I get inconsistent unRar checksum errors

[soccerballtux]

how do I determine the source of the error?

-passes 2 passes of Windows Memory Diagnostic (Advanced)
-one pass of Memtest+
-can Prime95 for 3 hours
-20 pass of IntelBurnTest (Standard)

and yet if I unRar a large archive I will get an error on one or several files in particular; reboot and they'll show up on a different file during unpack. Consistently on the different file though. Until I reboot, then maybe it passes without trouble.

similarly, uTorrent tossed a bunch of data due to hashfails, and would switch between 99.8 and 99.9% completed when I would force re-checks. At this point I copied it to another location for tests--for trying to complete the download, it kept downloading the same parts over and over and tossing them out. Or, one time I rebooted, forced a re-check, and the whole thing passes the check and says "complete". Since it's gone to "complete", I can't get it to fail.

rig in my sig.

powering down, waiting 20 seconds, and powering back up seems to 'improve' things. Doesn't seem to like it when I make BIOS changes to the CPU voltage, save/reboot, and go straight to Windows. If I do this 10 times in a row I can be guaranteed random program crashes all over the place, even with like...super high voltage 1.42v at 4.3ghz...(stock what the CPU came with)...until I power it down, let it sit 20 seconds, and power it back up. It's like there's a ghost of bad data that remains until power-off if the voltage has hypothetically been too low and caused data to be miscalculated.

 

[VirtualLarry]

Easy. Your rig is unstable. Test without the overclock.

 

[soccerballtux]

Easy. Your rig is unstable. Test without the overclock.

I forgot to say that if I return all settings to stock and reboot, I can get the same issues. I need a way to consistently replicate the issue/cause failure.

 

[tynopik]

maybe it's the drive?

 

[escrow4]

Something is failing, either RAM or HDD/SSD. Set everything to stock and thoroughly test both - run Memtest for 24hrs at least, and check the SSD. Don't forget torrents hammer the drive(s) they are on, I'd just use a separate HDD or SSD to stop thrashing.

 

[VirtualLarry]

I forgot to say that if I return all settings to stock and reboot, I can get the same issues.

In that case... that is troubling. I would try running HCI Memtest at 1000% coverage.

I wouldn't think it's the SSD, but possibly someone has some ideas to test it?

Edit: Have you applied the 1.01 firmware update for your SSD? It came out fairly recently.

 

[soccerballtux]

In that case... that is troubling. I would try running HCI Memtest at 1000% coverage.

I wouldn't think it's the SSD, but possibly someone has some ideas to test it?

Edit: Have you applied the 1.01 firmware update for your SSD? It came out fairly recently.

ooo hey maybe that's for me!

MANDATORY UPDATE (@#$&'ers)

Fixed a corner case issue with DDR corruption on 480GB capacity drives
Improved robustness of uncorrectable error handling
Improved read retry on bad block list
Numerous stability and reliability improvements

I literally installed it 2 weeks before this update was released.

how do you 'improve' read retry on bad blocks...sounds fishy... if (retry < 8 ); // 3); try increasing ?

 

[Idontcare]

-passes 2 passes of Windows Memory Diagnostic (Advanced)
-one pass of Memtest+
-can Prime95 for 3 hours
-20 pass of IntelBurnTest (Standard)

By "Memtest+" did you mean Memtest86+? Memtest86+ doesn't catch certain memory errors that Memtest can catch.

I recommend you use Memtest Ver 4.1 for memory testing. Run it within windows, and you can run multiple instances (1 per core or thread) to speedup the testing if you like.

Also, when you run IntelBurnTest are you setting the ram to "all physical" or maximum? If you don't use all the ram while running IBT then you can still have bad memory but it won't create an error in IBT.

 

[soccerballtux]

can I just say that Asus is a piece of crap?

when I wake the PC with a WakeOnLAN magic packet, either LLC temporarily turns to disabled (all I have to go on currently is CPU-z's voltage reading although this is tempting me to see if I can find a test point on the board itself) or the voltage drops by ~0.05v from what it should be, and then begins wandering all over the place, 1.392v->1.35v, back up, and drooping to 1.308v under load.

If you're outside the retailer return window, how do you claim warranty on something with crummy software? Between

• this WOL thing and
• having to seemly power off the PC and count to 20 before powering on and continuing testing after either 'encountering a Prime95 error', or 'changing the CPU voltage', and
• locking up sometimes after the Asus VRM Temp throttling code engages (clocks down the CPU and lowers its voltage) and sometimes before, but never always one or the other and never consistently at all

...leaves me very skeptical of Asus and I'd rather just dust my hands of it and try another brand.

Asus is one of the few manufacturers that goes to the trouble of maintaining separate code branches for their different product lineups-- most use the same branch and just disable features. Would prefer the latter as that would mean more testing for code bits like the thermal throttling thing, which is a cool feature! just terribly unstable

 

[soccerballtux]

By "Memtest+" did you mean Memtest86+? Memtest86+ doesn't catch certain memory errors that Memtest can catch.

I recommend you use Memtest Ver 4.1 for memory testing. Run it within windows, and you can run multiple instances (1 per core or thread) to speedup the testing if you like.

Also, when you run IntelBurnTest are you setting the ram to "all physical" or maximum? If you don't use all the ram while running IBT then you can still have bad memory but it won't create an error in IBT.

it was actually the MemTest.org 5.01 version, and yes I was scratching my head when I found there were 3 versions to choose from including the PassMark version, and you can't use the multi-threaded setting in one of them because it falsely reports errors sometimes.

Thank you for the stamp of approval on that one though, I feel like it's getting harder to find 'known-goods' these days...

I only did it on standard->1024MB as had already tried Memtest, but yes overlapping the tests could be useful. Will try Maximum now.

 

[soccerballtux]

AND OH MY GOSH NOW IT'S DOING THIS

this was after

1. send wake on lan packet
2. boot to windows
3. reboot to BIOS
4. exit without saving changes (saving changes does not 'enable' the LLC again)
5. Boot to windows
6. run Prime95 for a few seconds
7. stop Prime95. Now voltage is 1.58-1.6V!!!!!! Not ok!!!!!
8. Open one tab of Chrome. CPU temp hits 72C and my fan pegs to 100% in response

voltage was set to manual--1.418750 in BIOS, LLC Enabled. I think that 1.596v is a little bit more than 1.418750v

hm I should have clicked validate that would have been good proof.

 

[redzo]

-one pass of Memtest+

I would suggest to let memtest86+ do his thing for a few hours at least. Yeah, sometimes memtest86+ does not catch any errors, but most if the time it does its job right and sometimes it needs to run for hours in order to identify bad ram. Just fire it up in the evening, check for no red horizontal lines in the morning!

 

[ShintaiDK]

May be CPU degrade due to the overclock if it even fails at stock now. Else could be storage related.

 

[mikeymikec]

Memory testers that run within Windows - surely then they can't test the entirety of the memory?

 

[soccerballtux]

I returned the motherboard to Microcenter after filing a report with Asus. I believe a significant source of the instabilities I had over the past month or two were because I was using WakeOnLAN, which was causing haywire behavior in the BIOS. Now that I have something consistent to work with I've been able to narrow it down a bit.

 

[redzo]

Memory testers that run within Windows - surely then they can't test the entirety of the memory?

Yup. They can't access all of it.

This is straight out of memtest's f.a.q.:

...
To conclude, one successful pass of memtest will give you a pretty good idea
that your memory is ok, only in rare cases will there be errors showing
after the first pass. To be sure though simply have the test run overnight
or even for a couple of days depending on the level of importance of the system.
...

 

[soccerballtux]

Yup. They can't access all of it.

This is straight out of memtest's f.a.q.:

it's better than nothing, and beats staring at a blue screen of boredom

 

[redzo]

Of course that Win RAM testers are better than nothing. Sometimes they get the job done. Funny thing is that the OS itself may go BSOD while that windows ram tester is running.

 

[soccerballtux]

Of course that Win RAM testers are better than nothing. Sometimes they get the job done. Funny thing is that the OS itself may go BSOD while that windows ram tester is running.

I can deal with that. I can't deal with blue screen of boredom

 

[flexy]

>>
-passes 2 passes of Windows Memory Diagnostic (Advanced)
-one pass of Memtest+
-can Prime95 for 3 hours
>>

I strongly suspect a SATA issue, maybe old SATA cable. (Check windows event log whether it says something there)

But cannot hurt if you test using OCCT "large" FFTs as well, although the above suggests to me that the system is at least "semi-stable" and that memory SHOULD be alright.

 

[flexy]

May be CPU degrade due to the overclock if it even fails at stock now. Else could be storage related.

Who says his CPU fails? He said he runs prime etc. without problems....

Edit:

Memtest86++ can give you an idea whether your memory is defective or not but it's not really a good "stability test" for memory and timings. For that you should run HCI memtest from within windows. Or OCCT "large" works ok too for a good, first idea. (Also Intel XTU *benchmark* maybe 10 runs for a quick idea works too)

Anyway for debugging what's going on I would of course at first take out ALL overclocking from your system and see whether those unRAR errors still appear. (For me it's highly unlikely he'd pass prime, memtest etc. but RAR has a problem and throwing wrong checksums. This sounds like something different to me, problem with SSD or SATA interface maybe....)

 

[tweakboy]

Easy. Your rig is unstable. Test without the overclock.

Play with vcore and multi and bus .........


Whats your specs,,so we can help better.

 

[soccerballtux]

I don't know if I can replicate those errors.
I'll try more.

I have a different motherboard now and I'm still getting hard locks seemingly randomly. They're not entirely hard though...I cant.explain it it's like the Windows.scheduler is just bogged down with too much.


Maybe it's the SSD or power supply. I feel like I should be able to find a CPU problem in some testing or something if it were CPU related.

yes. it has to be the SSD. I will image it to a platter drive and continue from there. considering there was already one mandatory firmware update for stability purposes...I am skeptical to believe the first update fixed it.

this sort of stuff is exactly why I didn't want an SSD. even at $40 this definitely hasn't been worth the effort.

 

[soccerballtux]

wait that doesn't explain the RAR problems...unless it was unzipping to the SSD....hmm yes it could be unzipping to temp folder on SSD

the SATA cables have been swapped with the new ones from the new motherboard

 

[soccerballtux]

I will remove SSD from equation next.I bet you that fixes it