1. try catalyst 3.10 (seems to work for most people who had this problem)
2. downgrade drivers (3.4-3.7 were popular picks)
3. turn AGP fastwrite off in the BIOS and the display settings
4. change it to AGP 4x instead of 8x in BIOS and or the display settings
5. turn AGP write off in the advanced smartgart settings (start/run/smartgart) and leave everything else on
6. lower the hardware acceleration in the display settings under 'trouble shoot'.
7. up the AGP voltage a tad to 1.6 (dont go too high or youll get heat issues)

those are the only ones i remember.

For Jack Black:

1800+ @ 1926 MHz (11.5 x 167.5)
512 MB Mushkin PC2700 @ CL2-3-3-6
9600 Pro @ 485/320

= 12100 MK3D2001
= 3725 MK3D2003

do you have a score with your card at stock speeds?
i could go higher with my videocard overclock because i have an iceberq4 vga cooler on it but this card has to last me a long time so i dont want to risk frying it.

Hello again everyone,

I've tried setting my cpu/dram speed to 133/166 from 133/133 in the bios, apparently this allows my ram to run faster, i'm still not able to turn on either agp read or write. they seem to be locked to off no matter what i do..

Anyway i'm wondering what if any disadvantages i would have by running my system at 133/166 instead of 133/133?

I used a benchmarking program named Aida32 (its free so do a google for it if you'd like a copy) to test my ram and find some information about it.
Under Memory benchmark my memory read and write speeds have gone up about 200-400 mb/s after changing to 133/166
and heres some info that showed in aida32 up under the /motherboard/SPD tab that was absent when i was running at 133/133, makes me think my system should have been running at 133/166 all along, but you tell me

"Memory Module Properties
Module Name SpecTek
Serial Number 0400DE4Ah
Module Size 512 MB (2 rows, 4 banks)
Module Type Unbuffered
Memory Type DDR SDRAM
Memory Speed PC2700 (166 MHz)
Module Width 64 bit
Module Voltage SSTL 2.5
Error Detection Method None
Refresh Rate Reduced (7.8 us)
Highest CAS Latency 2.5 (6.0 ns @ 166 MHz)
2nd Highest CAS Latency 2.0 (10.0 ns @ 100 MHz)

Memory Module Features
Early RAS# Precharge Not Supported
Auto-Precharge Not Supported
Precharge All Not Supported
Write1/Read Burst Not Supported
Buffered Address/Control Inputs Not Supported
Registered Address/Control Inputs Not Supported
On-Card PLL (Clock) Not Supported
Buffered DQMB Inputs Not Supported
Registered DQMB Inputs Not Supported
Differential Clock Input Supported
Redundant Row Address Not Supported

Memory Module Manufacturer
Company Name Micron Technology, Inc.
Product Information http://www.spectek.com/menus/memory_module.asp
"

Hmm...I have no idea how to show quoted text on this forum, so I'll have to do it clumsily.

***
I've tried setting my cpu/dram speed to 133/166 from 133/133 in the bios, apparently this allows my ram to run faster, i'm still not able to turn on either agp read or write. they seem to be locked to off no matter what i do..

Anyway i'm wondering what if any disadvantages i would have by running my system at 133/166 instead of 133/133?
***

Your memory is indeed running faster - 166MHz instead of 133MHz. If you are using the SocketA platform, you will often get higher performance if you can run your memory (and therefore the bus connecting the memory to the northbridge of your motherboard) at exactly the same speed as the FSB (i.e. the connection between the northbridge and the CPU), even if the memory could run faster. This is usually referred to as synchronising the FSB and memory. Data transfer between CPU and memory goes through the northbridge (except on Socket754/950/939).

As an analogy, consider two people moving a pile of rocks from one place to another. Person A picks up a rock, turns round and passes it to person B, who turns round and puts the rock down. The most efficient way to move the whole pile of rocks is when Person A and Person B work in a synchronised way, at the same speed, so that Person B turns back to take the next rock just as Person A turns and holds it out.

A chain of people passing buckets of water to put out a fire would also be a good analogy.

This only applies to SocketA. If you're using Socket478, the FSB and memory bus are of different types (quad-pumped and DDR respectively) and therefore can't synchronise. If you're using Socket754/940/939, synchronising FSB and memory is irrelevant as memory isn't routed through the northbridge.

The bottom line is: benchmark with memory at both speeds and use the setup that gives higher performance in whatever you do most.

I am not aware of any disadvantages of running memory and FSB asychronously other than the potential performance reduction under SocketA (or, to be more accurate, not getting the potential performance increase that you can get by running them synchronously). Of course, that's assuming that your memory is capable of running at the higher speed without an increase in latency.

Just a quick half-awake thought about your problem with AGP read and write....have you installed the AGP miniport drivers for your motherboard? If so, have you tried checking for newer versions and installing them?

Re: Peter Howardson
Thanks, your message pretty much cleared everything up,
I'm not really sure what type socket i'm running but I tested the ram and it seems to run considerably faster at 133/166 so for now ill keep things that way unless i notice any new stability problems

and about the AGP drivers

That's a good thought, ill try installing/reinstalling the newest AGP drivers available, we'll see if that allows me to turn AGP read and write back on... of course AGP read and write being off may be what's making my computer run properly now, anyway we'll see

SocketA is for Duron/Athlon/AthlonXP, except the earliest Athlons (up to 650MHz, IIRC). Since you get better performance with 133/166, I suspect you're using something else.

I'd play with my settings, but my new motherboard is wildly temperamental on memory timings. Anything other than "by SPD" (i.e. automatic) stops my PC from booting at all. I have to clear the CMOS and start again.

ok.... catalyst 3.10 Doesn't work for Windows 2000 pro at all. Now with my nice 9600 pro I can't even load windows. I've played around with the setting for everything. Still locks up and shuts down. Or I can't even get into windows with some drivers. Anyone have any help on this?? Oh yeah, the only thing that allows me to go into Windows and unistall the drivers is entering in VGA safe mode, and turning down my hardware accelerator........ Anything I can do??????!!!??!!!??!


Jason

go back to an old driver that worked for you and wait until a new driver works.

Im happy for everybody whos problems are fixed with the 3.10 drivers. Unfortunatly 3.10 doesn't work for me. I had a reboot at two complete freezes just starting a game. 3.9 was s**t but at least I had a few minutes of problem free play. Back to 3.4.

my friends "GainwardFX5600Ultra performs exceptionaly well without any problems even with an old computer. very nice. ati ati ... ati, would you cry me a river.

I''ve just fitted a new 9600xt and having same problems as everyone else here. I've tried most the 'fixes' suggested but still get game freezes after 10 secs. I've used cat3.9 and 3.10. Freelancer often comes up with the message 'Cant recognise this video card' even after the game has been re-installed.

Interestingly my old ATI 8500 played all my games perfectly using cat 3.9 and 3.10 including COD, so I dont think its the drivers.

Gonna try a few more things then go back to old card. I resent having to waste time doing this cos like a previous subscriber ALL my other previous 8 graphics cards I've ever fitted have had no problems at all.

My mobo is a MSI KT3 ultra and only runs at AGP 4 anyway

Got to face it ATI - this card is PANTS!!!!!

Athlon 2100
512 mb pc2700
Maxtor 80g 7200 hard drive
Windows XP sp1

if your 8500 worked with cat 3.9 that doesnt mean that its not a driver problem... every card is made differently so there might be compadability problems with your card. All cat drivers worked with my old 7500, but they didnt with my 9600 pro. Now the 3.10 driver works so it was a driver problem.
As for the person who has a friend with a fx5600. The fx line is absolute garbage. The 5600ultra barely outperforms the ti4200 in dx9 games and i think the ti4200 actually outperforms it in dx8 games. I'd rather have a ti4200 over an fx5600 because even though the fx line is supposed to be a dx9 card, it performs like s**t in dx9 games.

Hello all, just thought id let you guys know the catalyst 3.10 Omega drivers are out, they may help some of the people having problems with 3.10.. i have no real reason to think they would but its something different and just might help..


also you can try turning off AGP read and AGP write in the smartgart shortcut in your ati directory, that seems to be what made my system stable..
good luck guys, and happy holidays

I'm using the Omega drivers based on Cat3.10. They fixed the problems with my card. They also have a rather nice advantage over the reference drivers - you *don't* need to uninstall an earlier version of the Omega drivers to install a newer version. You can just slap the new version straight over the older version. That's only Omega to Omega though - if you're moving from the reference Catalyst drivers, you'll need to go through the usual rigamarole of removing the older version. Which, btw, sticks like sand between your toes on a beach. I uninstalled, then used a driver cleaner, then ran NBGclean to check the registry - and it still found references to the old drivers.

As for the GeForceFX line, the only one that is really crap is the 5200. The 5600 is pretty poor too, but it's already been replaced by the 5700, which is a decent midrange card. The old GeForce4 Titanium line was just so good that it makes the next generation look bad - it will beat Radeon 9600s in DX8, too. The big advantage of GeForceFX5700 over Radeon9600 is that the FX installs cleanly and works straight away. Look around the net - the installation of Radeon 9600s is a major bodge job. The average PC user has damn all hope of getting it to work unless they are lucky enough to avoid the myriad problems so many people are having.

It's a fine card...if you can get it to work.

I tried to install the omega driver but after half of the installation it asked for a cd 2??
is it a bug? if not what is it?

I'm experiencing simliar problems as everyone else. However, I may have a unique perspective in that I recently bought two computers with the exact same hardware. One works fine (after going from AGP 8x to 4x). The other has hanging problems with games like RtCW and EverQuest. They're both MSI-km400-6734 motherboards, 9600 pro cards, on-board sound cards, running Windows XP SP1.

I've fiddled with all kinds of settings as suggested here with no luck (Catalyst 3.10, fastwrite off, agp 4x). Changing to AGP 4x made the one system stable (no crashes yet after a few days) and it helped a bit on the other system (crashing after 10 min in RtCW rather than immediately). It seemed very mysterious that the two "identical" boxes behave differently. So, I kept looking for that subtle difference.

The only difference that I've noticed between the two systems is that the "Video BIOS" from the 9600 is different on each. The working system lists its video BIOS as "BK-ATI VER008.011.003.000" whereas the non-working system lists the video BIOS as "BK-ATI VER008.015.028.000".

I wasn't able to find mention of people's video BIOSes in this thread so far, though I'll admit that I didn't read every post. If you're still having problems, I'd be curious to see if your video BIOS is the same as mine. You can find it via the display panel -> Settings -> Advanced -> Adaptor (look at Bios Information).

Depending how daring I feel I may try backing up the video bios from the working card and flashing it to the non-working card. I think that I'll first switch the cards in the machines to see if the current working one also works in the other box.

Steve

Well, step 1 of the experiment worked. I switched the two cards and now the previously working computer is having the same problems the other one was. And the previously non-working computer is now working. So, I've either got myself a bad card or that bios is no good. By the way, it turned out that the two cards are not identical. They're obviously made by different manufacturers though the only label one way or another is "sapphire" on the "bad" card's fan (which may be the fan brand, I don't know) and "ati" on the good card's fan.

Next I think I'll try backing up the BIOS from the good card and flashing it on to the bad card.

Steve

OK, when I tried to flash the good bios overtop of the bad bios, flashrom gave an error: "ASIC memory config does not match with BIOS file "good-ati.bin" memory config. Use -f to force flashing." I thought I'd see if anyone has experience with this kind of thing before I try forcing the flash. It seems kind of unlikely to me that if there are ASIC incompatibilities that the good BIOS will work on the bad card. But, I'm willing to give it a shot if the chances are low that I'll ruin the card (I already have a backup of the "bad" BIOS in case I need to put it back). Any thoughts?

Steve