cancel
Showing results for 
Search instead for 
Did you mean: 

Blender:BLF_lang_init: 'locale' data path for translations not found, continuing

MartenBE
Level 7
Hi,

I am trying to stress test my self build system at stock settings. When I run ROG Realbench stresstest for 30 min at 4GB, the test completes succesfully.
However, the test always fails with "Instability detected" when I try to run the test for 30 min at 8GB.

The log shows the following output:


...

Handbrake: E
Handbrake: ncoding: task 1 of 1, 51.17 % (7.81 fps, avg 0.92 fps, ETA 00h23m24s)
Handbrake:

Handbrake: Encoding: task 1
Handbrake: of 1, 51.17 % (7.81 fps, avg 0.92 fps, ETA 00h23m24s)
Blender:BLF_lang_init: 'locale' data path for translations not found, continuing
Writing: C:\Users\Somebody\AppData\Local\Temp\blender.crash.txt


blender.crash.txt contains the following:


# Blender 2.68 (sub 0), Revision: 58537


I've tested the RAM by 2 passes of MemTest86 and it passed with no errors. Are there any other programs which can help me diagnose the problem?
Is it a true instability or does "Blender:BLF_lang_init: 'locale' data path for translations not found, continuing" point to a problem with the Blender software in the benchmark?

The full log is available here: https://dl.dropboxusercontent.com/u/5550696/rog-realbench-log.txt
My specs are as follows:

i7-6700K (fixed at 1.265V because auto settings give too much voltage)
ASUS VIII Maximus Ranger
16 GB RAM
Cryorig R1 Ultimate cooler

Thanks in advance!

Marten
12,600 Views
15 REPLIES 15

MMikhail
Level 9
I am getting the same issue as you.

Usually RB reports instability after 3 goods passes (and fails at ~25 minutes)
with the same Blender:BLF_lang_init: 'locale' data path for translations not found,
but sometimes can run for a longer period. 16GB test seems to be the most repeatable at failing,
8GB seems to be more solid.

It has driven me nuts for the passed few weeks, and I am clueless what to try next and where might the problem be.

Setup: Win10, Z170-A (R1 board, bios 3007), 6700K, Crucial 32GB/2666, STRIX1070-OC (default gaming mode), HX750i
simple cooler master air cooling for cpu, Fractal R5 case with plenty of air flow, 4.0/4.2GHz
NVidia driver 376.33, 1070 micron bios updated, and all other updates should be up to date

Tried with different ram configs, 16 ,24, 32 GB and all speeds, 2133, 2400, 2500, 2600, 2666
Underclocked CPU to 3.5 GHz, gave plenty of power 1.35V to VCore, added power to RAM: up to 1.220V

Under most test CPU core temps upto 80C, gtx at 70C, ram at 52C

LuxMark can stress test for hours on it's own, iGPU and GTX together
CPU only stress test fails after first run, but it is broken imho, since it fails on all system I tried.

Other possible problems that can relate to GTX problems
Can play games for a very long time, no issues, but TC Division has crashed on me 3 times.
Complete system freeze, twice after 4 hours, and once after 13 hours
One more similar freeze happened during MPCHC video playback minutes after starting.

Otherwise no other major issues, can run 24/7 for weeks without rebooting.
That includes running games, stress tests/benchmarks (almost all tried by now), video playback, encoding, what ever.


And finally a small on the side question: Is it normal for shadow effects to flicker in some games/tests?
Looks like they get rendered every other frame. Not complete shadows, just small areas near the edges.

Now I am puzzled by more questions then answers.... How does RB check for instability ?
If it's SHA-1, then I don't understand it completely. Since the blender image is generated the same all the time, no matter the test was successful or failed.
That makes me come to a conclusion that it's not blender failure, but something else. But the logs are not very verbose as to what is going wrong from their point of view and it would be nice to see in new RB update a little bit more verbose explanation as to what has been spotted, not just a fail report.

Any ideas ?

meankeys
Level 13
MartenBE

Can you please list your full system spec's. We then may be able to give you more precise answers

ty

Win10 1607 + all updates (10.0.14393.576 (Win10 RS1))
Asus Z170-A (Rev1 board, bios 3007)
6700K running 4.0/4.2GHz
Crucial Ballistic Sport 4x 8GB/2666 currently running at 2400/1.2V
Asus STRIX1070-OC (default gaming mode) micron bios update, NVidia driver 376.33
Corsair HX750i PSU
Crucial SSD
WDC 2TB Green HDDs
simple cooler master tower air cooling for cpu (Hyper 212)
Fractal Define R5 case with plenty of air flow
Razer Keys+Mouse

Anything else?

Latest nvidia drivers changes log/fixed bug...
Display driver stopped responding while running benchmark LuxBall HDR (Simple Bechmark:217K triangles).

Might be our case solved?

Today I hope I have some what found the problem are.
At least when I set LLC=5, VCCIO=1.10 V, VCCSA=1.10 V and remove cpu current limit my Z170/6700K system finally passed RealBench tests without fails on all test 4 , 8, 16 GB for one hour. But that's at stock turbo of 4.2GHz for all cores.
So why this options effect the system so much and in AUTO mode it fails even at absolute defaults?
And what are the expected and better values for this option for better stability and lower temps ?

meankeys
Level 13
With LLC set to 5 you basically added voltage to the vcore. open cpuz and note the voltage at idle. run RB and watch the voltage increase.

Played a bit more, and this is the best result I can get so far.
RB eventually crashed, not always, so I guess we are somewhere on the edge of stability.

LLC=5 seems to be a key requirement, otherwise we are asking for trouble.

Also interesting, if in UEFI we select ASUS optimized default settings, system passes RB test
but VCore is at hi 1.344 and jumps to 1.420 with temps at close 90. VCCIO and VCCSA are set to 1.26 in Auto!
LinX 0.6.5 is out of the question, burns hell.

Setting default energy saving config, brings VCore down to 1.264, it still jumps to 1.38+ at load
but RB tests fail quickly.

So, this are the results for the current custom setting...

62725
Image

And this are the UEFI settings...

Any suggestions would be highly appreciated. I am loosing faith in getting this rig working stable ever, and really depressed.

I have spent days testing it with RB, but it usually fails at 30 minutes and this is very time consuming.
Luckily, I have discovered LinX 0.6.5 and with a small 5 minute test I can now do a quick test.
If anything is terribly wrong it will fail very quickly. If it passes 5 mins test, then RB is next.


62727
62728
62729
62730
62731
62732
62733