Stock 3950x crashes with prime95 small FFTs - Page 10 - Overclock.net - An Overclocking Community
Forum Jump: 

Stock 3950x crashes with prime95 small FFTs

Reply
 
Thread Tools
post #91 of 105 (permalink) Old 06-24-2020, 04:15 PM
New to Overclock.net
 
Join Date: Aug 2019
Posts: 311
Rep: 6 (Unique: 5)
Quote: Originally Posted by Nighthog View Post
Well at least I can say I have no issues on the 3800X even if I have tortured it with excessive voltage since day one.
100% stable on stock on Gigabyte X570 Aorus Xtreme F12f. Took the occasion as I updated my BIOS from F12a to F12f version.

Just loaded optimized defaults and it was all OK, even went and enabled XMP @ 4000Mhz and all was well as I only needed increase stock DRAM voltage from 1.200 -> 1.350V for it to work there.
So I wonder what the issue is with the higher core counts. I've used "excessive" voltage on my 3800X a long time and as far as I can tell it's not changed a bit since start.

What voltage do you run your cpu at?
2600ryzen is offline  
Sponsored Links
Advertisement
 
post #92 of 105 (permalink) Old 06-25-2020, 12:08 AM
New to Overclock.net
 
mongoled's Avatar
 
Join Date: Oct 2011
Location: Ayia Napa, Cyprus
Posts: 692
Rep: 24 (Unique: 19)
Quote: Originally Posted by Nighthog View Post
I had trouble with AVX2 LargeFFT when going for 4866/1900... It became utterly finicky for voltages on the VDDG/VDDP & vSOC. Lower speeds are more tolerable with most voltages, but if set wrong "work" but can give glitches. Not high Frequency like this, doesn't tolerate "wrong" settings.

Only ~950/950 VDDG_CCD/IOD works to make it pass, having enough VDDP is also critical, the recommended 900mv is to little, even 1000mv wasn't really enough.

I might be suspecting the 3950X is more taxing on FCLK? Maybe AUTO doesn't work? I know they are high when on AUTO settings. Above 1000mv on VDDG voltages are finicky in general from my testing. They make AVX2 fail more often than not.
Low VDDG_CCD voltage will make AVX2 fail also. My sample has a floor around 950mv to function without fail.

I noted a weird crossloading issue, AVX2 would work if I did nothing major but the minute I started a YouTube video they would crash. Had to adjust the CAD_BUS settings to fix that.
The issue I saw was even weirder, which was the reasoning i decided to investigate CCD voltage.

Basically, running 3800/1900 @ 4400 mhz, was wondering if the high fclk was effecting the stability of core 6 after extended prime95 load.

The issue occurs on both blend, large ffts and small ffts and around a similar time mark (after 30 mins).

So I simply dropped to 3733/1866 leaving everything else as they were.

What happened stumped me, prime95 fails almost instantly on core 6 !

So this got me thinking that something get put "out of sync" be it voltages, frequencies or what not.

Boom! Again, worker 12 at 35 minutes, temperature does not seem to be the issue as it happens when ambients are lower or higher!

Something ive noted, worker 11 is processing

"FMA3 FFT length 72k, Pass1=384, Pass2=192, clm2"

while worker 12 is processing

"FMA3 FFT length 64k, Pass1=256, Pass2=256, clm2"

The crash looks like it occurs when the 12th worker is going to pass to the same FFT length as worker 11.

Really, we should take this conversation to its own thread

Apologies to the OP

SeaSonic X750 Gold 750W
MSI X370 XPOWER Gaming Titanium | Ryzen 5 3600 2010SUS @ 4400 mhz @ 1.325v (prime95 load)
Viper Steel Series PVS416G413C9K @ 3800/1900 mhz | 1.52v | 14-15-14-14-28-42-252-16-1T
Asus GTX 780 DCII OC @ 1332/3522 mhz | 1.3v | Samsung 960 EVO 500GB
EK-FB MSI X370 XPower RGB Monoblock 1/2" ID | Swiftech D5 MCP655 | EK-XRES 100
Alphacool NexXxoS UT60 280 | 4 x Scythe GT AP-15 @ 5v
Thermochill PA120.3 | 1 x NF-A12x15 | 2 x ML120 | PWM
mongoled is offline  
post #93 of 105 (permalink) Old 06-25-2020, 09:23 AM
New to Overclock.net
 
Join Date: Feb 2017
Location: Sweden
Posts: 1,300
Rep: 56 (Unique: 43)
Quote: Originally Posted by 2600ryzen View Post
What voltage do you run your cpu at?
No static voltages for some time but EDC bug settings, High LLC + 0.050V+ offset for several months. Uses mostly above 1.400V.
My manual OC adventures where in the 1.450V range close to launch days but not used that for quite a while, but I tortured it quite a bit with high voltage trying to get frequency back then.

PBO EDC=10 works well enough and voltage makes it better. More voltage = more frequency.

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  
Sponsored Links
Advertisement
 
post #94 of 105 (permalink) Old 06-25-2020, 10:37 AM
New to Overclock.net
 
Join Date: Aug 2019
Posts: 311
Rep: 6 (Unique: 5)
I'm sure there's lots of vdroop when the cpu's asking for more than 1.4v so adding a little extra voltage through +offset is probably safe. Did you try increasing scalar instead?
2600ryzen is offline  
post #95 of 105 (permalink) Old 06-26-2020, 11:19 AM
Turtle Lives Matter
 
damric's Avatar
 
Join Date: Feb 2010
Location: Azalea City
Posts: 4,774
Rep: 433 (Unique: 333)
How old is that PSU? If the capacitors have aged, you could be having enough ripple to cause instability. The fact that raising LLC or adding a bit of offset voltage fixes that leads me to think it is ripple on the 12v rail from evaporated dielectric in the caps causing the ripple to not be filtered after it is rectified.

~~~~~~~~ instead of ----------------- feeding your CPU.

Borg Cube
(15 items)
Ryzen RGB AF
(17 items)
Polaris AF
(13 items)
CPU
R5 3600
Motherboard
X470 TAICHI
GPU
Gigabyte RX Vega 56 Gaming OC
RAM
Ripjaws V 4x8GB
Hard Drive
Silicon Power A80 NVMe
Power Supply
EVGA SUPERNOVA G3 750W
Cooling
TCP-612
Cooling
2x EVGA FX 120mm Fan
Cooling
2x Cooler Master MasterFan Pro 140 Air Flow
Case
HAF XB
Operating System
10-64 PRO
Monitor
Samsung 4K 60Hz
Keyboard
Logitech G105
Mouse
Devastator II
Audio
Logitech Z623
CPU
Ryzen 5 1600 (12nm)
Motherboard
X470 Pro Carbon
GPU
Vega 64
RAM
OLOy WarHawk RGB 4x8GB
Hard Drive
Silicon Power A80 NVMe
Hard Drive
Silicon Power A80 NVMe
Power Supply
SUPERNOVA G3
Cooling
Wraith Prism
Cooling
Metallic Gear Skiron 120mm x4
Cooling
Metallic Gear Skiron 140mm x2
Case
Phankteks Eclipse P400A
Operating System
10 Pro
Monitor
Samsung 58" UHD
Monitor
V7 L238E-2N
Keyboard
Rosewill NEON K85 RGB
Mouse
Logitech M570
Audio
Vintage Sony
CPU
R5 1600 (12nm)
Motherboard
B450 Steel Legend
GPU
XFX RX 570
RAM
OLOy WarHawk 2x8GB
Hard Drive
Silicon Power A80
Power Supply
Rosewill CAPSTONE
Cooling
Wraith Prism
Cooling
Cooler Master Jet Flow 120mm x2
Case
Cougar MX330-G
Operating System
10 Pro
Monitor
ASUS 1080p 23"
Audio
Xonar SE
Audio
Logitech
▲ hide details ▲


damric is offline  
post #96 of 105 (permalink) Old 07-02-2020, 09:55 PM
New to Everything
 
blitzkrieg666's Avatar
 
Join Date: Apr 2018
Posts: 18
Rep: 1 (Unique: 1)
3950X BF1947SUT batch

Was about to post a new thread, and found this! Im in the same club apparently.
Doesn't matter if XMP is off or on, Auto or manual SOC, Optimized defaults/Clear CMOS; Core 12 and/or 14 (2nd CCD) fails on the first minute of Small FFT (It was on length 36K). Also no difference whether on bios F20a or F11. Die temps ranged ~57°C and spiked 60°C during the run (water temps ~30°C) so definitely not heat issue.
The only thing that helped is just by changing CPU LLC to Medium. I'm on X570 Master and BF 1947SUT batch (memtest passed).

You guys think i should go through the warranty hassle and get it exchanged? All I'm looking for is stability.
blitzkrieg666 is offline  
post #97 of 105 (permalink) Old 07-02-2020, 10:48 PM
New to Overclock.net
 
rdr09's Avatar
 
Join Date: Mar 2011
Location: From the US but lives in Africa
Posts: 19,985
Rep: 892 (Unique: 643)
Quote: Originally Posted by blitzkrieg666 View Post
Was about to post a new thread, and found this! Im in the same club apparently.
Doesn't matter if XMP is off or on, Auto or manual SOC, Optimized defaults/Clear CMOS; Core 12 and/or 14 (2nd CCD) fails on the first minute of Small FFT (It was on length 36K). Also no difference whether on bios F20a or F11. Die temps ranged ~57°C and spiked 60°C during the run (water temps ~30°C) so definitely not heat issue.
The only thing that helped is just by changing CPU LLC to Medium. I'm on X570 Master and BF 1947SUT batch (memtest passed).

You guys think i should go through the warranty hassle and get it exchanged? All I'm looking for is stability.
If you can buy time and the cpu only fails during Prime, then wait when the XTs are out. They might give you one as a replacement.

By Built-in Benchmark Queen:
My 2.5 years old delidded 7980xe @ 4600mhz all core with 49ns memorylatency (4000c16 twekaed) is faster than 3950x max overclocked on water in every scenario. Pretty good for a ancient cpu
In gaming it has no chance vs my new 10900k @ 5400mhz and 4600c16 tweaked memory. ~35ns memorylatency.
2nd AMD Build
(10 items)
CPU
2700
Motherboard
470
GPU
5700
RAM
3466/14
Hard Drive
1000
Power Supply
700
Case
212
Operating System
10/64
Monitor
55/2160
Keyboard
M100
▲ hide details ▲
rdr09 is offline  
post #98 of 105 (permalink) Old 07-03-2020, 01:14 AM
New to Everything
 
blitzkrieg666's Avatar
 
Join Date: Apr 2018
Posts: 18
Rep: 1 (Unique: 1)
Quote: Originally Posted by rdr09 View Post
If you can buy time and the cpu only fails during Prime, then wait when the XTs are out. They might give you one as a replacement.
Woah, didn't know there'll be a 3950XT..
blitzkrieg666 is offline  
post #99 of 105 (permalink) Old 07-03-2020, 01:20 AM
New to Overclock.net
 
rdr09's Avatar
 
Join Date: Mar 2011
Location: From the US but lives in Africa
Posts: 19,985
Rep: 892 (Unique: 643)
Quote: Originally Posted by blitzkrieg666 View Post
Woah, didn't know there'll be a 3950XT..
There won't be but by the time the XTs come out, all the rest should be binned. You get one now you might end up with the same issue.

By Built-in Benchmark Queen:
My 2.5 years old delidded 7980xe @ 4600mhz all core with 49ns memorylatency (4000c16 twekaed) is faster than 3950x max overclocked on water in every scenario. Pretty good for a ancient cpu
In gaming it has no chance vs my new 10900k @ 5400mhz and 4600c16 tweaked memory. ~35ns memorylatency.
2nd AMD Build
(10 items)
CPU
2700
Motherboard
470
GPU
5700
RAM
3466/14
Hard Drive
1000
Power Supply
700
Case
212
Operating System
10/64
Monitor
55/2160
Keyboard
M100
▲ hide details ▲
rdr09 is offline  
post #100 of 105 (permalink) Old 07-03-2020, 01:46 AM
New to Everything
 
blitzkrieg666's Avatar
 
Join Date: Apr 2018
Posts: 18
Rep: 1 (Unique: 1)
Quote: Originally Posted by rdr09 View Post
There won't be but by the time the XTs come out, all the rest should be binned. You get one now you might end up with the same issue.
Ahh ok make sense. I got mine December last year anyways. Will wait till XTs come out then..
blitzkrieg666 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: 1 (0 members and 1 guests)
 
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