Help me get the extra mhz out plz on my 165 opty

[Candymancan]

I have a 165 opteron with the CCBBE 0617FPWM stepping, and everyone i see with that stepping yea everyone seems to be able to easily get 2900mhz and 3ghz stable on there cpu using relativly low voltage around 1.4-1.5v.

 

My cpu however doesnt want to be stable near there. I can only manage around 2800-2850mhz stable using 1.47v. I dont understand this at all, i can get 2960mhz to stable to run series of 3dmark tests and stuff without problems or any other test i want, yet when i run orthos i fail 2 seconds into it, i lower my clock slightly and only whe ni reach around 2850mhz i can run the test 24/7.

 

I can get 3ghz to boot take pics but again orthos fails 2 seconds into the test. I dont know why my cpu with the same stepping as everyone else seems to overclock so poorly compared to them. I am watercooling my temps on core temp are saying core 0 is 10c warmer then core 1 on idle and load its about 5c warmer. They are reporting @ 41c and 31c. I figure it was because my core was hotter then the other so i decited to put my radiator in ice water and it lowerd my temps to around 20c and my overclocks didnt change i couldnt get above 2850mhz not even 10mhz and be stable even with the temps change....

 

To me that sounds like something is wrong. You'd figure with a 20c drop in temps i could get 10 more mhz outa it but i cant. Anyway Im using the bios that came with my SLI-DR board the 6/23/05 bios so its dated when i boot up. I tried a couple newer ones on DFI's website and my overclock got worse.

 

I am @ a loss here can anyone help me ? My ram is good up to 250mhz and im not going near that when i overclock.

[jlqrb]

It seems like you hit the wall at the volts you are at.

 

It seem like you have a really good cooling system so you could up the volts and try for it again, with Opterons I wouldn't be afraid to try 1.55 volts or higher especially if you temps are that good. Don't up the volts endlessly but I would give them a little tweeking.

 

when I was using socket 754 I had my AMD 3000 stable at 2.5ghz and could prime it forever, but when I raised it even to 2.53 I couldn't get it stable.

I hit my wall and had to stop.

 

I understand where you are coming from but just because you have really good stepping doesn't mean that you will be able to pull off what the others are, I think that 2850 is a really nice overclock and would be happy with it.

My 170 has hit around its max at 2750, I think I can take it a little higher but will need better cooling.

[Candymancan]

Yea i tried 1.55v no go, it just really sucks that people with the same stepping pretty much everyone are able to get like 100-150 more mhz then me.

[jlqrb]

sorry, it seems that you hit your max overclock, at this point no matter what you do it most likely won't have an effect on your overclock, but it is a good time to go through and tweak your timings and try to get the best results possible with what you go it up to.

[laserred]

I'd be much more likely to believe your failure is your memory, not your processor. I know it took a bit of tweaking my settings on my RAM (and it's PC4000, not 3200) to get stable over 250MHz. I bet if you go to a 150 RAM divider to keep your RAM under 230ishMHz you will have much more luck. It's possible it might not have an effect, but your only other option is buying some PC4000 or PC4800 especially since you've only got 1GB of RAM. Try loosening your Max Async Latency and Read Preamble 1ns each and retry, up until you get to around 10ns M.A.L. and 8ns R.P. That might help on your 166 divider. Also, is 100MHz really that big of a deal for you? It's only 3.5% or so difference, nearly the margin of error for most tests. Otherwise, good luck.

[Candymancan]

I'd be much more likely to believe your failure is your memory, not your processor. I know it took a bit of tweaking my settings on my RAM (and it's PC4000, not 3200) to get stable over 250MHz. I bet if you go to a 150 RAM divider to keep your RAM under 230ishMHz you will have much more luck. It's possible it might not have an effect, but your only other option is buying some PC4000 or PC4800 especially since you've only got 1GB of RAM. Try loosening your Max Async Latency and Read Preamble 1ns each and retry, up until you get to around 10ns M.A.L. and 8ns R.P. That might help on your 166 divider. Also, is 100MHz really that big of a deal for you? It's only 3.5% or so difference, nearly the margin of error for most tests. Otherwise, good luck.

 

 

If you ever played Supreme commander yes that extra will help me even if its only 1fps lol.. That games just eats my computer alive..

 

Yea i tried 150 and a 140 and 133 divider didnt help i lowerd my memory down to 200mhz with dividers to eliminate my memory no go.. Thx tho guess i shoulda mentiond that, tried changing those settings also and the drive strength and skew level it cant possibly be my memory.

 

One odd thing iv noticed while testing my overclocks out are to where iv been testing for like 36 hours on orthos if my screen saver or if i search the internet while the test is going it fails on me, but if i just let the computer sit there it'll run forever. Its really odd... to be honest iv ran my computer on 3dmark06/05/03/01 and aquamark @ 2960mhz where orthos would fail 2 seconds into and i never had problems im gonna try and see if its game stable if it is im just delete orthos and prime cuss the last couple years iv noticed these programs have been doing these wierd things even on my newcastle

[laserred]

Try the M.A.L. and R.P. timings I mentioned... I ran into this myself and loosening those timings helped make it stable.

[Candymancan]

i edited my post ehehe, tend to do that alot

[kitfit1]

From everything i've read in this thread i've got to agree with laserred, Ram timings is the cause of your problems. It would be really helpfull if you could post up an OCDB template with all your bios settings. From that something may be obvious to us, but not to you. Also i know you are using Orthos(because of the OCDB requirements), but have you tried the same settings using dual Prime instead? I ask because all over the net people are having issues with Orthos(me included). Sometimes tried and trusted tests are the best way to start, then move on to whats needed for the OCDB latter when your sorted.

[Candymancan]

Here is what i am using atm, iv tried changing the scew to increase and 255, i tried changing the drive strength from 1-8, iv tried all the settings nothing helps. Regardless of my ram timmings even if diviers are set to make my ram run @ stock i get no overclock increase so how could it be my memory timmings? Maybe i wrong who knows all i know is this cpu in the pic shouldnt be stuck @ 2800mhz specially with this stepping..

 

Also i have 2GB of ram not 1gb, 2x 1gb sticks.. HX G.Skill ram built for low timmings they are the Infineon CE-5 chips with the date 0514 before G.skill started putting the crap ce-6 in the HX brand which dont overclock past 230mhz, they overclock up to 250mhz with 2.5-3-2-5-1t settings, iv already ran memtest up to that speed and passed without problem on a 24hr pass, any higher then 255mhz and i get error's

 

 

 

 

 

[Peanut2001]

May not help but you could try to up chipset voltage to 1.7 if your chipset temps are good. The most I can get out of my opty is 2925 @ 1.5vcore. I doubt it would do much more with more volts because it stepped out beautifully with the voltage.

[THEnumber337]

It looks like you only have your vcore at 1.425. Try upping it to somewhere between 1.425v and 1.5v as long as your temps stay within a good range. It might be that the integrated memory controller on the cpu just needs a little be more juice to get those few extra mhz. It is not unheard of for Opterons to clock up to 2.8Ghz or 2.9Ghz on stock or almost stock volts but need a real push to get to that 3.0Ghz mark.