Stock 3950x crashes with prime95 small FFTs - Page 3 - Overclock.net - An Overclocking Community
Forum Jump: 

Stock 3950x crashes with prime95 small FFTs

Reply
 
Thread Tools
post #21 of 105 (permalink) Old 04-26-2020, 09:06 PM
New to Overclock.net
 
Pavelow's Avatar
 
Join Date: Mar 2010
Location: Australia
Posts: 285
Rep: 10 (Unique: 8)
I finally got my hands on that B450 Board, it's actually a Gigabyte B450 I Aorus Pro WiFi with BIOS F50.

With all settings at their defaults, running prime95 SmallFFT I don't get failures immediately which is an improvement.
I instead consistently get failures on core 15 after about an hour and a half.

I also ran Prime95 SmallFFT overnight on the X570 Aorus Pro Wifi with LLC set to Medium, I had a failure on the first thread of core 14 after 10 hours and another on the second thread of the same core after a further 8 hours.
At defaults on that board I see failures immediately on the threads belonging to cores 8,13,14 and 15.

The threads that fail are always on the second CCD (even on my original chip) and I have seen failures on the same cores when testing with two different motherboards.
Maybe the binning of the second CCD is marginal for the samples I got (serial numbers are within 14 of each other), perhaps marginal enough that normally acceptable manufacturing variations in the motherboard are enough to markedly affect stability.

I'd love to know if anyone thinks my conclusion is wrong, I find it hard to believe that there is a systemic problem with Gigabyte's AM4 power delivery across multiple circuit designs and board models made years apart vs marginal binning on CPUs close to each other in the same batch.

'Never let your sense of morals prevent you from doing what is right" - Foundation - Isaac Asimov
Pavelow is offline  
Sponsored Links
Advertisement
 
post #22 of 105 (permalink) Old 04-28-2020, 05:13 AM
New to Overclock.net
 
DemonAk's Avatar
 
Join Date: Mar 2014
Posts: 28
Rep: 0
the same problem guys. mb x370 taichi. First i thinked it's defective cpu, yesterday i replaced to new one and nothing changed (not stable at stock). only help set: LLC1, PBO ON or cpu offset +0.05v. Asrock suport can't help. using latest bios 6.20b agesa 1.0.0.4b but trying 1.0.0.3abba same problem, not stable prime95 small fft's avx and occt power test with avx2.

Ryzen 3950x (batch 2008), ASRock X370 Taichi, 4x32gb [email protected], Palit GTX 1080ti JetStream, Windows 10 x64 Pro
DemonAk is offline  
post #23 of 105 (permalink) Old 05-09-2020, 08:24 AM
New to Overclock.net
 
smonkie's Avatar
 
Join Date: Sep 2013
Posts: 182
Rep: 4 (Unique: 4)
I've received an answer via ticket by Gigabyte and they just gave some model answer bull**** regarding overclock which is not related in any way to my issue.

At this point I've lost any hope for this to be solved, so my only chance is getting the board back, switch to Asrock or Asus and never, EVER going back to Gigabyte again.
smonkie is offline  
Sponsored Links
Advertisement
 
post #24 of 105 (permalink) Old 05-09-2020, 09:45 AM
New to Overclock.net
 
smonkie's Avatar
 
Join Date: Sep 2013
Posts: 182
Rep: 4 (Unique: 4)
Quote: Originally Posted by DemonAk View Post
the same problem guys. mb x370 taichi. First i thinked it's defective cpu, yesterday i replaced to new one and nothing changed (not stable at stock). only help set: LLC1, PBO ON or cpu offset +0.05v. Asrock suport can't help. using latest bios 6.20b agesa 1.0.0.4b but trying 1.0.0.3abba same problem, not stable prime95 small fft's avx and occt power test with avx2.
Very weird. Different brand, but exactly the same problem: I have to enable PBO + LLC Turbo OR +0.05 offset to make system stable with Prime 95. So it's getting clear this is a board issue, but it could apply to all Ryzen family. Taking into account brands are not providing help, we could try to investigate it by ourselves.

Quote: Originally Posted by Pavelow View Post
The threads that fail are always on the second CCD (even on my original chip) and I have seen failures on the same cores when testing with two different motherboards.
Maybe the binning of the second CCD is marginal for the samples I got (serial numbers are within 14 of each other), perhaps marginal enough that normally acceptable manufacturing variations in the motherboard are enough to markedly affect stability.

I'd love to know if anyone thinks my conclusion is wrong, I find it hard to believe that there is a systemic problem with Gigabyte's AM4 power delivery across multiple circuit designs and board models made years apart vs marginal binning on CPUs close to each other in the same batch.
As stated before, the issue could be extensive to other brands, so your guess about CPU batches might be the culprit.

Last edited by smonkie; 05-09-2020 at 11:09 AM.
smonkie is offline  
post #25 of 105 (permalink) Old 05-09-2020, 03:25 PM
New to Overclock.net
 
Join Date: Feb 2017
Location: Sweden
Posts: 1,300
Rep: 56 (Unique: 43)
Marginal binning process from AMD falling between the lines of "acceptable" with a tolerance to wide.

AMD is binning too tight leaving too little room for variance in power delivery.

Gigabyte X570 AORUS XTREME, Ryzen 7 3800X PBO boost, 2x8GB 4866/1900 18.27.24.48.90.tRFC 680 1T 1.650V [Micron Rev.J]
4x8GB 3800/1900 13.23.17.30.66.tRFC 600 1T 1.700V Kingston HX434C19FB2K2/16 3466C19 1.2V (Micron Rev.E 19nm [D9VPP]),
2x8GB Kingston HX436C17FB3K2/16 3600C17 1.35V (Micron Rev.J 17nm),
2x8GB 3866/1933 15.21.20.32.60.tRFC 448 1T 1.590V Kingston HX436C17FB3K2/16 3600C17 1.35V (Hynix DJR 17nm)
Nighthog is offline  
post #26 of 105 (permalink) Old 05-17-2020, 08:02 PM
New to Overclock.net
 
LetterKilled's Avatar
 
Join Date: Mar 2012
Location: Sacramento
Posts: 75
Rep: 1 (Unique: 1)
I just got my 3950x up and running a few days ago and I’m having the same problem. It’s also on a gigabyte board as well. The x570 Ultra for me. All stock every test from Prime95 results in an instant failure, small, smallest and custom. I haven’t tried the LLC or offset yet, but PBO set to 300 230 230 1x results perfectly fine. I however am running ram that is not certified so can’t even run xmp without it crashing (new ram coming soon). So I have 3 factors: RAM, CPU, MObo. I’m testing the ram right now with no errors on test 2/4. There a few people in reddit having the same problem as well Running Gigabyte. This is the first time going with them. I was pretty pumped about it. Now I’m sad. Setting the LLC or offset shouldn’t degrade the chip, it just sucks to do it.
LetterKilled is offline  
post #27 of 105 (permalink) Old 05-17-2020, 08:42 PM
Never Finished
 
airisom2's Avatar
 
Join Date: Aug 2011
Location: Within the Milky Way
Posts: 2,212
Rep: 279 (Unique: 216)
So, you say you're using a D15, but there is no D15 specifically designed for sTR4. The only Noctua coolers that do that are the U14S TR4 and its offspring (U12S, U9, etc.).

It could be possible that the edges of the chiplets are overheating due to having little to no cooler contact, causing system shutdown. Might be something to look into.

1680 V2 4.5GHz 1.33v | G.Skill Trident X 4X4GB 2400MHz | R4BE | Titan Xp + Morpheus II
Wasabi Mango UHD550 | Rosewill Rise | Thermalright LGM RT
Logitech G600 | G440 | Tt Meka G-Unit Cherry MX Black+Double O-Rings
JBL LSR308s | Temblor T10 | Custom headphones | TEAC-UD501+Passive Preamp | Emotiva BasX A-100
OCZ Vertex 4 256GB | 51TB DAS | SS Prime Ultra Titanium 1KW
airisom2 is offline  
post #28 of 105 (permalink) Old 05-17-2020, 09:40 PM
New to Overclock.net
 
LetterKilled's Avatar
 
Join Date: Mar 2012
Location: Sacramento
Posts: 75
Rep: 1 (Unique: 1)
The 3950x is a AM4 socket.
LetterKilled is offline  
post #29 of 105 (permalink) Old 05-17-2020, 09:41 PM
New to Overclock.net
 
LetterKilled's Avatar
 
Join Date: Mar 2012
Location: Sacramento
Posts: 75
Rep: 1 (Unique: 1)
Quote: Originally Posted by airisom2 View Post
So, you say you're using a D15, but there is no D15 specifically designed for sTR4. The only Noctua coolers that do that are the U14S TR4 and its offspring (U12S, U9, etc.).

It could be possible that the edges of the chiplets are overheating due to having little to no cooler contact, causing system shutdown. Might be something to look into.
Quote: Originally Posted by LetterKilled View Post
The 3950x is a AM4 socket.
there we go. The 3950x in an AM4 socket.
LetterKilled is offline  
post #30 of 105 (permalink) Old 05-17-2020, 10:13 PM
Never Finished
 
airisom2's Avatar
 
Join Date: Aug 2011
Location: Within the Milky Way
Posts: 2,212
Rep: 279 (Unique: 216)
I should've known. Naming still gets me.

1680 V2 4.5GHz 1.33v | G.Skill Trident X 4X4GB 2400MHz | R4BE | Titan Xp + Morpheus II
Wasabi Mango UHD550 | Rosewill Rise | Thermalright LGM RT
Logitech G600 | G440 | Tt Meka G-Unit Cherry MX Black+Double O-Rings
JBL LSR308s | Temblor T10 | Custom headphones | TEAC-UD501+Passive Preamp | Emotiva BasX A-100
OCZ Vertex 4 256GB | 51TB DAS | SS Prime Ultra Titanium 1KW
airisom2 is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 2 (1 members and 1 guests)
blitzkrieg666
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off