Ati card ID problem

[APOLLO]

I purchased a second hand Ati HD 3850 earlier this year and can't seem to install drivers for this card. When I attempt to install drivers whether it's from the CD or recent drivers from AMD's site, the installation aborts because it can't identify the card. I posted a thread on the Ati forum a while ago but never received a response. Has anyone experienced this kind of problem?

I have an identical card in another system running F@H that does not present any problems with identification but this card does. The problem appears in either system yet the 'good' card is recognized no matter what system it's in. When I swap this card to the system with drivers already installed for the other card, the F@H GPU client won't start because it can't recognize the card.

 

[extide]

Try reflashing the bios on the card. If the 2 cards are identical you can extract the bios from the good one and then put it on the bad one. Check out AtiWinFlash

 

[APOLLO]

OK, thanks for the information. I thought it might be a BIOS related issue. Maybe the original owner flashed the BIOS on the card to something unofficial like an OC BIOS or something, causing the drivers not to recognize it. I'll try to extract the BIOS from the operating card once it completes the current WU and attempt to flash the other card with it.

Another thing I noticed about both cards is the small differences that could indicate different revisions. The 'good' card has silver main cooler screws and black smaller screws, while the other card is reversed with the four main screws being black and the secondary cooler screws are silver. Moreover, the 'good' card seems to have an additional microchip that is missing from the PCB on the other card. Don't know what it's for. So, the cards although largely the same are definitely not identical, but their designations are exactly the same.

 

[Zero82z]

Could be different card revisions.

 

[APOLLO]

I'm a bit uncertain how to proceed with this utility. Can I save my video card BIOS in Windows or do I need to exit and save in DOS? The utility doesn't seem to work when I click on the save button.

 

[sirmonkey1985]

as long as the clocks, gpu, and shaders are the same.. it doesnt matter what manufacture made the card.. they should work together.. have you tried swapping the cards in the slots? possible the one not being recognized has a higher clock speed then the good one.. which you would want the faster one in slot 1 and the slower one in slot 2..

 

[APOLLO]

These are AGP cards and they are identical brand and model, just some minor differences mentioned further up. I have two AGP systems with one card purchased new a year ago working fine, and a second card purchased used earlier this year for another AGP system. The used card is not recognized by the drivers and therefore cannot be installed in any system. I tried 3 different AGP systems that I have here, so something is up with the card...

I got the AtiFlash utility to save the BIOS of the good card, but the problem with the other card is that I can't get above 640x480 resolution, and the flash utility requires a minimum of 800x600, I believe to work.

 

[[H]ugh_Freak]

Use Atiflash to flash the bios of the bad card. You'll need a DOS startup disk to boot to though. Instructions are in the zip file.

 

[Zero82z]

Is the card still under warranty? If it is, I would try and get it replaced before you start messing around with a BIOS flash.

 

[APOLLO]

Is the card still under warranty? If it is, I would try and get it replaced before you start messing around with a BIOS flash.

I bought the card used for about half the retail price and don't know if any warranty is left on it. I'm going to try the flash later this evening. The way I see it, it's a paperweight either way and might as well try something. If anyone has suggestions I'm open to anything at this point.

 

[extide]

I'd try the flash. Might be a little worried that the PCB's arent exactly the same, but still do the flash. I mean, it wont void the warranty if you hit it with a bad flash, so either RMA it now or try and fix it and then maybe RMA it later.

 

[Cheeptech]

I've had the same issue with a few AGP cards from ATI we sold to customers at work. Try the link here
http://support.amd.com/us/kbarticles/Pages/CatalystAGPHotfix.aspx

It has worked perfectly on 2 different 4650 AGP cards from Diamond. I can't understand why they cant just put the right drivers on the disc they give you, instead of the PCI-E drivers that go with every other card they sell. Crappy support as far as I'm concerned

 

[APOLLO]

I've had the same issue with a few AGP cards from ATI we sold to customers at work. Try the link here
http://support.amd.com/us/kbarticles/Pages/CatalystAGPHotfix.aspx

It has worked perfectly on 2 different 4650 AGP cards from Diamond. I can't understand why they cant just put the right drivers on the disc they give you, instead of the PCI-E drivers that go with every other card they sell. Crappy support as far as I'm concerned

Hi, thanks for providing the fix. It actually worked and I'm glad the card is operating. No need for a risky BIOS flash at this point.

The problem now is the card runs sluggish with delays and very slow window draws. Worst of all, the GPU client doesn't work. I get a message: F@H has run into a serious error running the core and will shut down. Is the hotfix compatible with the F@H GPU client? I'm going to try the card in another machine and see if it works better.

 

[ThreeDee]

maybe pull the heatsink off of the GPU and get the numbers off of the chip to make sure the card is what you think it is .. if it is .. I'd reflash it to factory default bios and that might get your card to be recognized without a hotfix and be able to run like it's supposed to ..

..maybe

 

[Cheeptech]

the hot fix will be needed regardless. The drivers ATI has listed for the newer series cards (2xxx and up) all seem to be only PCI-E. Since his card is AGP, he will need the hotfix driver for it to properly ID his card. Pulling the heatsink off will also allow to check to make sure the thermal paste is in good shape, and reapply as needed. But if it's running sluggish, and IS being ID'd correctly, there may be a reason someone was getting rid of that card.

 

[APOLLO]

the hot fix will be needed regardless. The drivers ATI has listed for the newer series cards (2xxx and up) all seem to be only PCI-E. Since his card is AGP, he will need the hotfix driver for it to properly ID his card. Pulling the heatsink off will also allow to check to make sure the thermal paste is in good shape, and reapply as needed. But if it's running sluggish, and IS being ID'd correctly, there may be a reason someone was getting rid of that card.

I was thinking the same thing. The only other reason the card could be running sluggishly is that older drivers from previous video hardware aren't fully or properly removed, and could be causing conflicts. It has happened to me on this system in the past. This is an old installation of XP...

maybe pull the heatsink off of the GPU and get the numbers off of the chip to make sure the card is what you think it is .. if it is ..

Later on today, I'm going to install the card in another system with a fresh copy of XP. That will determine if the card is faulty or not. Removing the HSF will certainly reveal the exact GPU model, however, if it wasn't a 3000-series card, the hotfix probably wouldn't install at all. There are different hotfixes for each series (2000, 3000, 4000). Lastly, the hotfix is the only driver that works so far. Even after installing the hotfix, other catalyst drivers will still refuse to recognize the card. So, it's either the hotfix alone or nothing.

I'd reflash it to factory default bios and that might get your card to be recognized without a hotfix and be able to run like it's supposed to ..

..maybe

Flashing the BIOS is still on the table as an option of last resort. I don't see why this card has problems and another one nearly identical to it purchased last year worked with the drivers on the CD from the box. I even tried installing those CD drivers for the other card and no difference. Something is very strange about this card.

 

[Cheeptech]

I was kind of hinting at that with the last sentence
It may just be buggy/bad. A clean OS install with clean driver install would help rule the software out of the equation.

 

[APOLLO]

I was kind of hinting at that with the last sentence
It may just be buggy/bad. A clean OS install with clean driver install would help rule the software out of the equation.

OK.. I pretty much eliminated most of the possibilities. The second system I tried the card with had a fresh OS and driver installation, but as soon as I reboot after installing the drivers I get a BSOD. So, that option is out.

Back to the first system with the old XP installation. I wiped out all old drivers using a driver cleaner and then reinstalled the hotfix. The stalls and sluggish window draws I was experiencing earlier seem to be gone, but F@H still errors. I get two error messages. In the client window, there's a 'client-core communication error' and I also receive an external message box stating again: F@H has run into a serious error running the core and will shut down.

I removed the OEM cooler and there's really nothing that indicates what generation the GPU could be. However, the card is identical to the other HD3850 I have with the exception of a small microchip and screw colors that can be accountable for a revision change. There's also decals stating it's a HD3850 and I do not have any doubts now.

Anyone has final suggestions? A flash is probably the only thing I haven't tried. I'm just wondering if there is something else I need to install to get this card running the GPU client...? It seems to be operating well otherwise...

 

[APOLLO]

I believe I have it working - finally. Before attempting a flash, I made one last try at getting the card to work. I uninstalled the drivers and then rebooted in safe mode and used driver cleaner to remove any remaining files. I rebooted into Windows, redownloaded the hotfix, reinstalled the drivers, downloaded and reinstalled the GPU client and it has been working fine for about 12 hours. I'm not sure what the problem was, but it could have been something as simple as a corrupt client installation, LOL.

 

[[H]ugh_Freak]

Good to hear you got it working (at least so far).

 

[Cheeptech]

Glad to hear ! keep on Folding

 

[sirmonkey1985]

I believe I have it working - finally. Before attempting a flash, I made one last try at getting the card to work. I uninstalled the drivers and then rebooted in safe mode and used driver cleaner to remove any remaining files. I rebooted into Windows, redownloaded the hotfix, reinstalled the drivers, downloaded and reinstalled the GPU client and it has been working fine for about 12 hours. I'm not sure what the problem was, but it could have been something as simple as a corrupt client installation, LOL.

damn you have way to much patience.. if it was me i would of already thrown that card across the room by now..

glad to see ya got it all working..

 

[APOLLO]

Good to hear you got it working (at least so far).

Thanks, it's still up and running but I have a few unrelated questions, which are at the bottom of this post...

Glad to hear ! keep on Folding

Thanks, if you hadn't mentioned the hotfix, I would likely have flashed the BIOS or just chucked the card. Your help is very appreciated.

damn you have way to much patience.. if it was me i would of already thrown that card across the room by now..

glad to see ya got it all working..

Thank you and yes, my patience is legendary among people who know me personally. I have what some people call an immortal patience, LOL. Trust me, on several occasions I had the temptation of determining the aerodynamic efficiency of a flat, rectangular-shaped object in an enclosed space...
_______________________________________________________________

On to another question/issue I have. This card is processing WUs significantly slower than my other near-identical card. The other card is in a P4 system running @3.4GHz. It is a far more recent 8X AGP system. The card in question is installed in a dual AMD Barton@2.3GHz 4X AGP system. The AMD system is producing only 1500-2000 PPD depending on WU, but the other card that never had a problem is producing 2000-2500 PPD in the P4 system.

I did not swap cards again after getting it working, so don't know if this may be caused by the card itself or the system it's installed in. If it's actually the card, it will be processing WUs slower whatever system it's installed in, and I don't intend to swap and test it just yet.

These cards are OC equally at 810MHz core, but the slower card does not show any increase in performance beyond a certain frequency, whereas the better card continues to increase performance that scales with frequency. Is this due to the specs of the AGP bus being 4X compared to 8X? I was under the impression that bus speed doesn't usually make much of a difference. If not bus spec, what other reason could there be for a performance ceiling??

 

[Zero82z]

These cards are OC equally at 810MHz core, but the slower card does not show any increase in performance beyond a certain frequency, whereas the better card continues to increase performance that scales with frequency. Is this due to the specs of the AGP bus being 4X compared to 8X? I was under the impression that bus speed doesn't usually make much of a difference. If not bus spec, what other reason could there be for a performance ceiling??

AGP 4x has about 1GB/s of bandwidth. PCI-E 4x has about 2GB/s, and is roughly equivalent to AGP 8x. I believe the minimum is 4x PCI-E to ensure no bandwidth issues, so it could be that the 4x AGP slot is limiting your performance. It could also be the CPU in that machine. Have you configured the environment variables to decrease the CPU usage? If not, give that a shot and see if it helps.

 

[APOLLO]

AGP 4x has about 1GB/s of bandwidth. PCI-E 4x has about 2GB/s, and is roughly equivalent to AGP 8x. I believe the minimum is 4x PCI-E to ensure no bandwidth issues, so it could be that the 4x AGP slot is limiting your performance.

I was afraid of that. Well, if it's the slot there's very little I can do, unfortunately.

It could also be the CPU in that machine. Have you configured the environment variables to decrease the CPU usage? If not, give that a shot and see if it helps.

I have not configured any environment variables yet. The only EV I was thinking of adding was the affinity variable to prevent affinity drift from one processor to another. Is that what you're recommending? Is there something else I can do? Honestly, in most situations, my OC Bartons are head-to-head with the Northwood P4 in my other system, even with the more than one GHz Intel advantage. I doubt it's the processors but I'll believe anything at this point.

 

[Zero82z]

I have not configured any environment variables yet. The only EV I was thinking of adding was the affinity variable to prevent affinity drift from one processor to another. Is that what you're recommending? Is there something else I can do? Honestly, in most situations, my OC Bartons are head-to-head with the Northwood P4 in my other system, even with the more than one GHz Intel advantage. I doubt it's the processors but I'll believe anything at this point.

Follow this guide: http://www.hardforum.com/showthread.php?t=1419852

 

[APOLLO]

Follow this guide: http://www.hardforum.com/showthread.php?t=1419852

Thanks, I'll go through this guide but don't know how many of these variables will help with a 3000-series card and legacy systems...?

Anyone know how well a HD4650 does with the GPU client compared to a HD3850?

 

[Zero82z]

The 3850 is clocked higher than the 4650, so it should theoretically perform better since both have the same amount of shaders.

 

[APOLLO]

I went through the entire guide but it stated the EV are only to be employed with Catalyst 9.4 drivers. The hotfix appears to be pre-version 9.4 drivers. Actually, from looking in the Device Manager, I see v8.64 reported. I tried each EV setting for the fun of it. Some settings showed a negligible improvement while others slowed the card drastically. I kept the settings that improved performance but we're talking about a 100 PPD improvement at most, bringing the average performance to ~1650 PPD. There are a few WUs that permit ~2000 PPD but I don't receive them often.

Now, all of the above testing is the results I get with the 'bad' card in the dual Athlon system. The 'good' card I had acquired last year installed in the more recent P4 system shows absolutely no improvement with the EV settings. This is to be expected with non-compatible drivers. Hopefully, Ati will release a newer driver set for these cards that will improve performance. I just wish I knew why supposedly identical cards demonstrate such a big difference in performance. At any rate, I'm glad this card is working.