Prime 95 small vs large fft and bios vs winxp vcore

[nealh]

I realize best solution is to test Prime95 small and large fft...blend test and memtest

will we often see small fft be ok but large fft fail? on my last Venice 3200@2750 I ran small and large fft for 15hrs each before I stopped it and memtest for like 6hrs

I have opteron that I trying to find max oc on setup..so testing is ongoing


Vcore and temps

For me it is harder to get assess in bios on temps as the temps drop quickly on reboot

but vcore is perplexing..I see most post the vcore in CPU-z 1.31...but for me this is always lower than bios reports on my DFI SLI DR

for example...Opteron 165 9x 289=2600 testing as we speak on DFI SLI-DR 623-2 bios
vcore 1.35*110%= 1.485 in theory
bios reports 1.46 vcore
XP w/sp2 via speadfan 4.27 and cpu-z 1.31 report 1.44

I know multimeter is best..but I am not sure how to use and frankly do not want to try

So why are people reporting the XP vcore if this seems to be less accurate..I am hoping to keep vcore at 1.5v or less in the bios readings....and if lucky get 2700...HAHAHA..LOL...doubt I will get there

Also what program do you guys use for screen shots??

 

[RichUK]

With regards to the question "Also what program do you guys use for screen shots" i use this However if you are enquiring into a BIOS screenie, the only possible way is a Digi cam for example, then just upload you pic.

When i was testing my 165 i finally ended up sticking with 2.6Ghz at 1.52v (see sig), i eventually ended up testing for around (if i can remember) 9 hours or so on small FFT?s, and so many hours on large FFT's, and also plenty of runs on memtest.

If you are looking to be fully satisfied with the stability of your OC, i would recommend the following.

12 hours of Small FFT's two instances

12 hours of Large FFT's two instances

12 hours of blend test

Also about 30+ passes of memtest "Bootable method only!! Aka use a bootable floppy disk or CD image. Test this without a single error, to me a single error is a no go in memtest. A lot of people just stress in tests 5 and 8 because they see these specific tests as more defining however I prefer to run all test, having unstable ram can be detrimental on bad writes to your hard drive, therefore corrupting your data a lot quicker than a unstable processor.

And that is about it really with regards to stress testing, some people in my eyes go over the top and test for like 48 hour but i dont deem this necessary, of course that is my opinion.

I'll be interested in you results, i was stable at 2.7Ghz on 1.52 volts, but decided to run 2.6 to make the most out of my RAM, which was purpose bought so that i could run at a high frequency at CAS 2.5

Also with regards to the fluctuating voltage i wouldn?t worry to much on that, my board almost fluctuates .7 of a volt which is tremendous. As far as I am aware CPU-z receives its voltage readings from the BIOS, so they should be somewhat the same. I remember reading the temperature from my BIOS and within windows the temps were greater, only by a few degrees though. I always like to refer to CPU-z for my voltage reports, although no software can really produce definitive results that is left down to a multimeter (CPU-z should be sufficient if you are using a suitable BIOS rev), but as you are running a DFI can you not use smart guardian.

Keep us posted on you results, also what stepping is your processor, was you able to catch that at all before you put it in?

 

[nealh]

With regards to the question "Also what program do you guys use for screen shots" i use this However if you are enquiring into a BIOS screenie, the only possible way is a Digi cam for example, then just upload you pic.

DUH!



FYI:
Opteron 165 0534 CCWBE MPMW
using SI 120 with Nexus fan

Rich..what are bios vcore settings..this is my big concern right now

BIOS vcore 1.35v *110%= 1.485

BIOS reports 1.46v

Speedfan/cpu-z 1.31= 1.44v in winxp

so what is the chip getting

last night I stopped at 6hrs 50min small fft 2500 with vcore of 1.43 in bios/1.41v in xp

I have created 2 folders on C Drive with Prime 95..open each and start prime 95 and go to task manager and set affinity for 1 core each....speedfan repots 100% cpu use..

funny when I died last night at 1.41v@2600..cpu0 stopped cpu 1 was still priming but speedfan reported 50% cpu use on on core 1...weird

 

[RichUK]

Its hard to say without a multimeter what voltage the CPU is actually being supplied. I would be more inclined to believe the BIOS, with my board the BIOS and CPU-z have always reported the same voltage value however. I would say it is not really a worry until you start hitting a voltage of ~1.53/1.55 as that is really the highest safest operating voltage on these DC processors on air. In your scenario i would suggest not to worry until you start to exceed 1.52v ish (which hopefully you wont have to).

And with regards to "1.35v *110%= 1.485", i wouldn't always expect the theoretical BIOS set voltage to be the same as the actual voltage, unfortunately it never works out that way. For instance my board is awful with working out a processor voltage, if i set CPUvid at 1.55v is supplies 1.4, and the percentage increase on the over volt never works out to be what it is supposed to be either. I just boot into windows and read what CPU-z states.

The main thing to observe when using voltage below ~1.53 is the load temperature, which you don?t really want exceeding ~55 Degrees .

Also with regards to one core failing and one core carrying on with prime, it is simply to do with one core being stronger than the other, I have found this to be the same with mine, as a lot of other people have with their dual core procs also. In fact I tried to prime at 1.48v at 2.6 on my setup and I found the exact same thing happened with core 0 failing after about 25 mins, but core 1 would just carry on priming. I just had to up the voltage to 1.52 which was sufficient at that speed, and also at 2.7Ghz.

I personally would be more inclined to believe what task manager reports, with regards to each cores CPU usage rather than speed fan, have you tried to replicate the problem with task manager?

 

[nealh]

Originally posted by: RichUK
Its hard to say without a multimeter what voltage the CPU is actually being supplied. I would be more inclined to believe the BIOS, with my board the BIOS and CPU-z have always reported the same voltage value however. I would say it is not really a worry until you start hitting a voltage of ~1.53/1.55 as that is really the highest safest operating voltage on these DC processors on air. In your scenario i would suggest not to worry until you start to exceed 1.52v ish (which hopefully you wont have to).

Thanks I figure..my bios reading is best..I fell I am at 1.46 @2600 with small fft for 2.5hrs...hopefully this will go for 12 hrs

And with regards to "1.35v *110%= 1.485", i wouldn't always expect the theoretical BIOS set voltage to be the same as the actual voltage, unfortunately it never works out that way. For instance my board is awful with working out a processor voltage, if i set CPUvid at 1.55v is supplies 1.4, and the percentage increase on the over volt never works out to be what it is supposed to be either. I just boot into windows and read what CPU-z states.

I own a Neo 2 Plat...loves this board...it is in my sons computer now

The main thing to observe when using voltage below ~1.53 is the load temperature, which you don?t really want exceeding ~55 Degrees .

my load temps with SI 120/Nexus fan are at 47/48C on cpu...PMW IC is at 56/57C

Also with regards to one core failing and one core carrying on with prime, it is simply to do with one core being stronger than the other, I have found this to be the same with mine, as a lot of other people have with their dual core procs also. In fact I tried to prime at 1.48v at 2.6 on my setup and I found the exact same thing happened with core 0 failing after about 25 mins, but core 1 would just carry on priming. I just had to up the voltage to 1.52 which was sufficient at that speed, and also at 2.7Ghz.

my only real comment was when the one core faild..speedfan only reported a 50% cpu utlization on core 1..not 100%, which I thought was strange

ersonally would be more inclined to believe what task manager reports, with regards to each cores CPU usage rather than speed fan, have you tried to replicate the problem with task manager?

crap never really checked task manager I need to do this...crap,crap...Task manager is reported each cpu is at 50%..yet affinity has been set for cpu o for one instance of prime 95 and another instance has been set cpu 1

What the hell am I doing wrong...great..I am just wasting my time

I assume each core shoould be running at 100%...oh no

 

[RichUK]

Try downloading Stress prime and using this instead of prime it is exactly the same just a different GUI. Download here ... Reference picture here

Notice it has a better GUI (heuristically better in every way, which is always a must ).

Now open up two instances of this application (you don?t need two different folders, just double click again on the same icon to open a second window or "instance"). Now untick the "CPU affinity" tick box in both windows (instances), then set in one Stress prime window CPU 0 and in the other Stress prime window choose CPU 1. The top drop down can be used to set whether you want to test using Small or large FFT's or the blend test (of course if testing both cores simultaneously you want to select the same test in both instances). Then click on start and it will start priming, it uses the same iterations and is exactly the same but uses a far better GUI (Graphical User Interface), i use this all the time, and think that everybody that uses prime 95 to stress test should aswell. Also if one core fails you here a constant beep noise to indicate a test failure (of course you can turn this off if you wish).

Also when testing i like to have task manager open so i can see the activates of both processing cores on the two graph?s, along with the net usage of both the processors on the left pain (Each core at full load equates to 50% net usage of that processor, which would be indicated in the left hand pain), and also a temp monitoring program to obviously monitor load temps, i would suggest using MBM5 or the DFI Smart Guardian, this can also be used to monitor several other values along with your PWM temps etc all in the same window.

You can also use the latest version of S&M, and run a loop test, this is a very good program also, but be aware this program will be able to produce the most heat from your processor, which is good for testing max load temps.

hope this helps

 

[imported_Sincity]

RichUK: Thanks for the link to the GUI. Nice to know the time progress.:thumbsup: