• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

The Opteron Skt 939 Overclocking thread !!!!

matt1 said:
I'll be having nightmares tonight.. yeesh. Did you put it in the wrong way initially or something? I've never had a problem with installing a socket 939 chip, just make sure the little lever is up, drop the CPU in (should see it fall into the holes) and push the lever down.



Nice results there mate. You ran SP2004 for almost the exact same amount of time I did for my CAB2E, freaky, spiritual bond? :p

Sold mine now, was a good chip, but I want 3G so hopefully getting my mitts on a 146. :D

Yeah i think initially i put it in the wrong way and must have bent the pins, it went in, but one side was raised a little, i ignored this (couldnt really distinguish it as a substantial "raise"). Mounted CPU block, booted but got a blank screen, LED's deiagnostics read "internal memroy error". At that point i was like "****" :p .

Took me ages to straighten the pins, you need huge amounts of paitence.
 
Well my 146 (CABYE 0540 FPMW) passed OCCT 30min test at 3Ghz, i put it to 3.1Ghz with the voltage at 1.55v and i'm testing under S&M right now - 22% completed.

...
 
Bossman said:
Just come back home from being at work all day, CPU is still going under SP2004, currently 10hours 16mins in and happily sitting under 40 degrees load.
This is a 144 CAB2E for the new comers, heres the OC setup.

311x9 = 2800 @ 1.55v (set to 1.6 in the bios)
Memory set-up as default with a 133Mhz divider running at 200Mhz (stock)
The LDT bus is at x3 which is slightly below the 2000Mhz target.

I'll get some screenies done of all the test results and post a healthier message in a bit.



Multiplier of 9 on a 146? :confused:

Yes 9 multiplier cause I was testing out my mobo's ability to hit 300fsb. Since then I swapped the bios to an older one and dialled in 280fsb * 10multi, +0.1v vcore still, seems happier at this setting. Will test this out, dont want to jump the gun and say this is stable. Its tricky overclocking your cpu when ** mobo is overclocked quite high to be able to stretch the cpu's legs.

...
 
here's my results

Opteron 144 CABYE 0536GPMWs one 2949mhz and 3044mhz max
http://i4memory.com/showthread.php?t=1536

Opteron 146 CABYE 0540FPBW one 3050mhz and 3072mhz max
http://i4memory.com/showthread.php?t=1594

Opteron 146 CAB2E 0540FPMW - 2800mhz at 1.31v and 3171mhz at 1.616v
http://i4memory.com/showthread.php?t=1662

average over 5 cpus = 3044-3057mhz at 1.57v

2800mhz at 1.312v
superpi-32m_tn.jpg


stressprime_1_tn.jpg


haven't tried lower vcore yet :D

2x 512MB Gskill PC4400LE TCCD 431
@282mhz 2.5-3-3-7 7-14-2222 on 180 divider


superpi-32m_tn.jpg


So far best 1M clock/time

superpi-1m_tn.jpg


room temp a bloody hot 32C right now hehe
 
I dont know why people are so worried about pushing your RAM to its upper limits, CPU speed is critcal above all else with the A64 chips as the memory controller is on the core. You'd be best to find the max speed of your CPU first then get your RAM as fast as you can (on dividers if need be) as the A64s dont take the performance hit that the XPs did when the RAM is on a divder.

I've started an all day test on my 146 now - SP2004, will see if it can do 10hrs+ at 3.1Ghz 1.55v.

Edit: it stopped already :mad: back down to 3Ghz it is then.

...
 
Last edited:
Increased RAM frequency still gives an increase in performance though, this cannot be disputed.

288mhz at 2.5-3-3-7 is going to absolutely waste RAM at 200mhz, even if it's at 2-2-2-5, just as my Redlines will running at 250mhz on 1.5-2-2-5.

For me it's important to get a good CPU clock and a good RAM clock.
 
Kesnel said:
Increased RAM frequency still gives an increase in performance though, this cannot be disputed.

288mhz at 2.5-3-3-7 is going to absolutely waste RAM at 200mhz, even if it's at 2-2-2-5, just as my Redlines will running at 250mhz on 1.5-2-2-5.

For me it's important to get a good CPU clock and a good RAM clock.

I half agree with you, but a couple of points (which you may or may not agree with):

With my opteron (see sig), I have had to use a pretty harsh divider to get the CPU to run at this speed (which I suspect points to the mem controller as the weak point). It is not worth sacrificing 50 Mhz CPU clock to gain 20 Mhz on the memory (for example). It does depend on the bench you are using to test this, and SuperPi will show the most memory-skewed weighting. Something like 3DMark '01 (which I believe represents a pretty good all-round benchmark now days), will show MAJOR CPU dependence and won't respond to memory increases much at all, apart from tighter timings.

The second point is that, IMO, running the memory @ 2-2-2-6 1T and low(ish) speed beats all other configurations. e.g. when I had TCCD which could run up to 320 Mhz 1T (IIRC), I still ran it at low speeds and low timings because benches showed this to be the best.

I think, in the example you gave, 200 Mhz memory @ 2-2-2 might actually be better in games benchmarks than 288 @ 2.5-3-3, although I'm pretty sure it would be close. Of course, it depends on your system usage. Mine is almost exclusively a games PC, so I use games benches to determine the best combination of CPU and memory clock.
 
I agree completely that CPU speed should not be sacrificed for the sake of the memory as you're not going to see any performance increase.

Obviously CPU speed should come first, however I like to try and optimise my clock so that I get the most out of my memory as well. Although, this will always be dictated by the maximum CPU clock as opposed to the max memory clock.

In an ideal world we'd be able to max out both componants, however as this simply isn't feasible or practical in most cases, CPU speed must take priority. :)
 
Just found out the reason my 148 wasn't 100% stable (only OCCT could find errors) - it seems my Corsair 4400 doesn't like running at it's rated speed with 1T or just doesn't like this board/chipset setup.

It's meant to run 275 @ 2.5-4-4-8 (2.75v) but I couldn't get it stable at ~250-260 at those timings (1T). Currently running it at 166/200 divider with the settings above and OCCT has been running perfect now for 3 hrs - failed after 20-25mins before.
 
Last edited:
Overclocking Database Fixed

Ok guys, I hold my hands up to being a lazy developer when I am at home. The database now accepts multipliers higher than 9, the memory divider now holds the speed your memory is running at rather than a decimal divider e.g. 0.83, 0.66 etc.

For those lucky people on phase the idle and load temps now accept negative numbers. If you use liquid nitrogen send me an email :D

Anyway here is the URL:-

http://www.skiltynet.co.uk

Sorry for the previous "features" :rolleyes:

If somebody sends me the current spreadsheet I will upload it into the database.
 
hmmm wondering if anyone can help me here - left SP2004 running last night, on mt CAB2E 0540 FPMW clocked at 2520mhz and it failed after half an hour or so, at this rate I'm going to end up running it at stock!!!!

I have an Epox 9NDA3J (NF3 Ultra mobo)
XP-120 with Panaflo 120mm L1A
Corsair 2GB DDR XMS3200C2PT
x800 XT PE
Akasa 460W PSU
xp64

I don't see general instability, however all my stress testing programs fail on me, I can't for the life of me figure out what the problem is. I'm sure the CAB2E is capable of more than 2.5.
 
Vader there is the chance that the CPU just isn't a good clocker. What's even worse is that the other one I sold to the guy on another forum is probably a good clocker. Oh well... can't win em all I guess.... luck of the draw and all that.

I suggest buying a 146 as soon as they come in stock as they are still likely to be CABYE 0540's and then stick that one on Ebay to get your money back. :(
 
I just had a call from the competitor who has them in stock today as I get on well with the guys there. There have cherry picked out 2 x 146's for me. I have coming tomorrow: -

CABYE 0540 FPMW -- Batch 0012
CABYE 0540 FPAW -- Batch 0019

Am I fcuking excited or what !!!!!! Thats the lowest batch number Ive ever had/seen. With the previous batch of CABYE, the lower the batch numbers, then better the overclocks were. I'm gonna be in processor heaven !! LOL
 
benk, dunno who told you that but its not true really, my 146 cabye 0540fpbw is batch 0564 and is prime stable @ 2.9ghz on stock volts & 3ghz with 1.45V which is amazing and yet to see others with such low volts
i also have a 148 exact same stepping batch 0169 yet it needs 1.5V for 2.8ghz & 1.6V for 3ghz
 
Its something to do with the beginning and ends of the wafer. My uber techie mate was trying to describe it to me last night. Apparently, the very beginning, the middle and the very ends of the wafers are the best for overclocking, and you can get a pretty good idea where the position of the wafer by a batch number. I'm not an uber techie, maybe he was just winding me up, but he generally knows his stuff. Perhaps I may have mis-understood. All I know is that I have 2 very very early batches for the particular steppings.

I suppose the proof of the pudding will be with the results I get tomorrow
 
Back
Top Bottom