Welcome Guest ( Log In | Register )

Outline · [ Standard ] · Linear+

 AMD Ryzen, AM4 / AM5 Platform

views
     
senzr
post Jan 28 2021, 11:30 AM

New Member
*
Newbie
8 posts

Joined: Jan 2006


QUOTE(SSJBen @ Jan 27 2021, 10:29 PM)
Think I found the issue, but I'm not sure.

So I got another Event 41 today and this time, the time stamp is exactly at the time Windows booted. Coincidentally, there is another system error (Event 29) which has the exact same time stamp as Event 41 error, this error is related to fast boot.
I'm on a R5 3600 + B550, and I've also been noticing more frequent WHEA. Taking your lead, I checked Event Viewer and also discovered a bunch of Event 41 critical errors whenever I had WHEA.

My fast start-up settings are off, so that's not the issue for me.

But I did some digging, and remembered I updated my mobo bios not long ago to version F11 (B550 Aorus Pro AC), and for some reason, it was removed from the official bios download page. Then I found this:

https://www.reddit.com/r/gigabytegaming/com...os_f11_missing/
https://www.reddit.com/r/gigabytegaming/com...s_pro_f11_bios/

Quoted from thread:
"There is a reason. F11 BIOS was generating WHEA errors in Event Manager in Windows if you were using RAM running at 3600 MHz. This caused instability for some people. Any slower RAM speed was fine."

"I still haven't been able to run my RAM at 3600MHz via its XMP profile without intermittently logging WHEA event 19 errors in Windows 10 and the equivalent errors in Linux, which occur seemingly at random."


I also remembered that I'm running my Trident Z Neo RAM at 3600mhz via manual overclock instead of XMP, because enabling XMP would cause crashing.

So it seems like a combination of Gigabyte 500 series mobo on F11 (or F31 on X570) with G.Skill Trident Z Neo RAM running at 3600mhz has been documented to cause crashes and random WHEA.

The Gigabyte B550 Aorus Pro AC bios page has been updated with bios version F13a as of 27/1, which updates to AGESA ComboV2 1.2.0.0. I'm gonna cross my fingers and update to this bios and see what happens.

This post has been edited by senzr: Jan 28 2021, 11:38 AM
senzr
post Jan 28 2021, 02:02 PM

New Member
*
Newbie
8 posts

Joined: Jan 2006


QUOTE(Bonchi @ Jan 28 2021, 01:36 PM)
usually such error is across all brands as its mostly got to do with agesa and the design of the ryzen IMC. pretty tough luck on zen2 because it seems like most support has already been halted so it’s better to use older proven stable bios if youre on zen2 as most of the newer bios fixes are meant for zen3 and pretty it will compromise zen2’s stability. Im using gigabyte’s F3 bios 1.0.0.2.. it seems to be the most stable.
*
I see...well, so far so good on F13a. I'm aware it's beta, so still might have issues. But try n see lor. No crashes yet, but I haven't enabled XMP, just manual overclock. The last few mobo bios haven't fixed the bad XMP profile, so I'm just manually overclocking RAM still.
senzr
post Feb 2 2021, 06:34 PM

New Member
*
Newbie
8 posts

Joined: Jan 2006


I've been experiencing the random resets, WHEA and BSOD for 3 months on a 6-month old build. I'd be gaming, then everything hangs, screen goes black, and all the fans ramp up to the max, requiring a hard reset. Listed as Event ID 41.

But I may have found a solution, hopefully it helps someone else out.

Rysen 5 3600
B550 Aorus Pro AC
RTX 3080 Aorus Master
G.SKill Trident Z Neo 3600 CL18 (4x8GB)
Aorus copper SSD 1TB
Adata SX8200 Pro SSD 1TB
Corsair RM850

I'm convinced the issue is with memory compatibility and the stock XMP profiles, along with BIOS. I spent the last week messing around with different BIOS versions, and RAM settings.

Previously, I could never run the stock XMP profile with this RAM. It would always crash for certain games. Initially, I got around it by just manually changing only the memory speed multiplier to 36.00, and that seemed to work. For awhile, at least. Also, my particular model of RAM is NOT on the QVL list for this mobo, even though it looks the same as other Trident Z RAM.

I started with 16GB, then added another 2 sticks for 32GB, and stability became an issue again. As I began to learn more about RAM, I discovered other settings in the BIOS to tweak, because without the XMP profile on, the timings were very loose (26-25-25-58 @ 3600mhz, 1.20v), and it wasn't delivering enough voltage to sustain 3600 mhz.

I also discovered another setting in Gigabyte BIOS called "XMP High Frequency Support". This is key, because until I messed around with this, it wasn't stable. Did some digging online and on Reddit, found there are 3 levels, which apparently correspond to RAM speed. Level 3 is 3600, 2 is 3400 and 1 is 3200. I chose Level 3 to support the max rated speed of my RAM. If you change your RAM speed, you also have to change your Infinity Fabric speed to half your RAM speed, to maintain 1:1 ratio. Usually, these settings should be done when set on AUTO, but somehow they don't work for me.

Also, I found out that for my particular RAM which is made with Samsung C-die, the max voltage is 1.35v. But when I set it to 1.35, it still crashes and shows about 1.4 in the monitoring apps. So I set it to 1.30v, and got more stability (showing 1.35v in monitoring.)

Right now, I've been pretty stable for a day with no crashes. My recipe:
BIOS F12 (AGESA 1.1.0.0 D)
XMP enabled
XMP High Frequency Support Level 3
Memory Speed multiplier 36.00
DRAM voltage 1.30v
Infinity Fabric speed 1800mhz

I had to manually input all of these in BIOS. Fingers crossed, everything seems fine so far, but it's only been a day with no crashes. I used to get 2-3 crashes a day.

YMMV, because it really depends on the particular combination of BIOS, RAM, mobo and CPU. This worked for my particular rig, might be different for yours. But I hope this helps someone somehow!

This post has been edited by senzr: Feb 2 2021, 06:35 PM
senzr
post Feb 2 2021, 09:35 PM

New Member
*
Newbie
8 posts

Joined: Jan 2006


QUOTE(Bonchi @ Feb 2 2021, 08:59 PM)
XMP High Frequency Support Level 3 is just a quick way to set the IF but at fixed steps. It’s better to use auto and set it to be coupled instead.  1800mhz IF at 3600mhz ram is not guaranteed to work on zen2 so gotta keep that in mind.
*
In my case, I only got the system stable if I set High Frequency Support to Level 3 AND manually set the IF speed, memory speed + voltage. Shouldn't have to double confirm like that in AUTO, but that's a quirk of Gigabyte boards, I guess. Got this tip from Gigabyte forums and Reddit, so I'm not the only one who had to do it.

I'm aware the R5 3600 is only officially guaranteed to run RAM at 3200mhz, but many folks are running at 3600mhz, and I already paid for the 3600 kit, so just gatal to get my money's worth. Just didn't anticipate it'd take so much effort to get it working properly...

Another upshot is that my Time Spy scores have improved by 200 points with the current XMP settings (1.30v, 18-22-22-42) compared to when I previously ran the RAM at 3600mhz manually and left everything else at auto (1.20v, 26-25-25-58). The auto settings are pretty loose and safe.

This post has been edited by senzr: Feb 2 2021, 09:40 PM
senzr
post Feb 3 2021, 09:17 AM

New Member
*
Newbie
8 posts

Joined: Jan 2006


QUOTE(Bonchi @ Feb 2 2021, 09:50 PM)
in most cases, xmp doesnt work out of the box even for 3600 as each board will intepret it slightly differently thus the instability varies, especially on the difference at the subtimming and SoC voltage.
*
Oh wow...you mean even if a RAM kit is listed in the mobo's QVL, it isn't just plug n play?
I was planning on avoiding this whole hassle in future builds by just checking the QVL...
senzr
post Feb 28 2021, 08:32 AM

New Member
*
Newbie
8 posts

Joined: Jan 2006


QUOTE(timeonce @ Feb 28 2021, 07:57 AM)
But nontherless. AMD is damn good. I run it at 100% performance for 72 hours straight and zero issue with it during my simulation task.
[attachmentid=10799773]
*
Wow, what monitoring software is that? Looks pretty slick!
senzr
post Jun 6 2021, 06:59 PM

New Member
*
Newbie
8 posts

Joined: Jan 2006


Hello, sifus!

I just got a 5900X, and wanna check with y'all what kinda temps you're getting.

My observations after 3 days, PBO off (HWinfo Tctl/Tdie):
Idle: 48-55C
Gaming: 70-84C

Room ambient temp: 27-28C

Ryzen Master shows temps about 5-6C less than HWinfo, not sure which is more accurate.

Case: Fractal Meshify C case, 3 intake fans (1x140, 2x120), 3 exhaust (1x140, 2x120).
Cooler: Arctic Freezer 34 with 2x120mm fans, push/pull, with Noctua NT-H2 thermal paste.
GPU: 3080 Aorus Master (67-70C, this thing is like a nuclear reactor, and exhausts 320W worth of air towards the CPU)

Just came from a 3600, so these temps are pretty high in comparison.

I've been looking up 5900X temps on Reddit to compare, and realise that most of those guys have much lower ambient temps in Western countries.
Just wondering what other Msians with our weather are getting.

Wondering if the Freezer 34 is up to the job, considering a Noctua D15s if it isn't.
Haven't even attempted PBO and undervolting with Curve Optimiser yet.

This post has been edited by senzr: Jun 6 2021, 07:04 PM
senzr
post Jun 6 2021, 09:19 PM

New Member
*
Newbie
8 posts

Joined: Jan 2006


QUOTE(JohnLai @ Jun 6 2021, 07:07 PM)
Getting the same idle and gaming temp range with same ambient temp for my 3950x and 5950x (both PBO on +200Mhz ).  yawn.gif

Difference is.....uncle is using Noctua NH-D15 for both.
*
QUOTE(llk @ Jun 6 2021, 08:24 PM)
5900 here, around 70~75 max but mine is full custom watercooling setup
*
Wow...then it really makes me wonder how much better a D15 or AIO could be. The Freezer 34 is a pretty cheap single tower cooler, but this makes it seem like really good value. But I haven’t tried PBO yet, maybe that’s the difference.
senzr
post Jun 6 2021, 11:21 PM

New Member
*
Newbie
8 posts

Joined: Jan 2006


I'm now attempting to enable PBO and undervolt with Curve Optimiser.

But when I set PBO limits to "Motherboard", I get these limits in Ryzen Master:
PPT: 720W
TDC: 420A
EDC: 215A

Mobo: B550 Aorus Pro AC

Don't those "limits" seem astronomically high???

user posted image
senzr
post Jun 15 2021, 07:36 PM

New Member
*
Newbie
8 posts

Joined: Jan 2006


When you guys are talking about temps, which monitor are you looking at?
There's a big discrepancy between Ryzen Master and HWinfo, at least 6-7C, and I'm not sure which is more accurate.

Gaming temps at 85C (HWinfo reading) on my 5900X are kinda worrying me, not sure if I need a better cooler.
senzr
post Jun 15 2021, 09:53 PM

New Member
*
Newbie
8 posts

Joined: Jan 2006


QUOTE(Bonchi @ Jun 15 2021, 07:59 PM)
Hwinfo is more reliable as ryzen master takes the average between ccx so it will look odd when your cpu is throttling but the temps are not that high. And yes, ryzen are actually way hotter than what everyone says especially for gaming so you definitely need something like a 360mm aio for that 5900X or a D15 for air cooling.
*
QUOTE(1024kbps @ Jun 15 2021, 08:47 PM)
if your cpu load is not 100% while gaming then you need a better cooler, a stress test benchmark will send your cpu to thermal throttle mode.
*
Thanks for the tips and feedback!
I'm currently running an Arctic Freezer 34 with push/pull fans in a Meshify C case.
Thanks to the great advice in this thread, I've got the 5900X tuned with PBO, CO -25 on most cores, Auto OC +200 and stock power limits.
I'm pretty sure I'm bouncing off the 90C thermal limit, as max temps are 89.9C.

I've been struggling to decide between a Noctua D15S + additional 120mm fan (case clearance issues) or a 240mm top-mounted AIO, because that's all I room for.

Which would give better overall cooling performance in a mesh case?



 

Change to:
| Lo-Fi Version
0.2285sec    0.63    7 queries    GZIP Disabled
Time is now: 17th December 2025 - 10:50 PM