Passes Memtest86 v2.11 fails IBT

[BTRY B 529th FA BN]

I know Memtest isn't exactly perfect but what about failing IBT?

 

[n7]

Memtest86+ pretty much tests the memory only, though it does find major mobo issues too.
It doesn't find minor mobo issues though.

It's a basic test IMO; HCI Memtest will find errors that it does not.

IBT stresses the CPU heavily, but also the RAM, & somewhat the motherboard.

I suspect you've gotta a really high FSB quad in this picture, right?
Basically, your FSB or NB isn't stable, more than likely, or you need more vcore.

It's easy to test this via lowering the multiplier.
If it's still unstable, you know it's not CPU.

If you can do Large FFTs for hours & hours, mobo is generally fine, though a final check w/ IBT or LinX is always good.

Small FFTs for CPU.

HCI really thrashes RAM & NB.

 

[BTRY B 529th FA BN]

Thanks again N7

I tested at 400fsb max stress with IBT and it failed off the bat. Something might be on its last legs but I don't know what. The time for my UD3R is closing in to being built.

 

[BTRY B 529th FA BN]

I just reset the cmos/bios all to default settings, ran IBT and it failed again.

Y error reports
1 max stress
5 itterations

=
--------
failed

 

[BTRY B 529th FA BN]

I just thought of something. If super PI and the Hyper PI were failing and now IBT fails - all at stock - could that mean a burnt out chip? I was running it at 1.43v

 

[Elixer]

Perhaps. I have seen that happen before.
Have you tried OCCT yet?

 

[BTRY B 529th FA BN]

Going to try HCI memtest & then OCCT - which OCCT test should I run?

note: i swapped out my memory sticks with my other pair of DDR2 1000 and same results

EDIT: my last option is to get a new cpu which I already had thoughts of coming from a not so good mix of an Asus MIIF board and a quad. I think a dual on this MIIF board will bring 500+ fsb results better than this quad.

 

[n7]

What is stock that's failing?
Stock CPU + RAM @ DDR2-1000?

 

[BTRY B 529th FA BN]

Everything. cpu ram all bios settings... it'll pass 1/2 stress test aka "2" but not "1" max stress

EDIT: this is with 4gb of ram; I might try sticking in the other two sticks i have and see if it fails at max. which could possibly lead back to the app using too much ram for the OS to operate. im using vista 64

 

[n7]

What RAM speed?
What RAM vDIMM?

And what happens with max stress? Error? BSOD? App locks up?

Honestly, i know you know what you are doing, but in your posts, there's a lot of info lacking, so it makes it kinda hard to figure out what you are asking.

 

[BTRY B 529th FA BN]

DDR2 1000 - it can function at 467 with 1.9v but at 1000 it needs 2.1v.

The error I get is a bunch of zeros within the app. I don't blue screen, lock up etc.

I ran 3+hrs of OCCT without an error...

let me try the full 8Gb and see if it'll make 1 run ***at max stress***

 

[n7]

OCCT (assuming you aren't running the LinPack test within OCCT) is a joke for testing RAM, especially for 8 GB.

Use HCI Memtest like i said.

Set four instances to utilize nearly all your RAM.

So say 4x 500 MB...run it till 500%+ (you'll wanna do this overnight).
Or for 8 GB i do 4x 1500 MB usually.


And why would you be @ 467?

I know how much of a PITA getting a quad stable @ 440+ FSB is.
I could write a novel with the amount of different settings i've tested with mine at higher FSBs, & the funny thing is, stability becomes ridiculously finicky.
Memtest86+ will pass, but HCI Memtest will error.
IBT/LinX will pass, but P95 will error.
Or everything will pass, but UT3 will freeze up.

It's a nightmare with 8 GB...

Anyway, if you are just testing your RAM only, do it at stock or lower than stock multiplier for now; please stay away from high FSBs, as that will add another possibly of error into your testing.

Do 375 FSB 3:4 (DDR2-1000), which uses 400 strap.

I say that specifically, because if you do 333 2:3 (DDR2-1000), you'll be forced to use the 266 strap, which isn't as ideal for 333 MHz FSB CPUs in terms of stability IMO.

Or do 417 FSB 5:6 (DDR2-1001), which uses 333 strap. (assuming you have known stable board voltages for that FSB)

If that can do HCI Memtest stably, your RAM ain't the issue.

I don't think it is anyway, i'm just trying to help you isolate the RAM as not being the issue.

 

[BTRY B 529th FA BN]

Windows reports 8Gb, CPUz reports 8Gb, IBT; selecting '1 max stress' asks for 5173mb, 1 iteration = well, it didn't fail so i'm expecting it to pass. Its a huge step for me even though it hasn't finished. Pass using 8Gb on Vista U 64

That's it, its the app requesting too much ram. Again, my system - 4Gb, Vistal Ultimate 64, failed max stress test IBT due to requesting too much RAM

EDIT: Looks like we uncovered an issue using IBT with 4Gb or Ram on Vista 64 U

I will try the HCI memtest which is probably a good idea. I backed off my fsb to run 4.2Ghz and am back to a 9 multi. 480 I was running 1.43vcore @ 4080Mhz; with this I'm back to 1.39v

 

[n7]

Oh, & i'd suggest LinX...much better interface that IBT.
http://www.xtremesystems.org/f...howthread.php?t=201670

See my screenie here with LinX: http://ense7en.com/pics/q9650/...%20passes%20STABLE.JPG

-You can set it to stop on error (instead of running it 50 times like IBT only to find out it's unstable).
-You can set precise memory amounts to test.
I usually use 6144 MB, as Vista & idling appz take up a bunch, so if i do 6144 MB, it leaves about 500-1000 MB free.

Run at least 10 passes with LinX if you want to consider it half stable.
I've had the app freeze, or errors, or BSODs, hours into testing.
I consider 50 passes good, as i've even had errors on pass 34 before.
That + P95 Large for at least 8 hours.

HCI Memtest should run overnight till at least 500%+...you'll hate it though with 8 GB.
Memtest86+ will tell you everything is great, & you'll have HCI erroring all over the place.

I can run Memtest86+ @ 1100 6-5-5, but HCI doesn't even consider DDR2-1050 stable with 8 GB

 

[BTRY B 529th FA BN]

WOW, lol, yeah, i can imagine 50 passes, thats like atleast 12hrs with 8Gb - OW! hehe...
500% sounds like a marathon too... gonna have to consider the electric bill
I've tried doing the LinX thing but I don't have WinZip for that to open up the files. What can I do?

 

[n7]

WinRAR trial?
7-Zip (it's free).

As for power bill, lol...you're overclocking your quad heavily instead of getting a low power CPU & undervolting but cannot afford to ensure the system is stable? :laugh:

Right...

 

[BTRY B 529th FA BN]

LoL

Ok, I do have WinRAR I can reinstall and see what happens; do i need Winzip also?

 

[RebateMonger]

IBT has been designed to intentionally overheat things, while that's only a secondary effect of Memtest. Chemical and electronic processes will behave differently at different temperatures.

 

[Yellowbeard]

Purely out of curiosity, does the rig do what you built it to do when you are OCed? As you noted, you may simply have found a software bug. But, over the years I have seen people getting hung up on some sort of stress test or benchmark when the system is actually operating perfectly for its intended purpose.

 

[BTRY B 529th FA BN]

Originally posted by: Yellowbeard
Purely out of curiosity, does the rig do what you built it to do when you are OCed? As you noted, you may simply have found a software bug. But, over the years I have seen people getting hung up on some sort of stress test or benchmark when the system is actually operating perfectly for its intended purpose.

It does, but what raises an eye brow is that i was having trouble running super PI and Hyper PI so strike 2 does not get overlooked and things need attention... so that's why I was trying other tests

 

[Yellowbeard]

Originally posted by: BTRY B 529th FA BN

Originally posted by: Yellowbeard
Purely out of curiosity, does the rig do what you built it to do when you are OCed? As you noted, you may simply have found a software bug. But, over the years I have seen people getting hung up on some sort of stress test or benchmark when the system is actually operating perfectly for its intended purpose.

It does, but what raises an eye brow is that i was having trouble running super PI and Hyper PI so strike 2 does not get overlooked and things need attention... so that's why I was trying other tests

That's logical. I was just curious also if there was any rhyme or reason or pattern to the failures that may indicate a software issue vs your hardware or OC being the culprit.

 

[BTRY B 529th FA BN]

I think I brought the same topic up in XS forums and a member mentioned that running the test also would always fail max stress test and that the alternative was to use 1/2 stress test and he or she commented that the app might be requesting more RAM than it can use. There might be memory reserved for the GPU or other apps. The error that would come up while I was using 4Gb with Vista U. 64 would a long string of zeros right off the bat. I never got a screen shot of it, but I did get a screen shot with 8Gb or RAM passing 1 iteration in the same OC settings (467fsb * 9multi)