC tests are equally cool. But when I'm idling on desktop, playing WoW or doing mundane tasks the CPU temps can suddenly spike to 75C or even slightly above. At that point I see the voltage hitting 1.5V or close to 1.5V and this makes me wonder what the hell is going on?

Also enabling PBO in AMD Overclocking, setting range to "Motherboard" and +200 has no effect. Bench, frequencies, temps are the same as on Auto.

 

[kundica]

I recently got my 5900X and am running F30 since so many have reported problems with F31. Frequencies (up to 4950MHz on CCD0 and up to 4800-4850MHz on CCD1) and benchmarks are mostly as expected but one thing is strange and that's my temps. In multicore stresstests like AIDA64, 3DMark TS or FS physics, CPU-Z and CB20, temps are quite low (64-67C) with frequencies hitting 4200-4250MHz and SC tests are equally cool. But when I'm idling on desktop, playing WoW or doing mundane tasks the CPU temps can suddenly spike to 75C or even slightly above. At that point I see the voltage hitting 1.5V or close to 1.5V and this makes me wonder what the hell is going on?

Also enabling PBO in AMD Overclocking, setting range to "Motherboard" and +200 has no effect. Bench, frequencies, temps are the same as on Auto.

You won't see a drastic difference unless you change the PBO Limits setting from AUTO to MOTHERBOARD or Manual limits and use the Curve Optimizer.

For example, PBO Limits on Auto with +XXX boost and CO not enabled give me roughly the same results as stock. Enabling CO with a negative offset boosts my cores to the set PBO boost and improves my single and multicore perf at slightly more power draw 137w for CPU package power. Setting PBO limits to Motherboard on top of that provides insane all core boost but comes at the cost of heat and power. I hit 210w CPU package power and had an all core CB20 boost to 4.575 when using CO neg 10 all core and +100 boost. You'll need to work through the settings and tweak for stability because using CO and PBO with boost can get hairy.

 

[kaoski]

Bus/Interconnect Error or Sound Stuttering Error A slower delay occurs when exceeding 3733Mhz.
For now, it is forced to focus on 3733Mhz

 

[nievz]

Bus/Interconnect Error or Sound Stuttering Error A slower delay occurs when exceeding 3733Mhz.
For now, it is forced to focus on 3733Mhz

View attachment 2466208 View attachment 2466209

what brand of RAM are you using?

 

[nievz]

@nievz about your Hynix DJR, can't your kit do lower tRFC?
I had a kit that could do 448 tRFC @ 3866/1933. Would lower your latency more.

tried tRFC 485 and i CTD on BFV using 1.4v or 1.42v.

 

[kaoski]

what brand of RAM are you using?

G.SKILL Flare X Series 16GB (2 x 8GB) Model F4-3200C14D-16GFX

 

[wirx]

I tried F31k, same WHEA errors when IF2000, nothing changed, no restarts or crashes, just whea errors. Only thing what changed is after F31k installation my windows can´t shut down or restart. After shut down, screen goes black and CPU fan starts spinning, but after 5 minutes nothing changed, motherboard shows no errors. Installed windows on separete SSD and it works there fine, with restarts and shut downs. I installed F31k when I was in Windows, so probably that broke somehow my win. If anybody have some idea how to repair windows without over installing, please let me know

 

[Ohim]

I can`t make 3800 CL16 stable like i did with my former 3700X, i do have 4 sticks of ram instead of 2 but i couldn`t do it even with 2 sticks ... but got fully stable 3600 CL14 with no WHEA errors or any mem tests

 

[Mullcom]

I tried F31k, same WHEA errors when IF2000, nothing changed, no restarts or crashes, just whea errors. Only thing what changed is after F31k installation my windows can´t shut down or restart. After shut down, screen goes black and CPU fan starts spinning, but after 5 minutes nothing changed, motherboard shows no errors. Installed windows on separete SSD and it works there fine, with restarts and shut downs. I installed F31k when I was in Windows, so probably that broke somehow my win. If anybody have some idea how to repair windows without over installing, please let me know

I get this symptoms when I use ryzen master.



Skickat från min SM-G973F via Tapatalk

 

[Mullcom]

I include now your board on TT forum.

Btw,

X570 series BIOS bump to F31k.

Is ther any chance that 3000 gett any love.
Or only focus with ryzen 5000?

Skickat från min SM-G973F via Tapatalk

 

[Yuke]

Call me crazy but i get more and more hints pointing towards CPU-cache related instabilities causing my RAM errors.

I significantly get more errors when using Karhu with cache stressing set to enabled (switched from default which only stresses it a bit).

Here is the FAQ about it straight from the website:

CPU cache:

The CPU cache mode to use during the test.

- Disabled:

The memory pages are marked non-cachable and the CPU cache is not
used during the test. The test will progress very slowly and it
will not pick up any CPU cache instability related errors.

- Write-combine:

The memory pages are marked write-combined and the CPU cache is
used only for buffering writes. This is a little faster mode that
might pick up CPU cache instability related write errors.

- Default:

The CPU cache is used, but flushed frequently. The test will
progress very quickly and it might pick up CPU cache instability
related read and write errors.

- Enabled:

The CPU cache is used without restriction. This is the fastest
mode and the probability of picking up CPU cache instability
related read and write errors is also the highest.

I got rid of my undervolting, that was always working with older BIOS versions, and switched to an +offset value with LLC Vdrooping to 1.32V and so far i am hammering my RAM with my old VSOC, VDDG, VDDP settings and still going strong...usually gave me errors in the first 5-20 minutes of the run when Cache option is set to Enabled.

Maybe have a look into that if you are still pulling out your hairs about "RAM" problems...