The Official Opteron 939 Overclocking Thread

[dogsx2]

I can't see how anyone could use the stock cooler that came in the retail box to oc very much. It is sure a cheap thing. Even after a few days at 1.4v under load it runs 42c with pretty good air.

I see your board will run 400, have you dropped the x and tried that route?

[dinos22]

Does this apply to all 939 chips? Well, with higher preamble you don't exactly lose much performance - I'm sure increased CPU mhz, if attainable, would easily outweigh any negligible performance hit of higher preamble...

PRTeamJapan was specifically talking about this Opteron.....so does anyone have actual figures of performance impact? I know that CPC to 2T loses roughly 2-3% in performance

[ozzimark]

I can't see how anyone could use the stock cooler that came in the retail box to oc very much. It is sure a cheap thing. Even after a few days at 1.4v under load it runs 42c with pretty good air.

I see your board will run 400, have you dropped the x and tried that route?

i should try that, just in case something is hating on the 9x multi for some reason. 400x8 = 3200mhz, more than enough headroom

[afireinside]

PRTeamJapan was specifically talking about this Opteron.....so does anyone have actual figures of performance impact? I know that CPC to 2T loses roughly 2-3% in performance

3% will make or break a record... Last I checked, xtremesystems was about pushing your hardware to the limits, not setteling with mediocrity because it's fast enough for daily useage.

Yeah, I changed my avatar! The other one was ghey but I'm lazy I look 200% more professional with an expensive thermometer in my avatar

[Absolute_0]

Oh noes his ram didn't break a record when his intent was solely to test an Opteron processor for compatability with BH-5.... can we get back to overclocking?

[asianguy80]

I just got my opty 146 from vandi this morning...omg the ride there was CRAZY traffic but WELL worth it I only have pics of the chip right now and it's going in the machine soon! will have some results later tonight. anyways here is the teaser for you guys

It's not going into the shuttle i dont think it can handle 300+HTT so its going in a spare neo4 plat/nonsli using my gskill LE

oh yea it's the magic stepping of CABNE 0530 APMW

[afireinside]

Oh noes his ram didn't break a record when his intent was solely to test an Opteron processor for compatability with BH-5.... can we get back to overclocking?

TeamJapan already said it works fine. They DID say that a higher preamble is needed. Thats what needs testing.

Hopefully USPS does their job and I get my CPU tomorrow. I'll find out how well BH5 actually works on them and show you guys a neekid 146

[ZoKi]

be sure to keep us update asianguy80 on ur results

[dinos22]

AFI take it easy man......I'm sure these issues will be ironed out.......DFI will need to optimise the bios a bit too I'm sure

[afireinside]

I'm not trying to be a jerk but team japan said that you need to run a higher preamble to reach max stable clocks on bh5 with an opteron. Is it to much to ask for someone with an opteron and BH5 to to test it out?

[dinos22]

I'm not trying to be a jerk but team japan said that you need to run a higher preamble to reach max stable clocks on bh5 with an opteron. Is it to much to ask for someone with an opteron and BH5 to to test it out?

^^^^^^^I'd like to know what he's asking ^^^^^^^^

[dogsx2]

I just got my opty 146 from vandi this morning...omg the ride there was CRAZY traffic but WELL worth it I only have pics of the chip right now and it's going in the machine soon! will have some results later tonight. anyways here is the teaser for you guys

It's not going into the shuttle i dont think it can handle 300+HTT so its going in a spare neo4 plat/nonsli using my gskill LE

oh yea it's the magic stepping of CABNE 0530 APMW

Crank the bugger up and lets see what it will do. How high will a neo go?

[ozzimark]

Is it to much to ask for someone with an opteron and BH5 to to test it out?

i tried 6.0ns.. no boot

[afireinside]

^^^^^^^I'd like to know what he's asking ^^^^^^^^

Can I make it any easier to understand?

TEAM JAPAN TESTED AN OPTERON WITH BH5. THEY SAID THAT BH5 REQUIRES A HIGHER PREAMBLE TIME TO REACH MAX STABLE CLOCK SPEEDS THAN IT DOES ON A DESKTOP A64. It would be nice if someone could prove/disprove this statement by testing it. If you have BH5 and an opteron I'm sure you've maxed the BH5s clocks using a desktop A64 chip. It would take all of 5 minutes to boot your max clocks with the tightest preamble you had before and see if it works.

Thank you ozzi, thats all I wanted to know

[dinos22]

i tried 6.0ns.. no boot

with BH5?

[dogsx2]

I'm not trying to be a jerk but team japan said that you need to run a higher preamble to reach max stable clocks on bh5 with an opteron. Is it to much to ask for someone with an opteron and BH5 to to test it out?

Nope, I think a guy with a 146 is going to do it tomorrow.

[goodcooper]

holy crap man tone it down a bit, i'm just as eager to see as you, just hold the friggin fone a second

you may be interested in that, but i got 4400LCs and the rumor is that certain brainpowers don't work right withum, but i'm just going to take it easy until these things are widely available

[ozzimark]

with BH5?

yes. i'll try again just to verify

the one problem, i need a lot more voltage to oc my bh-5 with this board. on the dfi nf3, 275mhz was doable with 3.5v, now i need 3.67v for 273mhz
either way, i'll be back in a bit with a second word to make sure i didn't do something stupid the first time

[afireinside]

NF3=single channel NF4-dual channel

[dogsx2]

NF3=single channel NF4-dual channel

[afireinside]

His previous OC was done in single channel while his new setup is dual channel. Running in DC will require higher voltage to achieve the same clocks.

[ozzimark]

His previous OC was done in single channel while his new setup is dual channel. Running in DC will require higher voltage to achieve the same clocks.

almost correct. i did not forsee switching to s939, thus one of my two bh-5 sticks is still at home in my parents computer. both results are single channel

with 3.26v and 260mhz in the yellow slot to give a few errors per memtest #5 pass, but not a lot. not the most scientific way, i tested each two times

Preamble	Test 1		Test2
4.0ns = 	48 errors	76 errors
4.5ns = 	55 errors	71 errors
5.0ns = 	54 errors	69 errors
5.5ns = 	33 errors	35 errors
6.0ns = 	54 errors	81 errors - tried a third time, memtest crashed
6.5ns = 	44 errors	77 errors
7.0ns = 	no boot		no boot

sorry about that, i lied. won't boot with 7.0ns.. though it seems that 6 isn't exactly reliable
either way though, seems like 5.5ns is the best in this case

[afireinside]

Lower works better? Ok I'm really confused now.

BTW use orange slots for BH5

[ozzimark]

BTW use orange slots for BH5

trying. this board is somewhat new to me


also, the defintion of read preamble..

The Read Preamble time is a function of the asynchronous round trip loop latency, the clock
frequency, DIMM type and process type. These settings are initial recommendations only. They can
be adjusted to compensate for signal integrity issues, bus length mismatches, or any other empirical
data that proves that other settings are more reliable.

check page 172:
http://www.amd.com/us-en/assets/cont...docs/26094.PDF

there's a nice chart. 133mhz = 7.0ns, 166mhz = 6.0ns, 200mhz = 5.0ns. seems to be 1000ns / mhz = read preamble, thus as you scale up the mhz a lower value SHOULD be optimal, but for some reason, other people find the opposite


edit2: the orange slot will have to wait, i've gotta test out the 8x cpu multi first

[afireinside]

interesting... I hope I can test this tomorrow