HL2 and 9800 pro.

  • Thread starter Thread starter Lee
  • Start date Start date
L

Lee

Hello I have a Powercolour 9800 pro graphics card and for some reason after
about 1 hour play of HL2 the computer just locks up, I don't think the card
is overheating as I can run 3dmark05 for about 10 hours non stop with no
crashes or lockups, could it be the new ati drivers using the NGO ATI
optimised driver based around the catylst 5.7s.
Anyone experienced this as its really annoying when getting into HL2 then
suddenly boom! locks up :(

TIA.
 
Lee said:
Hello I have a Powercolour 9800 pro graphics card and for some reason
after
about 1 hour play of HL2 the computer just locks up, I don't think the
card
is overheating as I can run 3dmark05 for about 10 hours non stop with no
crashes or lockups, could it be the new ati drivers using the NGO ATI
optimised driver based around the catylst 5.7s.
Anyone experienced this as its really annoying when getting into HL2 then
suddenly boom! locks up :(

TIA.

I only had one lockup with HL2, but that was with earlier drivers and I
haven't played it since.

I have a similar problem with Battlfield 2 demo (5.6/5.7 ATI drivers)
though. I left it long enough once and it reported an infinite loop error.
No other games do this, so it could be the ATI drivers, or hardware failing.

I am going to try the full version of BF2 and see what happens.

Good luck with it.
 
abc said:
I only had one lockup with HL2, but that was with earlier drivers and I
haven't played it since.

I have a similar problem with Battlfield 2 demo (5.6/5.7 ATI drivers)
though. I left it long enough once and it reported an infinite loop error.
No other games do this, so it could be the ATI drivers, or hardware failing.

I am going to try the full version of BF2 and see what happens.

Good luck with it.

A memory leak problem was reported starting with Catalyst
5.x (don't remember which version exactly, it could have been
5.2 or 5.3). Try reverting to a known good version (4.12
WHQL) and see if the problem goes away.
 
Lee said:
Hello I have a Powercolour 9800 pro graphics card and for some reason after
about 1 hour play of HL2 the computer just locks up

With the latest drivers I get a lot of this sort thing. The sound loops,
or stutters, and everything goes unresponsive for tens of seconds at a
time. It then breifly works for about half a second before happening again.

I 'fixed' it by setting the resolution down to 1024x768. It's a bummer to
play at that resolution, but at least it all works smoothly again.

I'm now at that bit where those freedom-fighter muppets try to 'help'.
Hmm. Maybe I should just shoot the lot of them!
 
I'm now at that bit where those freedom-fighter muppets try to 'help'.
Hmm. Maybe I should just shoot the lot of them!

Unfortunately, you can't. I actually wrote the only e-mail I've ever
written to a game developer because of the "friendly" NPC AI. A
blemish on what was an otherwise perfect game.Villain

"Bravery is not a function of firepower"
-J.C. Denton-
 
Lee said:
Hello I have a Powercolour 9800 pro graphics card and for some reason
after
about 1 hour play of HL2 the computer just locks up, I don't think the
card
is overheating as I can run 3dmark05 for about 10 hours non stop with no
crashes or lockups, could it be the new ati drivers using the NGO ATI
optimised driver based around the catylst 5.7s.
Anyone experienced this as its really annoying when getting into HL2 then
suddenly boom! locks up :(

TIA.

the best and most stable driver certainly for a 9x00 series card is the
4.12s - the 5.6 and 5.7 while claiming to boost performance do run cards
hotter and that can easily cause instability. just go back to the 4.12s - I
bet
every game you've got runs well on them. Ive got dozens of games and the
only recent one that needed a specific driver ther than 4.12 was KOTOR 2
which ran great with 4.3s OpenGL driver.
 
I ended up doing that, but it still locked on 5.7 drivers. Went back to 5.2
and everything seems OK.
All other games till now worked fine for me except Battlefield 2.

I really hate having to turn down the AGP speed because someone somewhere
can't write more robust software.

Like being told you can't use 4th gear on your car, because the engine might
stop...
 
abc said:
I really hate having to turn down the AGP speed because someone somewhere
can't write more robust software.

AGP8x is of absolutely zero performance benefit, as seen from the Sudhian
benchmarks. There's no reason to even *want* to run at 8x, aside from
perhaps padding up 3DMark scores.
Like being told you can't use 4th gear on your car, because the engine
might stop...

The difference? 4th on your car actually gives you a higher top speed /
better fuel economy. AGP8x is 90% marketing bullshit and 10% pumping
synthetic bench scores (really more marketing bullshit).
 
First said:
AGP8x is of absolutely zero performance benefit, as seen from the Sudhian
benchmarks. There's no reason to even *want* to run at 8x, aside from
perhaps padding up 3DMark scores.

Some people claim that games are loading about 50% faster with 8xAGP
compared to 4xAGP because textures are transferred faster to vram. Don't
know about the difference myself because I have always kept AGP at 8x...
 
de Moni said:
Some people claim that games are loading about 50% faster with 8xAGP
compared to 4xAGP because textures are transferred faster to vram. Don't
know about the difference myself because I have always kept AGP at 8x...

Most benchmarks I've seen show a small performance boost with 8x, still the
point is I didn't buy this card not to use the features...
 
Nope, just tried it. Ran Doom3 v1.3 Demo1 with a console command suffix
(like "timedemo demo1 b") so it precaches the textures at the beginning of
the demo. Precaching occurs after the demo is loaded from the hard drive, so
disk performance has no influence. I then timed the precaching duration with
a stopwatch.

AGP 4x:
11 sec
11 sec
11 sec

AGP 8x:
11 sec
11 sec
11 sec

No difference. I'm running 2x 512 MB dual-channel DDR400 on a Soltek K8TPro
Socket 939 board, so the system RAM interface is not the bottleneck. Fast
Writes and SBA are enabled in both cases.
 
Some people claim that games are loading about 50% faster with 8xAGP
compared to 4xAGP because textures are transferred faster to vram. Don't
know about the difference myself because I have always kept AGP at 8x...

the bandwith of AGP4 is about 1GByte/s. In theory this should fill 256MByte
graphic-memory in 0.25 seconds. Using AGP8 could fill it in 0.125 seconds
so in theory, games should realy load faster with AGP8. Unfortunatly I never
see a game loading that fast ;-)

regards

winfried
 
the bandwith of AGP4 is about 1GByte/s. In theory this should fill 256MByte
graphic-memory in 0.25 seconds. Using AGP8 could fill it in 0.125 seconds
so in theory, games should realy load faster with AGP8. Unfortunatly I never
see a game loading that fast ;-)

regards

winfried


Clearly First of One's loading times were not sufficiently accurately
clocked, then. They should have read:

AGP 4x:
11.125 sec
11.125 sec
11.125 sec

AGP 8x:
11 sec
11 sec
11 sec

Pretty earthshaking improvement when you look at it that way! ;-)

Patrick

<[email protected]> - take five to email me...
 
Actually AGP4x is about 2 GB/s (66 MHz x 4 x 64 bit); and AGP8x is 4 GB/s.
:-)

However, your point still stands.
 
Isn't AGP capable of xferring a LOT more data than even dual-channel at
200Mhz? Even with dual-channel, at 200 Mhz the maximum theoretical bandwidth
is 200Mhz * 8bytes or 1600Mbytes/sec. The theoretical bandwidth for AGP 8x
is 533Mhz * 4 bytes or 2128Mbytes/sec. Therefore your RAM is definitely
bottlenecking at AGP 8x.
 
1. It's 200 MHz *DDR*, so 400 MHz effective.
2. The bus width *per channel* is 64 bits, or 8 bytes; this has stayed the
same since the PII days.

So for single-channel it's 400 MHz * 8 bytes = 3200 MB/s, hence "PC3200".
Dual-channel adds another 64-bit interface, so bandwidth doubles to 6400
MB/s. This is why you see S939 Athlon64 CPUs being advertised with a 128-bit
memory bus.

3. AGP interface is also 64 bits wide, or 8 bytes. 533 * 8 = 4264 MB/s.

So the AGP is still the bottleneck. For reference, the local memory
bandwidth on a Geforce2 GTS is 5300 MB/s. Really, a card from five years ago
still has 25% bandwidth advantage over AGP8x!
 
Uh, wrong. AGP bus is 32-bits wide or 4 bytes, furthermore, your memory has
a LOT more latency than the AGP bus does so you'll NEVER realize those kind
of xfer rates. I'd like to see any benchmarks where dual-channel doubles the
effective bandwidth. The GeForce 2/GTS also has a slower GPU that could
NEVER utilize AGP 8x because you can't clock-in data faster than the clock
speed of the GPU.
 
Back
Top