Prime stable, but Windows bluescreens?

[Falkentyne]

Greets,

Here is my hardware:
Abit IC7 max3,
Pentium 4 3.4C
Kingmax PC4000 2x512 MB
Windows XP Pro
Lian-Li PC-6089A case
Thermalright SLK-947U + vantec tornado 92mm fan

What is supposed to happen when Prime isn't stable?
Isn't it supposed to report some error?
Or is windows XP simply supposed to crash and burn with a STOP bluescreen error? Or do the errors only occur in win98se ?

Reason I'm asking is, when I run prime 95, it, so far, has NEVER ever reported an error. Usually, windows blue screens with a stop error after a few hours of it ( at default vcore @ 3.7 ghz).
(The last stop error was IRQL_NOT_LESS_OR_EQUAL).

I've never gotten a bluescreen or crash at 3.7 ghz @ 1.675v, with the case side cover off (runs 2 prime instances all day), though problems do start appearing after awhile, if the case cover is on.

Unreal Tournament will usually GPF, or bluescreen crash at 3.7 ghz @ stock voltage, but Prime95 has never reported an internal error.... either windows bluescreens, (most of the time), locks up (very rarely, usually only happens at 1.7v or more), or reboots.

Does anyone know why this is happening?
I'm using a SLK-947U + Tornado fan, so I would think that this chip would have at least a LITTLE headroom in it...unless I simply got the WORST overclocking 3.4C in existance...

Note: reducing RAM timings, removing 1 stick of RAM or setting divider to 5/4 doesn't help. However, turning OFF hyper-threading does help (with HT off, I can run 3.7 ghz all day, with case cover off, @ stock vcore), BUT, I take a thousand point drop in 3dmark 2001se; enough that its better to run HT enabled at 3.7 ghz than HT disabled at 3.8 ghz.

Anyway, does anyone have any idea why Prime never reports errors, but the computer just bluescreens? Are bluescreens when running Prime, usually caused by CPU or Memory?

BTW, the max limit for this chip seems to be 3750 mhz @ 1.675v--220 fsb, case cover off. Seems extremely weak for a 3.4C D1, especially considering the high end air cooling i have....

I'm considering trying my luck with another chip, but I'd definitely like to know if I would have the same problems.....

[Overbrazil]

whick prime95 did you tested ?
try small FFt one.
Maybe heat issue, limit of the cpu, psu, etc...
try memtest for 10 hours. If passes without errors, your memory are fine.

[Lancelot]

IRQL_NOT_LESS_OR_EQUAL Is usually caused by a driver-conflict somehow/somewhere. What happens is that a certain driver is trying to access (mostly memory) resources that are already occupied by another driver. This is a very annoying stop screen cuz it can sometimes happen after 5 minutes but also after 5 hours. Check to see if there are any new windows-certified drivers for your hardware, and also make sure there is no shadowing (system/videoBIOS etc.) of any kind enabled in BIOS.

[Edward2]

The IRQL error could be due to either CPU or memory problems. What memory speed and timings are you running? Also, you mention that you have more problems with the case cover on. What is your CPU temp and system temp while running at 100% load? You may not have good enough air flow through your case.

Like Overbrazil said, I would recommend that you run Memtest for a couple of hours to see if it is your memory. If your memory passes, then it may be your CPU or NB is overheating.

Do you really need 1.675 volts to run at 3.7 GHz? That seems very high for such a small overclock. Too much voltage can be just as bad as too little voltage.

[Lancelot]

Stop 0x0000000A or IRQL_NOT_LESS_OR_EQUAL
The Stop 0xA message indicates that a kernel-mode process or driver attempted to access a memory location to which it did not have permission, or at a kernel interrupt request level (IRQL) that was too high. A kernel-mode process can access only other processes that have an IRQL lower than, or equal to, its own. This Stop message is typically due to faulty or incompatible hardware or software.

Interpreting the Message
This Stop message has four parameters:

Memory address that was improperly referenced.
IRQL that was required to access the memory.
Type of access (0x00000000 = read operation, 0x00000001 = write operation).
Address of the instruction that attempted to reference memory specified in parameter 1.
If the last parameter is within the address range of a device driver used on your system, you can determine which device driver was running when the memory access occurred. You can typically determine the driver name by reading the line that begins with:

**Address 0xZZZZZZZZ has base at <address>- <driver name>

Partially pasted from: HERE (http://www.microsoft.com/resources/documentation/Windows/XP/all/reskit/en-us/Default.asp?url=/resources/documentation/Windows/XP/all/reskit/en-us/prmd_stp_hwpg.asp)

[Falkentyne]

Actually the stop errors are random. I Just reported the last one I saw.

If it happens while windows is loading, it's anything from an error in ACPI.SYS, to NTFS.sys (I think), and the aforementioned IRQ_L.

GAT settings are all auto (disabling the last 2 made no difference), ratio is set to 1:1 (5/4 made no difference).

By the way, I ran MEMTEST86 @ 3.91 ghz, 1.675 vcore, it passed all tests for 4 hours, until I got bored of sitting there, and rebooted the computer. Then, windows bluescreened as soon as it accessed the HD to load the OS :P. So, the RAM seems up to the task for sure, unless this board doesn't like high FSB's with this RAM. (on my roommate's old Abit BF6, using Kingmax RAM causes errors in MEMTEST.exe (NOT memtest86, memtest is an older, different (and rather useful) at 140 FSB, while using Siemens = 100% stable at 150 FSB. MEMTEST Was able to pick up errors that memtest86 didn't catch on that system.

Case temp with the case panel off is around 30 C idle, 34 C load
with the panel on, case temp usually reaches 41 C load.

CPU temp HAS to be completely wrong.
At 1.7 ghz, with HT on, its 42.5 C idle, and 54 C load @ 1.675v (the board undervolts down to around 1.62v when in load).

With HT off, idle temp is 48C idle, 54C load.
Only a 6C rise... (interesting that the idle HT off temps are higher than HT on....).

I know the MAX3's CPU temps are totally hogwash, because I ran the CPU at 1.4 (!) ghz @ default (1.55) vcore, and my IDLE temps were 42(!!)c, and Load temps 46 C (Prime), with HT turned off. (I turned HT off, to match HSF reviews with the older 1.4-1.8 ghz P4's). With games running @ 1.4 ghz, load temp only increased by 2C.

There's no way a 1.4 ghz CPU is going to idle at 42 C.
Especially when I've seen reviews tested on 1.8 ghz CPU's, with my same HSF combo, which reported 32 C idle, 34C load.

Anyway, I think I just have the worst 3.4 C ever made. Unless others with poor 3.4C's are simply not posting here.

Has anyone else with a 3.4 C required a voltage increase past 3.6 ghz with high end air?

The reason I think my chip was just a dud is, when I was using the retail HSF, even 3.6 ghz required a vcore increase(!)
That's like NO headroom at all....how could the chip pass strict 3.4 ghz testing like that ?

[Falkentyne]

Originally posted by Edward2
The IRQL error could be due to either CPU or memory problems. What memory speed and timings are you running? Also, you mention that you have more problems with the case cover on. What is your CPU temp and system temp while running at 100% load? You may not have good enough air flow through your case.

Like Overbrazil said, I would recommend that you run Memtest for a couple of hours to see if it is your memory. If your memory passes, then it may be your CPU or NB is overheating.

Do you really need 1.675 volts to run at 3.7 GHz? That seems very high for such a small overclock. Too much voltage can be just as bad as too little voltage.

MEMTEST86:
passed 4 hours of testing @ 3.91 ghz, 1.675v.
Win XP wouldn't even begin to load (immediate STOP error) unless I went over 1.7v, which I'm not doing. (1.725v, which is 1.69v real, almost loaded the desktop before bluescreen in acpi.sys).

Cooling:
2 case fans sucking in, 1 fan at top sucking out, 1 fan, + PSU fan in rear, going out. there's no way the cooling is bad with the side of the case off, not with a Lian-Li case ;)

Voltage: Believe it or not, yes I do.
using stock voltage @ 1.7 ghz (with high end air, mind you) causes bluescreens after running games or Prime for awhile, or causes Unreal Tournament 2003 to GPF or just bluescreens windows. (Oddly, again, prime95 NEVER crashes..it just bluescreens). This is with the case cover off, case temp 30 C. Remember, 1.675v is actually 1.62v under load, since the board undervolts.

Sigh, I should have created a FAT32 partition and installed win98se, just to see if there's a difference in howfar I can o/c, and how Prime95 would operate. Don't know if 98se would like a 3.7 ghz CPU, however @_@

You think I should just buy another chip and sell this one? They aren't cheap... I'm sure that almost anything I buy can beat this overclock ...there's no 3.4C in the world that is worse than mine...

[KrisMCool]

Man, I've been having the EXACT same problem with BSODs. I haven't narrowed it down yet, but I think it's a driver issue or HD corruption. I don't think the issue is your CPU because I'm 99% positive that it's not the root of my problem. When I first got mine (3.0C) it made it to over 4GHz at 1.5875 vcore. Then I backed off and started stability testing. Memory ratios and FSB settings are where I've defined when the BSODs start. At 1:1, they occur at FSB 251 and above. At 5:4, it's FSB 256 and above. I've adjusted every reasonable BIOS setting, reinstalled my OS and multiple drivers (recently upgraded to Cat 4.3 video and most recent EAX 4.0 sound drivers), all to no avail. Below those settings, I can run the CPU at default voltage at 100% load 24/7 (2xFAH) without errors. Haven't found the breaking point of either set of memory I'm using (Mushkin PC4000 & OCZ PC3700 Gold Rev 2) - they test out error-free in Memtest86+ for over 24hrs at well above their rated speed and timings. I thought it might be a mobo issue because removing a stick & running single channel seemed to eliminate the problem (tested all dimm slots individually - all OK), but I didn't test for too long that way - single channel sucks. But since you have a different mobo and are experiencing the same problems, maybe it isn't. And I just ordered an Abit Max3 this morning to find out. Oh well.

I hope someone finds a solution to our problem soon. I'm still testing. One suggestion for you though - I'd really lower your vcore. Mine actually produced errors faster at higher vcores (up to 1.7v). Try to find your max stable OC at default vcore first - I'll bet the FSB setting will be the same regardless of vcore. At least mine was. Good luck and don't blame your CPU just yet.

[situman]

my friend had the same problem. He flashed his bios to the newest version on his IC7 and the problem is solved. If not then u should go into the bios and set the IRQs manually. There wont be conflicts if everything is asigned an IRQ. Not sure which IRQ is for which hardware but u can always go ask in the software forums or MB forum section.

[Falkentyne]

Originally posted by KrisMCool
Man, I've been having the EXACT same problem with BSODs. I haven't narrowed it down yet, but I think it's a driver issue or HD corruption. I don't think the issue is your CPU because I'm 99% positive that it's not the root of my problem. When I first got mine (3.0C) it made it to over 4GHz at 1.5875 vcore. Then I backed off and started stability testing. Memory ratios and FSB settings are where I've defined when the BSODs start. At 1:1, they occur at FSB 251 and above. At 5:4, it's FSB 256 and above. I've adjusted every reasonable BIOS setting, reinstalled my OS and multiple drivers (recently upgraded to Cat 4.3 video and most recent EAX 4.0 sound drivers), all to no avail. Below those settings, I can run the CPU at default voltage at 100% load 24/7 (2xFAH) without errors. Haven't found the breaking point of either set of memory I'm using (Mushkin PC4000 & OCZ PC3700 Gold Rev 2) - they test out error-free in Memtest86+ for over 24hrs at well above their rated speed and timings. I thought it might be a mobo issue because removing a stick & running single channel seemed to eliminate the problem (tested all dimm slots individually - all OK), but I didn't test for too long that way - single channel sucks. But since you have a different mobo and are experiencing the same problems, maybe it isn't. And I just ordered an Abit Max3 this morning to find out. Oh well.

I hope someone finds a solution to our problem soon. I'm still testing. One suggestion for you though - I'd really lower your vcore. Mine actually produced errors faster at higher vcores (up to 1.7v). Try to find your max stable OC at default vcore first - I'll bet the FSB setting will be the same regardless of vcore. At least mine was. Good luck and don't blame your CPU just yet.

Ugh... @_@
My max STABLE overclock, at default voltage, with the case cover ON, is about 3.6 ghz. That definitely isn't much of a jump from 3.4 ghz.

My max stable overclock, def voltage, with case cover OFF, is around 3.65 ghz. Going higher requires more voltage, or I'll start getting bluescreens while under load, or games crashing to desktop. (like, if I try 3.7 ghz @ default voltage). Prime95 NEVER crashes, it just bluescreens with STOP error in something.

That's it..
I'm stable as a rock at 3.7 ghz, 1.675v w/case cover off, with 2 games, or a game+prime running (max CPU temp is 56 C with 1 game running, though I have no idea what the REAL cpu temp actually is, cuz this board definitely reports CPU temps too high).

see my above post: at 1.4 ghz (!!) and 1.55v, my IDLE temps with hyper threading off (to match the older cpu's) are 42 C(!), load 45C.
So I suspect at least a 5-10C temp inaccuracy, as others with 1.4 ghz P4's and SLK heatsinks+tornado fans, were getting 32 C idle, 34 C load.

With case cover on, it's stable for awhile, but I eventually start getting errors or bluescreens if I try running 2 games at once, or game+prime. (1 prime, or 1 game is 100% stable at 3.7 ghz, 1.675v). Upper limit is about 3.75 ghz, 1.675v, case cover off, as long as only 1 cpu hogging process is run.

*REMEMBER this board undervolts under load, so my 1.675v is 1.62 under load*

I think this chip is just "bad", though it runs at its rated speed *Cough*. I'm really about to shell another $400 and get another chip. ANYTHING is better than this one. And no one has reported having a 3.4c that needs extra volts above 3.6v :(

BTW I'm using the newest BIOS for my IC7 max3. (Dated February).

I doubt its an IRQ issue, otherwise I would get crashes, regardless of voltage. And I only have a few things installed:

the ONLY cards are the Audigy2 zs,and the radeon 9800 pro, no other PCI slots are in use.