Help me overclock my opty 170

Nazbit · 2 Mar 2006 at 21:31

Hi all,

I wonder if some of you could help me overclock my new cpu. I was posting in the official opty 170 overclocking thread but my noob questions were getting in the way of the big guns

Ok here, it goes:

I have tried 210 x 10 = 2100, this works fine
Then I tried 215 x 10 = 2150, this also works fine

Then there is a brick wall at 216 x 10

I have tried everything that I know in order to get past it (but I dont know all that much

)

1)Increasing the voltage to 1.375 (the next option in the bios) results in the machine failing to boot, when it gets to the xp loading screen it just restarts, this continues on a loop.

2)Lowering the LDT multiplier from 5x to 4x causes the machine to post and pass the windows loading screen normally, but then instead of the logon screen its just a black screen with the windows cursor on it.

3) Lowering the multiplier and raising the FSB 240 x 9 the system posts and then when the windows xp boot screen is about to appear, it restarts, this continues on a loop until stopped.

The stepping is:
OSA170DAA6CD
CCBWE 0551UPMW
1392568L50&&3

Memory is at ddr400 - This is equivalent to 1:1 right? AGP is fixed at 66mhz.

My specs are in my sig, im running bios version 26 and have downloaded and installed the amd dual core driver from the amd site. let me know if theres any other info I can give you.

Can anyone suggest what is causing my severe lack of overclocklability or suggest how to improve it?

Cheers

Nazbit · 2 Mar 2006 at 21:50

Hi, thanks for the reply,

The bios is the latest version, I updated to this version when installing the opty as its the only version of the bios that supports the amd dual core range (and even then theres no mention of optys just x2's) There were problems detecting it properly before the bios update, I didnt try any overclocking, just sorted out the detection issue first.

On older bios versions my 3500 newcastle reached 2.5 once, but It wasnt stable, I didnt really do much overclocking then, so I dont have any cast iron figures, but I bought the 170 thinking I could overclock it to the hills.

I suppose theres one thing I havnt tried, thats overclocking the opteron whilst on older bios (when detected as an AMD hamer family cpu - model unknown) although this bios clearly offers even less support than my current version.

Apart from that is the solution to buy a new board, new ram or what? I cant afford a new mobo right now because I am still on agp graphics, and I would have to fork out for both new mobo and a new gfx card

Cheers

Nazbit · 2 Mar 2006 at 22:27

Hi IzaLearnin, Im just testing out that idea on the desktop now, running the memory at ddr333. Will report the results back as soon as Im done.

To ajgoodfellow again, I do have another 939 board in the family machine in the other room, The problem is though I built it for everyone to use for office applications, so it has the power of a mars bar, meaning I would have to take my gaming rig, and that machine, de-assemble the major part of them both and make a new, 3rd machine from them, It will take a while and isnt something I particularly want to do, but if IzaLearnin's suggestion doesnt work then I will have to do exactly this.

Nazbit · 2 Mar 2006 at 22:43

Ok here goes, Dropping the DRam clock to DDR 333 has enabled me to sucessfully boot at a higher fsb (htt)

Everything seems fine at 235 x 10 = 2351Mhz with the ldt multi at 5 x although I have not put the machine under any great stress or tested at great length. Temps are 46 - 48C Idle

I went up to 240 x 10 = 2400Mhz with the ldt at 5 x but it blue screened after a couple of minutes of use. Seemed to be a memory error as it showed a couple of numeric memory addresses but wasnt on long enough to read.

What do you think I should do next? I was thinking trying the 240 x 10 with the ldt dropped to 4x.

Nazbit · 2 Mar 2006 at 23:31

Ok my thanks to you both, you have been very helpful.

Its now stable at 2401Mhz
Thats with 10 x 240 and the LDT dropped to 4x FSB, so far so good, I ran 3dmark03 (seeing as it had previously crashed at this speed with 5xfsb) just to make sure and it all ran fine, put 400 points on my previous best too

I am off tomorrow so I will come back and try to push it further in the morning some time, the higher it goes though, the more stability testing its going to need. Ideally I want to get it over 2.6Ghz, but we will see what happens. (I have promised myself watercooling if I can get it that high

Cheers

Nazbit · 3 Mar 2006 at 15:04

Ok, we're off again

Running at 2420mhz.........242 x 10 with the ldt multiplier set at 4 x, the ram is still at ddr333 and the voltages are all stock. This works fine.

Raising to 2430mhz.........243 x 10 with all the other settings kept the same, is unstable.

Threfore I raised the voltage up by 0.025v. And it was still unstable.

What do I do now? Do I raise the voltage again, or what.

Cheers

Nazbit · 3 Mar 2006 at 21:28

Ok, CPU-Z screenies:

Here

and here

Nazbit · 3 Mar 2006 at 22:27

Hi again mate.

The ram is at 2.65v and it is the BH Set that I am using.

Edit:

Hi ajgoodfellow

I have been playing with your suggestion for the last 15 - 20 mins, and all seems good, thats with the cpu core now at 1.4v

Edit again - This isnt stable afterall so its off to ram voltages for a solution I think.

Nazbit · 3 Mar 2006 at 23:12

The bios is only offering options up to and including 2.8v for the ram. The DRam clock can be set to ddr500, but I am guessing that without the ability to provide it with the 3.4v -3.5v that you mentioned it will not run at this?

IzaLearnin said:
If not I would leave the ddr333 divider in place and try 2.8v-2.9v for the ram at 2-2-2-5 timings and continue to raise the htt 5mhz at a time and test with a 1mb superpi run each time until you encounter instability again. I am sure 2.6ghz is acheivable with your 170 as long as you can keep it cool.

I'll get going on that then.

Cheers

Nazbit · 4 Mar 2006 at 00:10

Ok cheers.

Testing the stability now with the ram clocked at 2.8v. @ 243 x 10

Hopefully it will prove stable but its getting pretty hot in there, so I might leave it until the morning now when I can rig up a temporary fan near the ram and fiddle with the airflow a bit.

As soon as I can sort out the sale of my 3500, I will be buying some watercooling, I got my eye on the Swiftech H20-220 Apex "Ultra" Watercooling Kit.

You say that you think 2.6 is possible as long as its kept cool, do you think thats possible with this kit?

Nazbit · 4 Mar 2006 at 00:29

I have an xp-120 with a 120mm akasa fan.

At the moment the cpu is getting pretty hot, idle its in the 40's and under load its anything from 50 - 75, occasionally it can report in the 80's.

My motherboard (abit av8 3rd eye) has it self a huge stonking great reputation for overkilling temps by as much as 20C higher than they actually are, but I want to play it safe.

The ram is certainally warmer since the voltage increase to 2.8 but I dont think its anything major.

Im hoping the water kit will help me reach these higher voltages while keeping the temps down and get approx 2.6 out of the cpu

Nazbit · 4 Mar 2006 at 11:28

Another problem:

I am now at 2450, 2 x 245, ldt 4 x, ddr333, 1.4v core, 2.8v ram.
It seems ok here

But If I go any higher, (2 x 246) the cpu tests run fine, but 3dmark03 and 3dmark05 freeze and lock up part way through, mainly on the 2nd cpu test, but sometimes on other tests aswell.

Any ideas?

---------------
Edit: Based on the fact that it failed while in the CPU tests, I am assuming it is a problem being caused by the overclock, and the cpu being unstable (I may be wrong....you tell me) so I have tried upping the core voltage again, and the problem still continued, I also tried lowering the ldt to 3x fsb, same result. I dont really see how else I can make it more stable.

Is this cpu related, do I need more volts?

Nazbit · 4 Mar 2006 at 19:12

I moved the ram divider down to 266 and the same problem occured.

This means that another increase on the cpu will take it to 1.45v is it me or does this seem quite a lot, just to get 2470mhz?

When you look at the voltages some people on these forums are using to get 2.8ghz, putting 1.45v through the cpu to get 2.47 seems a little over the top.

How much further should I go before stopping?

Nazbit · 4 Mar 2006 at 19:25

Ok it failed again at 1.475v. Im now at 1.5v and the temps are pretty high. I will try and put it under some load but I think it might just get too hot and I might have to abandon, but I am going to give it a go.

-----------

Edit: It failed at 1.5v aswell. Did exactly the same thing. These last few lockups have been during the game tests and not the cpu tests as before.

I am going to have a quick go at dropping the ram divider even lower than I did before, just to make absolutley sure its not that, but I dont think it is.

This might be the highest I can get I suppose?

Would love to get to 2.5, from the 2.47 im at now though, if I could do that I would be happy, then I could buy that nice watercooling setup we mentioned

Nazbit · 4 Mar 2006 at 22:09

Yer its a shame its maxing out this low, but as you say I still have the equivalent an x2 4800 which isnt bad for the money

I would like to just stretch it that extra bit higher if I Can though.

IzaLearnin said:
I wonder if it is a heat issue, what load temps are you getting with two instances of prime etc with 1.425v?

I will do this test now and let you know. I will report back. I have thought long and hard about it being a heat issue and I am undecided. Earlier when on 1.5v I was about to shut down due to extreme temps when just as I moved my arm to go for the button, 3dmark03 "got stuck" the way it has been doing, yet other times it has locked up at very reasonable temps.

lay-z-boy said:
There is the culprit, 0551’s are mediocre, I have the same stepping on my x2 4400, the only way to make it clock high is to shove volts though it.

If I can get 2.8ghz with 1.6 vcore then you should be looking around the same, I need a 0.2v boost to the core at 1.4v to get it dual sp2004 stable from 2.5+

Its good to know that there is hope for the chip if I can really shove the volts through it. In about 10 days I should be able to afford the swiftech 220 watercooling kit so I think I may wait until I get it, get the temps nice and cool and then shove all the voltage I can through it to try and get it above the 2.5 mark. Hopefully if I can keep it cool enough then I can run enough volts through it to get it up there.

Cheers

Nazbit · 4 Mar 2006 at 23:09

ooooook..

The abit av8 has a reputation for having the poorest temp sensors ever marketed on a mobo, so take them with a pinch of salt. When I first bought the mobo the temps were way out, since then bios updates have been needed to recalibrate he sensors on about 8 different occasions, they just couldnt get it right, anyway, running the latest version is better, but I still dont trust them too much....

Temps at 1.425v

Idle = 25 - 29 C

50% load for 60 seconds = 60C

90% + load for 60 seconds = 75C

Nazbit · 5 Mar 2006 at 00:14

My feelings exactly.

I think I am going to try and get the water cooling setup asap, (probs about 10 - 14 days) then once thats installed I can maybe try a few more volts through the cpu hopefully something like lay-z-boy talked about. I dont think I will get as high as him, but I think that my cpu has more to give yet.

Dont really want to push it any further until I can properly keep the temps under control.