Sign in with
Sign up | Sign in
Your question
Solved

GA-Z68XP-UD4 build unstable

Last response: in Motherboards
Share
June 27, 2013 11:20:41 PM

Hi all. Experienced builder here looking for ideas on freezes/crashes from web browser.

Used parts combined with proven parts that I already own.

parts in question:
GA-Z68XP-UD4 v1.0 BIOS U1L
Patriot DDR3 1600 RAM
i5 2500k with CM Hyper 212 Evo (runs super cool)
running Windows 8 Pro 64bit.

All other components have been working in my Core2Q Q6600.
Corsair TX650v2, GA GeForce 560 Ti, LG SATA DVD-RW, Corsair Force GT 120GB SSD (5.05 fw), WD 1TB Black HDD.

NOT OC'd at this point, RAM running at 1333Mhz.

System has the occasional cold start reboot loop issue where it restores the BIOS to default after ~5 cycles. This is not the problem. I can attempt to live with that. Seems to be a common plague with Z68 boards.

The weird thing is that the system will crash when just browsing the web (IE or FF). It has never crashed while stress testing or gaming. Rock solid while under load. Multi-game DOTA games are regularly played and it is super solid.

Usually I have music playing via google music, but not always. When it crashes the sound glitches (if I have music playing) and the mouse and keyboard are completely unresponsive. Sometimes it just take 10 seconds then it restores itself, other times I have to restart it.

The RAM has been tested with memtest86 4.2 for 8 hours without an error.
The processor has been stressed with Intel Burn In Test for maximum stress/RAM usage for 10 iterations multiple times. Temps are in the mid 50s under load at stock clock.

I have tried using both the latest drivers from the Gigabyte web site as well as grabbing the latest drivers from component's respective manufacturer. No different outcome.

Ideas besides RMAing MB? (Which I am free to do, I have a Gigabyte RMA number in hand, I just want to avoid CAN to USA shipping and back).

My top possible answers:
Windows 8 unstable with driver's in use? Try Win 7?
Still possible that RAM is bad even after memtest ok?

I always have too many tabs open when it freezes. Can't tell if flash might be somehow crashing the whole system.

Suggested next steps? Or just bite the bullet and ship my motherboard off for RMA?

**Update: I forgot. No event logs are ever created for the problem. Only that an unscheduled shutdown occurred. Of course. :-(
a c 1581 V Motherboard
June 27, 2013 11:30:12 PM

Z68's sort of became poor stepchilds, between the P67 and the Z77, make sure you have the latest BIOS, and mobo drivers. In the BIOS enable XMP and select profile 1, should set your DRAM to 1600, next, raise VCCIO up+ 0.06 from current voltage, also your VCCSA should be at 0.925, thake that to 1.0 or 1.1 (closest it allows) and try that
m
0
l
June 28, 2013 7:37:00 AM

VCCIO = QPI/Vtt?


I have made these the settings:

XMP enabled. RAM running at 1.6V, 1600.
CPU LLP = enabled
CPU Vcore = 1.3V
QPI/Vtt = 1.12V
System agent voltage = 1.135V

CPU LLP left at 1.8V
Left at stock multiplier, etc.

These are the settings that I was using when I first built the system last week and started with a 43x OC. I used the i5 2500k guild at bit-tech. Once I hit a few snags I went back to stock to eliminate issues. They did not stop instability at 43x. It behaved the same then as now. We'll see how it goes.
m
0
l
Related resources
a c 1581 V Motherboard
June 28, 2013 3:55:42 PM

Yes, VCCIO = QPI/VTT, it's the memory controller voltage
m
0
l
June 28, 2013 9:37:08 PM

After running all day I thought that extra .05v for the ram solved it.
Crashed during dinner out with the family running google music player in ff and prime95 blend.

I guess I will have to source some different ram and try it out. If that doesn't fix, different mb. If that doesn't work, new CPU. Yuck.
m
0
l
a c 1581 V Motherboard
June 28, 2013 9:46:32 PM

Getting closer QPI/VTT to 1.15 and DRAM up +0.03
m
0
l
July 2, 2013 4:22:57 PM

I tried some alternate ram at 1333mhz. CPU at 4.3ghz. One of the workers died after 4 hours. Then put it to 4.2ghz. It was good for 2 days including 12 hour run of prime95 blend.

Then had cold boot cycle issue.

Can we safely say not ram? No mem test errors, two different sets unstable.

Or run original with small boosts as suggested?
m
0
l
a c 1581 V Motherboard
July 2, 2013 4:54:39 PM

I'd try the original as I suggested, but sounding like maybe a mobo/CPU problem, so yeah, give those changes a shot
m
0
l
July 2, 2013 9:36:25 PM

OK, trying with the following:

Multiplier: 42x
QPI/Vtt: 1.154V
System Agent: 1.135V
DRAM: 1.68V
SATA: AHCI

Everything else optimized defaults.
m
0
l
July 3, 2013 11:37:10 AM

Well it survived a 13 hour Prime95 blend run, then a 10 iteration maxium stress IntelBurn Test.

Highest core temp reported was 68. Usually they were in the mid 60s.

So everything looks good. As long as I don't get any more cold boot loop problems.

I've never had so many issues with a system, even running at stock. If I don't have any further issues my best guess is that that RAM just needs a little extra V push even when at 1333mhz.
m
0
l
a c 1581 V Motherboard
July 3, 2013 12:26:52 PM

Could well be, they brought those out, then almost immediately were into the Z77 and Ivy Bridge, so the 68's didn't get a lot of attention BIOS wise for updates or testing, think you can prob close this out
m
0
l
July 20, 2013 6:17:10 PM

So I continued to struggle along with this hardware. Idle crashes, boot loops. Then came to the conclusion that I have the methods right, but incomplete support in bios to get the results I want.

Although the uefi bios has been through a ton of revisions I figured that it still doesn't have LLC support. As with every latest bios on gigabyte's official pages it was listed as beta. So I took the steps to go back to the last non beta bios: F6 efi hybrid.

For the last few days the hardware has behaved like a champ.

I haved oc'd to a nice, comfortable 44x multi using both the offset method and the volt locked method. I seem to be getting cooler results with the static method, which you would think the the opposite of what I expected. Especially since right now my static volt profile has increased vtt and vccsa.

Whatever, I'm pretty happy as long as this proves stable. Then to reactivate the powersaving techs.

So here is how it looks now with static vcore specified:

Gigabyte Ga-z68xp-ud4 v1.0 F6
Multiplier: 44x
Real time os changes: enabled
LLC: level 5
CPU vcore: 1.27
QPI/vtt: 1.12
Sys agent: 1.135

No change to spd ram setting: 1.5v, 1333mhz.
All Sata controllers on Ahci
Hpet: 64 bit

This setup gets me a 1.308v on idle, 1.272 on full load. Prime95 blend has the highest reached core temp at 70C. I'll probably turn the vtt and sa voltages to stock and see if she'll still fly, if not i'll try setting at 1.1 each. I have all the power savers off right now. But they will be going back on when everything is proven solid.

Moral of the story: don't start with a damn beta bios! Use the production one!

I just wanted to thank tradesman1 for the assistance of a fellow builder. You are a gentleman and a scholar sir.
m
0
l

Best solution

a c 1581 V Motherboard
July 20, 2013 6:38:54 PM

Glad to see all is well, don't you just love builds like this, think of them as PITA, no not the bread but Pain In The .....think you get my drift ;)  Might want to close out the thread with the Pick as the Solution , Best Answer box...catch ya next time ;) 
Share
July 20, 2013 6:58:22 PM

With all of the power saving BIOS items on I now sit at 1.284V at idle and 1.284V at load.

No vdroop!

The processor is also scaling it's multiplier to the load nicely. I settles down to a 16x idle with an estimated 7W power usage at the processor. Temps 30C-32C at idle. Up to 70C under load.

It looks like I finally got this thing locked up.

We'll know when I get a 12hr prime95 run completed.
m
0
l
a c 1581 V Motherboard
July 20, 2013 7:00:35 PM

Have some fun with it first, Prime can run while YOU sleep ;) 
m
0
l
!