prime95 small FFT's passed but large and blend error. - Overclock.net - An Overclocking Community

Forum Jump: 

prime95 small FFT's passed but large and blend error.

Reply
 
Thread Tools
post #1 of 24 (permalink) Old 01-22-2019, 11:25 AM - Thread Starter
New to Overclock.net
Assirra is an unknown quantity at this point Assirra is an unknown quantity at this point
 
Assirra's Avatar
 
Join Date: Jul 2009
Posts: 3,067
prime95 small FFT's passed but large and blend error.

Greetings

I got a i7 8700k overclock to 4.8Ghz with 1.260volt and it worked flawless for while now. Curious about stability i decided to run prime95 and see if it could handle it.
Now the results i got are making me question my memory might be bad.
Small FFT test went 24hours without any issue.
Large FFT test got rounding errors around the 17hour mark
blend test got rounding errors after about an hour.

Now considering the RAM usage in comparison when each test started to having errors i think my ram might be bad.
I also did a test with putting everything to stock without any OC or XMP and it still spewed out rounding errors.

Is my memory bad and do i just need new or is there something deeper going on?

Assirra is offline  
Sponsored Links
Advertisement
 
post #2 of 24 (permalink) Old 01-22-2019, 11:39 AM
New to Overclock.net
tostitobandito
 
Join Date: Sep 2018
Posts: 276
Increase CPU voltage to like 1.3v and see if the errors go away. 1.3v should be overkill, though still safe, and if you're still getting errors after that I would focus on the RAM. If the errors go away at 1.3v then step the voltage back down until you find the line of stability and run something above that.

The only other thing that comes to mind is VCCSA/VCCIO voltages if you have them set to something other than auto.
tostitobandito is offline  
post #3 of 24 (permalink) Old 01-22-2019, 11:48 AM
New to Overclock.net
Falkentyne is on a distinguished road Falkentyne is on a distinguished road Falkentyne is on a distinguished road Falkentyne is on a distinguished road
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 5,268
Quote: Originally Posted by Assirra View Post
Greetings

I got a i7 8700k overclock to 4.8Ghz with 1.260volt and it worked flawless for while now. Curious about stability i decided to run prime95 and see if it could handle it.
Now the results i got are making me question my memory might be bad.
Small FFT test went 24hours without any issue.
Large FFT test got rounding errors around the 17hour mark
blend test got rounding errors after about an hour.

Now considering the RAM usage in comparison when each test started to having errors i think my ram might be bad.
I also did a test with putting everything to stock without any OC or XMP and it still spewed out rounding errors.

Is my memory bad and do i just need new or is there something deeper going on?
This is 100% either RAM or IMC related.

[email protected] ghz, RX Vega 64, 32GB DDR4, Gigabyte Aorus Master, Seasonic Platinum 1000W, Corsair 760T
Alt: MSI GT73VR Throttlebook with 7820HK @ 4.7 ghz, GTX 1070 MXM TDP mod to 230W, 32 GB RAM
Falkentyne is offline  
Sponsored Links
Advertisement
 
post #4 of 24 (permalink) Old 01-22-2019, 12:07 PM - Thread Starter
New to Overclock.net
Assirra is an unknown quantity at this point Assirra is an unknown quantity at this point
 
Assirra's Avatar
 
Join Date: Jul 2009
Posts: 3,067
Quote: Originally Posted by tostitobandito View Post
Increase CPU voltage to like 1.3v and see if the errors go away. 1.3v should be overkill, though still safe, and if you're still getting errors after that I would focus on the RAM. If the errors go away at 1.3v then step the voltage back down until you find the line of stability and run something above that.

The only other thing that comes to mind is VCCSA/VCCIO voltages if you have them set to something other than auto.
I used 1.3 before for a while and it still gave me errors. it was also a lot hotter.
I don't know what either of these settings are but i am pretty sure the only voltage i changed was the CPU one.

Assirra is offline  
post #5 of 24 (permalink) Old 01-22-2019, 12:07 PM - Thread Starter
New to Overclock.net
Assirra is an unknown quantity at this point Assirra is an unknown quantity at this point
 
Assirra's Avatar
 
Join Date: Jul 2009
Posts: 3,067
Quote: Originally Posted by Falkentyne View Post
This is 100% either RAM or IMC related.
IMC?

Assirra is offline  
post #6 of 24 (permalink) Old 01-22-2019, 12:56 PM
New to Overclock.net
Falkentyne is on a distinguished road Falkentyne is on a distinguished road Falkentyne is on a distinguished road Falkentyne is on a distinguished road
 
Falkentyne's Avatar
 
Join Date: Dec 2007
Location: Riverside
Posts: 5,268
Quote: Originally Posted by Assirra View Post
IMC?
The memory controller of the CPU (mostly controlled with Input output (VCCIO) voltage and system agent (VCCSA) voltages. Default are 0.95v and 1.05v but this is for 2133 mhz-2400 mhz memory. Officially they can go to 1.05v and 1.15v and be in spec, but everything past that is overvolting. But for higher speed memory, it has to go up higher sometimes. When you pass the small FFT size range, the prime95 data has to move out of the L1 and L2 caches and start using the L3 cache and system memory access. I think this starts at the 128K-384K FFT range. Usually, for testing RAM, you're supposed to set "Blend" to custom size, 512K-4096K, and 75% of your total system memory in mbytes (e.g. 24576 if you have 32 GB). I do not know how to purely test IMC however. I never did that. But I'm guessing 128K-768K range.

[email protected] ghz, RX Vega 64, 32GB DDR4, Gigabyte Aorus Master, Seasonic Platinum 1000W, Corsair 760T
Alt: MSI GT73VR Throttlebook with 7820HK @ 4.7 ghz, GTX 1070 MXM TDP mod to 230W, 32 GB RAM
Falkentyne is offline  
post #7 of 24 (permalink) Old 01-22-2019, 01:05 PM
Programmer
BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about BradleyW has a spectacular aura about
 
BradleyW's Avatar
 
Join Date: Feb 2009
Location: United Kingdom
Posts: 28,171
Quote: Originally Posted by Assirra View Post
Greetings

I got a i7 8700k overclock to 4.8Ghz with 1.260volt and it worked flawless for while now. Curious about stability i decided to run prime95 and see if it could handle it.
Now the results i got are making me question my memory might be bad.
Small FFT test went 24hours without any issue.
Large FFT test got rounding errors around the 17hour mark
blend test got rounding errors after about an hour.

Now considering the RAM usage in comparison when each test started to having errors i think my ram might be bad.
I also did a test with putting everything to stock without any OC or XMP and it still spewed out rounding errors.

Is my memory bad and do i just need new or is there something deeper going on?
The issue lies with the RAM, memory controller or the uncore frequency.
Set the uncore speed to stock! Enable XMP for the RAM and retest. If it fails, you'll probably need to reduce the RAM speed or increase the VCCIO/VCCIA voltages.....or both combined.

Intel Core i9 9900K HT 5GHz
GIGABYTE Z390 Ultra
AMD ASUS STRIX OC VEGA 64
G-Skill Trident 2x8GB DDR4 4000MHz
Samsung EVO NVM.e 250GB
SeaSonic 1250W Gold
Corsair 900D + Custom Loop.
BradleyW is offline  
post #8 of 24 (permalink) Old 01-22-2019, 01:45 PM
New to Overclock.net
spooklexity Level 1
 
spooklexity's Avatar
 
Join Date: Jan 2017
Posts: 34
If you have fast RAM or did manual tuning after the OC, you might have insufficient core voltage. Decreasing latencies and simultaneously increasing throughput will lead to higher AVX throughput once store/load commands are affecting the RAM as well. I used to easily pass all prime95 tests at 4.2 GHz until I tightened my timings significantly. That in turn has lead to this OC becoming p95-unstable. Reducing multiplier by 1 and increasing vCore has fixed that for all non-in place FFT tests, the latter might require even more input/core voltage from my side, with gSAT being stable for 5+h, and a custom p95 blend-type test being stable for 7+h. Maybe try running gSAT or Karhu. If they run through without issue, it is most likely that you have a slightly unstable OC.

X99-Longterm
(9 items)
CPU
Core i7-6800K
Motherboard
Gigabyte GA-X99 Gaming 7 WiFi
GPU
Gigabyte R9 Fury Windforce 3X OC waterblock retrofit
RAM
4x G.Skill F4-3000C16-8GISB
Hard Drive
Samsung 970 Evo
Hard Drive
Samsung 850 Evo
Cooling
Custom Water loop: 2x140 slim internal, 4x180 thicc external
Case
Nanoxia Deep Silence 5 Rev. B
Operating System
Windows 10 Education
▲ hide details ▲
spooklexity is offline  
post #9 of 24 (permalink) Old 01-22-2019, 03:03 PM - Thread Starter
New to Overclock.net
Assirra is an unknown quantity at this point Assirra is an unknown quantity at this point
 
Assirra's Avatar
 
Join Date: Jul 2009
Posts: 3,067
I will repeat this since it seems to have flown under the radar since people mentioning OC and voltage tweaks.
I had blend testing failing on a complete stock machine, no XMP, no CPU OC, straight up restored bios to defaults as safety to make sure i didn't mess a setting up.

Assirra is offline  
post #10 of 24 (permalink) Old 01-22-2019, 03:24 PM
New to Overclock.net
spooklexity Level 1
 
spooklexity's Avatar
 
Join Date: Jan 2017
Posts: 34
Wooops. Yes, most likely bad memory, also makes sense given the time-delta between the LFFT and Blend tests, respectively. I keep my advice in general, run memetest86, gSAT, or Karhu to investigate and confirm the RAM failure.

X99-Longterm
(9 items)
CPU
Core i7-6800K
Motherboard
Gigabyte GA-X99 Gaming 7 WiFi
GPU
Gigabyte R9 Fury Windforce 3X OC waterblock retrofit
RAM
4x G.Skill F4-3000C16-8GISB
Hard Drive
Samsung 970 Evo
Hard Drive
Samsung 850 Evo
Cooling
Custom Water loop: 2x140 slim internal, 4x180 thicc external
Case
Nanoxia Deep Silence 5 Rev. B
Operating System
Windows 10 Education
▲ hide details ▲
spooklexity 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)
 


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