Socket 939 Discussion Thread

[zarathrustra]

Nice entry... That 7950 sure pulls some nice 3d scores.

[likewhoa]

http://www.dfi-street.com/forum/showpost.p...2&postcount=312

 

AMD X2 4800+ with IHS removed, OC's stable to 2868Mhz.

Will boot to 2900mhz + but not stable in Prime95 and VERY hot.

 

Idles about 30 degrees, load is about 38-39.

Note: TIGHT RAM timings and almost DDR500 out of DDR400

 

damn I'm jealous!! :s

can't wait for my 7950X2 to arrive

[soundx98]

There is updated information in the OCDB Rules sticky.

http://www.dfi-street.com/forum/showthread.php?t=10192

Please make sure you re-read Section #4 - Requirements for OCDB Entires

[cuoficr]

There is updated information in the OCDB Rules sticky.

http://www.dfi-street.com/forum/showthread.php?t=10192

Please make sure you re-read Section #4 - Requirements for OCDB Entires

three cheer's for soundx, the rule's finally actually reflect the rule's

[soundx98]

Same rules, just a consolidation of threads and some more links for you dyslexic alzy fuggers

[supertux1]

Something that concerned me was that there wasn't any specification for which kind of Prime95 tests to do. So I did 'Maximum FPU Stress' on both cores which is the one that focuses on the CPU and less on RAM. Can we put that in the rules?

[Roger_D25]

Good point supertux1, I just assumed we were suppose to use the "In Place Large FFT's (maximum heat, power consumption, and some RAM tested"? There may be a way to know what test people have been using from their screenshots but I wasn't able to?

[soundx98]

You should be using "Blend" (auto sets up RAM memory usage).

You should use "Custom" and set RAM memory to use to 400 for 1024MB and 800 for 2G per core for Dual Cores.

This has been covered earlier in an attempt to standardize.

Not a question of which is better but what is the standard that has been used.

Blend is what was used for the original toaster sigs.

We'll get to updates at a later date.

[Praz]

Something that concerned me was that there wasn't any specification for which kind of Prime95 tests to do. So I did 'Maximum FPU Stress' on both cores which is the one that focuses on the CPU and less on RAM. Can we put that in the rules?

Something probably should be put in specifying the correct way to run Prime95 for SC processors. For DC processors a link is provided that clearly shows the correct way to set up Prime95.

 

Like the Toaster sigs, you MUST complete the following benchmarks, in the following manner

(Dual Core Prime95 Set Up is here [http://www.dfi-street.com/forum/show...18&postcount=1])

 

Also notice the following towards the bottom.

DO NOT GET CREATIVE AND ADD ADDITIONAL SCREENSHOTS OF ANYTHING!

 

edit: I guess I'm to slow.

[soundx98]

You're doing just fine Praz.

[supertux1]

Good point supertux1, I just assumed we were suppose to use the "In Place Large FFT's (maximum heat, power consumption, and some RAM tested"? There may be a way to know what test people have been using from their screenshots but I wasn't able to?

 

I read somewhere that the 'maximum fpu stress' test is best for testing strictly the

CPU and not much RAM, so that's what I use to find maximum CPU clock.

IMHO memtest86 is much better at testing RAM than Prime95 is, so I use that.

Right tool for the right job sort of thing. The inplace large FFT's stress the CPU/RAM controller,

and blend focuses more on RAM. Ideally the best test would be to run all three of these for 24 hours.

 

I see the logic in wanting to run a balanced torture test because although CPU and

RAM may function independently well, the real test is to see if they can work together

at overclocked speeds for practical use; ie games. That's what the 3DMark requirements

are for I guess. When both are going like mad the PSU is stressed and the case heats up

more and that can have an influence on stability.

 

I actually got my overclocking to the point where if I ran the Prime95 test during the day it would fail

after an hour or so (higher ambient temperatures) and if I ran it during the night with a window open in

my computer room it would be fine (lower ambient temp).

 

So this is why I recommend a 24 hour stability torture test, cuz it's easy to cheat and fake a stable

OC that only works at night or with the AC cranked.

[Nightwish68]

You should use "Custom" and set RAM memory to use to 400 for 1024MB and 800 for 2G per core for Dual Cores.

That´s what i used for my prime test custom with 800 MB per core for almost 9 hours, or was it 900MB?

I also added the stepping of my cpu to my OCDB entry.

For the rest my entry seems to apply to the rules.