PDA

View Full Version : HD3850 black screen of death


nsx
06-08-2008, 02:13 AM
I get frequent black screen of death* with my Gigabyte HD3850 ultra durable. The frequency of this happening is dependent on how high AA/AF levels are set. I can play Racedriver Grid for hours while only cat 8.5 are installed (no ccc or att), everything at default. That time reduces dramatically if I max out all the settings. I.e 8xAA/16AF. Also, after a black SOD, AA level goes down to 0x and cannot be altered using CCC until drivers are reinstalled. I could set the level to whatever I liked with no effect.
*A black screen of death = monitor loses signal, sound loops, vpu recovery displayed. With 8.0-8.4 cats it'd die without even displaying the recovery msg.
Not once have I seen a bluescreen, which is frustrating because that would atleast give me a hint as to whats going on.
Specs:
Asus p5k P35 <-any known issues with this and 3850? Bios options?
Apacer cheapo 800 mhz ddr2 dual interlaced 2x1024MB (no memtest errors)
19" lcd, vga converted
2xIDE drives, 1 USB drive (makes no difference if turned off)
550w chieftec
E4400

Maximum temps:
CPU-60, Gpu diode-65, HDD1/2-45, mobo-40 - so that's not it
voltages are stable, ranging 11.894-12.08V @12v idle-load
there is significant vdroop on the cpu core, but is not the cause

Other notes:
1)Tried the card on another p5k system which was heavily overclocked, but stable using 8800gt. Had the same symptoms with my card in it. Still, this test isn't conclusive towards the card being faulty, as it may be some freak bios option continually escaping my attention.
2)The "ultra durable"cooling solution on the card is slightly stupid. It has a zalman cooler but no heatsinks on vram. Under full load the chips feel hot to the touch, but as that's too subjective and I have no laser thermometer, nothing can be concluded from this. Specially because I am yet to find one post on the net blaming black screens of death on vram temps.
Can I hear you say RMA? :banghead: Any ideas/suggestions?

OzzmanFloyd120
06-08-2008, 02:16 AM
Are you getting any artifacts before the black screen?

dark2099
06-08-2008, 02:16 AM
Sounds like either you have a faulty card or a bad PSU.

tkpenalty
06-08-2008, 02:17 AM
Issues of overheating bare ram chips are very common fyi, and yes memory overheating causes this (personal experience on both X1950PRO and 8800GT when running memory bare with low airflow)- with AMD's cards, they go into VPU recover, on the other hand Nvidia's GPUs just lock up and the audio loops. If the memory chips are hot enough that you cant keep your finger on it, at idle, its way too hot, I would invest in some ramsinks from zalman (they aren't expensive, less than 10 bucks).

BIOS options would not cause such problems...

Note: Generally heat will cause VRAM to crash and not artifact, artifacting VRAM is due to physical damage or a lack of voltage.

nsx
06-09-2008, 05:42 PM
Got this reply from AMD support

We have responded to your issue.
Solution:

Hello

Thank you for contacting AMD Customer Care for ATI products.

Based on our experience, the error "The drivers for the display adapter got stuck in an INFINITE LOOP" is a result of a system communication problem. On some systems, the error "Unable to complete a drawing operation", or an ATI VPU recover error may also occur.

This error does not occur on all system configurations.

This issue can occur using any motherboard or graphics chipset along with the Windows XP Operating system. It appears that in most cases this problem is due to a BIOS setting, specifically options affecting the way the RAM, CPU and graphics card communicate with XP and DirectX. Improper timing settings can result in system lockups or random system reboots.

This issue also affects different makes and models of video cards from various manufacturers.

In an attempt to gather more information on why this error occurs, ATI developed a component to catch when the INFINITE LOOP occurs. This component is called VPU Recover. If the display driver detects that the graphics processor has hung, VPU Recover will attempt to reset the graphics processor, eliminating the need for a system reboot.

VPU errors can occur mostly while using 3D applications and games. This error has also been known to occur in 2D applications as well.

The following suggestions may help to correct this issue:

* Update the display drivers. Ensure that you have installed the latest ATI CatalystT software for your ATI product. Drivers for your video card can be found at http://ati.amd.com/support/driver.html

* Ensure that all three DirectX Features are enabled. Run the available tests to ensure that each of the three components pass. To start the directX utility, type dxdiag.exe at the run prompt.

* Update your motherboard chipset driver, AGP, GART or PCI-Express software. Updating your AGP or PCI-Express chipset drivers can add stability to the PC and may help to prevent the Infinite Loop error from occurring.

* Use the ATI SmartGartT utility to adjust AGP Reads/Writes. In newer revisions of the CatalystT Drivers, a feature called SmartGartT was introduced. With the latest ATI CatalystT software installed, you can use the ATI SmartGartT settings to adjust the Fast Writes and AGP BUS speed. This feature is only available with AGP based cards. To use the SmartGartT utility, start the CatalystT Control Center in ADVANCED MODE and go to the SmartGartT section. Ensure that both the AGP Read and Write settings are On. If these are set to OFF, then enable them and restart Windows as prompted. As a test you should try disabling the AGP Write setting. For stability, the "Fastwrites" feature should be turned OFF.

NOTE: AGP Writes may not be enabled on all chipsets.

* Update the motherboard BIOS. Any updates can be found on the motherboard manufacturer's website.

* Make sure you have an adequate power supply.

Other Recommended Hardware Troubleshooting:

* Test system with a single memory module if multiples are being used.
* Test with different memory modules.
* Test with a different Power supply.
* Disable/Remove ALL non-essential hardware i.e. sound card, network card, additional HDD etc.
* Test the video card in a different PC.


Regards,
AMD Customer Care for ATI Products
http://ati.amd.com/

Most of it is default dell-style blabber, but do look at the underlined text. Let's assume for a minute that it isn't the bare VRAM in my case. Is one to deduce from this that all 3x00 aTI cards are undocumentedly and randomly bugged using the XP operating system? And also that my troubles might go away if I "upgraded" to Vista? I already RMA'd the card and bought an x1550 as replacement for the duration, not a single crash, I was even able to play COD4 and GRID @800x600 low :)

eidairaman1
06-09-2008, 05:54 PM
ok NSX, Refer your Problems to Gigabyte.

eidairaman1
06-09-2008, 05:55 PM
another point, if the card is being problematic in another machine when properly installed then the card is bad.