Division By Zero

[Seekermeister]

I just upgraded my processor, and have been booted up for several minutes, but it just popped a dialog saying:

Exception EZeroDivide in module AsmiTHFA.dll at 00002906. Floating point division by zero.

Since there was no BSOD or crash, I wondering if this is some kind of software error, or if it relates to the new processor?

 

[Seekermeister]

When I closed out the previous dialog, it was followed by two more, calling out VCL35.BPL and BORLNDMM.DLL.

 

[SparkyJJO]

Sounds like your CPU has a defective FPU. Considering you started getting the errors after changing the CPU that sounds like the culprit. You may have to RMA it.

 

[Seekermeister]

Is there any way to verify this?

 

[pcgeek11]

Originally posted by: Seekermeister
I just upgrade my processor, and have been booted up for several minutes, but it just popped a dialog saying:

Exception EZeroDivide in module AsmiTHFA.dll at 00002906. Floating point division by zero.

Since there was no BSOD or crash, I wondering if this is some kind of software error, or if it relates to the new processor?

What program was running when this happened. It sounds like a software issue as the PC was still running etc.

BORLNDMM.DLL ( Borland mem manager ) and VCL35.BPL ( also a Borland package library ) are involved I would say you have a poorly written program or corrupted setting for a program causing you pain.

This has been an issue with WinDVD and some of the Direct show filters...

****************************** Here is more info:
I think this is from the Asus MB you are using in combination with the Asus Probe application. Here is a link to a discussion:

http://hardware.mcse.ms/printthread.php?threadid=34392

http://forums.extremeoverclocking.com/showthread.php?t=154597

http://www.motherboardpoint.com/t21701-p4p800-power-fan-speed.html

http://www.mailarchive.ca/lists/alt.com...riphs.mainboard.asus/2005-10/1089.html

Seems you are not alone!

Try it without running the AsusProbe application if you are using it.


pcgeek11

 

[Seekermeister]

I have a number of programs that start at bootup, but nothing that I haven't always used. I looked in the Event Viewer...there were some errors, but none that seemed to be related to the problem above. I don't recall any programs using anything from Borland, but without simply going through every program directory, is there a way to see which might? Nothing has happened since that described above, so I guess I might just have to wait it out.

EDIT: I definitely am not using WinDVD or DirectShow, but I just saw you mention of Asus Probe, and yes it is running. It's never caused this problem before, but I really don't need it, since I'm using SpeedFan, and will shut it down.

 

[Seekermeister]

I didn't notice until after posting, that Asus Probe is no longer running in the system tray, so that is probably the problem.

 

[pcgeek11]

I edited the above post a lot as I was searching for a solution to your problems. Asus Probe is causing lots of people issues it seems with the same errors you had. Uninstall it and find something better if need be.

These types of errors are very rarely hardware related. I see everyone jumped right on the CPU due to the first Pentiums that came out with the floating point errors due to a hardware problem in the CPU, that was one of the very rare incidents.

Peace Out
pcgeek11

 

[Seekermeister]

If it had not been for the timing, I wouldn't have been so fast to suspect a processor problem, but since I hadn't had a problem with the probe before, I never even thought of it. Don't need it anyway with SpeedFan. But, I love get a specific answer to a question...thanks.

 

[pcgeek11]

Originally posted by: Seekermeister
If it had not been for the timing, I wouldn't have been so fast to suspect a processor problem, but since I hadn't had a problem with the probe before, I never even thought of it. Don't need it anyway with SpeedFan. But, I love get a specific answer to a question...thanks.

It was a normal response as you had just replaced the CPU, The Borland library file errors gave it away. They have been troublesome for many people on various applications.

You sir are very welcome and just glad I could get you going in the right direction.

pcgeek11

 

[oynaz]

You can get a Linux live CD (for example Ubuntu), and see if your computer boots that. If it does, you can more or less rule out a defective processor.

 

[Seekermeister]

I'm curious, since it did boot with the MCE CD, does that mean the same thing? But then it only saw the first CD and not the second.

 

[Mr Fox]

Originally posted by: Seekermeister
I just upgraded my processor, and have been booted up for several minutes, but it just popped a dialog saying:

Exception EZeroDivide in module AsmiTHFA.dll at 00002906. Floating point division by zero.

Since there was no BSOD or crash, I wondering if this is some kind of software error, or if it relates to the new processor?

Did you do the Dual processor Hotfix ? and /or set the processor affinity ?

There is nothing wrong with the CPU... they are very heavily tested.

worst case try uninstalling and reinstalling affected app.

 

[Lemon law]

As someone who dabbles at programming number crunching utilities, I have some experience with these type of errors. But to make a long story short, I was programming a circle line intersection program using long doubles in C. And the test for a tangent line circle intersection was having the square root of b squared minus 4AC be exactly equal to zero. But with nothing wrong with the processor, instead of being exactly equal to zero, a tiny error would be made, way out at the 19'th decimal place past the decimal in base 10. If this error worked out to be positive, the result would not show in the output, but if the error showed as less than zero---the program would bomb because a negative square root is an imaginary number. Simply writing an error handler in the case fixed the program---but the lesson is clear---computers do make tiny errors in base 2 to base 10 conversions---and in some case---especially when close to zero---these errors can become huge.

For the OP---you need to test your processor against known benchmarks---perhaps in dos and not windows. A local math or computer science resource near you might be the answer.

 

[oynaz]

Originally posted by: Mr Fox


There is nothing wrong with the CPU... they are very heavily tested.

That does not mean it cannot be faulty, though I agree a CPU doing wrong calculations would be very rare indeed.

 

[Seekermeister]

Originally posted by: Mr Fox

Originally posted by: Seekermeister
I just upgraded my processor, and have been booted up for several minutes, but it just popped a dialog saying:

Exception EZeroDivide in module AsmiTHFA.dll at 00002906. Floating point division by zero.

Since there was no BSOD or crash, I wondering if this is some kind of software error, or if it relates to the new processor?

Did you do the Dual processor Hotfix ? and /or set the processor affinity ?

There is nothing wrong with the CPU... they are very heavily tested.

worst case try uninstalling and reinstalling affected app.

If the application that you are referring to is Asus Probe, I don't need it. But, I'm not sure what you mean by the "Hotfix...and or set the processor affinity. I have installed the AMD Optomizer, CPU Driver and CPU Setup. Does these take care of what you were referring to? If not, please explain.

 

[IEC]

Originally posted by: SparkyJJO
Sounds like your CPU has a defective FPU. Considering you started getting the errors after changing the CPU that sounds like the culprit. You may have to RMA it.

Running Prime + Memtest will iron out that possibility.

 

[pcgeek11]

Seekermeister

Have you had any more errors since getting rid of Asus Probe? If not I wouldn't worry about all the doom sayers. This is a well documented issue with the Borland libraries and poor code writing.

pcgeek11

 

[xtknight]

This would not happen if the CPU had a FDIV bug.

This is a handled exception in which the code contained an operation that equated to dividing by zero.

 

[Seekermeister]

Originally posted by: pcgeek11
Seekermeister

Have you had any more errors since getting rid of Asus Probe? If not I wouldn't worry about all the doom sayers. This is a well documented issue with the Borland libraries and poor code writing.

pcgeek11

Not a one. I intend to run Prime95, just on GP, but not until I'm going to be away from the computer for awhile. Even without the errors, it seems like a good idea.

 

[JEDIYoda]

Originally posted by: Lemon law
As someone who dabbles at programming number crunching utilities, I have some experience with these type of errors. But to make a long story short, I was programming a circle line intersection program using long doubles in C. And the test for a tangent line circle intersection was having the square root of b squared minus 4AC be exactly equal to zero. But with nothing wrong with the processor, instead of being exactly equal to zero, a tiny error would be made, way out at the 19'th decimal place past the decimal in base 10. If this error worked out to be positive, the result would not show in the output, but if the error showed as less than zero---the program would bomb because a negative square root is an imaginary number. Simply writing an error handler in the case fixed the program---but the lesson is clear---computers do make tiny errors in base 2 to base 10 conversions---and in some case---especially when close to zero---these errors can become huge.

For the OP---you need to test your processor against known benchmarks---perhaps in dos and not windows. A local math or computer science resource near you might be the answer.

I also dabble om programming using any number of computer languages. I was programming a rectangular line sectional program using long quituplets in C. And the perpetuated affirmation for a qutient line circle subdivision was having the square root of g triangulated minus 2squared be exactly equal to g. But with nothing wrong with the processor, instead of being exactly equal to zero, a tiny error would be made, way out at the 101st decimal place past the decimal in base 43. If this error worked out to be positive, the result would not show in the triangulated matrix, but if the error showed as less than zero---the program would implode because a negative square hectidecimal is an imaginary number. Simply writing an error handler in the case fixed the program---but the lesson is clear---computers do make tiny errors in base 39 to base 101 conversions---and in some case---especially when close to zero---these errors can become huge.