Overclocking frustration, I'm pulling my hair out

[CrashX]

Specs in title.
First let me say, I know the Corsair isn't the favorite, and I wish I would have gone with something different. But it's specs are high, and I've been happy with Corsair in the past. I will return it if need be.

Now, I've tried so many combinations in the past 24 hours, I feel like I've gone in circles blaming the ram, motherboard, cpu, and cooler!

Here is the quick and dirty version, with more info to come as I can think of it.

System will NOT pass Orthos blend when OCed to 3.2GHz (400MHz FSB). It EITHER fails at around 30 sec or around 5 min reliably +/- 30 seconds.

I thought it had to be the RAM, I even ran it at 6-6-6-20 and that wouldn't do it. But I just did a 10 minute Orthos test with the system NOT OCed, but the ram running at 400MHz 4-4-4-12 no prob. So it seems the ram CAN handle it's rated speeds just fine.

But as soon as I set the FSB to 400MHz, it doesn't matter if my times are way loose, the CPU set to 1.4v, Ram set to 2.1v, I can't pass Orthos blend.

TAT temps are around 50-55c (depending on the voltage I am trying) under load, so I think that is good. Motherboard is around 40-43c.

So is this the problem I have read about where the motherboard just won't go above 350-380? I will keep reading, but I have been reading for HOURS now. Looking to see if anyone has had a similar problem and/or solution.

 

[n7]

401+

In all your reading, i'm not sure how you didn't see that.
Do not mess with this mobo below 401.

You need to start isolating the potential issues.

Is RAM set at rated vdimm? If so, i'd bump it up one step.

For timings, please try 5-5-5-15-6-42-10-10-10-10
I don't care what it's rated for, try those.
Use the 1:1 divider, so your RAM should be @ 802

Now while i use lower values for some of these myself, try these mobo voltage settings:

vFSB = 1.4
vNB = 1.45
vSB = 1.6
vICH = 1.2

Now set CPU to 7x401.
NOT 8x401

7x401 with 1.45V vcore in bios.

That will result in 1.42V idling; 1.4V load

Do not worry about temps unless they go over 70C.

Now run Orthos small FFTs for an hour @ 7x401.

Then if that passes, try Orthos Blend for a hour.

Do not attempt to use the 8x multi for now.

 

[Duvie]

Try 8x401!!!

Blend test in orthos is not ISOLATION!!! It test some ram and some cpu....Run small FFT to isolate cpu...

Also ISOLATION means dont jack everything up at one time...You should if you read an ocing guide worth a shite have gone with looser timeings right out of the gate....4-4-4-12 is not loose timings...stock doe not mean looser times....

your frustration is from your lack of research and preparation before you started this...

 

[CrashX]

I agree with what both of you are saying. I shouldn't be jumping straight to Blend.
Although I did jump straight into trying a 400MHz FSB, I have gone all the way back to the drawing board and done a little at a time. I have run all my OC tests with the ram at 5-5-5-15 or even 5-6-6-18. I was just stating that the ram is rated for 4-4-4-12 DDR2-800, and it tests out at that just fine if I don't overclock the FSB. But when I OC the FSB and run my ram at 5-5-5-15 Blend fails.

I just started seeing information on running at 401 as I was posting my original post. I am currently testing 7x401. Small FFTs pass at this (40 min, I know, I should go longer, but it's a start). Large FFTs seem to pass as well (haven't run it nearly as long yet, 20 min, but will post back). BUT, Blend still fails within 5 minutes. Should I be using memtest in place of Large FFTs?

This is with all of n7s suggested voltages.

BTW, the "overclocking guide" I read was a post by, Yoxxy called "The Definitive, Unbiased Core 2 Duo Overclocking Guide." He basically stated exactly what would need to be set to get an easy 3.2GHz out of a e6400. Suffice to say, this has not been the case for me. So I have gone back to the beginning and started testing everything out one at a time.

 

[Duvie]

Small test is just cpu and cache

Large FFT is cpu and cache with a little bit of memory...

Blend is more baklnaced between cpu and memory subsystem in general...

It is clear to me there is an issue with your ram


What vdimm are you running? Take the stock rating and dont be afraid to try up to 2.3v if the board has it...anything after 2.3v should be actively cooled as well..

Have you notched up any of the volatges for the chipset? May want to at least bump it .1v to .2v.,...maty also want to actively cool the NB chipset if it isn't already...

 

[CrashX]

It is definitely pointing to Ram.
Memtest errors out at 1831MB at 7x401 and 8x401 (I know same speed for the memory, just wanted to try it) within 15 seconds 3 out of the 4 times I tried it.
That would explain why sometimes an orthos test could pass and sometimes fail. It's not a very consistent failure. So I guess that really could be voltage then, just fluctuating right?
I am currently running 2.15v. Corsairs sight says 2.1v. So I will try your suggestion of 2.3v then.

Other voltages are:
CPU 1.45
FSB 1.4
NB 1.45
SB 1.6
ICH 1.2

I am going to try the ram right now though and see if that's what it was all this time. I have been afraid to go higher since "stock" is technically 1.8v.
"Fingers crossed"
BTW, thanks a LOT for the help, I will be back in 20

 

[CrashX]

I don't get it.
Ram at 2.3v
Memtest errors after just 10 seconds. 7x401, 8x401, 7x433 (just wanted to get way out past the 400 area to see if that makes a difference). Ram at 5-5-5-15.

BUT, if I don't OC the FSB, and just set the DDR speed manually, it is fine. Currently running at DDR2-889 5-5-5-18, stock voltage, cpu at stock 266FSB, 2.13GHz. Does not error in Memtest and just passed the 10 minute mark in Orthos blend. (remember Orthos usually fails for me at the 5 minute mark)

Now, maybe my testing methodology is way off base here, so somebody tell me if my logic is SCREWED up...
If the ram can pass memtest at DDR2-889 on a non-OC CPU, but fails when the FSB is OCed you would think it's the CPU.
If the CPU can pass 40 min of Orthos small fft when OCed to 400FSB but Blend fails and memtest fails, you would think it's the Ram.

I feel like I am missing something really big and obvious here! And if I am, CALL ME ON IT. I'd rather find out that I am stupid than to keep going on like this.
But what is different about combining the two actions? The ram is still running at less than speced speeds when the FSB is OCed, then why does it fail memtest when it can pass memtest without the FSB OCed but still at spec speeds?

 

[cmdrdredd]

Make sure you got the latest BIOS...it helped a few people I know get rid of pesky problems

 

[CrashX]

I upgraded to 804 as soon as I got it.
But then I noticed a lot of people were using 711, so I downgraded to that early in my testing so my setup could be as close to others as possible.
I will try going back up to 804 though and retesting.

I see you have a very similar setup to me as far as motherboard, CPU and ram. Did you have any trouble getting it running? Any tips?

 

[The-Noid]

804 Seems to work great for me. It doesn't seem like it was built off of 711 though. 711 fixed CPU uCODE not found for ES Kentsfields, but the 804 still has the same problem. Maybe Bios made by different people. It still seems to work well though.

 

[CrashX]

Update.
Running 6x401 is so far perfectly fine.
I am running high volts, and will back them down as I test, but this proves that the ram runs fine at 401 and at SPD settings (4-4-4-12, 2.1V) to boot! So the Ram CAN'T be the problem.
So what is it with the higher multiplier that causes memtest to fail?

One important thing I noticed. At stock speeds and stock voltages CPU-Z shows 1.38v. But when I've been setting my voltage to 1.375 for examply, it's actually been running at 1.34v as per CPU-Z, so that appear to SERIOUSLY undervolting it. So I am manually setting it to 1.45 so I can get it to run at 1.40v.

So with that knows, I will try bumping up one multiplier at a time.

 

[CrashX]

Second update.
Now 7x401 is running fine.
Not sure what I did differently besides going to 1.45v. That seems higher than most people have needed for something so mild.
I will try 8x401 next.

I read on a forum one guy stating he couldn't get his stable the first couple of days. Then the more he played with it the more stable it got. Like a burn in. Maybe that's what I am seeing?

 

[cmdrdredd]

Originally posted by: CrashX
Second update.
Now 7x401 is running fine.
Not sure what I did differently besides going to 1.45v. That seems higher than most people have needed for something so mild.
I will try 8x401 next.

I read on a forum one guy stating he couldn't get his stable the first couple of days. Then the more he played with it the more stable it got. Like a burn in. Maybe that's what I am seeing?

I dunno what to say I just put 1.4v in the Bios for CPU, bumped every voltage up one notch from the lowest setting available, and put memory at 2.1v and 5-5-5-15 and FSB at 400, DDR2-1000, 8x multi, spread spectrum off C1E off, vanderpool tech on, and that's it.

 

[CrashX]

Well so far so good 8x401. Over 30min of Orthos.
I've even slowly been dropping my volts back to what I've been seeing others use. So I don't know what to think. Was it just a burn in issue like I posted above.

One NEW question (if anyone is still reading this). I've seen several posts, but no definitive answers. Sometimes I notice CPU-Z shows my PCI-e at x1 instead of x16 (third tab at the bottom). I can't find any setting that seems to be causing this. In fact, last reboot it was x16, then I dropped some volts and it's x1. 3dMark runs at a quarter of the speed with it at x1, so I know it's not just CPU-Z detecting it wrong.

Is this a common problem? I searched the term x1 and only came up with a few posts.

 

[imported_Buckster]

hi there - my Corsair RAM has exactly the same issues, its 6400 C5 stuff, but it won't run at DDR800 if the FSB is pverclocked - no matter what volts I throw at it.

Also it HATEs with a passion 8x401 or above .... no matter what people say here about it being more stable .... and a friend of mine with same RAM also has the same problem.

PC will just about boot with 0711 at 8x401 but is very unstable, even windows crashes, Orthos not stable for more than 1 minutes. Sometimes if I set 8x401 I can't even post.

I put it down to the RAM - it really doesn't like the high FSB - as I can run my CPU at higher clocks at lower voltages at 9x multiplier.

I've stuck to 9x356 at the moment, which is plenty of performance, and rock solid stable

Mark.

 

[The-Noid]

Try 403 sometimes there is a hole at 399-402.

 

[Regalk]

The issue of 1X vs 16X has been addressed a few times. I encountered the same issue and at one time was running my PCI-e at 110. I increased my SB Volts to 1.6v, changed the SB to - PCIEx16 2/PCIEx 1Force to X/4 disabled mode fast and changed back the PCIex to to 100 and the problem seems to be gone now.


SB to - PCIEx16 2/PCIEx 1Force to X/4 disabled mode fast - NOW AT AUTO

 

[CrashX]

I have found increasing the SB and NB by one notch seems to fix the x1. I could probably back one of the two down, but for now, it's easier to leave them both up.

I am REALLY close to being stable. Last night it failed Orthos blend after 2 and half hours. So I bumped the core up to 1.4v (1.376v in Windows). And am at almost 2 hours.
I am running FSB of 401, and PCI-e 101.
But the 403 suggestions seems to be a good one.
Buckster, try some of the suggestions here like the 403 FSB, you may be able to get it after all. In my case, I have a max multiplier of 8, so I need to get to 400 FSB if I want to get anywhere near what the CPU is capable.

Thanks ALL for the suggestions.

 

[n7]

Good to hear you've got it somewhat figured out.

I really don't know why the RAM acts up when 1:1 OCed, but it seems somehow it's stressed more, & it seems Corsair's cheapo chips don't like that.

I've seen near duplicates of your situation posted on forums all over, so you're not alone.

 

[CrashX]

Originally posted by: n7
I've seen near duplicates of your situation posted on forums all over, so you're not alone.

That's good to hear at least. I was going crazy thinking that it seemed like such a no brainer for everyone else, but I was having such goofy problems.

 

[de8212]

Just skimmed the thread as I am thinking of getting the p5b deluxe.
Is this your RAM?

I will re read this thread and do a search but is there a better RAM to use woth this board? It is some pricey RAM and Newegg has it bundled with the mobo for $486. Just don't want to spend that much $ and have issues.

 

[cmdrdredd]

Originally posted by: de8212
Just skimmed the thread as I am thinking of getting the p5b deluxe.
Is this your RAM?

I will re read this thread and do a search but is there a better RAM to use woth this board? It is some pricey RAM and Newegg has it bundled with the mobo for $486. Just don't want to spend that much $ and have issues.

I would get crucial 10th Anniversary DDR2-667 and put the voltage at 2.2 and it will run DDR2-800 at 4-4-4-12 and at 2.3 can do 5-5-5-12 at DDR2-1000

 

[de8212]

thanks cmdrdredd
Is this the RAM?

SO it will definately run 800?

Anyone else here run that?

 

[CrashX]

I was THIS close to getting the Crucial 10th Anniv. Ram. But was cautious about it being rated at DDR2-667.

And yes, that Corsair link is my Ram. I am currently running it at DDR2-800, 4-4-4-12, 2.0v.

I just finished an 8 hour run of Orthos blend, so I think I am finally happy.

 

[cmdrdredd]

Originally posted by: CrashX
I was THIS close to getting the Crucial 10th Anniv. Ram. But was cautious about it being rated at DDR2-667.

And yes, that Corsair link is my Ram. I am currently running it at DDR2-800, 4-4-4-12, 2.0v.

I just finished an 8 hour run of Orthos blend, so I think I am finally happy.

All memory ICs are rated at 667Mhz and actually overclocked and reprogrammed to run 800/1000/1100Mhz

Basically there is no standard rating for DDR2-800 and higher. And even the memory in the OCZ Titanium Alpha VX2 which can run DDR2-1200 is actually the same ICs used in the crucial 10th Anniversary memory. The difference is in the binning. OCZ takes the best of the best ICs ans only those that can accept that type of speed and voltages.

Anyhow yes the crucial will do DDR2-800 without issues. Not at 3-3-3-12 but 4-4-4-12 is reasonable to expect if you use 2.2v

I don't have a link handy but there was someone (I think on xtremesystems) who ran this Crucial memory at 2.4v and got over DDR2-1100