ATi Radeon 9700 AE2 Stepping AGP 8x Bug

[Mr.Tweak]

Since receiving our Sapphire Radeon 9700 graphics card a couple of days ago during testing we have been experiencing a few strange problems – such as continuous system reboots and programs and games crashing back to Windows desktop when used on all SiS 648, VIA KT400 and P4X400 based motherboards with AGP 8X support operational.

We cannot be 100% sure if this is an isolated problem however can be sure in saying we received confirmation from our contact at Epox Taiwan (who got first hand information direct from ATi) saying the problem is only occurring in AE2 stepping Radeon 9700 chips, and not the newer AE3 chips which are now shipping from the Canadian company, who say fix the problems discussed here. The problem is, the AE2 stepping chips are vast, and thus many made it into the retail market where end-users are no doubt going to suffer. It’s important to note that we had no problems what-so-ever running the AE2 stepping Radeon 9700 on VIA KT333 and Intel 845E/G based motherboards operating in either AGP 2x or 4x modes.

[ragingWS6]

How do you know what stepping stone you have? Is there some kind of identifier on the board? Please let me know.

[Wiggo]

Usually the chips themselves have it marked on them and sometimes they're put spec stickers but never actually handlin' one of those I can't say for sure. :smokin:
<center>:cheers:</center>

[saidin]

So how do we go about ordering the correct one....i do not want to get a defective part.......:no:

[Wiggo]

Now that depends on if ya can find a honest sales person (yeah I know) who is well informed (yeah, yeah, lay off) but there maybe a few around here who can give ya an idea on what to look for. ;)
<center>:cheers:</center>

[hivez]

Is this isolated to the ATI cards only or will this effect the gigabyte cards? BTW I have an ASUS A7V8X

Thanks
Hivez

[Mr.Tweak]

All cards using the AE2 stepping chip.

[Bathrone]

How do you query the VPU for what stepping it is?

I need to know what stepping mine is.

Not sure based on looking at the card.

[saidin]

Hopefully ATI will officially recognize this problem to the consumers and except an exchange for one that works with 8x AGP.

[lancero]

P4 2.533 B
Abit SR7-8X latest bios revision B
Ati radeon 9700 Pro bulk (sapphire)
Latest driver from Ati web site
AGP111.exe from SIS
Fast write AGP disabled in bios.

The computer work about 10 minutes in windows and after that he change the screen resolution from 1280X1024 32 bit to 800x600 in 4 bit "Windows has to correct a problem with an Hardware ...". I can still work but only in this resolution.
I tried some games MHAA, NHL 2002 they hang up after 2 minutes of playbut i think for that its a direct x problem. Anyone tried direct x 9 beta ?
Same problems with 3D software test after 10 minutes ...

I don't know what to do maybe its an Ati driver problem or an SIS AGP driver problem. I don't think its a power problem (PSU) because if it was the system will directly hang up or refuse to start ...

got this from kyle at hardocp

This is one of the things that comes with being first to market with new technology and you guys need to be aware of it. The Asus KT400 is the only board we have found so far that it does not work with.

Our buddies over at Asus had their two cents to add as well.


The problem is caused by a combination of power supply and motherboard +5V voltage ramp curves creating a situation where certain process corners in the ASIC will intermittently return an incorrect value for the ASIC ID after a cold boot.

Fix:
* Short-term solution is diligent screening prior to shipping.
* Long-term there are 2 options:

1. A modification to the motherboard BIOS (system BIOS) to create an extra read of the ASIC ID, the return value of which is ignored. The second read of the ASIC ID will then return the correct value

2. A board modification to monitor the read of the ASIC ID using programmable logic which will generate a NAK command back to the Northbridge controller while allowing the Radeon 9700 Pro to act as if the read command were executed successfully in order to cause the Northbridge to reissue the read command.


Any help would be appreciate ...