Wierd behaviour of Pentium D
 

I am having trouble running the Torture Test on a Pentium D 3.2G, I get the following error:

FATAL ERROR: Rounding was 0.5, expected less than 0.4

However, this error is only occuring when I "Let program run on any CPU". If I specify it runs on a specific core (either), there are no errors. If I run two instances of Prime95, one on each core, there are no errors.

I have tried reducing the memory speed, and also Memtest returns no errors. Temperatures appear okay.

Any suggestions will be much appreciated!

What are the CPU and mainboard temperature readings in BIOS?

Also, what voltage readings do you see? i.e. CPU, memory, 3.3v, 5v and 12v

What are the mainboard's manufacturer and model?

Don't you mean weird?

[quote=paulunderwood;120088]What are the CPU and mainboard temperature readings in BIOS?

Also, what voltage readings do you see? i.e. CPU, memory, 3.3v, 5v and 12v

What are the mainboard's manufacturer and model?[/quote]

Thanks for your reply Paul, here is the info':

Temp at idle = 40C
Temp at full load = 60C

CPU voltage (from BIOS) = 1.3V
Memory voltage and others = no entry in BIOS
(Should I be using a software tool for this? SiSoft SANDRA?)

Motherboard manu' = Gigabyte
Motherboard model = GA-8I865G775-G-RH

Since my previous post, I have replaced the RAM with DIMMs from another PC, but this made no difference. I have also replaced the PSU with another unit, and this made no difference either.

I suggest that from BIOS you set the "Load Fail-Safe Defaults" option and then retest with Prime95/mprime. If this fails again, under "MB Intelligent Tweaker(M.I.T.)", up the "DIMM OverVoltage Control" by "+0.1v" and retest with mprime/Prime95 and let us know whether you system is more stable...

[quote=paulunderwood;120131]I suggest that from BIOS you set the "Load Fail-Safe Defaults" option and then retest with Prime95/mprime. If this fails again, under "MB Intelligent Tweaker(M.I.T.)", up the "DIMM OverVoltage Control" by "+0.1v" and retest with mprime/Prime95 and let us know whether you system is more stable...[/quote]

Thanks for your time with this Paul!

Okay, I went with "Load Fail-Safe Defaults" and errors are still occuring. I increased the DIMM voltage by 0.1, and this didn't help either.

Btw, I also have the latest BIOS revision (F3).

Ok, thanks.

Can you identify your processor? This [URL="http://processorfinder.intel.com/List.aspx?ProcFam=2112&sSpec=&OrdCode="]list[/URL] might help.

What are your memory modules? Manufacturer(s), model(s), frequency?

[quote=paulunderwood;120138]Ok, thanks.

Can you identify your processor? This [URL="http://processorfinder.intel.com/List.aspx?ProcFam=2112&sSpec=&OrdCode="]list[/URL] might help.

What are your memory modules? Manufacturer(s), model(s), frequency?[/quote]

This is the processor (it is a 3.4, not a 3.2 as I wrote earlier):
[URL]http://processorfinder.intel.com/details.aspx?sSpec=SL9QQ[/URL]

I have a single DIMM fitted:

Manu': TwinMOS
Model: M2G9J16A-TT (512MB)
Freq': PC3200
CAS: 2.5

Okay, I am running out of ideas. My final guess is: at a temperature of 60C the chip is close to its maximum. Do you have adequate case cooling? If you can, try cooling the chip more, possible by running with case side removed and, further, with a fan blowing on the CPU area somehow. Are you using a standard Intel CPU heatsink and fan? You just might need a better heatsink and fan.

Ps. try [I]underclocking[/I] the CPU, say 3.0GHz i.e. "CPU Host Frequency" at 171MHz

[quote=paulunderwood;120141]Okay, I am running out of ideas. My final guess is at a temperature of 60C the chip is close to its maximum. Do you have adequate case cooling? If you can, try cooling the chip more, possible by running with case side removed and, further, with a fan blowing on the CPU area somehow. Are you using a standard Intel CPU heatsink and fan? You just might need a better heatsink and fan.[/quote]

I am using a Zalman CNPS-7000B cooler, which is usually a very effective cooler. I have been running it with the side removed and in a cool room.

I will try reseating the cooler with some fresh paste.

The error is occuring almost immediately after starting the Torture Test, and only when I set Affinity to "Let program run on any CPU". It works, if I assign Prime95 to a specific core. It's all very odd!

I'm starting to wonder whether I have a motherboard problem.

Many thanks for your assistance!

[quote=paulunderwood;120141]Ps. try [I]underclocking[/I] the CPU, say 3.0GHz i.e. fsb at 171MHz[/quote]

Sure, I will try this now.

[quote=paulunderwood;120141]Ps. try [I]underclocking[/I] the CPU, say 3.0GHz i.e. "CPU Host Frequency" at 171MHz[/quote]

I have set it running at 171MHz and 150MHz, and the errors are still generated - but, they take several minutes to occur.

Hmm. You have case fans? The rear one is important. When you apply the thermal paste, do you get a full and even, but also thin, covering? Is the paste a cheap brand? You might need better, silver-based paste, such as Artic Silver.

[quote=paulunderwood;120146]Hmm. You have case fans? The rear one is important. When you apply the thermal paste, do you get a full and even, but also thin, covering? Is the paste a cheap brand? You might need better, silver-based paste, such as Artic Silver.[/quote]

I am using Artic Silver, do you think it worth me reseating the cooler with fresh paste?

Yes. It is important to use a thin layer to ensure good heat conduction. I assume you know how to fit a socket 775 heatsink -- its best to remove the mainboard to make sure the legs of the heatsink are put correctly through it.

Okay, I have reseated the cooler, and I'm still getting errors.

Oh dear! Is your Windows patched up from Microsoft's site?

Do you have a in-line fan speed controller for the Zalman?

With the CPU Host frequency rebooted at 150MHz, what is the temperature reading in BIOS?

Maybe you should also try setting the memory manually to its specification...

Do you have another memory stick you could try? Make sure you first reset the memory to auto/SPD before fitting it...

What is the version number of Prime95 are you using?

[QUOTE=mjbaxter;120128]

CPU voltage (from BIOS) = 1.3V
[/QUOTE]

Are you sure about this value? It is too low...
Your CPU is based upon a Prescott core, isn´t it? I think the standard voltage for these cores is around 1.5V (at least my Northwood, which is pretty similar, runs at at 1.55). Try increasing the CPU voltage to 1.5V. You may wish to improve cooling, in case the temperature rises due to this increase in voltage.

See the link to the processor specification in post #8. It is 1.225V-1.325V

Could it be a bad CPU? It works on either core alone, but not on both cores. The only reason I can come up with is some kind of fault in the cache coherency circuitry. I've never seen this behavior before.

Thank you all for your posts!

[quote=Prime95;120226]Could it be a bad CPU? It works on either core alone, but not on both cores. The only reason I can come up with is some kind of fault in the cache coherency circuitry. I've never seen this behavior before.[/quote]

I take your point, any suggestions on how I can test for this?

[quote=paulunderwood;120161]Oh dear! Is your Windows patched up from Microsoft's site?[/quote]

Yes, a fresh install with all updates, and the latest drivers for MB/video etc.

[quote=paulunderwood;120161]Do you have a in-line fan speed controller for the Zalman?[/quote]

No, it is running stright from the MB.

[quote=paulunderwood;120161]With the CPU Host frequency rebooted at 150MHz, what is the temperature reading in BIOS?[/quote]

37C

[quote=paulunderwood;120161]Maybe you should also try setting the memory manually to its specification...[/quote]

I tried this, still getting errors

[quote=paulunderwood;120161]Do you have another memory stick you could try? Make sure you first reset the memory to auto/SPD before fitting it...[/quote]

I tried this, still getting errors

[quote=paulunderwood;120161]What is the version number of Prime95 are you using?[/quote]

The latest version (2414)


This is a very frustrating problem!

Have you tried running the memory at 333MHz i.e "x 1.66" of 200MHz ?

general question
 

I don't have any PentiumD 9**. Does this processor have four CPUs, two of them virtual HT ones?

[QUOTE=paulunderwood;120244]I don't have any PentiumD 9**. Does this processor have four CPUs, two of them virtual HT ones?[/QUOTE]There is no HT in the PentiumD's. Standard dual core, single thread per core.

[quote=paulunderwood;120240]Have you tried running the memory at 333MHz i.e "x 1.66" of 200MHz ?[/quote]

Yes, and I have also tried slackening the timings, and I still had errors.

Can you try running small FFT and large FFT torture tests? If it's failing on small FFT, then it's the CPU. If it fails on large FFT, then it's the memory.

To isolate a memory problem, I would try running:
1. memory at CAS3.
2. run memory at reduced speed and CAS3
3. run memory at reduced speed, CAS3 and bump voltage up 0.1V

What about double the RAM bandwidth with another DDRAM if you have access to some ?

Upgrade your chipsett driver ?

just to give some idea.