HD 3850 AGP - infinite loop in MoHAA \ VOGONS

HD 3850 AGP - infinite loop in MoHAA

Topic actions

Post a reply

First post, by XTac

Posted on 2018-09-16, 21:59

XTac Offline

Rank Newbie

Rank: Newbie
Posts: 24
Joined: 2017-09-03, 13:58

Hi, I have a problem with my card and I'd like to have some help with diagnosing it.

I am running a HD 3850 AGP by Sapphire, running under Windows XP SP3. The driver I am using is Catalyst 8.3

I've had some issues with other games, for example KOTOR 2 with newest patches used to occasionally cause VPU Recovers, but it didnt stop me from finishing the game.

However, with Medal of Honor Allied Assault i was getting some random black screen and sound loop, but after a couple of seconds i was able to continue playing. But then I stumbled upon areas that caused these errors and didnt recover, it kept looping with the black screens. If I let it loop for a while, i get a BSOD caused by ati2dvag.dll. If I quit the game while its doing it, MoHAA wont start up due to "could not load opengl subsystem", and any other game i run either wont run or will run extremely slowly. This behavior continues until i restart. All is fine after that. As soon as i run MoHAA and approach the exact same spot, it loops again and craps out the entire graphic driver.

So by this point I am sure something is wrong. I hope its not the card. I can run benchmarks and other games (including MoHAA) for hours and nothing bad ever happens. The temperatures are in norm (60 celsius max), and i never see any artifacts at all. So I suspect the driver, Catalyst 8.3. Unfortunately, this is the driver that supposedly squeezes out the best performance out of this card... but oh well. So here is my question. Does anyone know what the exact issue is? Did someone else encounter a similar problem with a HD 3850 AGP (or perhaps any ATI card that uses a bridge chip?) And if someone here also has a HD 3850 AGP, you could try running MoHAA on Cata 8.3 drivers, and play through the second tank section in mission 5 from start to finish. For me it always crashes right behind a corner with the observation tower and a panzer tiger that comes at me.

Any help is appreciated.

Reply 1 of 5, by agent_x007

Posted on 2018-09-16, 22:48

agent_x007 Offline

Rank Oldbie

Rank: Oldbie
Posts: 1538
Joined: 2016-01-19, 11:06

Why 8.3 ?
Older or newer (9.1/10.2) drivers don't work with those games at all ?

Reply 2 of 5, by XTac

Posted on 2018-09-16, 22:51

XTac Offline

Rank Newbie

Rank: Newbie
Posts: 24
Joined: 2017-09-03, 13:58

Used 8.3 because that's what everyone seemed to recommend. I should perhaps do some testing with newer drivers to see if it runs any more stable and how much performance is lost (if any)

EDIT: Tested it win 9.3. Same issue, same place.

Reply 3 of 5, by Imperious

Posted on 2018-09-18, 12:43

Imperious Offline

Rank Oldbie

Rank: Oldbie
Posts: 686
Joined: 2012-06-18, 14:02
Location: Brisbane, Australia

I have a Powercolor HD3850 AGP. It used to do the same thing until I figured out that the Bridge chip needs a heatsink on it.
After I did that and volt modded the gpu for 1.4v it was good for up to nearly 900mhz.
It's amazing it still works after all that abuse. The heatsink I had stuck down with sikkaflex but it came loose so I just
put one on top with some heatsink compound. You have to remove that useless pad 1st.

Atari 2600, TI994a, Vic20, c64, ZX Spectrum 128, Amstrad CPC464, Atari 65XE, Commodore Plus/4, Amiga 500
PC's from XT 8088, 486, Pentium MMX, K6, Athlon, P3, P4, 775, to current Ryzen 5600x.

Reply 4 of 5, by swaaye

Posted on 2018-09-18, 19:03

swaaye Offline

Rank l33t++

Rank: l33t++
Posts: 8131
Joined: 2002-07-22, 21:24
Location: WI, USA

ATI dropped a bunch of old OpenGL extensions after Catalyst 7.11 and this causes problems with old games. KOTOR seems to work perfectly with Catalyst 7.11 on a Radeon X1950. You even get the bloom effect. But I don't think HD 3850 can run 7.11.

With Quake3 based games you can try renaming the executable to quake3.exe. This stunningly sped up a Radeon HD 5770 for me with RTCW. I imagine a driver application profile is used when Quake3.exe is run.

For some reason ATI's OpenGL driver varied a lot in its compatibility over the years. Seemingly with every other release. The Bioware games are perhaps the most troublesome. There are only a few OpenGL game engines. You'd think they could've kept things working better than they did. My favorite debacle was Rage, which barely worked on ATI cards for the first 3 months while ATI/AMD knee jerked out ~4 half baked hotfix driver releases. Those occasional bouts of piles of hotfixes also showed how the Catalyst program was flawed because its staggered release of parallel driver code trees caused long delays and regressions.

Reply 5 of 5, by leileilol

Posted on 2018-09-19, 00:34

leileilol Offline

Rank l33t++

Rank: l33t++
Posts: 11763
Joined: 2006-12-16, 18:03

typing /r_primitives 2 in the console or starting with +set r_primitives 2 in an unrenamed idtech3 game will get you the same speedup. 😀 but it'd do nothing about this agp bridge crash though. There's reasons why I don't suggest DX10+ hardware for AGP machines

The slowdown's due to r_primitives 0 (the autodetection setting) not finding the compiled vertex array extension and falling back to a slower path to push polys because it couldn't (often from the extension list being too long to read, hence the compatibility profiles shortening it). To force this slow path, try /r_primitives 1

Makes you wonder how many old skewed benchmarks, "id is in bed with nvidia" scandals, "ha ati suck at opengl" and "enhanced driver" controversies came from this extension list bug blunder that is completely id software's fault...

(in the old old Geforce256/Geforce2 pre-det12.41 days, r_primitives 2 was SLOWER, so yeah there used to be a reason for it)

long live PCem