Abysmal HL2 Benchmarks

[MDE]

Running the Anandtech Canals timedemo, I got 71 FPS at 1024x768 (no AA\AF) on my 6800GT @ Ultra when AT got 110FPS with the GT at 1280x1024. The framerate tanks when there's water on the screen as well as in a couple closeup firefights in the demo, and my hard drive light flashes like crazy. I've tried backing off the overclock to the stock 370\1000, disabling F@H, and I'm about to upgrade to the 67.02 Beta drivers (from the 66.93). Any other suggestions as to why my scores are so low? My CPU is slower than the Athlon 64 4000+ they used, but not that much slower.

Pentium 4 3.2 Northwood @ 3.6 (225 FSB)
1GB Mushkin PC3200
BFG GeForce 6800GT @ Ultra (with NV Silencer 5)
2x 36GB WD Raptors (RAID-0)

EDIT: I tried disabling my anti-virus program and installing the 67.02 drivers with the same results

 

[Viper96720]

I think it's the CPU. Because I get 61fps with a 9500Pro sound high all video settings high water set reflect all with no aa/af at 1024x768.
Using an Athlon 64 3000+. I would think your 6800GT would be more than just 10fps difference.
Otherwise maybe something happened to the video card when you changed the heatsink.

 

[klah]

My CPU is slower than the Athlon 64 4000+ they used, but not that much slower.
...
Pentium 4 3.2 Northwood @ 3.6 (225 FSB)

http://www.firingsquad.com/har...cpu_shootout/page4.asp

The 4000+ is 25% faster @ 1280x1024 than a P4-3.6GHz in HL2.


If the water is causing slow down are you using the lower water detail level like Anand?
http://www.anandtech.com/video...oc.aspx?i=2278&p=2

Here is a comparison of the setting Anand used and the higher level: http://www.hardocp.com/image.h...VYNGNfNl8xM19sLmpwZw==

 

[MDE]

Originally posted by: klah

My CPU is slower than the Athlon 64 4000+ they used, but not that much slower.
...
Pentium 4 3.2 Northwood @ 3.6 (225 FSB)

http://www.firingsquad.com/har...cpu_shootout/page4.asp

The 4000+ is 25% faster @ 1280x1024 than a P4-3.6GHz in HL2.

I have a Northwood which should be faster than a Prescott, and it's running a 900MHz FSB which should make it faster than the Prescott 3.6. There's still a about a 60% gap between my numbers and AT's. I'm using the exact same settings as them except for resolution.

This is the only game that runs slow. Doom 3 benchmarks are good (don't remember the exact numbers), 3DMark03 and 05 are spot on also.

 

[kogase]

Sure, Northwood is faster, like 60MPH is faster than 59MPH. I guess I could have made a more potent analogy with FPS, but oh well. Looking at the benchmarks here an Anandtech, I'd say your CPU isn't really performing much differently from theirs. P4 is just a hell of a lot slower in games than A64.

 

[Viper96720]

What drivers did the review use? I just looked at the results so I didn't notice what driver they used.

 

[Gamingphreek]

Is that 1 gig in dual channel config or is it only 1 stick.

-Kevin

 

[nitromullet]

Originally posted by: Viper96720
What drivers did the review use? I just looked at the results so I didn't notice what driver they used.

Both ATI and NVIDIA have supplied us with beta drivers for use in our Half Life 2 testing. ATI's driver is the publicly available Catalyst 4.12 beta, which is specifically targeted to improve performance under Half Life 2. NVIDIA's driver is the most recent internal build of their ForeWare drivers (version 67.02).

Sounds like the 67.02's.

 

[carlosd]

Originally posted by: MDE
Running the Anandtech Canals timedemo, I got 71 FPS at 1024x768 (no AA\AF) on my 6800GT @ Ultra when AT got 110FPS with the GT at 1280x1024. The framerate tanks when there's water on the screen as well as in a couple closeup firefights in the demo, and my hard drive light flashes like crazy. I've tried backing off the overclock to the stock 370\1000, disabling F@H, and I'm about to upgrade to the 67.02 Beta drivers (from the 66.93). Any other suggestions as to why my scores are so low? My CPU is slower than the Athlon 64 4000+ they used, but not that much slower.

Pentium 4 3.2 Northwood @ 3.6 (225 FSB)
1GB Mushkin PC3200
BFG GeForce 6800GT @ Ultra (with NV Silencer 5)
2x 36GB WD Raptors (RAID-0)

EDIT: I tried disabling my anti-virus program and installing the 67.02 drivers with the same results

They use an Athlon 64. That's the reason. And this A64 CPU is much faster than yours.

 

[MDE]

Yes my RAM is in dual channel. I KNOW my CPU is slower than the one they used but it's not slow enough to justify a 110 FPS\71 FPS gap, especially with mine running at a lower resolution.

 

[nitromullet]

I have a Northwood which should be faster than a Prescott, and it's running a 900MHz FSB which should make it faster than the Prescott 3.6.

Not necessarily, the Prescott scales better than the Northwood at higher clockspeeds quite often. While a P4 2.8C performs better than P4 2.8E in most cases, a P4 3.6E will most likely top a P4 3.6C. In some cases, the Prescott is superior at any clockspeed. In Anand's CPU test of Doom3 the Prescotts beat the Northwoods across the board.

http://www.anandtech.com/cpuch...oc.aspx?i=2149&p=7

Like it or not, HL2 is pretty much an A64/ATi game... That doesn't mean that I haven't been able to enjoy it with my Intel/nVidia rig, but I bet it would run better on the former.

 

[Gamingphreek]

Uh no.... Nvidia is VERY close behind. THis is not like DIII, it is much much closer with the 6800GT beating the X800Pro and the 6800U coming in 5-10% slower than the XTPE. DIII the XTPE is around the speed of the NU.

-Kevin

 

[nitromullet]

That really depends on the settings you play with. Above 1280x1024 with the the "reflect all" setting chosen instead of "reflect world", the X800's are the top performers. Throw in some AA/AF, and it seals the deal.

IMO, when the X800Pro beats the 6800Ultra in almost any situation, I would have to say that ATi has a solid edge. Likewise, in a situation where the 6800NU gives the X800 XT a run for its money, nVidia has the edge.

The cool thing for us nVidia owners is that you can select the "reflect world" setting, and the game plays quite well and still looks phenomenal. This gives us back the performance that we want, and while the screenshots look different, I really haven't missed "reflect all" when playing.

 

[MmmSkyscraper]

I hardly noticed any difference between 'reflect world' and 'reflect all' when doing the CS:Source stress test. I lost maybe 1fps... Or is this a bad indicator of game performance?

 

[LocutusX]

To be fair, while the Reflect All water does indeed look better than the Reflect World - and thus is what I prefer - you must realize that the entire game does not take place in the "Canals". At most you would spend 20% of the game in or around those water-heavy environments. I am being liberal with my estimate so it is probably closer to 15%.

But if you're someone who has only been playing the game for about 4 hours, it'll seem like you spent half the game there.

 

[LocutusX]

Originally posted by: MmmSkyscraper
I hardly noticed any difference between 'reflect world' and 'reflect all' when doing the CS:Source stress test. I lost maybe 1fps... Or is this a bad indicator of game performance?

pretty much (IMO)...


the actual difference between Reflect World/All is much more significant (in terms of framerate impact) and is apparent on both x800 and 6800 based systems.

 

[jer0608]

Do you have Vsync disabled?

Reflect world or reflect all? Would probably make a big difference in canals.

For comparison, my system is significantly slower than yours and I get ~77 fps in the canal demo with the following settings:
1024x768
No AA/AF
Trilinear Filtering
Everything High
Reflect World

System:

P4 2.4C Northwood @ 2.87
1GB Kingston HyperX
eVGA 6800GT stock
66.93 drivers