b-die tweaking, timings ok?

[Fry178]

Thanks for not spending the time to lock at my info nor the specs,
you can save even more time, by not posting irrelevant answers here, thanks.

 

[freeagent]

TRFC at 282 might get you fixed up. I can run 262 easily with 2T, or needs more vdimm with gdm on..

 

[Deleted member 185158]

1.45v Vdimm for only 3600mhz cl14's seems high. I guess the kit came that way???

Saw mention of "cold boot trouble" ?
Is that the restart deal at cold post?
If the board is equipped with memory training algorithm, the the restarts at cold post are normal. The board is training 2nds and 3rds

Should check out what taiphoon burner says about the PCB. I know my A2 kits don't like tighter than CL13 on 1T command rate.

At any rate, some screen shots from the OP would be grand. I've only encountered screen shots from those trying to help....

GL!

 

[freeagent]

1.45v Vdimm for only 3600mhz cl14's seems high.

Both of my 3200C14 kits need 1.45 for 3600C14. I am assuming sticks used to do it with less voltage? That would be nice..

Welcome to the new normal I guess

 

[Taraquin]

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..

Spread spectrum is enabled by default and often works with OC, but it's eadier to OC without it

 

[Deleted member 185158]

Both of my 3200C14 kits need 1.45 for 3600C14. I am assuming sticks used to do it with less voltage? That would be nice..

Welcome to the new normal I guess

Mine are fine at 1.4v. Where it's set now actually. Though I'm not home to do any proving it, but on the old 8700K and Maximus X hero... I guess that's a different story there.

 

[tabascosauz]

1.45v Vdimm for only 3600mhz cl14's seems high. I guess the kit came that way???

High if your standard is what, 4400CL16 XMP bins? 1.45V is pretty reasonable at 3600/14 for most run of the mill B-die.

Mine come with that XMP but I can manage down to about 1.42V at 3600/14. Rather middle of the road for this bin

Also Gigabyte boards always overvolt VDIMM without exception.

 

[freeagent]

I think I can squeeze 1.425 on one set, the other might be 1.45.. I will have to check when I get home lol.. all good Jon I believe you

 

[Deleted member 185158]

High if your standard is what, 4400CL16 XMP bins? 1.45V is pretty reasonable at 3600/14 for most run of the mill B-die.

Mine come with that XMP but I can manage down to about 1.42V at 3600/14. Rather middle of the road for this bin

Also Gigabyte boards always overvolt VDIMM without exception.

Who can afford that? Oh I do have a set of 4267mhz.... cl19, A2 pcb...
But also 3600 CL 16s B-Die 1.3v, only takes 1.40v to run em cl14s which would be no better or worse bin than 3200 CL 14s.

I've found the PCB revision has a bit more concern rather than a rated frequency at X voltage. My A2s do not like CL12 at 4000mhz 1.90v at all. But A0 and A1 kits seem to pull it off.

But let's ignore the mention of the PCB revision and pick on Shrimps voltage statement instead. Yes, my silly opinion is 1.45v seems high.

In fact running 1.2v SOC seems silly for any frequency under 4000mhz.

However I don't run AMD for my personal use at this time. So I should digress from the thread and leave it alone. I know nothing.

 

[freeagent]

I know nothing.

You both know more than I do.. Not afraid to say it

And I am pretty sure memory and it characteristics would be slightly different when running on Intel or AMD..

 

[tabascosauz]

But let's ignore the mention of the PCB revision and pick on Shrimps voltage statement instead. Yes, my silly opinion is 1.45v seems high.

I know nothing.

Hey, no need for that. We both know that you know your stuff. Just saying that 1.45v isn't so unreasonable. And because it's Gigabyte it's like, impossible to tell where VDIMM is without a multimeter.

Never seen A0 outperform A2 at just about anything, but there's a first time for everything......maybe A0 is tighter at lower speeds so you're probably right

 

[freeagent]

You guys are all awesome, just so you know.. be kind rewind

 

[Deleted member 185158]

Hey, no need for that. We both know that you know your stuff. Just saying that 1.45v isn't so unreasonable. And because it's Gigabyte it's like, impossible to tell where VDIMM is without a multimeter.

Never seen A0 outperform A2 at just about anything, but there's a first time for everything......maybe A0 is tighter at lower speeds so you're probably right

I wasn't saying unreasonable. Just a little high/er than I'd expect.

You guys are all awesome, just so you know.. be kind rewind

It's all good.

I don't start OCing till 1.6v lol

Totally untuned, set 1.40v - reads 1.408 (which I like, droop is bad) -
Cpu is stock.
Memory timings only manually 14-14-14-34, the rest is auto.

 

[Taraquin]

Who can afford that? Oh I do have a set of 4267mhz.... cl19, A2 pcb...
But also 3600 CL 16s B-Die 1.3v, only takes 1.40v to run em cl14s which would be no better or worse bin than 3200 CL 14s.

I've found the PCB revision has a bit more concern rather than a rated frequency at X voltage. My A2s do not like CL12 at 4000mhz 1.90v at all. But A0 and A1 kits seem to pull it off.

But let's ignore the mention of the PCB revision and pick on Shrimps voltage statement instead. Yes, my silly opinion is 1.45v seems high.

In fact running 1.2v SOC seems silly for any frequency under 4000mhz.

However I don't run AMD for my personal use at this time. So I should digress from the thread and leave it alone. I know nothing.

On overclock.net the general rule us that you often can run lower timings on A0, but speed is limited vs A2. Some A0 kits can't go much past 3600, but happily do 3200cl12. There are always exceptions though

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..

For many it seems agesa 1.2.0.3b or c was the best. Tried any of them?

 

[basco]

what sometimes helps on A2 pcb or not so strong Bdie if going for lower voltage without loosing much is cl14-15-15

 

[Fry178]

running ram stock/auto settings (jedec), as well as with mentioned timings, passed 1200% in hci (dos + win).
i wasnt too sure about some timing "ranges", and/or "connection" between certain timings, hence this post to get some insight,
as i know lower doesnt always equal "better", just not much clue past primary timings/voltage, especially on Ryzen...


@freeagent
no trouble with lower rfc
why i would want to go higher (as it seems to be an important non-primary).
dont remember perf diff between them, didnt write down results

@ShrimpBrime
its set because its sold as CL14/1.45v kit, and all bdie gets min 1.35, usually 1.45v recommended anywhere i looked,
and im tweaking timings right now, ram voltage isnt my focus rn.

there is no restart issue nor memory training (turned off).
with everything stock, any soc below 1.03v will lead to no-post (on next boot),
after power was removed (no difference if 10s or 10 days).

@Taraquin
yeah, back on .b, after i experienced another no post and had to reflash anyway (or risk bclk boost again).

@basco
looking at timings first, will lower v once i know what i can run...

 

[Fry178]

seems to be fine with 1.03 for stock, gonna see that i start testing tweak this weekend.

 

[Taraquin]

seems to be fine with 1.03 for stock, gonna see that i start testing tweak this weekend.

If it's fine stock the it will be fine once tweked, but what subs you can tweak can vary a bit. Big variance in tolerance for rcdrd, rp, rc and rfc on B-die. Most others like cl, cwl, rrd, faw, wr, rtp, wtr rdwr acts similar between good and bad bin and just need more or less voltage.

 

[Fry178]

so checking board site to see if latest bios is still beta, i saw this added by GB:

• Major vulnerabilities updates, customers are strongly encouraged to update to this release at the earliest.
Credits to "Assaf Carlsbad and Itai Liba from SentinelOne"
• Introduce capsule BIOS support starting this version.

just wanna see if anyone can chime in

 

[tabascosauz]

so checking board site to see if latest bios is still beta, i saw this added by GB:

• Major vulnerabilities updates, customers are strongly encouraged to update to this release at the earliest.
Credits to "Assaf Carlsbad and Itai Liba from SentinelOne"
• Introduce capsule BIOS support starting this version.

just wanna see if anyone can chime in

Gigabyte adopted an inherently vulnerable way of packaging their BIOSes, so now they're moving to the .CAP format that Asus has used since forever. It's not got anything to do with performance and iirc it's really not a matter of critical importance for home users

It's AM4, so every time you change AGESA you are pretty much spinning the lottery wheel in hopes that new AGESA doesn't randomly make your performance worse. AGESA 1.2.0.0-1.2.0.3b were pretty good in my experience on Asus/Gigabyte/MSI, 1.2.0.5 has not been regarded so favourably but ymmv

 

[Fry178]

thanks, ill stick to F34 then, just wanted to check before i start tweaking.

so usb doesnt work for sticks, and volt for vddp/vddg are stuck on some form of auto/normal,
and wont change to what i set, so going back to 1.2.0.4a