X2 3800 instability in Prime95

[geekified]

I should start this out by saying I have only limited experience with overclocking and that most of that is further limited to reading forums and guides. My luck with this X2 3800+ has not been great at all. I can run applications fine and I'm not getting any crashing -- except in Prime95's torture test. Prime95 will run for many hours when I limit it to one instance. However, when I run two instances of it (one on each core, set from Task Manager in XP Pro), I get an error message that can be seen farther down the page. The slogan I read on one board said "if it isn't prime stable then it isn't stable". My first question unto you all is: is this true? After running for usually less than a minute on the one core, the Prime95 error is the following:

FATAL ERROR: Rounding was 0.5, expected less than 0.4.
Hardware failure detected, consult stress.txt file.
Torture Test ran 1 minutes - 1 errors, 0 warnings.
Execution Halted.

In most cases the CPU temperature according to the ASUS Probe tool is far under 50C. I'm thinking it might be a power draw issue or possible something with memory divider, but I'm asking here in case you guys and your experience can guide me in any manner.

In case it helps, I have my BIOS settings and hardware specs listed below. I have currently changed it back to default BIOS settings to avoid possible hardware damage from my incompetence on overclocking.

Should I be worried about Prime95 crashing like this? If so, what can I do to correct above error?

All help and/or advice and/or flames are greatly appreciated

Hardware specs:

X2 3800+ Manchester core BH-E4 revision
ASUS A8N-E mobo
4x512MB PC3200 DDR400 (Nanya+generic RAM from ebay that has no discernable brand)
Antec 500W Neo HE
WinXP Pro SP2 (patched with AMD driver and Windows dualcore fix)





BIOS settings I was using:

Hyper Transport Frequency = 4x
Cool 'n Quiet = disabled

DRAM:

Memclock frequency = 200 Mhz (lowest frequency supported, uncertain if this is also a 'memory divider'
CAS latency= 2
Min RAS = 5T
RAS to CAS = 3T
Row precharge = 3T

FSB/HTT = 240
DDR voltage = Auto
CPU Multiplier = x10
CPU Voltage = 1.375
PCI clock synchronization = 33Mhz

 

[FelixDeCat]

Have you consulted the x2 overclocking thread in this catagory for help?

As an longtime Intel overclocker who has dabbled with AMD a few times, Ive learned one expensive lesson : Dont overclock something that wont take stress reliably for a prolonged period of time.

Of all the chips Ive OC'd only the AMD chips crapped out and died after awhile. Of course youll have vociferous AMD zealots who always shout the Troll word to anyone who doesnt prefer that brand. But that will never prevent me from letting people know what might happen if they OC AMD, based on my experience.

Even though my 3.2c is capable of 3.7 ghz long term, I run it stock to prolong its anticipated lifespan, which Ive read can reach 20 years!

Goodluck, but play it safe.

 

[Viditor]

Originally posted by: FelixDeKat
Have you consulted the x2 overclocking thread in this catagory for help?

As an longtime Intel overclocker who has dabbled with AMD a few times, Ive learned one expensive lesson : Dont overclock something that wont take stress reliably for a prolonged period of time.

Of all the chips Ive OC'd only the AMD chips crapped out and died after awhile. Of course youll have vociferous AMD zealots who always shout the Troll word to anyone who doesnt prefer that brand. But that will never prevent me from letting people know what might happen if they OC AMD, based on my experience.

Even though my 3.2c is capable of 3.7 ghz long term, I run it stock to prolong its anticipated lifespan, which Ive read can reach 20 years!

Goodluck, but play it safe.

Which is why you find so many "AMD zealots on this board...who wants to risk an Intel Prescott/Smithfield OC for a sustained period (except in winter)?
I am truly sorry you never figured out how to OC an AMD chip Felix, but if you still have your old 3.2c then I can see you haven't tried for awhile...
You should really try again and come back from the Dark Side!

 

[stevty2889]

Probably need to up the Vcore a bit. Also will want to run memtest. If you can't change your mem from 200mhz, then it's overclocked a good bit, and most likely the cause of instablity. I have my 4200+ at 2.618ghz, needs 1.475 vcore for that. Running at 51c with the stock heatsink. If you get an error in prime95, then you aren't 100% stable.

As for Felix, just ignore him. If he's refering to Athlon XP's or older, they didn't have thermal protection to shut them down if too hot, or he could have overvolted them, so if he burned them up it's his own fault. The newer AMD chips have thermal protection as well, to prevent that problem. And I am hardly an AMD zealot, I have an X2, because it's the better chip. I also have a Pentium-D, which is a slow excessivly hot waste of money. I have 2 p4's as well, and a pentium-m.

 

[PingSpike]

From the looks of your bios settings, your ram is probably unstable. Are you sure that generic ram can run timings that tight? If you're using the 200 setting in the bios then your ram is also overclocked signifigantly. If there's no other setting for ram speed...you might be out of luck. You could try lowering the timings, and you should definately increase the dimm voltage but you're pushing cheap stuff well beyond spec and they probably just plain don't have it in them to go that far.

The ram would be my first and best guess, based on your specs and settings. You could try increasing the vcore some, but I doubt thats what's really causing the problem.

To take the cpu out of the equation, you can drop the processors multiplier to its lowest setting and try the stress test again.

And yeah, just ignore felix.

 

[Jojo1971]

looks like it's a ram issue...

 

[professor1942]

Originally posted by: FelixDeKat

Even though my 3.2c is capable of 3.7 ghz long term, I run it stock to prolong its anticipated lifespan, which Ive read can reach 20 years!

...and I'm sure you will have many great uses for that cpu in 20 years. I wouldn't even want it now

 

[Allio]

Why are you running four sticks of generic memory at 2-3-3-5? Try 3-3-3-8 and seeing if your problems go away.

Edit: assuming that 200mhz setting is a 1:1 divider and your HTT is at 240mhz, are you telling us you're running four sticks of generic memory at 240mhz at 2-3-3-5? If it's actually 100mhz x 2 and you're running your memory at 120mhz, that's pretty stupid, but it could probably handle those timings. Can't you set a 166 divider to get your ram to run at 200mhz?

Edit: Help us out and tell us whether it's a CPU problem or a memory problem by dropping the multiplier and keeping your HTT as it is to give you the CPU's stock speed, or near to it, then running Prime again.

You mentioned that you're not getting any kind of instability other than in Prime, but the NeoHE 500w is pretty incompatible with Asus A8N boards and a lot of (most?) people with that combo get spontaneous shutdowns.

 

[Wentelteefje]

AMD is the light side... The good side...

 

[Markfw]

Originally posted by: FelixDeKat
Have you consulted the x2 overclocking thread in this catagory for help?

As an longtime Intel overclocker who has dabbled with AMD a few times, Ive learned one expensive lesson : Dont overclock something that wont take stress reliably for a prolonged period of time.

Of all the chips Ive OC'd only the AMD chips crapped out and died after awhile. Of course youll have vociferous AMD zealots who always shout the Troll word to anyone who doesnt prefer that brand. But that will never prevent me from letting people know what might happen if they OC AMD, based on my experience.

Even though my 3.2c is capable of 3.7 ghz long term, I run it stock to prolong its anticipated lifespan, which Ive read can reach 20 years!

Goodluck, but play it safe.

That is the biggest bunch of crap I have ever read you troll felix !

The the OP: Ignore felix, he craps on threads all the time. I have both Intel and AMD chips OC'ed and the AMDs are more stable, and perform better, cooler, and are more stable. You just need to raise your vcore to 1.45, and then you will be fine. You can try 1.425 and see if its stable, then OK, but you may need 1.45.

 

[Wentelteefje]

Originally posted by: FelixDeKat
Of all the chips Ive OC'd only the AMD chips crapped out and died after awhile.

Couldn't agree less... I bought myself one of those fabulous 2.4GHz P4s and OCed it to 3.2GHz on moderate voltages... At the same time I bought an AMD AXP 2000+ and overclocked that one too, to 2.1GHz (it needed quite a lot of voltage to stay stable...)

Guess what? My AMD is still running, while the P4 crapped out whilst gaming... I'll never take Pentium again (except for their mobile parts)...

X2 3800s aren't the best overclockers around (so are the Asus A8N-Es...), but you should be able to get at least a respectable 2.4GHz out of it... My guess would be that your Cas Latency is pretty low for generic RAM...

Setting it to 2.5 will prolly solve it... Also don't hesitate to up your voltage a bit... As long as you stay in acceptable ranges, you should be perfectly fine...

 

[geekified]

To clarify a point, the memory divider/memory frequency settings in BIOS also goes to 266 Mhz and 333Mhz and well beyond -- although I'm not sure why. It can go no lower than the 200 Mhz. How is the memory divider calculated? I'm sort of a neophyte to all of this OC stuff so please do inform me of its meaning.

I changed memory in BIOS to 3-8-3-3. Voltage is now at 1.4125 and (for 5 minutes at least) Prime95 appears to be working. I also bought a heatsink that wasn't stock and it is saving me roughly 3-8 degrees celsius. Still running at all of the same BIOS settings above minus the changed memory settings and the upped voltage (240 FSB, 4x HT, 10x multiplier).

New questions:

How long should I let Prime95 do its thing before I deem it stable? What are the chances of turning my poor generic RAM into pwnsauce with this procedure? Do I need to change something other than DRAM frequency in BIOS in order to change the memory divider? If so, what?

Thanks muchos for all the help so far, gentlemen. All of my base are belong to you. Seriously.

 

[DrMrLordX]

Geek, looks like you've stabilized your overclock. Good move.

You should ideally let Prime95 run for 12-24 hours(24 is ideal) to deem it "prime stable".

As long as you aren't overvolting your RAM, it will likely not fry unless the DIMMS themselves are duds(i.e. unstable at stock speeds)

 

[lopri]

Who keeps her/his CPUs for 20 years? I'm serious. Anyone?

 

[FelixDeCat]

Originally posted by: lopri
Who keeps her/his CPUs for 20 years? I'm serious. Anyone?

Why are people still running 386s as firewalls, etc? Because they can. Im planning on keeping this a secondary box to the one I build from scratch either in late 2006 or eary 2007.

I just bought a x850 pro to extend its lifespan a tad.

 

[geekified]

Originally posted by: geekified

How is the memory divider calculated? I'm sort of a neophyte to all of this OC stuff so please do inform me of its meaning.


What are the chances of turning my poor generic RAM into pwnsauce with this procedure?

Do I need to change something other than DRAM frequency in BIOS in order to change the memory divider? If so, what?

Not to be tehrude, but can anyone answer my questions above?

 

[The Blue Factor]

im not sure if you figutred out your problem due to the fact you did 2 things and then it worked. so it could have been the vcore or the ram. if you are up to it i would set one or the other back up and just do one at a time. your ram speeds are fairly low due to using the 200 divider. i would set the vcore to 1.4 and then run the ram how you had it with the 2-3-3-5 and the prime again. if it passes then you know it was the vcore. after you figure that one out i would bring your ram divider to 333 and try and prime again. if it doesnt pass then start loosing the timings.

as for the prime question on how long. its pretty much up to you. you will get mixed answers from people. some will say 8hrs some will say 24. i myself only prime for 12hrs. that is what i feel is good enough for me.

hope this helps you out and good luck.

 

[geekified]

Originally posted by: The Blue Factor
im not sure if you figutred out your problem due to the fact you did 2 things and then it worked. so it could have been the vcore or the ram. if you are up to it i would set one or the other back up and just do one at a time. your ram speeds are fairly low due to using the 200 divider. i would set the vcore to 1.4 and then run the ram how you had it with the 2-3-3-5 and the prime again. if it passes then you know it was the vcore. after you figure that one out i would bring your ram divider to 333 and try and prime again. if it doesnt pass then start loosing the timings.

as for the prime question on how long. its pretty much up to you. you will get mixed answers from people. some will say 8hrs some will say 24. i myself only prime for 12hrs. that is what i feel is good enough for me.

hope this helps you out and good luck.

Can I ask you how big of a performance gap you see between 2.4 and 2.7 on your computer and would you try it if you had el cheapo generic RAM instead of your uber Corsair?

 

[The Blue Factor]

to tell you the truth if i see a performance difference between 2.4 and 2.7 i really dont know. i really dont do numbers and benchmarks. i overclock because i like to get the most out of it. its fun and i like the challenge. would i try it if i had your set up?? yeah i would. see what i could get. you never know. but if your really not into the overclocking scene and you just want to stay at 2.4 then by all means just go with what you know. have fun