Prime95 question: (Prime95 fails but 3DMark2003 doesn't)
 

I've been incrementally overclocking my 2700+ to see how far I can push it
and yet still be able to have it be stable in 3D games.

At certain OC settings, Prime95 will fail within 2 minutes, but 3DMark2003
doesn't (and I executed it several times).

How critical are the Prime95 errors, I mean, can they be ignored if other
benchmarks don't crash? I get one of those:
"Fatal Error: rounding was 0.499999 expected less than 0.4

The only CPU intensive tasks I'll be using this machine for are 3D games.

And if a 3D benchmark doesn't crash, but Prime95 shows an error, how stable
can I expect the system to be when I actually play games?
And if having Prime95 find an errors can be ignored, how long should I aim
to have Prime95 run before an error occurs if I still want to system to be
stable in games?
Is anybody running their system with Prime95 errors and not having it crash?
Why is Prime showing errors yet 3DMark2003 isn't crashing or showing video
artifacts?

I know, I'm asking a lot of questions here

Any of your input/experiences are much appreciated

In article ogers.com,
said...
I've been incrementally overclocking my 2700+ to see how far I can push it
and yet still be able to have it be stable in 3D games.

At certain OC settings, Prime95 will fail within 2 minutes, but 3DMark2003
doesn't (and I executed it several times).

How critical are the Prime95 errors, I mean, can they be ignored if other
benchmarks don't crash? I get one of those:
"Fatal Error: rounding was 0.499999 expected less than 0.4

The only CPU intensive tasks I'll be using this machine for are 3D games.

And if a 3D benchmark doesn't crash, but Prime95 shows an error, how stable
can I expect the system to be when I actually play games?
And if having Prime95 find an errors can be ignored, how long should I aim
to have Prime95 run before an error occurs if I still want to system to be
stable in games?
Is anybody running their system with Prime95 errors and not having it crash?
Why is Prime showing errors yet 3DMark2003 isn't crashing or showing video
artifacts?

I know, I'm asking a lot of questions here

Any of your input/experiences are much appreciated



Chances are its just the FPU which is crapping out at that speed.

There is still quite a lot of floating point math done in games by the
CPU, chances are you not going to have any major problems but you might
get some strange behaviour/glitches in some games you otherwise wouldnt
get.

If prime is failing straight away or within 10 minutes or so then
chances are that increasing vcore wont help (meaning the FPU is well and
truely beyond its limit).

If however prime only fails after half an hour or more a little increase
to vcore *may* just be enough to stabalise it completely.

Personally I wouldnt trust any machine which fails Prime95 or any of the
other known tools, I'd rather lose a little speed and know any problems
arent being cause by the CPU rather than having it in the back of my
mind it could be the CPU crapping out causing that strange behaviour.

On Thu, 11 Mar 2004 13:11:56 GMT
"Daniel Czajko" wrote:

I've been incrementally overclocking my 2700+ to see how far I can push it
and yet still be able to have it be stable in 3D games.

At certain OC settings, Prime95 will fail within 2 minutes, but 3DMark2003
doesn't (and I executed it several times).

How critical are the Prime95 errors, I mean, can they be ignored if other
benchmarks don't crash? I get one of those:
"Fatal Error: rounding was 0.499999 expected less than 0.4

I had a similar error - increasing Vcore slightly solved it. Prime95 errors DO
mean just what they say: your computer is not stable. This is not some
rinky-dink software.

Larry Gagnon, A+ certified tech.

How critical are the Prime95 errors, I mean, can they be ignored if
other benchmarks don't crash? I get one of those:
"Fatal Error: rounding was 0.499999 expected less than 0.4


very
if it can't run prime95 torure test for at least 1 hour, your pc will be
unstable
games no worky in english :)

On Thu, 11 Mar 2004 13:11:56 +0000, Daniel Czajko wrote:

I've been incrementally overclocking my 2700+ to see how far I can push it
and yet still be able to have it be stable in 3D games.

At certain OC settings, Prime95 will fail within 2 minutes, but 3DMark2003
doesn't (and I executed it several times).

How critical are the Prime95 errors, I mean, can they be ignored if other
benchmarks don't crash? I get one of those:
"Fatal Error: rounding was 0.499999 expected less than 0.4

The only CPU intensive tasks I'll be using this machine for are 3D games.

The bottom line is that if you run your unstable machine, sooner or later,
something is going to get corrupted. And the longer you run it, the more
corruption that will occur, eventually getting to the point that it won't
boot because of system file errors. In the menatime, there's no telling
how much else is corrupted, maybe your database, spreadsheets, etc.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.htm

"Wes Newell" wrote in message
news:pan.2004.03.11.22.29.47.743898@TAKEOUTverizon .net...
On Thu, 11 Mar 2004 13:11:56 +0000, Daniel Czajko wrote:

I've been incrementally overclocking my 2700+ to see how far I can push
it
and yet still be able to have it be stable in 3D games.

At certain OC settings, Prime95 will fail within 2 minutes, but
3DMark2003
doesn't (and I executed it several times).

How critical are the Prime95 errors, I mean, can they be ignored if
other
benchmarks don't crash? I get one of those:
"Fatal Error: rounding was 0.499999 expected less than 0.4

The only CPU intensive tasks I'll be using this machine for are 3D
games.

The bottom line is that if you run your unstable machine, sooner or later,
something is going to get corrupted. And the longer you run it, the more
corruption that will occur, eventually getting to the point that it won't
boot because of system file errors. In the menatime, there's no telling
how much else is corrupted, maybe your database, spreadsheets, etc.

Can I just add to this some personal experience:

Since last year, I have had a corrupted registry. Strange symptoms: my PC
appeared to work perfectly, but whenever I accessed 1 particular key in the
registry, my PC would *instantly* blue screen stop on me. I managed to
avoid this happening most of the time, by not doing registry scans and
things, so its not been a day-to-day problem. But I have known it was there
and it has prevented me from doing various things.

I tried *everything* to fix it. I even raised some support entries with
Microsoft, sent them minidumps and full 1GByte core dumps. They (and
I)spent many hours trying to fix it. But there was nothing we could do to
remove or repair that damaged key. Any access to it would BSOD instantly.
Even a repair install didn't fix it.

In the end, we had to give up and do a complete fresh install, and reinstall
all my apps and data, security settings, services configuration, firewall
etc etc etc. Its taken me the last 2 days!

And all because of 1 corrupted registry key, lurking there unseen. And the
punchline is, MS said that 99% of the time, these sorts of errors are caused
by overclocking and CPU instability!

My lessons from this are twofold:

1. NEVER experiment with overclocking settings on a LIVE system. Do it on a
ghosted backup, or on a spare disk, or have a backup you can restore when
you have established what overclock settings are stable.

2. Any errors (Prime95 etc) whatsoever are unaccepable to me. I want 100%
stability. If you get errors, then trouble may be around the corner. And
the problem is, if you get corruption without realizing its there, by the
time you find out, it may be too late. (i.e. all your restore points and
backups have the corruption in them too.)

Chip.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.htm