| video card failing? I am using an old Radeon X800 All in wonder AGP card and I am noticing a few things.
I am not seeing any POST screens anymore, no video output until the XP boot splash.
Once and awhile I see an 'out of range' error or my lcd screen before the XP boot screen.
OpenGL seems to crash my system and make tons of weird lines on my display.
Direct3D is starting to act choppy.
These video issues did not happen in the past.
Could raising the AGP voltage correct the problem?
And yes the power supply is still good.
This system is just an old spare parts box that I use. It's an LGA 775 3.20ghz p4, 2GB DDR2, 160GB IDE HDD, XP Pro SP3 32bit. |
|
| said by robman50:ICould raising the AGP voltage correct the problem?
Unlikely, though you can't lose that much by experimenting with an old box.
--
Wacky Races 2012! |
|
| On my old socket 478 celeron it ran more stable with a slightly raised DDR voltage.
Oh well if the raising AGP voltage kills the card I can all ways toss in my PCI Express X800.  |
|
| Card is running at AGP 8x on AGP 3.0 mode maybe I should just leave the voltage setting alone on 'Auto'. Aperture is set up 128MB and Fast Writes are enabled. |
|
 koitsuPremium,MVM
join:2002-07-16
Mountain View, CA
kudos:19 | reply to robman50
The "out of range" message you see is your LCD indicating the horizontal or vertical frequency clock rate, which are used to select a video mode (resolution) -- not that these are not the same thing as refresh rate! -- is too high or too low for the LCD to display.
This is probably explainable because some older (circa past 10-15 years) BIOS screens tend to use a unique video mode that some LCDs might not like (I believe the resolution is something strange like 720x400 -- you're asking me to go back to 1992 and remember custom H/V VGA frequency register details  ).
The other problems sound more like driver bugs (as in, OpenGL features that may not be implemented in ATI's driver, or may be implemented badly/buggy). The D3D performance "acting choppy" is inconclusive; too many things could cause this (ranging from driver bugs to some power-saving features of chipsets to the VGA card actually failing in some way).
This might sound crazy, but: are you sure the hard disk is in good health/condition (please provide screenshot of HD Tune Pro's Health stab)? Reallocated LBAs could result in a driver having N number of bytes as zeros instead of code, and could cause oddities once the driver is loaded (some code attempts to get executed but it now contains zeros -- behaviour will vary depending on what the code does or if its actually just a data table/data section).
Do not screw with the AGP voltage -- period. I don't know why people resort to thinking "lets play with voltage!!!" when there is some anomaly. Don't. Just don't.
AGP aperature size has absolutely nothing to do with this issue.
As for AGP fast writes -- always turn them off. They are known to cause problems with some chipsets.
If you want to test your video card for bad on-board memory, I can direct you to a very specific Windows application that can test this, and it needs to be configured in a very specific way (different than out-of-the-box) for it to work correctly (otherwise it'll spew tens of thousands of errors). I believe the program should only be used in Windows 2K or XP, but it may work under Vista or 7 (though possibly oddly since the program uses DirectX surfaces for testing, and Vista/7's entire UI is one big DirectX surface).
--
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer. |
|
| said by koitsu:This is probably explainable because some older (circa past 10-15 years) BIOS screens tend to use a unique video mode that some LCDs might not like (I believe the resolution is something strange like 720x400 -- you're asking me to go back to 1992 and remember custom H/V VGA frequency register details ).
Well this motherboard is not that old. I think it might be around 2006. Could it be something to do with having the LCD screen on the DVI port instead of the VGA port? |
|
Reviews:
 ·Frontier Communi..
 ·WildBlue
| reply to robman50
yes the card might be on the way out.
These old cards still have two separate pipelines for 2d and 3d. If the 2d pipeline is marginal you can see what you are getting.
This can also be caused by the ramdac not working correctly anymore so it is not outputting a usable signal during boot.
However you mentioned you are now using an LCD use a regular CRT monitor if you have one and see if that displays video during post if it does the issue is actually your lcd not liking the signal being sent hence the "out of range" message you are getting.
If the lcd has multiple inputs try one of the other ones if your video card supports it.
My recommendation would be to swap it with a newer x1600 to 4000 series video card. You can get a new one here on newegg. »www.newegg.com/Product/ProductLi···20series
If your mainboard has pcie slots go with one of those instead such as a 4870 or so. |
|
| said by Aranarth:yes the card might be on the way out.
I don't know if it's possible but can I run it until it quits all together? Benchmarks maybe? |
|
|
|
| reply to koitsu
said by koitsu:If you want to test your video card for bad on-board memory, I can direct you to a very specific Windows application that can test this, and it needs to be configured in a very specific way (different than out-of-the-box) for it to work correctly (otherwise it'll spew tens of thousands of errors). I believe the program should only be used in Windows 2K or XP, but it may work under Vista or 7 (though possibly oddly since the program uses DirectX surfaces for testing, and Vista/7's entire UI is one big DirectX surface).
Sure it wouldn't hurt. I have tried to run Vista and 7 on this machine but wow that was painful. LOL |
|
koitsuPremium,MVM
join:2002-07-16
Mountain View, CA
kudos:19 | reply to robman50
said by robman50:said by koitsu:This is probably explainable because some older (circa past 10-15 years) BIOS screens tend to use a unique video mode that some LCDs might not like (I believe the resolution is something strange like 720x400 -- you're asking me to go back to 1992 and remember custom H/V VGA frequency register details ).
Well this motherboard is not that old. I think it might be around 2006. Could it be something to do with having the LCD screen on the DVI port instead of the VGA port? Unlikely. In fact, I would think the behaviour you describe would be significantly *more* likely when using the VGA port.
The root cause of the message you see from the LCD screen, however, is what I described in my previous paragraph. Some monitors (LCD or CRT; doesn't matter which) have different sensitivities to frequency changes, so some will show that message (or similar messages like "out of sync" or "no sync") for 3-4 full seconds, others will only take 1 second, etc... It just varies.
--
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer. |
|
koitsuPremium,MVM
join:2002-07-16
Mountain View, CA
kudos:19 | reply to robman50
»mikelab.kiev.ua/index_en.php?pag···S/vmt_en
Run this (specifically vmt.exe -- you don't need to use the .bat files), and make the following adjustments and do not change anything else please:
* Uncheck: "File log"
* Uncheck: "Ignore colors bit mask"
* Check: "Use onscreen buffer"
* Make sure test type is "DirectX"
* After clicking "Start", you will be prompted to change your video mode to 640x480x16. Answer ***NO*** to this question.
* While the test is running, DO NOT DO ANYTHING WITH THE PC. Do not move any windows around, do not foreground/background any applications, etc..
* You are welcome to use something like GPU-Z at the same time (make sure the two application windows do not overlap however) if you wish to check GPU temps, etc. (if available on your card) -- there will be no conflict. Make sure to run GPU-Z first, then afterward, launch Video Memory Tester.
* If using GPU-Z at the same time, under the Sensors tab, make sure to CHECK "Continue refreshing this screen while GPU-Z is in the background"
Finally: let all the tests run. Even if the tests get 30% through and there are no errors, DO NOT interrupt it. There are many tests which can pass with flying colours (pun intended) yet latter tests which fail (such as walking-ones tests). You need to let at least 1 pass complete.
Error counts and anomalies are shown somewhat near the top of the window under the "Satistics" section (yes, "Satistics" -- author needs to fix that typo).
--
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer. |
|
| reply to robman50
said by robman50:On my old socket 478 celeron it ran more stable with a slightly raised DDR voltage.
Raising voltage for a memory module is very different from raising the voltage of a whole subsystem.
If your card is failing the 2 most likely component types are either the caps (yes, it could have bad caps), or some chip that runs hot.
A close visual inspection may be enough to spot bad caps, if that's the case.
If you haven't done it yet, you may want to make sure that the heatsink(s) are not clogged with dust bunnies, and there is sufficient airflow.
But as others already mentioned, some of the issues may be display related.
--
Wacky Races 2012! |
|
| The GPU fan failed a long time ago and now it is running with an Vantec Iceberg addon GPU fan. |
|
| reply to robman50
Should I pull the video card and post some pictures of it? |
|
| Can't hurt. Make sure you have good focus. |
|
| reply to koitsu
How long will vmt run for? 30 min? an hour? |
|
koitsuPremium,MVM
join:2002-07-16
Mountain View, CA
kudos:19 | You only need to let it run for a single pass, or as many passes as you want. The longer the better of course, but usually in the case of bad cards or bad RAM (I speak from experience) it happens usually within the first pass.
Things that can cause these kinds of failures:
* Bad RAM tends to result in failures at repeatable memory locations (either individual locations or large batches (in the case one module or one section of a module is bad)),
* Power/voltage-related issues, such as things like this,
* A card overheating / not getting proper cooling (troubleshooting this is difficult on older cards since most of them at that time did not provide on-board thermistors nor offer DTSes),
* Actual transistor damage inside of the GPU (caused by overheating or other kinds of issues)
The final thing you need to do is figure out if it's the card that's bad or the AGP port (or some other circuitry/chip on the motherboard). That's easy to do: replace the card with another (same model would be best, but if not pick something that does 3D or has similar specs + power draw + etc).
I will admit I have seen AGP busses go bad (I'm looking at you VIA, you bastards!), but it's very very rare.
--
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer. |
|
| said by koitsu:I will admit I have seen AGP busses go bad (I'm looking at you VIA, you bastards!), but it's very very rare.
This motherboard has the VIA PT880 Pro chipset. :S |
|
koitsuPremium,MVM
join:2002-07-16
Mountain View, CA
kudos:19 | That doesn't really mean anything at this point in the troubleshooting process.
I should be clear though: I absolutely have a stick up my ass when it comes to VIA. I've dealt with so many silicon-level and driver-level bugs over the years WRT their crap that I'm beyond biased. Their CPUs are supposedly not half bad (for little embedded CPUs anyway; at least until the Atom came out), but the rest of their stuff should be melted down. I go as far as to avoid any of their chips on motherboards I buy (including their sound chips, USB chips, and Firewire chips). So yes, I am completely opinionated when it comes to VIA, but it's based on experience. |
|
| reply to koitsu
Which RAM are you talking about? Video card or system?
Since this system is just a box of old spare parts I will not be surprised if something goes bad soon.
In the past I have all ways wondered if the motherboard was failing because once and awhile something very strange happens and then the problem is gone and everything is fine again. Could that be possible? |
|
