Can't even get 2200MHz stable on a 3200

[MoPMatrix]

I can not get my 3200 Winchester (CBBHD 0447UPEW) with the system in my sig to run any type of overclock stable, not even 2200MHz. At stock clocks its 100% stable. But any type of CPU OC and its BSOD's all over the place.

And the CPU does not seem to respond to voltage at all, doesn't make a difference if it’s at 1.36v or 1.6v, the chip will not run stable when it has any type of OC. Though at voltages over 1.55v it seems to make it even more unstable. Temps accroding to corecenter are 37-40c idle and 45-48c load, though I think they are way off.

I currently have the system running at:
275x7 (1925MHz) 3x HTT / 1.36vcore (Bios 1.5x) Bios 1.4 had the same performance
Ram @ 2.5-3-3-10 1t 1:1 @ 2.65v
High Performance mode = Manual
Aggressive timings = Disabled
NV/ATI speed up = Disabled
Dram Drive Strength = Normal
Dynamic OCing = Disabled
Spread Spectrum = Disabled
HT Frequency = 3x
Cool 'n' Quiet = Disabled
PCI Clock Auto Detect = Enabled
AGP Freq. = 66Hz (Clockgen confirms that it does stay at 66Hz)
AGP Voltage = 1.50v (Tried 1.5v to 1.6v and makes no difference for the CPU OC)

That other menu:
AGP APP Size = 128MB
AGP Speed = 4x/8x
AGP Fast Writes = Disabled
AGP Side Banding = Disabled (Though CPUz shows it enabled in the MB tab)

And so far it is 100% stable. At these settings it passed 10hours of prime95 a few times. Games and 3dmark are stable and it passes 10 hours of memtest86 with 0 errors and passes every SuperPI test fine.

But as soon as I bring the CPU Multi up to 8 making it 275x8 (2200MHz) it is unstable again. The first time this happened I figured ok it needs a little more vcore, even though I didn't think a 200MHz increase would need more vcore I tried it anyway. But raising the vcore from 1.36v all the way 1.5v testing it in every increment in between made no difference. It would still only pass a few hours of prime95, BSOD once in a while, and SuperPI would error out once in a while. (Though at 272x9 (2448MHz) I got it to pass 10 hours of memtest86 with 0 errors, but still not stable in windows)

I tried OCing with a higher multi/lower FSB before all of this also, It doesn't seem to make a difference
220x10, 240x10, 250x10, 275x8, 275x9, All of these settings are equally unstable. 2500MHz isn't anymore unstable then 2200MHz (vcore has no affect for stability on any of it). The only thing that is stable is 2000MHz.

Going by all of this testing I figure that I just didn't get a lucky chip?
275x7 is stable and 275x8 is not, and vcore has no affect, what else can it be? Is there any other settings that would have an affect on this?

I don't feel that it is my PSU because the system is still stable when I have my GT @ Ultra clocks and both of my CCFL's on. Also I'm only running 1 5400RPM pos HDD right now because the BSOD'ing killed my WD 120GB JB drive and I also took out my 250GB JB drive while OC testing so that wouldn't die also.

What else do u guys think it could be? If it is the CPU u'll see it up in the market place real soon

--------------------------------------------------------------
Screen shots of the stable settings....
http://www.matrix.gamingunited.com/OC/1.jpg CPUz
http://www.matrix.gamingunited.com/OC/2.jpg CPUz ram timings
http://www.matrix.gamingunited.com/OC/4.jpg Prime 10 hours
http://www.matrix.gamingunited.com/OC/6.jpg Corecenter

At 272x9 (2448MHz) I got it to pass 10 hours of memtest86 with 0 errors, but still not stable in windows.
http://www.matrix.gamingunited.com/OC/3.jpg Memtest86

 

[Mr. Anderson]

Maybe its the 64 chips don't like going that low of a multiplier, Just a guess because I ran into the same prob and thats what I thought.

 

[MoPMatrix]

Like I said above it wont OC using the stock multi of 10 either, 250x10 gives the same instability that 275x9 does

 

[el]

I think it is your mobo the cbbhd should overclock to about 2.6 with 1.55volts. Your stepping is good I would lower everything in the bios and try to see how high you can post/get into xp and then go from there. Is the Neo2 flashed to the newest bios?

 

[apu318]

RAM? Try using a devider or something.. or pumping a bit more voltage into your memory.. Give it a shot. Maybe put it to 220mhz HTT and toss the dimm voltage to 2.7-2.8?

 

[MoPMatrix]

I think it is your mobo the cbbhd should overclock to about 2.6 with 1.55volts. Your stepping is good I would lower everything in the bios and try to see how high you can post/get into xp and then go from there. Is the Neo2 flashed to the newest bios?

The highest I've got it into windows so far is 2500, I tried for 2600 but it wouldn’t even show the windows loading screen at 1.6Xv, just went into a post loop. I haven't tried anything between 2500 and 2600 yet though, probably because I was so upset it wouldn’t even handle a weak 2600 . When I bought this chip I was expecting ATLEAST 2600 maybe 2800 if I got lucky.

Bios is 1.5x

There’s no chance I could RMA the board cuz it doesn't OC good right? Since its stable at stock clocks. Or can u since it’s supposed to be an enthusiast board? Sell it and buy a new one?

RAM? Try using a devider or something.. or pumping a bit more voltage into your memory.. Give it a shot. Maybe put it to 220mhz HTT and toss the dimm voltage to 2.7-2.8?

If it’s stable at 275x7 and not 275x8 I do not see how it can be the ram if all I'm increasing is the CPU clock. The more voltage I give my ram the more unstable it becomes. At 2.65v is will run memtest forever, at 2.70-2.75 is will get a couple errors with in 5 hours, at 2.80-2.85 it will get a couple dozen errors in the first pass.

 

[rseven]

Sadly, after reading over this thread twice, it is looking like the limits of the cpu. The problem is I find it hard to believe that there is so little headroom in an AMD. That you got nothing more with extra vcore is also hard to believe. I have one idea that's probably a long shot, but easily done so, you may want to give it a try. Lower your RAS to CAS to 4. I read on the AMD forum that the A64 memory controllers had problems syncing up with low latency ram at high htt's causing BSOD's. Since you mentioned getting some perhaps you could give it a try. I have also read on this forum http://www.ocforums.com/showthread.php?t=365963&page=1&pp=30
that some late 04 steppings were poor overclockers, but it was CBBID's that were mentioned. Your HD is frighteningly close to the time those were made however that's total conjecture.
I just had another grasping at straws idea. What are your vcore readings like in MBM. Do they dip when the cpu is stressed or do they go up? The reason I ask is when I first built my system I remember having some difficulty getting the 20 pin power connector to make contact even though it clicked into place. If that were to happen with the four pin connector then the cpu wouldn't get the extra juice it needed under load and would fail? It's a bit far fetched, but soetimes we think it's something complicated and it turns out to be something simple.
Well, good luck with it. Report back with any findings.

 

[MoPMatrix]

I just tested the vcore under load, right now I have it running
275x9 - 2475MHz - Set to 1.425 3.3% which really volts to 1.44v
2.5-3-3-7 1t 2.65v
Passes memtest at these settings but its not windows stable

While idle the vcore reads 1.424v to 1.440v and under load with SuperPI and Prime it reads 1.440v to 1.456v
That sounds normal right?

I'll try cutting the ram timings down tonight to see if it has any affect

 

[rseven]

Yes, that sounds normal. It was just a stab at it. I run mine at 9X275, stock voltage and 5:6 divider and it purrs, but I have a 3000 CBBFD 0442.

 

[MoPMatrix]

Yeah my main concern is if it is the chip, I dont know where I'm going to find one to replace it with that has older stepping then 0450
I got this one from gameve (0447) while newegg was already handing out 0501's
I'm sure gameve has to be into 05 chips by now also

 

[rseven]

You might be best off to just wait till the Venice core comes out next quarter. They are supposed to have one as a replacement for the current Winchester 3000 so; it shouldn't be uber (No, I'm Italian not German!) expensive. Of course how it OC's is anyone’s guess. Something else you could consider is a small computer shop where you can see the steppings before you buy, but you'll pay mega bucks.

 

[MoPMatrix]

Ya ur right I probably could wait a while, its not like this thing isn't a beast at stock clocks anyway
Just it will probably be harder to get a good price for my winchester once the venice is out

 

[hitechjb1]

I just tested the vcore under load, right now I have it running
275x9 - 2475MHz - Set to 1.425 3.3% which really volts to 1.44v
2.5-3-3-7 1t 2.65v
Passes memtest at these settings but its not windows stable

While idle the vcore reads 1.424v to 1.440v and under load with SuperPI and Prime it reads 1.440v to 1.456v
That sounds normal right?

I'll try cutting the ram timings down tonight to see if it has any affect

At least the G. Skill 4400 is running at least at specification of 275 MHz 2.5-3-3-7 1T.

CPU is running at 2475 MHz stable.

In case you have not tried the following:
With these two stable operating points in mind, I would try go back to CPU multiplier x10 and starting with

HTT = 247, 248, 249, 250, ... MHz
memory_HTT_ratio = 1:1
2.5-3-3-7 1T
starting with whatever Vcore that got CPU stable at 2475 MHz, and increase one small step if needed, say 1.5 V.

1.425 V + 3.3% = 1.425 * 1.033 = 1.47 V

This should let you test the CPU limit up to 2.7 GHz as memory is not limiting at this range under 275 MHz.

If this CPU test gets you no where, another long shot, is to see whether you can borrow another motherboard to try. But I think the above CPU test should be focused first until the CPU limit can be established first, be it Yes or No.


BTW, I have a Winchester 3000+ CBBHD 0447 UPCW which is same week as yours, and it does 2.7+ GHz 1.52 V Prime95 stable.
SuperPI 32 M, 3DMark01 stable at 2.85 GHz, 1.55 V.
It is running on a DFI LP UT NF4 Ultra-D board.
For details:
Overclocking sandbox for A64 939 systems with Winchester and beyond

 

[MoPMatrix]

I haven't been able to get the CPU stable at 2475MHz in windows yet, Only Memtest.
It’s not super unstable at this clock while running 1.44v though, SuperPI will pass a few tests, fail a few. I can game for a while. But it will eventually BSOD.
It is more stable at 1.44v then it is at 1.36-1.42, so that’s a sign that it is responding to voltage but past 1.44v it doesn't seem to get any better.
I get this every time I work my way through the voltages "1.44 is more stable then 1.42!, maybe 1.46 or 1.47 will make it stable!" But then I find my self working all the way past 1.55v with no improvement in stability.

And last night I made an attempt to find the CPUs max,, The highest I can get it to even attempt to load windows is 255x10 2550MHz. That took 1.61v and was super unstable. Only lasted a couple min, SuperPI failed the 16k then BSOD.

I got it to post and load memtest at 265x10. Though at 265x10 and 260x10 it doesn't even attempt to load windows, it just post loops even with 1.61v.

It wouldn't post at 270x10, even with the voltage maxed at 1.61v

Also just to make it clear I just didn't set it to 1.61v as a first try to make it post at these settings, I worked up to it, testing everything inbetween just to make sure it wasn't the voltage it didnt like and not the clock.

btw thanks for the help everyone

EDIT:
Btw the BSOD I see the most is "Page Fault in Non-Paged area"

 

[glock19owner]

Btw the BSOD I see the most is "Page Fault in Non-Paged area"

This is a memory error...and with the memory passing MT and the CPU still being unstable...then the odds just went right back to the CPU as being the bottleneck...not the motherboard or memory...

 

[jrb357]

Well I dunno but it seems that some of these 0451's are kinda weak. I have one too in a Neo4 I've been beating on for 3weeks w/ some Corsair pc4400. The bloody thing won't post over a 215fsb, but if i use Clockgen the best it will do stable is 260x9. It just plain doesn't like anything over a 2.4GHz. So I'm kinda bummed with it too.

 

[MoPMatrix]

well 2.4 is better then the nothing I have

 

[strappleberry]

my 3200 blows, im at 278*9 = 2.5 at 1.76v :S i got the ****ty chip of the draw...i was only prime95 stable at 2.45 at 1.76v but running BurnIn (for ever) was a savour

 

[MoPMatrix]

strapple do u know what week ur chip is?

 

[el]

well atleast you still got a 20%+ overclock. I know hitting 2.7 is my goal and I am said to see the newest chips not hitting the highest speeds.