Solution : reason for failing conroes at prime

**uOpt** · 08-29-2006, 11:37 AM

The code coming up with this doesn't use any of the flags or exceptions. It just checks what the maximum error from an expected result was doing nothing but normal FP instructions.

I also have to say that the tolerance chosen seems very high. I wonder what happens if I tighten it up by an order of magnitude...

So no, this CPU bug cannot have to do with mprime.

I don't think SuperPi has source code available, that might be worth a look, too.

**typo_pl** · 08-29-2006, 11:52 AM

Looking at CPU errata for previous Intel CPUs shows that the same erratum exists for the Core Duo (not Conroe == Core 2 Duo) and the PentiumIII CPUs, but not for the P4.

Care to guess where C2D comes from?

**Ruhnie** · 08-29-2006, 12:53 PM

Originally Posted by AnjinSan1966

It says:
FATAL ERROR:Rounding was 0.49609375, expected less than 0.4
Hardware failure detected,consult stress.txt file.

That is the exact error I get whenever I run Prime95 or Orthos blend, and I get it anywhere from 10 seconds to 4 mins in. I get this even at stock settings on my E6400, so does that mean I have this bug? What does this mean for someone who primarily wants to OC and use their rig for gaming? How exactly do I know if I'm stable?

**mds47** · 08-29-2006, 02:24 PM

I think uOpt is right on this one. This bug has no effect on prime torture testing.

The errors you're seeing are ordinary instability errors. Either your voltage is too low, your heat is too high, or the chip simply can't handle the hertz.

**Gautam** · 08-29-2006, 02:34 PM

What might be hard for people to grasp is that there is a very large gap between prime stable speeds and superpi/3D stable speeds. 1M pi -> dual prime has been around 300MHz for me on two CPU's. If 3DMark06 and PCMark05 can be completed at a certain speed, then chances are that Prime can be within 50MHz or so.

However, I was never personally able to get A64's prime stable. I ignored it altogher, but have used it for Intel systems.

**Vric** · 08-29-2006, 02:50 PM

Prime fail for me after 2 hours. Dual prime, one of the 2 simply crash.
I tried rising Voltage (I'm at stock voltage) with no better result.

I'm only at 320x9 on my E6600 (making it at 2.9Ghz) CPU can go much higher, but the OCZ Plat memory simply hate any divider beside 2:3 or 4:5.

I'm sure it's the memory that caused this problem. While testing them with memtest, they would fail randomly at any speed (even under stock speed).

Beside the Prime crash after 2 hours, I have no problem at all. Running all bench and games. Computer run 14 hours a day without problem.

**Btrice** · 08-29-2006, 04:13 PM

Originally Posted by toledo

Great

. Let me know your results as soon as possible .... . You can post here , i bet i am not the only one who is worried about it .

Sorry couldn't get it all installed till tonight. Results exactly the same as my work machine down to 1x10^-9 (probably even more but that is as far as my data printed out for that run).

**lawrywild** · 08-30-2006, 02:07 AM

I think it's all bull

tbh..

This is what happens when lots of people don't get the clocks they were expecting. People can't handle the fact that their cpus (or mobos) can't clock as high as some so feel the need to make excuses. I think people need to get over it tbh

**toledo** · 08-30-2006, 02:22 AM

Originally Posted by Btrice

Sorry couldn't get it all installed till tonight. Results exactly the same as my work machine down to 1x10^-9 (probably even more but that is as far as my data printed out for that run).

Thanks Btrice , good to know

Originally Posted by lawrywild

I think it's all bull

tbh..

This is what happens when lots of people don't get the clocks they were expecting. People can't handle the fact that their cpus (or mobos) can't clock as high as some so feel the need to make excuses. I think people need to get over it tbh

My concern was the CPU doing wrong calculations regardless of the overclock . I haven`t got my C2D yet

**lawrywild** · 08-30-2006, 02:43 AM

Originally Posted by toledo

Thanks Btrice , good to know

My concern was the CPU doing wrong calculations regardless of the overclock . I haven`t got my C2D yet

well mine is fine at stock and 3.2 1.385v

It's only when I push it further that it fails.. i.e. it won't clock that far (or the mobo/chipset won't let it)

**kadinho** · 08-30-2006, 03:04 AM

I think the prime bug it's different that cpu bug... My prime bug ocurrs at diferent times. 3s - 5s - 5 hours and at diferent speeds, 2.88 - 3 - 3.2 mhz .... Its just a feeling prime bug it's not the cpu bug...

cumps

**pcdoc1** · 08-30-2006, 04:16 AM

Originally Posted by kadinho

I think the prime bug...

The Prime Bug? Who started this BS... Years after development, disappointed users are now blaming the app...

**herderien** · 08-30-2006, 05:54 AM

Originally Posted by pcdoc1

This is an excellent thread for people that can't get stable enough to complete prime... It's just one more excuse in the arsenal…

well then tell me why am i priming stable with bios 0502 and not 0507 with exactly same settings? this cannot be related to OC stability

**pcdoc1** · 08-30-2006, 06:12 AM

Originally Posted by herderien

well then tell me why am i priming stable with bios 0502 and not 0507 with exactly same settings? this cannot be related to OC stability

That just further makes the point! It maybe a bios or chipset issue, but not a "Prime Bug" as being stated...

**astaris** · 08-30-2006, 06:14 AM

Originally Posted by herderien

well then tell me why am i priming stable with bios 0502 and not 0507 with exactly same settings? this cannot be related to OC stability

Different Bios use different hidden settings...

Originally Posted by kadinho

I think the prime bug it's different that cpu bug... My prime bug ocurrs at diferent times. 3s - 5s - 5 hours and at diferent speeds, 2.88 - 3 - 3.2 mhz .... Its just a feeling prime bug it's not the cpu bug...

That's a good one, Prime Bug...
Uhm i think here the bugged ones are the users that fail to admit their system is not 100% stable.

**herderien** · 08-30-2006, 06:22 AM

Under certain rare conditions, this bit may not be set when this rounding occurs.

memory settings + whatever hidden setting + demanding software ?

so my point is still valid

ps : i'm not saying it's a prime bug i'm just on the first statement that it is a cpu bug

**Lestat** · 08-30-2006, 06:50 AM

logic would dictate that even at stock even at 100mhz a conroe would cause prime failures if there is a bug in the cpu.
but the fact is thats not true. not for 99% of the people.

so that throws the bug theory out the window does it not ?

if its overclocking thats causing the failures. then that right there, without a doubt is the answer.

why debate it. if it runs fine at stock and yet fails when you overclock... what does that say ? to deny it is just silly.

UNLESS the bug is DIRECLTY related to higher clockrates of the core 2 duo chip. which i havent seen anything yet to denote that the higher clock rate causes the bug to surface

**crspyjohn** · 08-30-2006, 07:03 AM

I am having this problem at stock speeds as well

Should i return my chip?

**Pyr0** · 08-30-2006, 07:06 AM

check your memory and psu first

**crspyjohn** · 08-30-2006, 07:19 AM

ah, it can memory or psu. I'll check those out. thansk!

**Ruhnie** · 08-30-2006, 07:41 AM

Yeah, in my case, I think the memory is the problem, using the OCZ Plat 6400. Anytime I run a divider other than 4:3 I get the errors, even with the CPU/mobo at stock speeds and volts.

**Lestat** · 08-30-2006, 08:49 AM

4:3 equals what ? whats your ram running at ?

**Fred_Pohl** · 08-30-2006, 09:43 AM

There is no 'C2D/965 Prime bug'. This is an 'overclocked too high bug', plain and simple.

**Btrice** · 08-30-2006, 09:53 AM

Originally Posted by Fred_Pohl

There is no 'C2D/965 Prime bug'. This is an 'overclocked too high bug', plain and simple.

Originally Posted by crspyjohn

I am having this problem at stock speeds as well

**Fred_Pohl** · 08-30-2006, 10:16 AM

No offense meant to crspyjohn but he's had an amazingingly long list of problems with his new build.

If there was any relation between this errata and Prime rounding errors, it wouldn't be isolated to a handful of C2Ds, with most of them OC'd by ~50%. It would affect every C2D.

Thread: Solution : reason for failing conroes at prime

Thread Tools

Search Thread

Rate This Thread

Display

Bookmarks

Bookmarks

Posting Permissions