Appcrash
 

I just recently ran across a weird problem. I was noticing that P95 was giving Memory allocation errors during S2 P-1 on any value over 4GB. I did searches and found an article talking about how XP had a 2 GB limit and then I searched the internet to see if Vista had a similiar problem. I found that there was one for the 32 bit, but for the 64 bit the option had been removed. This got me to thinking since I used to have 32 bit Vista on here and sure enough, I was running the 32 bit P95. So, I switched tot he 64 bit version and it started using my entire 6GB (out of 8 on board) allotment with no more allocation errors. Now, however, every time the program completes a S2 P-1 I get:
[Code]Problem Event Name: APPCRASH
Application Name: prime95.exe
Application Version: 26.6.1.0
Application Timestamp: 4d9f7bf8
Fault Module Name: prime95.exe
Fault Module Version: 26.6.1.0
Fault Module Timestamp: 4d9f7bf8
Exception Code: c0000005
Exception Offset: 0000000001831c00
OS Version: 6.0.6001.2.1.0.768.3
Locale ID: 1033
Additional Information 1: 468f
Additional Information 2: 4cfcce410b3d87a1e7632947e28a6400
Additional Information 3: 6c75
Additional Information 4: 86236f39369b9888cfc0e77f26c32500
[/code]

Once the program crashes I cannot restart P95 until I reboot. Even though the system is only a couple weeks old, I downloaded memtest last night and ran it for 12 hours. It completed 8 passes on the memory with no errors, so I am now checking with the people who understand programming far more than I do to see if a possible solution can be found.

I thought that P95 got renamed to P64 when running in 64-bit windows....or did I miss something?

[QUOTE=Christenson;280715]I thought that P95 got renamed to P64 when running in 64-bit windows....or did I miss something?[/QUOTE]

Uh... no? I'm running 64-bit everything.

bcp19, 32 bits means the processor and operating system can only address ~4GB; what you get is actually closer to 3. If you're using 32-bit Vista, you can't run any 64 bit programs. P95 can issue all the commands it wants, but
Windows simply cannot handle more than 4GB of memory. Either upgrade to 64 bit Vista or Windows 7. (You also need to have a 64-bit proc, but if you bought it in the last few years it should be.)

Some quotes from [url]http://en.wikipedia.org/wiki/64-bit[/url] :
[QUOTE]Most CPUs are designed so that the contents of a single integer register can store the address (location) of any datum in the computer's virtual memory. Therefore, the total number of addresses in the virtual memory — the total amount of data the computer can keep in its working area — is determined by the width of these registers. Beginning in the 1960s with the IBM System/360 (which was an exception, in that it used the low order 24 bits of a word for addresses, resulting in a 16 MB [16 × 1024^2 bytes] address space size), then (amongst many others) the DEC VAX minicomputer in the 1970s, and then with the Intel 80386 in the mid-1980s, a de facto consensus developed that 32 bits was a convenient register size. A 32-bit address register meant that 2^32 addresses, or 4 GB of RAM, could be referenced. At the time these architectures were devised, 4 GB of memory was so far beyond the typical quantities (4 MB) in installations that this was considered to be enough "headroom" for addressing. 4.29 billion addresses were considered an appropriate size to work with for another important reason: 4.29 billion integers are enough to assign unique references to most entities in applications like databases.[/QUOTE][QUOTE]64-bit computing started to drift down to the personal computer desktop from 2003 onwards, when some models in Apple's Macintosh lines switched to PowerPC 970 processors (termed "G5" by Apple) and the launch of AMD's 64-bit x86-64 extension to the x86 architecture, itself a response to Intel's Itanium gaining early operating systems support.[/QUOTE][QUOTE]A change from a 32-bit to a 64-bit architecture is a fundamental alteration, as most operating systems must be extensively modified to take advantage of the new architecture, because that software has to manage the actual memory addressing hardware.[17] Other software must also be ported to use the new capabilities; older software is usually supported through either a hardware compatibility mode (in which the new processors support the older 32-bit version of the instruction set as well as the 64-bit version), through software emulation, or by the actual implementation of a 32-bit processor core within the 64-bit processor (as with the Itanium processors from Intel, which include an IA-32 processor core to run 32-bit x86 applications). The operating systems for those 64-bit architectures generally support both 32-bit and 64-bit applications.[18][/QUOTE]And more info on current '64-bit' processors:[QUOTE]Most 64-bit microprocessors on the market today have a physical limit on the amount of memory they can address, considerably lower than the 64-bit limit which would suggest a memory size of 16 exabytes. For example, the AMD64 architecture currently has a 52 bit limit on physical memory and supports a 48-bit virtual address space.[4] This is 4 PB (4 × 1024^5 bytes) and 256 TB (256 × 1024^4 bytes), respectively. A PC cannot contain 4 petabytes of memory (due to the size of current memory chips if nothing else) but AMD envisioned large servers, shared memory clusters, and other uses of physical address space that might approach this in the foreseeable future, and the 52 bit physical address provides ample room for expansion while not incurring the cost of implementing 64-bit physical addresses. Similarly, the 48-bit virtual address space was designed to provide more than 65,000 times the 32 bit limit of 4 GB (4 × 1024^3 bytes), allowing ample room for expansion in the near future without incurring the overhead of translating full 64-bit addresses.[/QUOTE]

Good info, a small observation: numbers as 10245 should be read as 1024^5 and so on. Copy/Paste miss (wiki uses its own tex-like editor for formulas, which is copy/paste-dumb).

@Dubslow: you still have time to edit, about 45 miutes, if you are around.

Thanks; I think I got them all. (?)

Edit: Question:
[quote]32 bit limit of 4 GB (4 × 1024^3 bytes)[/quote]
4*1024^3 = 2^32, and it has bytes as the units. Wolfram Alpha [URL="http://www.wolframalpha.com/input/?i=2%5E32+bits+to+GB"]confirms that 2^32 [u]bits[/u] is only .5 GB[/URL].

Conclusion: Each address identifies 8 bits, not one 1 bit. Is this correct? If so, is there no useful information stored on computers that is less than 8 bits = byte?

And [URL="http://www.youtube.com/watch?v=35LJy-Z3-78"]while we're on the topic...[/URL]

Correct. See the memory like a matrix, "A x D", with A address-lines and D data-columns (wires). For a 2^20 bytes of memory, you have 8 data wires running from north to south, and 1048576 (2^20) addresses wires, running from east to west. At the intersection of each 2 wires is a small flip-flop that can store a zero or an one. All addresses wires are connected to ground (0) except one which is connected to plus (1), selecting only ONE line at a moment. Then you read the data wires and get your byte. Imagine a village with streets running from east to west (millions of them) and 8 streets running from north to south. At each intersection is a small house with a villager. There is a wolf which can appear at the east on one of the streets, and only the 8 villagers from that street can see him, and they will start running north all 8 at a time.

Do not confuse these (internal) address-lines here, with the 32 (external) address-lines coming out from your processor. That 32 external lines are "[URL="http://en.wikipedia.org/wiki/Demultiplexer"]demultiplexed[/URL]" inside of the memory chip, into 2^32 lines, to select ONE and ONLY ONE byte. then you read that byte using the data lines. Well, this in theory. In practice they are multiplexed all together and written/read in a time fashion (same external lines can be used for address writing and data reading, this is where the latency times come, the chip need some time for "internal cooking" when switching from one to the other, etc).

Uhhhhh.....
okay. One address is a byte, got it.

The address is a byte, until you get into modern embedded processors...where the word size, especially the program word size, might be 18 bits or so (Microchip). Also, I think modern RAM modules and memory busses (on DDR3 for example) return a lot more than 8 bits at a time, like maybe 64 or 128 bits, even if the CPU inside can actually address 8 bits. The rest is cached.

[QUOTE=Dubslow;280740]Uh... no? I'm running 64-bit everything.

bcp19, 32 bits means the processor and operating system can only address ~4GB; what you get is actually closer to 3. If you're using 32-bit Vista, you can't run any 64 bit programs. P95 can issue all the commands it wants, but
Windows simply cannot handle more than 4GB of memory. Either upgrade to 64 bit Vista or Windows 7. (You also need to have a 64-bit proc, but if you bought it in the last few years it should be.)[/QUOTE]

To repeat myself: [I] I used to have 32 bit Vista on here. [/I]Since I upgraded to 64 bit Vista but forgot to upgrade to 64 bit P95, I was figuring that was causing the 4GB problem. Regardless, the problem is the APPCRASH not the memory issue.

Oh. My bad. Sorry :blush:
It did seem like too easy an answer.

Some new information to add, the program only crashes if I have alternating P-1's. I had set the program up to run a 45M P-1, 7M P-1, 45M, 7M... after each 7M it would crash trying to start the next 45M. So I set it up to run the last 7 7M I had and then do all the 45M. The program has completed all the 7M and done 3 45M and not crashed once.

If it's that weird I suspect we'll require George to answer this one. It's certainly way beyond me.

bcp,
Looks like it might actually be a problem with Windoesn't...you are up to the latest patches, right?

[QUOTE=Christenson;281044]bcp,
Looks like it might actually be a problem with Windoesn't...you are up to the latest patches, right?[/QUOTE]

I know SP2 has been installed, and every day or 2 there seems to be more, so if not, it must be close.

[QUOTE=bcp19;281099]I know SP2 has been installed, and every day or 2 there seems to be more, so if not, it must be close.[/QUOTE]SP3 came out in 2008.

Testing the new 27.1 on my 2500 and it came back:
[code]Problem signature:
Problem Event Name: APPCRASH
Application Name: prime95_271.exe
Application Version: 27.1.1.0
Application Timestamp: 4edf1c41
Fault Module Name: prime95_271.exe
Fault Module Version: 27.1.1.0
Fault Module Timestamp: 4edf1c41
Exception Code: c0000005
Exception Offset: 01c4821a
OS Version: 6.0.6002.2.2.0.768.3
Locale ID: 1033
Additional Information 1: fd00
Additional Information 2: ea6f5fe8924aaa756324d57f87834160
Additional Information 3: fd00
Additional Information 4: ea6f5fe8924aaa756324d57f87834160
[\code]

System is a build, P67B3 motherboard, 2500K, 8GB ram, Vista (completely up to date, check for updates comes up empty). From what I have been noticing, I think it has something to do with memory allocation/drallocation. Since this is a 32 bit program I had to drop down to 3600 on memory, but upon completion of a 47M P-1/start of another it crashed. I'm going to reboot and set memory to 2000 and see how it fairs.

Have you tried running Memtest? That's pretty much guaranteed to pinpoint whether or not it's a memory issue.

[QUOTE=Dubslow;281443]Have you tried running Memtest? That's pretty much guaranteed to pinpoint whether or not it's a memory issue.[/QUOTE]

Did you not read bcp19's original post? Obviously not...

"Even though the system is only a couple weeks old, I downloaded memtest last night and ran it for 12 hours. It completed 8 passes on the memory with no errors, so I am now checking with the people who understand programming far more than I do to see if a possible solution can be found.

My suggestion, and it's not meant as flippant... Have you tried running Linux and mprime?

Whoops. Read the first paragraph -- not the stuff below the quotes :blush:

Perhaps, bcp19, you should just ignore everything I post in this thread

[QUOTE=chalsall;281444]Did you not read bcp19's original post? Obviously not...

"Even though the system is only a couple weeks old, I downloaded memtest last night and ran it for 12 hours. It completed 8 passes on the memory with no errors, so I am now checking with the people who understand programming far more than I do to see if a possible solution can be found.

My suggestion, and it's not meant as flippant... Have you tried running Linux and mprime?[/QUOTE]

I know absolutely nothing about Linux and in a lot of ways do not have the time to learn it.

Program just completed it's last run of a 47M exp and crashed again with 2GB memory selected. I recently installed MVS10 and clicked the debug option when it crashed and it came up with "Unhandled exception at 0x0204821a in prime95_271.exe: 0xC0000005: Access violation reading location 0x04332000." My knowledge of programming is woeful, so this means nothing to me, but it may be usuable to others.

In with my deallocation theory, I noticed that there is 1.6 GB of memory in use which + the 2 GB I allocated for the P-1 means that if the program for some reason could not reuse the memory it just used, that there is not enough below 4GB (being a 32 bit build) for it to allocate another 2 GB. Rebooting and trying again with 1GB allowed.

FWIW, on my father's computer we recently encountered a pair of 4 GB DDR3-1333 RAM sticks which survived 1 hour of memtest without any errors (a memtest triggered after a number of blue screens in Win 7), but miserably failed at the mprime torture test (4M-8M FFT, use 7000 or 7500 MB of RAM) under Linux after about five seconds of such treatment.
We returned the faulty sticks to the reseller, and bought other sticks instead, which survived hours of the same treatment, and also hours of msieve with a 4.6 GB matrix - so we declared the new chips good enough :smile:

[QUOTE=debrouxl;281478]FWIW, on my father's computer we recently encountered a pair of 4 GB DDR3-1333 RAM sticks which survived 1 hour of memtest without any errors (a memtest triggered after a number of blue screens in Win 7), but miserably failed at the mprime torture test (4M-8M FFT, use 7000 or 7500 MB of RAM) under Linux after about five seconds of such treatment.
We returned the faulty sticks to the reseller, and bought other sticks instead, which survived hours of the same treatment, and also hours of msieve with a 4.6 GB matrix - so we declared the new chips good enough :smile:[/QUOTE]

Always a possibility, but it seemed to run fine on the 64 bit version for several days and the testing of this new 32 bit version is causing the crashes again.

I got another appcrash overnight, midway through the stage 1 and debugging says "Unhandled exception at 0x0204821a in prime95_271.exe: 0xC0000005: Access violation reading location 0x0274b000."

I had picked up another 8GB ram, planning to expand the system to 16GB, so I have installed this ram in place of what was in there for further testing on this.

Beginning to look like one of my memory sticks is causing the problem, I had set the memory allocation back up to 3600 when I swapped the ram out and it has completed a P-1 stage 2 and kept going onto a new one.

It just crashed again, with debug reporting "Unhandled exception at 0x0204821a in prime95_271.exe: 0xC0000005: Access violation reading location 0x04660000." I highly doubt I have 2 sets of bad memory sticks, so I am at a loss. In looking at the debugger, 0204821A says mov dl,byte ptr [edx]

After a LOT of troubleshooting, it appears I have narrowed down the problem to the part of the program that talks to primenet. The program crashed again in the middle of a stage 1 run last night, but when I started it back up, I saw it report a result, but primenet sent back an error 40, not needed. After a few hours, I see the rpogram crashed again, and again when I restarted, I saw the error 40 report. So I clicked manual reporting to check it out and the program crashed. Repeated it 4 more times, the program crashed each time it tries to talk to primenet.

Since this is only occuring on this one machine, I have to conclude that the problem is most likely (>99%) not in the software, or (<1%) I have some weird setup that does not agree with the coding that works with all the other computers out there. Problem is, how do I determine if this is a CPU or motherboard problem given what I have discovered so far? I only have 4 more days to return it under their 30 day no hassle policy for replacement, so any ideas on this would be greatly appreciated.

There's a lot of reason to think it's the 32 bit mode in Vista that's the problem..the CURL library does the communicating...can you ask it not to communicate when running 32 bit mode?

[QUOTE=bcp19;281820]Since this is only occuring on this one machine, I have to conclude that the problem is most likely (>99%) not in the software, or (<1%) I have some weird setup that does not agree with the coding that works with all the other computers out there. Problem is, how do I determine if this is a CPU or motherboard problem given what I have discovered so far? I only have 4 more days to return it under their 30 day no hassle policy for replacement, so any ideas on this would be greatly appreciated.[/QUOTE]

If none of your other computers are doing this, I would get rid of the problem child and try something else.....especially so since this is still under a return policy.


EDIT: Let me qualify. I did go back in the thread and see that this is a build. I am guessing that the return policy applies to various components. The most obvious candidate for return might be the motherboard. You've already tested the memory.

QUOTE (edit): Since this is a 32 bit program I had to drop down to 3600 on memory, but upon completion of a 47M P-1/start of another it crashed. I'm going to reboot and set memory to 2000 and see how it fairs.

On a 64 bit Linux system with 3 GB of physical RAM I can run P-1 with 2400 MB allocated between 2 cores, on a machine that does nothing else. I'm only guessing, but Windows might want more.

The communication aspect is puzzling. For me, that leads back to the mobo and the network components, since these symptoms are unique. Could the problem be with a network adapter driver?

[QUOTE=Christenson;281907]There's a lot of reason to think it's the 32 bit mode in Vista that's the problem..the CURL library does the communicating...can you ask it not to communicate when running 32 bit mode?[/QUOTE]

Both the 32 and 64 bit versions were crashing when the communication occurred.

[QUOTE=kladner;281912]If none of your other computers are doing this, I would get rid of the problem child and try something else.....especially so since this is still under a return policy.


EDIT: Let me qualify. I did go back in the thread and see that this is a build. I am guessing that the return policy applies to various components. The most obvious candidate for return might be the motherboard. You've already tested the memory.

QUOTE (edit): Since this is a 32 bit program I had to drop down to 3600 on memory, but upon completion of a 47M P-1/start of another it crashed. I'm going to reboot and set memory to 2000 and see how it fairs.

On a 64 bit Linux system with 3 GB of physical RAM I can run P-1 with 2400 MB allocated between 2 cores, on a machine that does nothing else. I'm only guessing, but Windows might want more.

The communication aspect is puzzling. For me, that leads back to the mobo and the network components, since these symptoms are unique. Could the problem be with a network adapter driver?[/QUOTE]

I have begun to think along the driver lines myself, since P95 was working flawlessly right after the build, but now that multiple updates from microsoft have been applied, it is having problems. I have also noticed that when you go to shutdown, it takes an extremely long time for the shutdown to complete and the computer to power down (often in excess of 5 minutes, more than 10 minutes a few times). I know the 'stock' version of Vista did not recognize the network adapter until I ran the disc that came with the motherboard. I also noticed something else today... I decided to put 27.1 on my 2400 and unlike this system, it had a definite speedup to the LL's, so I started checking everything between them and found that the AVX is not being recognized on this system either, and 27.1 actually runs 20% slower (which is probably the 64/32 bit difference)

I think my next move is going to be to locate the motherboard disc and reinstall it and see what that does.

[QUOTE=bcp19;281952]I decided to put 27.1 on my 2400 and unlike this system, it had a definite speedup to the LL's, so I started checking everything between them and found that the AVX is not being recognized on this system either, and 27.1 actually runs 20% slower (which is probably the 64/32 bit difference)[/QUOTE]

I don't think Windows Vista supports AVX. Windows 7 did not support AVX until a Service Pack was released about a year ago.

[QUOTE=bcp19;281952]I think my next move is going to be to locate the motherboard disc and reinstall it and see what that does.[/QUOTE]One thing you can try is looking at the updates that have been installed and see if there are any that are related to hardware.

I have an XP PC that showed an optional update for the "nVIDIA Network Bus Enumerator". I installed it and it killed the network completely. I removed with Add/Remove and it went back to connecting.

So you could check the updates first, because if Windows installed something that is messing with the adapter, re-installing the mobo drivers might not mitigate it, since the update would still be applied.

The slow shutdown sure sounds like something is having trouble. Either something is being polled and doesn't respond or something's crashing and being terminated by Windows. Do you get any "program not responding" dialogs during shutdown?

[QUOTE=Prime95;281958]I don't think Windows Vista supports AVX. Windows 7 did not support AVX until a Service Pack was released about a year ago.[/QUOTE]

Thanks, I'll disregard the new program on here then uless I decide to get Win7.

[QUOTE=schickel;281960]One thing you can try is looking at the updates that have been installed and see if there are any that are related to hardware.

I have an XP PC that showed an optional update for the "nVIDIA Network Bus Enumerator". I installed it and it killed the network completely. I removed with Add/Remove and it went back to connecting.

So you could check the updates first, because if Windows installed something that is messing with the adapter, re-installing the mobo drivers might not mitigate it, since the update would still be applied.

The slow shutdown sure sounds like something is having trouble. Either something is being polled and doesn't respond or something's crashing and being terminated by Windows. Do you get any "program not responding" dialogs during shutdown?[/QUOTE]

I did reinstall the drivers for the motherboard, and the problem has become less than it was. Not every communication attempt, but about every 3rd or 4th. I'll have to look through the updates and see if I can figure out what may have caused it.

On the shutdown, it closes all the programs, and the desktop goes to the 'default' screen and I see nothing other than 'shutting down' on the screen for several minutes. I am tempted to backup the files used in P95 and wipe the drive and start fresh with no updates to see how it works and then decide what to do from there.

[QUOTE=bcp19;281973]
On the shutdown, it closes all the programs, and the desktop goes to the 'default' screen and I see nothing other than 'shutting down' on the screen for several minutes. I am tempted to backup the files used in P95 and wipe the drive and start fresh with no updates to see how it works and then decide what to do from there.[/QUOTE]

A clean start might be the easiest at this point.

However, I have seen the prolonged hang on the "Shutting Down" screen in XP Pro. In that case it seemed to be something holding a Registry key open. There was a patch from MS for that condition. Some evidence of what was happening was in Event Viewer --in the Applications log, I think.

A search for "vista slow shutdown" turned up this TechNet page:

[url]http://social.technet.microsoft.com/Forums/en-US/itprovistadesktopui/thread/270b76c8-ecc9-45b5-b8ad-68161ff694e6/[/url]

It recommends backing up the registry and then applying a tweak to shorten the wait time for ending services.