My overclock is deemed stable when: ...

[UltraTaco]

You know what i find frustrating?
I was both stable at my 3.9 and 4ghz overclocks.

BUT, as I switched to newer prime95 version, from25.11 to 29.4, none of my overclocks are stable anymore! I mean they now fail within 2.5hour timeframe, whereas, they would be fine for over 20hours on previous version!

I should join this "my OC is 2 hours prime stable" club I'd be a lot happier and have more hair left..

Edit:speaking of which, look what I found in read me
View attachment 195587

 

[EarthDog]

Its because of the added avx testing from the different versions.

 

[UltraTaco]

What's that? Does my 920 have that? Do I need it?

 

[mackerel]

Recent versions of Prime95 will use the best instructions available for compute throughput. This means if it supports AVX, it'll use that. I think AVX was introduced with Sandy Bridge, and AVX2 with Haswell, each giving a significant bump in performance (and stress) compared to before. How much software uses it, is another question. Even if other software uses it, it may not use it in such an aggressive manner. I think from Kaby Lake onwards, they put in an additional bios setting "AVX offset" so you can run AVX workloads at a lower clock than other things.

BTW my main desktop (6700k) runs other prime number finding software which uses the same math library as Prime95, so effectively the loading is comparable. As all work is double checked, I find I get one detected error, typically every 3 to 6 months. This is still higher than I like... but is practically impossible to diagnose at such a low rate. Other systems don't have this error rate so I doubt it is a software problem.

 

[Kenrou]

Had to dig this from another thread, keep forgetting which versions are which For reference i always use the latest version with with the command lines :

26.6 = SSE = gaming, office work
27.7 = SSE/AVX = x264 encoding, image editing, heavy multitasking
28.5+ = SSE/AVX/AVX2/FMA3 = x265 encoding, folding, important work, ~100% stability

You can also add command lines to "local.txt" in Prime95 directory to simulate older versions:

CpuSupportsAVX=0
CpuSupportsAVX2=0
CpuSupportsFMA3=0

 

[UltraTaco]

Mackerei, so it slowly uh, degrades in as sense and prime is catching it? I'm a little confused sory. Could be ram due to not having ecc? Sorry, I don't know what I'm talking about..

Kenrou, thank you, I guess I jumped all in with this recent update

 

[mackerel]

Short version: Prime95 is one of the worst case stress tests, especially in the area it tests.

As a thought, is the OC system the i7 920 in sig? That pre-dates AVX anyway... it may still be that later versions of Prime95 have more optimisations for other instruction sets leading to some difference between recent and older versions.

 

[UltraTaco]

Yup, that's the one I'm interrogating.
It worked prefect witth earlier prime

 

[UltraTaco]

I'm sorry to keep beating.
Does this mean arlier prime versions make stability invalid? I mean since 26.6 is for office work stable, then mine(25.something) is "able to boot" stable?

I feel like I lost identity..

 

[EarthDog]

I have Deja vu...

Was it stable for YOU? If it is, it doesn't matter which you use. It was said before, a couple of times, there are going to be multiple opinions and they are ALL correct for their uses. You just need to figure out what you want to test with and be stable. Older P95? AIDA64? IBT? Realbench? So long as you are stable for your USES... you are set.

 

[UltraTaco]

Thank you, sir. And I am sorry bsod124 is up vcore. I just r,an into it. Turned on ht and while stuff was stable without it, I got 124. Increased vtt, didn't help. Went in and increased vcore, it runs!! sorry

 

[Nebulous]

Thank you, sir. And I am sorry bsod124 is up vcore. I just r,an into it. Turned on ht and while stuff was stable without it, I got 124. Increased vtt, didn't help. Went in and increased vcore, it runs!! sorry

See? It's all trial and error no matter the system. Gotta be patient and you'll get this

 

[UltraTaco]

I can see why all of a sudden EE cpu becomes appealing to some folks

 

[UltraTaco]

My new stress test. Fsx+video rendering. Poor thing will take 20 hours to render 20minute video

View attachment 195603

 

[RJARRRPCGP]

I sometimes got errors on 18th hour on prime95 blend, so 2 hours may not work

"sometimes" ->sounds like too high of temps at times... Reminds me of Prime95 suddenly giving me an error at 23 hours on my Barton 10 years ago, IIRC, LOL. It was because of the sun shining on the same window I pull cold air from...

Thank you, sir. And I am sorry bsod124 is up vcore. I just r,an into it. Turned on ht and while stuff was stable without it, I got 124. Increased vtt, didn't help. Went in and increased vcore, it runs!! sorry

I know that Windows XP's equivalent to "0x00000124", "0x0000009C" definitely meant more Vcore required or I hit a core brick wall, on my grumpy old 2002 Athlon XP T'bred 2000+ AIUCB...
Under XP, 9C means the same thing as 124 on Vista and later...

 

[UltraTaco]

Okay, so I got meself some RealBench2.56 from ROG asus.
Mama mia, it looks great, as soon as I fired her up she started resizing and doing a bunch of image stuff, then some encoding and then it halted because Lexmark.exe was not found..do I need it?

Anyways, I'm running infinite benchmark. I read somewhere it's more stressful than stress test, is it ture?

Thank you, men.

 

[UltraTaco]

Little finding thought I needed to share.

i7 920 c0 184x21 HT off vcore 1.218
Asus p6t deluxe v1

I did a test. I purposely dropped my stable vcore down 3 notches, making it 1.2volts. it is definitely unstable.

Prime95 25.11 small fft VS OCCT 4.5.1. large data set

Occt crashes bsod 010 or 124 within 2-3 minutes
Prime keeps going 10 minutes +...maybe eventually it would bsod too. (Blend bsod for sure, I tried..takes a while maybe an hour don't remembee)

 

[Alaric]

My overclock is deemed stable when: ...

A) No Magic Smoke appears
B) You don't get the dreaded "BOOTMGR.EXE File is missing. Please insert your installation DVD to fix this. Press F8 for Advanced Start up options" error.

 

[EarthDog]

Little finding thought I needed to share.

i7 920 c0 184x21 HT off vcore 1.218
Asus p6t deluxe v1

I did a test. I purposely dropped my stable vcore down 3 notches, making it 1.2volts. it is definitely unstable.

Prime95 25.11 small fft VS OCCT 4.5.1. large data set

Occt crashes bsod 010 or 124 within 2-3 minutes
Prime keeps going 10 minutes +...maybe eventually it would bsod too. (Blend bsod for sure, I tried..takes a while maybe an hour don't remembee)

Yup. Known. Different dataset sizes etc... makes sense.

 

[UltraTaco]

ED, but only cpu was affected by underclock and small fft in preme tests cpu primarily...what is small fft test for then?

Alaric been there. I was pushing my 1333 ram yesterday trying to hit 1900 and get into windows hoping those 2 things don't happen! Wanted to beat your winsat score. Max so far:
View attachment 196238