Major issue: H61N-USB3 with HDMI HD3000 i3 2105 : Crashes...

[jaredkipe]

So I've had intermittent screen tearing with this computer. I had a couple freezes as I was getting it setup but thought I had them narrowed down and started doing a bunch of server related stuff including transferring a lot of data around over the network.

Well I started having intermittent 'freezes' where the computer would clearly quit working (like activity monitor would show that terminal is not responding while rsyncing) but you could move your mouse but no windows would respond.

Remember passes ram, and cpu check was fine after a pretty sustained 15min at 100% on all cores.

Various CoreImage /Quartz things can and will cause tearing and other buffer related events. Like the one pictured.

Also, after crashes I sometimes have fun stuff like this second one in my Console complaining of *** Debug infer for apparent hang in Main graphics engine ***

It is somewhat frustratingly difficult to reproduce.
I haven't completely ruled out north bridge overheating, but the cpu is certainly fine.
If I was using a normal graphics card I would say no brainer, graphics card or MAYBE motherboard. But with the CPU integrated graphics card, I'm just not sure.

One last thing, I've messed around with default and other various settings in the BIOS but I can't think of anything that would cause this. (384MB VRAM for my 4gb ram)

 

[jaredkipe]

I am no longer so sure there is a hardware problem. I have run pretty extensive burn-in in Windows 7 and have not managed to get it to crash. I can sometimes get some screen tearing/glitches but they generally go away by moving the window.

Anything I should check or download in windows? (like DSDT or frame buffer related)

Anyone know any core graphics or core animation stress tests I can loop?

 

[jaredkipe]

I am hopeful that attempting a DSDT edit like this is working.

viewtopic.php?f=15&t=42418

Oddly, this is at least SIMILAR to the DSDT edit I made earlier to enable HDMI audio out. viewtopic.php?f=162&t=31948

EDIT--------------------
Nevermind. I got a crash. With the same odd output spammed to the kernel.log



Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: stampWait: Overflowed checking for stamp 0x4e79d on Main ring: called from
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: timestamp = 0x4e796
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: **** Debug info for apparent hang in Main graphics engine ****
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: ring head = 0xbee02a94, wrap count = 0x5f7
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: ring tail = 0x00003660 ring control = 0x00003001 enabled, auto report disabled, not waiting, semaphore not waiting, length = 0x004 4KB pages
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: timestamps = 0x4e796
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: Semaphore register values:
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: VRSYNC: (0x12044) = 0x4e796
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: BRSYNC: (0x22040) = 0x0
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: RVSYNC: (0x 2040) = 0x0
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: BVSYNC: (0x22044) = 0x0
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: RBSYNC: (0x 2044) = 0x0
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: VBSYNC: (0x12040) = 0x0
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: kIPEHR: 0x7a000003
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: kINSTDONE: 0xeffeffff
Jan 22 18:34:28 Locals-Mac-Pro kernel[0]: kINSTDONE_1: 0xbbfffffd
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: stampWait: Overflowed checking for stamp 0x4e7b4 on Main ring: called from
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: timestamp = 0x4e796
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: **** Debug info for apparent hang in Main graphics engine ****
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: ring head = 0xbee02a94, wrap count = 0x5f7
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: ring tail = 0x00003a18 ring control = 0x00003001 enabled, auto report disabled, not waiting, semaphore not waiting, length = 0x004 4KB pages
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: timestamps = 0x4e796
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: Semaphore register values:
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: VRSYNC: (0x12044) = 0x4e796
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: BRSYNC: (0x22040) = 0x0
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: RVSYNC: (0x 2040) = 0x0
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: BVSYNC: (0x22044) = 0x0
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: RBSYNC: (0x 2044) = 0x0
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: VBSYNC: (0x12040) = 0x0
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: kIPEHR: 0x7a000003
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: kINSTDONE: 0xeffeffff
Jan 22 18:34:33 Locals-Mac-Pro kernel[0]: kINSTDONE_1: 0xbbfffffd
Jan 22 18:34:38 Locals-Mac-Pro kernel[0]: stampWait: Overflowed checking for stamp 0x4e7a1 on Main ring: called from
Jan 22 18:34:38 Locals-Mac-Pro kernel[0]: timestamp = 0x4e796
Jan 22 18:34:38 Locals-Mac-Pro kernel[0]: **** Debug info for apparent hang in Main graphics engine ****
Jan 22 18:34:38 Locals-Mac-Pro kernel[0]: ring head = 0xbee02a94, wrap count = 0x5f7
Jan 22 18:34:38 Locals-Mac-Pro kernel[0]: ring tail = 0x00003a68 ring control = 0x00003001 enabled, auto report disabled, not waiting, semaphore not waiting, length = 0x004 4KB pages


Goes on like that every second or so for a long time. I think I may have some simple steps to reproduce..

 

[jaredkipe]

I have finally been able to reproduce the problem. And more importantly reproduce it on a reasonable time scale <10 min from boot.

The way I reproduce it is basically as follows:
Open safari and spam open a variety of things that will take a few minutes to finish (when downloading several of them. Open 10 (max in the download list) and be prepared to start downloading more as they finish. I've been using Cinebench and mac osx client updates as well as some downloads from a local computer I setup to share a large file over http.

Anyway, open the downloads drop down item, and behold 10 items racing to redraw their progress as quickly as possible. For me this causes a fair amount of graphics glitches and blinking.

Optionally, open various instances of Cinebench and start them all on GPU related tasks.
(I've gotten it to hang or kp without using Cinebench, only with the downloads, but it is my opinion having more windows and other Quartz/ Core Graphics drawing helps)

Eventually the computer will either freeze, but allow you to drag the mouse around still. Or KP (at least once, but no log was written indicating the KP)

I have ruled out NorthBridge overheating using this method. I will admit that this does seem like it could be a network or SATA/HDD problem (well SDD, but I've tried other drives too) The spamming of the "**** Debug info for apparent hang in Main graphics engine ****" to kernel.log and the graphics tearing does seem to indicate a graphics component.

I will start testing other network drivers and dsdt edits.

 

[jaredkipe]

http://www.youtube.com/watch?v=p49ENDX8q0Q

I made an application that uses a background thread to draw a bunch of constantly updating semi transparent status indicator bars and text fields. That application does NOT seem to have the same problems as apple's built in download manager when downloading a lot of things.

If you watch the video you'll see 2 instances of my application, one that is in focus and inactive after finishing a test loop (I thought I'd let the pretty blue lines go for a while) and one that is chugging away (grey).
You'll also see the 'blinking' of the download manager. I left it downloading like that for 10 min or so and it froze.

Anyone want to run a similar test on their HD3000 enabled computers??

 

[jaredkipe]

I have been able to recreate the problem with the stock DSDT.aml from this fine website, and the one I edited for HDMI audio. The stock one SEEMED to take longer to crash, a lot longer... but it still crashed in exactly the same way.

I can recreated it with the stock 2.0.6for10.7 network drivers AND Lnxman's(sp?) network driver.

I've restored defaults and done other bios stuff, to no avail. I could REALLY really use some help with this one...

EDIT: Oh and fresh installs each time for the following combos (3rdPartyATA and MacMiniSMBIOS being the only other things selected and the SMBIOS is required and the 3rdPartyATA is not the cause, unless not having it will also cause it):
EditedDSDT and stock network
EditedDSDT and lnxman's network
StockDSDT and lnxman's network

 

[jaredkipe]

After unsuccessfully trying to recreate the issue on a friend's extremely similar build, I can conclude it to be one of the following.

#1, graphics ram. Friend had 8GB in there and thus 512 dedicated to gnu. I will pull some ram from my main machine to bring it up to 8GB and test.

#2, hardware problem as originally suspected (CPU or MLB most likely)

#3, random difference between setup/drivers.

 

[netcastle]

I am installing the same system today. I was wondering if you could post your precise setup.

Also, I would be happy to try and re-produce your problem (assuming I don't have it upon first install).

 

[jaredkipe]

I really thought it was fixed with 8GB RAM. A few oddities, the 8GB sticks were correctly registered as 1333MHz in About this Mac, previous RAM is reported as 1600MHz despite being setup as 1333MHz in BIOS.

BIOS would not allow me to select above 480MB VRAM as an option. So this is what I chose, MacOS still reports it as 512MB VRAM.

It seemed to be fine after quite some time of running my ... rather interesting test. So I did a sanity check and put back in the first 4GB of RAM but left the VRAM at 480MB in BIOS. Mac OS reports it once again as 384MB RAM. So either MacOS is wrong or the settings in BIOS don't matter.
I got it to crash like this.

After putting in the 8GB of RAM I was smiling to myself figuring I had just solved the mysterious problem. But after an hour or more of testing, I got it to crash with 8GB.

So I'm on to hardware:

------------------------------------------------
Thank you very much for your support and reply.
Here is my hardware.
H61N-USB3 F3 bios
i3 2105 (HD3000)
4GB RAM
random HDD's but will final install on SSD.

Thats it, HDMI out to TV.
My current working installation is to install 10.7.2 combo updater (from a USB Stick)
Install UserDSDT in MultiBeast, system utilities (permissions and cache) with 3rd party SATA, and Lnxman's network. Use the Macmini5,1.smbios.

The DSDT.aml I was using was the one from this site, but I (just now) have gone back to one I edit for HDMI audio out.

 

[jaredkipe]

My current testing process.

Phase 1:
Open http://www.maxon.net/downloads/cinebench.html
http://support.apple.com/kb/DL1459
and
http://support.apple.com/kb/DL1461

up in new tabs to prepare.

Download 3x 10.7.2clients, 3x 10.7.2servers, and 4x Cinebenches.

This and end at the top of the Cinebench page. (it has a graphic of a cars that I use to visually see when it crashes really quickly) Open the dropdown/sheet that lets you see all 10 downloads going at once.

You can basically leave it at this point. I usually pull up console and sometimes Activity monitor. Don't let it go to sleep obviously. And I have disabled the.

It May crash during Phase1.

Phase 2:
You enter phase 2 if one of your cenibenches finishes downloading. Queue up another Cinebench as they finish. Then open the cinebench that just finished and get it doing the GPU tests. After each one go back into Safari with the Cinebench cars graphic and your downloads menu open.

Each Download that finished queue up a new one. Keep 10 going in your little drop down.
Each cinebench that finishes, open it up (as a new instance). And make it do the graphics test.

I've never had my computer make it through Phase 2.
Usually Cinebench will launch fine, and maybe run the test and go fine. But if I get it back in Safari with the new downloads going, it has always crashed at some point.