screen artifacts on AMD 6790 HD

[Lynn McGuire]

I am getting screen artifacts on AMD 6790 HD
on a 16 GB windows 7 x64 pc. I tried the
latest AMD driver and things got way worse.
Anything to try here ?

I am running dual screens with a 27.5" and
a 19" samsung LCD. My motherboard is a
Gigabyte Z68XP-UD5 with a Intel I7-2600 cpu
and 160GB Intel SSD drive.

Thanks,
Lynn

 

[Primordialooze]

Have you made sure that the card is not running too hot? The Catalyst
Control Center will show you. You can also download & run GPU-Z.
http://www.techpowerup.com/downloads/SysInfo/GPU-Z/

Next step would to completely remove the Driver and Reinstall. Download
"Driver Sweeper" http://www.guru3d.com/category/driversweeper/ Uninstall
the Driver using Remove Programs in the System Folder. Once done Reboot
into Safe Mode. Run "Driver Sweeper". Reboot Windows and Reinstall the AMD
Drivers.
http://support.amd.com/us/gpudownload/windows/Pages/radeonaiw_vista64.aspx

Hopefully that will work for you.

 

[Flasherly]

I am getting screen artifacts on AMD 6790 HD
on a 16 GB windows 7 x64 pc. I tried the
latest AMD driver and things got way worse.
Anything to try here ?

I am running dual screens with a 27.5" and
a 19" samsung LCD. My motherboard is a
Gigabyte Z68XP-UD5 with a Intel I7-2600 cpu
and 160GB Intel SSD drive.

Thanks,
Lynn

Probably a combination of the Intel chipset's boost and visuals
coupled to Gigabyte performance settings. You'll have to take it down
at the hardware level for basic compliance and start adding the BIOS
switches until you can duplicate the problem for limits to prior
settings, comparison to possible driver update revisions, as well as
software compliance issues. Not necessarily easy to actually learn
the character of feature emphasis to a new system, provisionally
having researched first for suitably pairing components in known
stable arrays. If it feels like standing on the edge and
simultaneously bleating and bleeding, that is.

 

[Paul]

I am getting screen artifacts on AMD 6790 HD
on a 16 GB windows 7 x64 pc. I tried the
latest AMD driver and things got way worse.
Anything to try here ?

I am running dual screens with a 27.5" and
a 19" samsung LCD. My motherboard is a
Gigabyte Z68XP-UD5 with a Intel I7-2600 cpu
and 160GB Intel SSD drive.

Thanks,
Lynn

The cards are made by different video card manufacturers.

One I could spot on Newegg, the owners needed a new VESA
BIOS for the video card, if shipped before January.

If you have the model number of the video card, please post it.
Maybe there's something about that specific design.

*******

Sometimes the artifact type is important. Lines ? Blocks ?
"Colored noise" like snow on a TV set ? Colored noise is
caused by a DVI or HDMI traveling over too long a cable,
or by the data rate on the cable being too high for some
reason (like, cheap cable that came with monitor).

*******

The motherboard may have an adjustment for PCI Express clock,
but I don't think people these days, waste time adjusting that.
(It should just stay at the nominal 100MHz setting. Video cards
aren't usually starved for bus bandwidth, unless they're on an
x1 slot or something.)

Other than that, I don't know if there are too many settings
available on the motherboard, that influence the video card.

ftp://download.gigabyte.ru/manual/mb_manual_ga-z68xp-ud5_e.pdf

I looked at the motherboard block diagram, and it has bifurcation
logic ("switch") to choose 1x16 or 2x8 for the two video card slots.
That means, no matter which of the two main video card slots you go through,
eight of the lanes go through the bifurcation logic. I suppose it could be
a motherboard issue, like noise or errors in the switch solution, but
I also suspect PCI Express packets are protected by CRC, so at least
some of the errors would be picked up if that were the case (retransmitted).
Then the question would be, does Windows present such an error count if
available or not ?

You could try the card in the bottom-most x4 slot, but then there might not
be room for the cooler on the thing.

At this point, I'd have to assume it was a video card issue of some
sort, rather than motherboard. Or, if the power supply was weak,
or had excessive ripple, perhaps that is causing a problem on
the video card. The card appears to have one 2x3 PCI Express power
connector, and most of the current for the 12V rail should be
flowing through that, rather than through the 12V pins on
the PCI Express slot.

The max power draw for the card is listed as 150W, which is unlikely
with the one 2x3 power connector showing on the end of the card.
It looks like Xbitlabs is no longer measuring these cards properly,
so that's one source of good info no longer available.

The 6790 was introduced about a year ago, and you'd think drivers would
be sorted out by now. That leaves a bad video card BIOS (like, selecting
wrong clocks or something, for the GPU). Or, the video card could just
plain be defective.

You could boot a Linux LiveCD, and see if there were still artifacts.
Not much of the card logic would get used by the default driver, but
if there was potentially a hardware problem, you might still see it under
Linux (like, if there was colored snow on the screen indicating a DVI
or HDMI problem, the colored snow would still be there).

Paul

 

[Lynn McGuire]

The cards are made by different video card manufacturers.

One I could spot on Newegg, the owners needed a new VESA
BIOS for the video card, if shipped before January.

If you have the model number of the video card, please post it.
Maybe there's something about that specific design.

*******

Sometimes the artifact type is important. Lines ? Blocks ?
"Colored noise" like snow on a TV set ? Colored noise is
caused by a DVI or HDMI traveling over too long a cable,
or by the data rate on the cable being too high for some
reason (like, cheap cable that came with monitor).

*******

The motherboard may have an adjustment for PCI Express clock,
but I don't think people these days, waste time adjusting that.
(It should just stay at the nominal 100MHz setting. Video cards
aren't usually starved for bus bandwidth, unless they're on an
x1 slot or something.)

Other than that, I don't know if there are too many settings
available on the motherboard, that influence the video card.

ftp://download.gigabyte.ru/manual/mb_manual_ga-z68xp-ud5_e.pdf

I looked at the motherboard block diagram, and it has bifurcation
logic ("switch") to choose 1x16 or 2x8 for the two video card slots.
That means, no matter which of the two main video card slots you go through,
eight of the lanes go through the bifurcation logic. I suppose it could be
a motherboard issue, like noise or errors in the switch solution, but
I also suspect PCI Express packets are protected by CRC, so at least
some of the errors would be picked up if that were the case (retransmitted).
Then the question would be, does Windows present such an error count if
available or not ?

You could try the card in the bottom-most x4 slot, but then there might not
be room for the cooler on the thing.

At this point, I'd have to assume it was a video card issue of some
sort, rather than motherboard. Or, if the power supply was weak,
or had excessive ripple, perhaps that is causing a problem on
the video card. The card appears to have one 2x3 PCI Express power
connector, and most of the current for the 12V rail should be
flowing through that, rather than through the 12V pins on
the PCI Express slot.

The max power draw for the card is listed as 150W, which is unlikely
with the one 2x3 power connector showing on the end of the card.
It looks like Xbitlabs is no longer measuring these cards properly,
so that's one source of good info no longer available.

The 6790 was introduced about a year ago, and you'd think drivers would
be sorted out by now. That leaves a bad video card BIOS (like, selecting
wrong clocks or something, for the GPU). Or, the video card could just
plain be defective.

You could boot a Linux LiveCD, and see if there were still artifacts.
Not much of the card logic would get used by the default driver, but
if there was potentially a hardware problem, you might still see it under
Linux (like, if there was colored snow on the screen indicating a DVI
or HDMI problem, the colored snow would still be there).

Paul

Mostly horizontal lines.

I tried the 2011 driver and the 2012 driver. The
2011 driver has less artifacts but both drivers
are horribly annoying.

Here is the info from my video card:
Graphics Card Manufacturer Powered by AMD
Graphics Chipset AMD Radeon HD 6700 Series
Device ID 673E
Vendor 1002
Subsystem ID 3170
Subsystem Vendor ID 1682
Graphics Bus Capability PCI Express 2.0
Maximum Bus Setting PCI Express 2.0 x8
BIOS Version 013.012.000.009
BIOS Part Number 113-679XZDF50-110719
BIOS Date 2011/08/14
Memory Size 1024 MB
Memory Type GDDR5
Core Clock in MHz 840 MHz
Memory Clock in MHz 1050 MHz
Total Memory Bandwidth in GByte/s 134.4 GByte/s

I do not see a way to update the video BIOS.

There is only a single pair of slots that the video
card fits in and it is in the middle of the PCI
slots.

I am buying a AMD 6850 card to replace it:
http://www.amazon.com/Sapphire-Radeon-6850-Video-100315L/dp/B0047ZGIUK/

Thanks,
Lynn

 

[Paul]

Mostly horizontal lines.

I tried the 2011 driver and the 2012 driver. The
2011 driver has less artifacts but both drivers
are horribly annoying.

Here is the info from my video card:
Graphics Card Manufacturer Powered by AMD
Graphics Chipset AMD Radeon HD 6700 Series
Device ID 673E
Vendor 1002
Subsystem ID 3170
Subsystem Vendor ID 1682
Graphics Bus Capability PCI Express 2.0
Maximum Bus Setting PCI Express 2.0 x8
BIOS Version 013.012.000.009
BIOS Part Number 113-679XZDF50-110719
BIOS Date 2011/08/14
Memory Size 1024 MB
Memory Type GDDR5
Core Clock in MHz 840 MHz
Memory Clock in MHz 1050 MHz
Total Memory Bandwidth in GByte/s 134.4 GByte/s

I do not see a way to update the video BIOS.

There is only a single pair of slots that the video
card fits in and it is in the middle of the PCI
slots.

I am buying a AMD 6850 card to replace it:
http://www.amazon.com/Sapphire-Radeon-6850-Video-100315L/dp/B0047ZGIUK/

Thanks,
Lynn

This is a picture of the output from an HD6790 under Linux.

https://bugs.freedesktop.org/attachment.cgi?id=49463

( http://comments.gmane.org/gmane.comp.freedesktop.xorg.drivers.ati/19824 )

If the card is "AMD" brand, it could be contract manufactured by Sapphire.
But the warranty should be handled by AMD, if any warranty is left on it.
For computer hardware, warranties can begin from date of manufacture,
or from date of purchase. If date of purchase, it's vital to have a receipt
to present, to prove when you bought it. Also, some video card manufacturers
insist you "register" with them immediately, in order for the warranty to
be valid, which is presumably another "date of purchase" type warranty.

I'm not having any more luck, tracing down any details on this. Search engines
these days suck, being perfectly willing to bury you in garbage hits, than to
admit a series of search terms have no hits.

The picture in that Linux forum, kinda reminds me of uninitialized memory.
Perhaps the GPU was supposed to flush that memory area, when it locked up.

The video card BIOS, contains clock settings, but it also contains memory
timing information. For example, when I flashed a Macintosh video card, for
usage in a PC, I needed to find a BIOS file with matching memory characteristics
(something like CAS4). There was a web site, with a ton of archived BIOS files.
I picked one, flashed my card, and used that card for gaming for three or four
years. Worked perfectly. So back then, the deal was, the BIOS file had to match
the memory type and speed, soldered to the card. A BIOS file from
one brand of card, wouldn't necessarily lead to error free operation of some
other brand of card - they could be using different memory, with different timing
requirements. I expect even the fancy GDDR5, still has timing to deal with.

Paul

 

[Lynn McGuire]

This is a picture of the output from an HD6790 under Linux.

https://bugs.freedesktop.org/attachment.cgi?id=49463

( http://comments.gmane.org/gmane.comp.freedesktop.xorg.drivers.ati/19824 )

If the card is "AMD" brand, it could be contract manufactured by Sapphire.
But the warranty should be handled by AMD, if any warranty is left on it.
For computer hardware, warranties can begin from date of manufacture,
or from date of purchase. If date of purchase, it's vital to have a receipt
to present, to prove when you bought it. Also, some video card manufacturers
insist you "register" with them immediately, in order for the warranty to
be valid, which is presumably another "date of purchase" type warranty.

I'm not having any more luck, tracing down any details on this. Search engines
these days suck, being perfectly willing to bury you in garbage hits, than to
admit a series of search terms have no hits.

The picture in that Linux forum, kinda reminds me of uninitialized memory.
Perhaps the GPU was supposed to flush that memory area, when it locked up.

The video card BIOS, contains clock settings, but it also contains memory
timing information. For example, when I flashed a Macintosh video card, for
usage in a PC, I needed to find a BIOS file with matching memory characteristics
(something like CAS4). There was a web site, with a ton of archived BIOS files.
I picked one, flashed my card, and used that card for gaming for three or four
years. Worked perfectly. So back then, the deal was, the BIOS file had to match
the memory type and speed, soldered to the card. A BIOS file from
one brand of card, wouldn't necessarily lead to error free operation of some
other brand of card - they could be using different memory, with different timing
requirements. I expect even the fancy GDDR5, still has timing to deal with.

Paul

I changed out to a AMD 6850 video card. All
is well now. The AMD 6790 is just not working
with my pc for some weird reason.

Lynn