Ryzen Owners Zen Garden

[thesmokingman]

Here's a tip, if you use hwinfo set the whea error tracker to the taskbar. That way you immediately know when the dreaded whea occurs so you don't waste time.

 

[RJARRRPCGP]

Here's a tip, if you use hwinfo set the whea error tracker to the taskbar. That way you immediately know when the dreaded whea occurs so you don't waste time.

That only applies to non-fatal WHEA errors (correctable, but correctable ones are bad, too!) Fatal WHEAs mean Windows crashes out.

 

[thesmokingman]

That only applies to non-fatal WHEA errors (correctable, but correctable ones are bad, too!) Fatal WHEAs mean Windows crashes out.

Well no duh? Why's everyone got to make a redundant counter all the time, can't take a tip?

 

[RJARRRPCGP]

Well no duh? Why's everyone got to make a redundant counter all the time, can't take a tip?

That was according to the HWiNFO author.

 

[thesmokingman]

That was according to the HWiNFO author.

When you get a Fatal, the game is over, jig is up. You can however get the non-fatal kind and not know it. Thus being able to see as soon as you get one, it saves you time. Duh.

 

[RJARRRPCGP]

When you get a Fatal, the game is over, jig is up. You can however get the non-fatal kind and not know it. Thus being able to see as soon as you get one, it saves you time. Duh.

In another words, it doesn't sound like HWiNFO reads the event log for past fatal WHEAs.

 

[Octopuss]

So what are you guys using for stress testing?
I sometimes get AMD Wattman (or whatever is it called now) reset to defaults (annoying as hell) due to some unspecified error (thanks AMD), and since it's been several BIOS versions since I considered my system stable, I thought it could potentially be related to the CPU, and thought I should retest everything.
Of course it's been over a year and I forgot what I did there I most probably used Prime95 just like always, but since I was primarily undervolting, it's pretty complicated.

 

[oobymach]

I still stress with at least 2 passes in p95 blend, I don't want to subject my machine to hours of stress, if all workers pass the second round I'm happy with it.

Just went through undervolting my static 4.65ghz to 1.2v which dropped my temps by a few degrees, barely breaking 60 degrees now in p95.

 

[freeagent]

I let my Intel CPU roast in the 90s for weeks, might have lost a bit of weight but she still rips

My AMD CPU's didn't get away so easy either.. OTP comes shortly after 100

They are pretty tough.. within reason..

 

[RJARRRPCGP]

Is there any evidence of notable degradation from having an all-core OC at 4.5 or 4.4 on a Vermeer? Mine looks stable at 4.00 at 1150mV -1% to -5% Vdroop roughly.

CTR indicates that I possibly can go for 4.5!

 

[freeagent]

Is there any evidence of notable degradation from having an all-core OC at 4.5 or 4.4 on a Vermeer? Mine looks stable at 4.00 at 1150mV -1% to -5% Vdroop roughly.

CTR indicates that I possibly can go for 4.5!

Its fine. 1.25v is fine, 1.35 is probably my all load limit.

Edit:

my 5600X is all load stable at 4650 with 1.25v, but for 4700 it needs 1.35v. On my 5900X 1.25v is good for 4500, and 1.35 for 4600 is my hard limit on that CPU..

 

[RJARRRPCGP]

Its fine. 1.25v is fine, 1.35 is probably my all load limit.

Edit:

my 5600X is all load stable at 4650 with 1.25v, but for 4700 it needs 1.35v. On my 5900X 1.25v is good for 4500, and 1.35 for 4600 is my hard limit on that CPU..

I remembered thinking that even for single core, 1.3xV was pressing my luck!

 

[Mussels]

New chipset drivers have landed
X570 Drivers & Support | AMD

 

[Chomiq]

Wouldn't read too much into it, time spent pre-POST is largely memory training which is dependent on what you have. When I was on B450I Aorus and B550M TUF, it wasn't until around I switched to B-die that I ever saw any proper "fast boot" whether it was enabled/disabled in BIOS or Windows. 4Gb E-die takes a very long time to train, CJR not so much but not "fast".

And even afterwards it has never been consistent either. Give it a couple months and if it starts up quick every single day without exception, maybe there's something going on. Sometimes I'd change some red herring and temporarily believe that the problem was "fixed" by something else, but best not to read too much into this stuff.

And just like that today I'm back to 12+ seconds on post instead of 7. Didn't change anything at all.

 

[YautjaLord]

Just to keep this 4k-everything-liquid-cooled beast alive, RTX 3080 Master (rev. 3.0) done a week or so ago, Dark Base Pro 900 rev. 2 - yesterday. I tell ya, it's night & day difference even aesthetically, the liquid cooling thingy.

 

[Chomiq]

First 5700X GB results are out.

 

[tabascosauz]

4000/5000 APU owners - the "new" GPU PBO and GPU curve optimizer in I think AGESA 1203+ BIOSes is pretty interesting and worth checking out if you're on iGPU.

Originally the stock iGPU behaviour was dynamic freq with dynamic voltage (VDDCR_GFX). If you ever OC'd then it would only accept a fixed freq @ VSOC.

The GPU PBO part now allows scaling up to +200MHz with dynamic freq and dynamic voltage (but VDDCR_GFX increases depending on boost override). The Curve Optimizer part I haven't delved into yet because it appears to work the same way as CPU Curve Optimizer but nowhere have I seen described how big a CO step is for GPU (millivolts).

For example, at 2200MHz IF (which increases VSOC requirements):

• Stock - dynamic 2000MHz @ ~0.7-1.0V (?)
• OC - fixed 2200MHz @ 1.2V
• OC - fixed 2300MHz @ 1.23V
• PBO - dynamic 2200MHz @ ~0.7-1.205V

The PBO part at least works very well - Valley score for 2200 fixed and 2200 PBO are identical. Been on 2200 more lately as 2300 is tough and I don't game as much on this one, so PBO is good.

 

[freeagent]

So this guy right here thought he would be smart and update to the latest bios to see what changes there would be to my boost clocks.. Wow it was just as terrible as on my other B550 board. Wow. huge cuts to power and boost. Then I flashed back to what I was using before.. and voila.. TDC bug.. wouldn't budge past 139.. booooo. I tried flashing back to the bios it shipped with.. and nope no can do.. weird I thought.. So I re flashed again to the second newest and plugged in my settings and voila, TDC bug is gone, back into the 150s

That was too close

 

[Mussels]

Yeah, some boards get some really shitty BIOS's
1.2.0.3c was pretty solid for just about everyone

 

[Mussels]

 

[tabascosauz]

View attachment 243945

This new auto-test CO function is about as reliable as my 3700X was........certainly not a compliment.

I ran the full per-core test which took about 1.5 hours. RM told me that every core was -27, which was strange since even in its test, the cores were topping out at different freqs in about the same order as I've come to expect from this 5700G. So I applied its recommendations and restarted.

Made it through 1/8 of an AIDA run before hard rebooting. AIDA.

I haven't thoroughly CO tested the 5700G as I have my 5900X (it took 6 weeks!) - all I know is that the best two cores are pickiest (as I have come to expect from 5000), so I left it at -8 for all cores as it seemed to be good enough for default config Corecycler. No practical instability since then (~8 months ago?).

So you can imagine my surprise when RM told me initially that -27 was good.

Looks like RM is still comical a software as ever, and custom corecycler is still the proper way to test.

 

[freeagent]

I never did like that program.

 

[Chomiq]

This new auto-test CO function is about as reliable as my 3700X was........certainly not a compliment.

I ran the full per-core test which took about 1.5 hours. RM told me that every core was -27, which was strange since even in its test, the cores were topping out at different freqs in about the same order as I've come to expect from this 5700G. So I applied its recommendations and restarted.

Made it through 1/8 of an AIDA run before hard rebooting. AIDA.

I haven't thoroughly CO tested the 5700G as I have my 5900X (it took 6 weeks!) - all I know is that the best two cores are pickiest (as I have come to expect from 5000), so I left it at -8 for all cores as it seemed to be good enough for default config Corecycler. No practical instability since then (~8 months ago?).

So you can imagine my surprise when RM told me initially that -27 was good.

Looks like RM is still comical a software as ever, and custom corecycler is still the proper way to test.

1usmus doing coding for AMD now?

 

[R-T-B]

1usmus doing coding for AMD now?

Did I miss where his rep took a hit somewhere? A bit out of the loop here.

 

[Mussels]

Did I miss where his rep took a hit somewhere? A bit out of the loop here.

his programs are hit and miss with some hardware, and some people demand perfection