Win7 Hang while L2 filling up. Topic is solved

[qweqwe321]

And today I can't repeat test condition - it start to work flawless.
CrystalDiskMark 32GB test passed for 10 times.
How I would heroically fight the challenge if it start to work without bugs

It's seem to be normal for programs now-days to act like virus.
Marcium Reflect - create 3 services, some task in scheduler, automatically silently reenable and reinstall it's useless services then I try to disable/delete them. As some kind of virus it keeps 2 programs in ram - then I kill 1 process it's been relaunched by another (and so on).

[InquiringMind]

I'd second Jaga on Macrium - it's legitimate and widely used. If you want a "lean and clean" disk imaging utility, then I would suggest you take a look at Drive Snapshot - it installs no services, just a driver (and that only temporarily). It is commercial, but the trial download will provide 30-day backup and unlimited restore functionality. It also doesn't fill PrimoCache with backup data, unlike some other software.

Looking at your hardware Qweqwe321, the one thing that sticks out is that you are using 2 different types (and likely different specs) of RAM which could result in problems for memory intensive applications (like PrimoCache). You may wish to run a memory test benchmark to verify this - something like Memtest86+ or the "Large" Torture Test option of Prime95 (the latter also acts as a good test of your CPU and cooling).

PrimoCache itself works via a driver - rxfcv.sys (its predecessor - FancyCache - did run as a service). This shouldn't conflict with any widely used chipset - Windows would almost certainly have reported a driver name if it had.

(edit: Changed link to point to the GPLed Memtest86+ rather than the original, and now commercial, Memtest86 - either should do the job but the commercial version free download seems to be more feature-restricted. Apologies for any confusion).

[Jaga]

Looking at your hardware Qweqwe321, the one thing that sticks out is that you are using 2 different types (and likely different specs) of RAM which could result in problems for memory intensive applications (like PrimoCache). You may wish to run a memory test benchmark to verify this - something like Memtest86 or the "Large" Torture Test option of Prime95 (the latter also acts as a good test of your CPU and cooling).

I didn't read that deep, it's a good catch and point. Usually motherboards adjust RAM timing when they see this, but not always. And sometimes when RAM doesn't run at one of it's native speeds, it isn't happy.

You may want to do just what InquiringMind suggests - test the RAM for both benchmarking and reliability. And not the quick type of tests either - the ones that run overnight and use different patterns on the RAM.

[Support]

I agree with InquiringMind. Do a memtest first because bsod 0x00000013 appears very infrequently.

[qweqwe321]

10 minutes after my last post - I start to get same BSOD again. Just needed to slightly change test condition - instead of launching CrystalDiskMark - I copy 35 GB RAR archive from SDD to RAID0.

I tested with default BIOS settings with only one 8 GB Nanya memory stick in slot A2 - BSOD was still there.

OFFTOPIC
I prefer TESTMEM 0.12 with 1usmus preset.
Since I use this test for about 2 year I start to knowledge which memory test error is related to which RAM settings.

Memory testing is work by running some CPU intensive program, moving big amount of data (too exceed L3 CPU cache size). Thus, it's only test CPU-RAM path.

I have some interesting time catching bugs - then video driver got crashed after running Crysis3. TESTMEM show no error. But in the end it was related to - too low tRRDS; tRRDL; tWTRS; tWTRL timings. This was System RAM-> PCI-e -> GPU RAM path.

So any components which use DMA - can get affected by RAM if they use RAM heavy enough.


Yesterday I test condition - then all SATA cable's of RAID0 HDD are connected to PCIe extension card. Instead of BSOD at writing - I get error what RAID get disconnected and immediately Reconnected (5-10 seconds after), with same sound which Windows use when you insert/unplug USB flash drive.

Later, while write testing each disk separately with Victoria 5.28 - I reveal (or rather kill half dead) Maxtor500 GB HDD - one of it 2 heads is dying (NoramRead-SlowRead-NormalRead-SlowRead...).

So for now I see situation like this (BlackBox analyze):
1. One of RAID0 disk get too busy with doing self-test/smart update/remapping sector... and don't response it time
2. Windows detect it as hardware failure
3. If all SATA cable's connected to same PCI-e сard - RAID0 gets disconnected.
4. When faulty HDD done with remapping - it remove busy status and RAID0 gets reconnected
5. When SATA cable's connected to different hardware - PCIe card and motherboard, things get really messy (maybe conflicting drivers, or wrong error handling). And instead of disconnecting RAID - windows wait for some time 30-180 seconds and give BSOD. This waiting look similar for me - then you kill wrong svchost.exe - Windows will continue to work normally but will give you a message what it will reboot after 1 min.

And it's all started from my decision to get extra 16GB RAM.
I has removed MaxtorHDD 500GB, and my RAID0 now consist of 7 HDD.
No BSOD for now in CDM and copy big file from SSD to RAID. I will continue testing it for a while...

[InquiringMind]

...But in the end it was related to - too low tRRDS; tRRDL; tWTRS; tWTRL timings...

If you are making amendments to your UEFI/BIOS for these settings without knowing exactly what you are doing, then you are very likely to encounter memory problems whatever RAM you use. I would suggest you use a standard memory profile, test that heavily and only start playing with these options (and the RAM voltage settings - you can expect to have to increase voltage to get away with tighter timings) once you have a several hour (and preferably more) successful run on a memory test.

Yesterday I test condition - then all SATA cable's of RAID0 HDD are connected to PCIe extension card. Instead of BSOD at writing - I get error what RAID get disconnected and immediately Reconnected...

If memory is unreliable, then it can show up as problems elsewhere. You don't say if you are overclocking, but if you are - reverse those settings (in particular, *do not* overclock the PCI-E bus) and see if that fixes the problem. If it does, then you know that your overclocking is the problem.

[qweqwe321]

...after some time heroically fighting the challenge (0x0000001E BSOD)...
... After inserting my head inside PC case - I noticed what one of HDD generate SlideDown sound (Glissando from upper to lower note) continuing with short clicking sound (HeadParking?). Looks like that HDD was thinking what power was lost - and begin ShutDown procedure, after voltage drop became normal - HDD continues to work.
So in the end it was related to cheap SATA power multiplier connected to long IDE power cable. After rewiring with good SATA power cable - BSOD 0x0000001E and related erroneous behavior (sudden RAID speed drop) - gone.
My RAID0x8 became RAID0x4 BTW.

Anyway latest Primocache version - 4.1.0 - still hang whole system while filling L2 cache. For clean test purpose I lowered RAM speed from stable 2933MHz to 2133MHz.

Test Procedure is the same:
0. TEST big WinRAR archive
1. Then PrimoCache decide what system is idle
2. and start to read HDD contents to write it to L2 cache
3. It writes 4-10 GB to L2 cache (free L2 cache size goes down, L2Write parameters goes UP).
4. And Whole system Hang.

Looks like subprogram to fill L2 cache executed with maximum (RealTime priority), and does not allow any disk activity while it's working. After writing to L2 cache 4-10Gb it hangs (conflicting with itself or some lowlevel RealTime Windows disk process?).

If hit PauseCache button while it just start to fill L2 cache - I can avoid hanging of whole system.

Can you add some test option - limit L2 filling speed, or better:

Make L2 filling thread work for N milliseconds, then pause it for M milliseconds, and so on (Work,Pause,Work,Pause...). I hope while L2 filling is paused - Windows will do its disk activity.

[Support]

Sorry, I misunderstood. I thought this problem had been resolved after you unplugged the problematic HDD. Regarding the L2 filling, actually PrimoCache does this when it detects the system is idle. Even the filling has started, PrimoCache still monitors the system status. And if PrimoCache sees that the system is busy, it will slow down the filling according to the setting "Gather Interval" which is available in the Level-2 Advanced Settings dialog.
I wonder that this hang issue might be related to the level-2 device. When the device has problems, the write request to the device might get hang, thus PrimoCache also get hang. Is it possible to try another device as level-2?

[qweqwe321]

I must apologize for not responding it time.

At end of this story:
random BSOD 1E and 7E was related to HDD not responding in time, and PCI-e Sata controller decided to generate low level error (some Marvell chip 8sata port controller).

At first I was thinking it was related to bad PowerSupplyUnit with thin wires but now I think HDD just don't like each other then it's crouded.

I have ST32000542AS HDD which very power sensitive. If this HDD don't like something in it's power - it thinks what PC being shutdown and start emergency parking program (HDD engine sounds like glissando from upper to lower note).

So then multiple HDD (4 and more) connected to one power line - user must solder some capacitors for each HDD connector (1000uF electrolite + 5uF ceramic would be enough). Or rewire power cable - giving problematic HDD it's own "power line".
Usage of ferrite clamp bead is questionable.

Sorry for wasting your time.

[Support]

Thank you very much for your follow-up and share the results! I'm glad that you have found the root cause and the problem is finally resolved.