FATAL ERROR: Rounding was 0.49xxxxxx, expected less 4 - Page 2

tediber · 2003-01-18, 12:46

hello everyone !!

i'm new to prime95 and have been having the same problem

i have an athlon XP1700+ T-bred A @ 1.9ghz temps are only 48 to 52C max

EPOX 8K3A at 200FSB

i am able to test my system using sandra's burn simultaneously with 3dmark ........ still passes the test ...... but when i tried prime95 ..... that is what i get rounded error 0.49999.... > 0.4 ???? is there a solution to this ...... i tried even at all default settings of my system

Prime95 · 2003-01-18, 15:19

Quote:

Originally Posted by tediber

i'm new to prime95 and have been having the same problem
i have an athlon XP1700+ T-bred A @ 1.9ghz temps are only 48 to 52C max

Ignore this thread. Look at some of the others about failing the torture test.

Your CPU, Ram, or motherboard is not functioning properly.

QuintLeo · 2003-01-18, 21:31

> EPOX 8K3A at 200FSB

I suspect you're pushing the FSB of the motherboard, or pushing the RAM, too fast.

Try backing off the FSB some.

outlnder · 2003-01-18, 21:34

The standard FSB for an XP processor is 266. I would think that you should up it to what it should be and stop playing with overclocking.

Stable by Sandra's definition is not error free by Prime95's.

tediber · 2003-01-19, 05:28

i have tried evrything at default speeds ...... the problem is still there ....

tediber · 2003-01-19, 05:53

i just solved a stability problem yesterday ...... apparently my 500w AVR couldn't handle the load well enough ..... am using 2 AVRs now 1 for the monitor and 1 for the rest .... this eliminated games crashing back to the desktop ...... been solid running sandra and 3dmark ...... right now ... i'm thinking the culprit of my CPU unable to pass prime95 could be the generic PSU i'm using ...... although as i mentioned above everything has been pretty stable enough ....... my mobo voltages aren't as "perfect" at all .... the 5Volt rail is around 4.7 to 4.8 ..... gets lower with higher Vcore..... so i'm thinking this could be the culprit ..... what do you guys think ?? i'm not inclined to buy an expensive PSU though .....

tediber · 2003-01-19, 15:42

i think i've found the culprit ........ my memory stick is 512mb Geil DDR333 ultra ... supposedl rated CAS2-2-2-1T at 166mhz .......... though as mentioned above it's fairly stable up to 200FSB at CAS2-3-5-3-2T ........ i decided to test it each component ....... now my XP1700+ runing @ 1.988ghz temps are closing in to 60C but hasn't failed prime95 yet ........ it's been the ram stick all along ...... i decided to use SPD in the BIOS and it is detected as 166MHZ cas 2.5 ...... and anything higher than that prime95 fails ...... damn GEIL using overclocked chips !!!!! and rating them as CAS2-2-5-2-1T ..... this would be cheating the customer right ...... unfortunately i can't return this stick as it would cost me more ..... i had my bro buy it in the united states and had it sent to me via someone coming home =(

outlnder · 2003-01-19, 20:14

Try to get the 1T command. This is the biggest tweak for increasing speed.

QuintLeo · 2003-01-19, 22:35

> The standard FSB for an XP processor is 266.

133, actually. Then "double-pumped" to the equivllent of 266. I'm used to thinking of the actual FSB clock speed on Athlons, though I realise a lotta folks call it by the "effective" rate. I suspect the original poster was runng "400 FSB" by that way of calling it.

> Stable by Sandra's definition is not error free by Prime95's.

No doubt. I've had a few boxes that ran dnet for years stable and error-free that I had to down-clock a little before they would run Prime at all.

Ian_H · 2003-02-24, 17:26

This is interesting. I'm running an Athlon 1.1, no changes to the hardware for two years and it's been running Prime95 24/7 for months, before that maybe 14 hours a day.

Today, for the first time, during a new FFT test for a new exponent range:
[code:1]FATAL ERROR: Rounding was 0.4997728097, expected less than 0.4
Hardware failure detected, consult stress.txt file.[/code:1]
Version 23.2.1

It is definitely not a hot day and the room isn't heated excessively. Now back to running LL and CPU temp shows as 63.

Any thoughts anyone? A coincidence that it's the same "0.49xxxx" message (same range of FFT tests?) See also a post in the thread "Errors during Torture Test"? Does a particular range make demans of the system that earlier ones didn't?

lycorn · 2003-02-24, 22:37

I also own an Athlon (XP1800+, Palomino core) that is running P95 for over 1 year, 24/7. Some 2 or 3 months ago, 2 roundoff errors showed up in the very same day, while testing a 33M exponent (I had already tested another one before). It had never happened, and it didn´t happen again. The CPU temp was somewhere in the low 50´s (ºC!...). The only explanation I can think of is that I had just put in place the side covers of the box, that didn´t use to be there, and the temperature had risen 3 to 4 degrees, but in just a couple of minutes. This means that, although staying within perfectly reasonable limits, the gradient was rather sharp, and I suspect that this sudden change in the working conditions of the CPU induced the error.
Now, Ian, is 63ºC the usual temperature your CPU works at? I would recommend a lower value, 58 or 59 at most.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
FATAL ERROR: Rounding was 0.5, expected less than 0.4	glal12	Information & Answers	2	2016-09-21 22:59
FATAL ERROR: Rounding was 0.5 les then expected 0.4,	sekramer10	Hardware	6	2015-11-16 13:31
Prime 95: FATAL ERROR: Rounding was 0.5, expected	etuckram	Hardware	1	2011-05-27 00:11
FATAL ERROR:Rounding was 0.5, expected less than 0	Unregistered	Information & Answers	20	2011-05-17 02:41
FATAL ERROR: Rounding was 0.49xxxxxx, expected less 4	tester	Software	1	2003-01-17 04:45

2003-01-18, 12:46	#12
tediber Jan 2003 100₂ Posts	hello everyone !! i'm new to prime95 and have been having the same problem i have an athlon XP1700+ T-bred A @ 1.9ghz temps are only 48 to 52C max EPOX 8K3A at 200FSB i am able to test my system using sandra's burn simultaneously with 3dmark ........ still passes the test ...... but when i tried prime95 ..... that is what i get rounded error 0.49999.... > 0.4 ???? is there a solution to this ...... i tried even at all default settings of my system

2003-01-18, 21:31	#14
QuintLeo Oct 2002 Lost in the hills of Iowa 2⁶·7 Posts	> EPOX 8K3A at 200FSB I suspect you're pushing the FSB of the motherboard, or pushing the RAM, too fast. Try backing off the FSB some.

2003-01-18, 21:34	#15
outlnder Aug 2002 2·3·53 Posts	The standard FSB for an XP processor is 266. I would think that you should up it to what it should be and stop playing with overclocking. Stable by Sandra's definition is not error free by Prime95's.

2003-01-19, 05:28	#16
tediber Jan 2003 4₈ Posts	i have tried evrything at default speeds ...... the problem is still there ....

2003-01-19, 05:53	#17
tediber Jan 2003 100₂ Posts	i just solved a stability problem yesterday ...... apparently my 500w AVR couldn't handle the load well enough ..... am using 2 AVRs now 1 for the monitor and 1 for the rest .... this eliminated games crashing back to the desktop ...... been solid running sandra and 3dmark ...... right now ... i'm thinking the culprit of my CPU unable to pass prime95 could be the generic PSU i'm using ...... although as i mentioned above everything has been pretty stable enough ....... my mobo voltages aren't as "perfect" at all .... the 5Volt rail is around 4.7 to 4.8 ..... gets lower with higher Vcore..... so i'm thinking this could be the culprit ..... what do you guys think ?? i'm not inclined to buy an expensive PSU though .....

2003-01-19, 15:42	#18
tediber Jan 2003 2² Posts	i think i've found the culprit ........ my memory stick is 512mb Geil DDR333 ultra ... supposedl rated CAS2-2-2-1T at 166mhz .......... though as mentioned above it's fairly stable up to 200FSB at CAS2-3-5-3-2T ........ i decided to test it each component ....... now my XP1700+ runing @ 1.988ghz temps are closing in to 60C but hasn't failed prime95 yet ........ it's been the ram stick all along ...... i decided to use SPD in the BIOS and it is detected as 166MHZ cas 2.5 ...... and anything higher than that prime95 fails ...... damn GEIL using overclocked chips !!!!! and rating them as CAS2-2-5-2-1T ..... this would be cheating the customer right ...... unfortunately i can't return this stick as it would cost me more ..... i had my bro buy it in the united states and had it sent to me via someone coming home =(

2003-01-19, 20:14	#19
outlnder Aug 2002 2×3×53 Posts	Try to get the 1T command. This is the biggest tweak for increasing speed.

2003-01-19, 22:35	#20
QuintLeo Oct 2002 Lost in the hills of Iowa 2⁶·7 Posts	> The standard FSB for an XP processor is 266. 133, actually. Then "double-pumped" to the equivllent of 266. I'm used to thinking of the actual FSB clock speed on Athlons, though I realise a lotta folks call it by the "effective" rate. I suspect the original poster was runng "400 FSB" by that way of calling it. > Stable by Sandra's definition is not error free by Prime95's. No doubt. I've had a few boxes that ran dnet for years stable and error-free that I had to down-clock a little before they would run Prime at all.

2003-02-24, 17:26	#21
Ian_H Aug 2002 Europe 30₁₀ Posts	This is interesting. I'm running an Athlon 1.1, no changes to the hardware for two years and it's been running Prime95 24/7 for months, before that maybe 14 hours a day. Today, for the first time, during a new FFT test for a new exponent range: [code:1]FATAL ERROR: Rounding was 0.4997728097, expected less than 0.4 Hardware failure detected, consult stress.txt file.[/code:1] Version 23.2.1 It is definitely not a hot day and the room isn't heated excessively. Now back to running LL and CPU temp shows as 63. Any thoughts anyone? A coincidence that it's the same "0.49xxxx" message (same range of FFT tests?) See also a post in the thread "Errors during Torture Test"? Does a particular range make demans of the system that earlier ones didn't?

2003-02-24, 22:37	#22
lycorn "GIMFS" Sep 2002 Oeiras, Portugal 2×7×113 Posts	I also own an Athlon (XP1800+, Palomino core) that is running P95 for over 1 year, 24/7. Some 2 or 3 months ago, 2 roundoff errors showed up in the very same day, while testing a 33M exponent (I had already tested another one before). It had never happened, and it didn´t happen again. The CPU temp was somewhere in the low 50´s (ºC!...). The only explanation I can think of is that I had just put in place the side covers of the box, that didn´t use to be there, and the temperature had risen 3 to 4 degrees, but in just a couple of minutes. This means that, although staying within perfectly reasonable limits, the gradient was rather sharp, and I suspect that this sudden change in the working conditions of the CPU induced the error. Now, Ian, is 63ºC the usual temperature your CPU works at? I would recommend a lower value, 58 or 59 at most.