Half-Life 2 Graphical Performance Evaluation

Articles

Event Coverage | Editorials | Guides

Author:
Editor:
Sponsor:
Published:

Kurtis Kronk
Brian

Nov. 17, 2004

Specifications

<< Home

Next >>

Introduction

Just over one year ago, on September 30th, 2003, one of the greatest games of all time was to be released. As you probably know, the game I speak of is Half-Life 2, and unfortunately it was not released on time. Despite the fact that Valve missed the deadline by over a year, HL2 still breaks a lot of ground. For five years this game has been under development, and for good reason. HL2 promises to be a revolutionary new game that delivers more than any other game ever could. One of the most exciting things about HL2 is the new physics system which allows you to interact with everything in the environment. Couple that with some amazing characters, weapons, and vehicles and you're in for one hell of a good time.

Today we will be examining Half-Life 2's performance with two different systems and several different video cards. ATI, feeling pretty good about their situation, took it upon themselves to run through the entire game and find the most graphically intensive locations suitable for graphics benchmarking. We also checked out several levels of the game and for the most part, agreed with ATI's findings. We will be using two of ATI's timedemos as well as one we recorded ourselves, just for good measure.

Our testing consists of running each timedemo three times at 1024x768, 1280x1024, and 1600x1200 with No AA/No AF & 4xAA/8xAF. All in-game graphics were set to Highest. Each result shown in the graphs to follow are an average of three runs at each setting. Let's get to the benchmarks!

All ATI cards were tested with Catalyst 4.11 drivers, and NVIDIA cards were tested with ForceWare 66.93 drivers.

Test Setup

AGP System:
Pentium 4 2.4C
ABIT IC7-Max3
2x512MB PC-3200 XL Pro in Dual Channel
80GB 7,200 RPM 2MB Samsung HDD
Video Cards Tested:

ATI Radeon X800 Pro 256MB
XFX 6800 GT 256MB
Leadtek FX5950 Ultra 256MB

PCIE System:
Pentium 4 3.4E
ABIT AA8 Dura-Max
2x512MB PC2-5400 Corsair XMS Pro Series (Dual Channel)
160GB 7,200 RPM 8MB Seagate SATA HDD
Video Cards Tested:

ATI X800 XT 256MB
PowerColor X600 XT 128MB
Leadtek 6600 GT 128MB

<< Home

Page 1 of 5

Next >>

Page 1: Introduction & Test Setup
Page 2: TimeDemo: riverboat_01
Page 3: TimeDemo: d1_trainstation_02
Page 4: TimeDemo: d3_c17_12
Page 5: Conclusion

Subscribe to Editorials [more info]

Add Comment

31 User Comments

1 - Posted by Guest on November 17, 2004 - 8:12 am

The CPU in the AGP system is too slow to get a good idea of performance. I imagine people with a 6800GT or X800 pro are going to have at least a 3ghz CPU. You can see by a lot of the results that cranking up the Res or the AA/AF has little effect on the benchmarks scores. This is a sure indicator of a CPU bound benchmark.

2 - Posted by Guest on November 17, 2004 - 9:26 am

i agree with the first guess comment, besides with no 6800ultra to compare this performance benchmark seems useless

3 - Posted by Guest on November 17, 2004 - 10:01 am

USELESS!

I will never get back the 10 mins of my life that I have lost reading this review.

4 - Posted by Guest on November 17, 2004 - 10:31 am

For example, with my athlon 2,2ghz, 512Mb memory and radeon 9800 pro@xt clocks,i get 49,5 fps compared to 42.15 fps with x800 pro using d1_trainstation_02 demo. Seems hl2 is really cpu hungry.

5 - Posted by Guest on November 17, 2004 - 11:13 am

If you want to see a real comparison, go to HardOCP as they used their own timedemos, not ones given to them by ATI. I hope Kurtis at least got a some cash for this trash.

6 - Posted by Guest on November 17, 2004 - 11:14 am

http://www.hardocp.com/article.html?art=Njky

In case you need a link....

7 - Posted by Guest on November 17, 2004 - 11:25 am

What??? "With some of the newer middle-of-the-road cards such as the 6600 GT, you can experience HL2 without constant annoying sputtering, with lower AA/AF/Resolution settings that is."

Is this site some automated translation from German or something? Or has the author not yet attended high school?

8 - Posted by Guest on November 17, 2004 - 2:29 pm

Hardocp links to Driverheaven:
"Holy cow, I haven’t even got my copy of Half-Life 2 unlocked yet and performance articles are already hitting the internet. DriverHeaven is the first site to have a quickie Half-Life 2 performance article posted today, the crew tested the game using a variety of video cards under various conditions. No surprises here, where NVIDIA dominated DOOM 3, ATi is kicking booty in Half-Life 2." - Hardocp home page

Hardocp links to The Tech Lounge:
"Kyle's Note: It is also worthy to mention that these benchmark numbers look to be gathered using timedemos supplied by ATI."

Hmm, Hardocp has no problem complaining this article uses ATI time demos, but they didn't even mention that Driverheaven ONLY used ATI time demos? I can see where this bunch of flamers came from.

9 - Posted by Guest on November 17, 2004 - 3:26 pm

Let's just say that TheTechLounge sucks. You ain't got nothing but a pretty looking site.

10 - Posted by Guest on November 17, 2004 - 3:49 pm

You are all losers! Stop fighting and go play the damn game!

11 - Posted by Guest on November 17, 2004 - 4:52 pm

The fact that one the CPU tested with the AGP system was a 2.4GHz Northwood, and the CPU in the PCI-E system was a 3.4GHz Prescott surely skews the results in favor of all of the PCI-E systems unfairly. Perhaps if you did the review on similarly clocked systems (with the same core? =/) you would have a better gauge of the performance of these cards.

12 - Posted by Guest on November 17, 2004 - 5:05 pm

Oh, ok. No problem then. =)

13 - Posted by Guest on November 18, 2004 - 1:01 pm

/. 1s teh w1nn3r

14 - Posted by Guest on November 18, 2004 - 4:25 pm

Personally i found the review to be useful in that it shows FPS in regardto to HL2 which can loosely translate to playability on midrange through to midrange/high system. This is practical in the sense that not every gamer has high end kit and the review clearly shows that top of the range kit isn't mandatory to enjoy and appreciate the gaming experience that is HL2.

15 - Posted by Guest on November 18, 2004 - 4:33 pm

^^^ I agree. Woulda been nicer with more cards but good enough.

16 - Posted by XoRRoX on November 21, 2004 - 3:48 pm

I just ran your riverboat_01.dem demo and compared my result with the ones in the stats in the article "TheTechLounge - Half-Life 2 Graphical Performance Evaluation"..

With these HL2 settings:

- Resolution: 1024 x 768

- Model detail: High
- Texture detail: High
- Water detail: Reflect All
- Shadow detail : High
- Antialiasing mode: 6x
- Filtering mode: Anisotropic 16x
- Shader detail: High
- Wait for vertical sync: Disabled

System details:

- Asus P4P800 Deluxe;
- Intel Pentium 2.8 GHz HT;
- 1GB (4x256MB DDR400 Corsair TwinX);
- Asus v9999/TD GeForce 6800 (Std. C325, M700 @ C376, M802);
- Creative Audigy 2

Driver details:
- Forceware: v66.93 WHQL

Score:
3097 frames 46.404 seconds 66.74 fps 6.582 fps variability

nice eh?!

17 - Posted by Kurtis on November 21, 2004 - 6:42 pm

wow. nice indeed. and with 6xAA/16xAF nonetheless! :)

18 - Posted by Guest on November 22, 2004 - 11:43 am

I disagree strongly with the comment about not recommending using FX cards to play HL2.
My slightly overclocked 5900XT @450/750, Had no problems whatsoever providing a continuous framerate over 40fps, even while the action on screen was intense.
Ok so this is backed up by an amd 64 3200 and 1gb of corsair xms, But HL2 was on the highest setting with 8x AF.

19 - Posted by Guest on November 22, 2004 - 12:50 pm

Image quality still sucks though. Image quality on any FX card in any game sucks. Your card sucks.

20 - Posted by SoylentGreen on November 23, 2004 - 5:51 pm

I for one appreciated the article, although I am now thoroughly confused. I'm seeing the same level of performance, as seen hear, from my x800 XT, but I'm running it on a tad bit faster machine. The thing I'm having trouble with is the large discrepancy between the numbers here and the numbers at HardOCP. Here I see about 64 fps @ 1024x768 with 4xAA & 8xAF. At HardOCP it's 105 fps. That's a huge difference.

Anyone know why the difference is so big?

If HardOCP is correct, I've got a big system bottleneck. He's my spec's if anyone has some helpful feedback.

Processor: Intel 3.06, 533 FSB
Memory: 512 MB PC2100 (Dual DDR setup)
Hard disk: 120 GB SATA150
Mother board: MSI GNB MAX
Video: x800 XT 256 MB (no O.C.)
Sound: SoundBlaster Audigy
Monitor: ViewSonic P810

21 - Posted by SoylentGreen on November 24, 2004 - 12:42 am

Thanks for the info, but I've tried their timedemo and the 4.12 drivers. I still get the same results, about 66 fps. Don't get me wrong, that is still a pretty darn good frame rate. I'm just concerned that my system is choking off my video cards performance, which blows after how much it cost.

On the other hand, if I run out and buy a new motherboard and processor with a 800 MHz FSB, and of course faster memory, and if my marks don't improve I'd hate blowing the money.

Just wish I knew if there was really a problem to fix, before I start mucking with things.

Thanks again.

22 - Posted by Guest on December 3, 2004 - 11:02 am

will my geforce 5700 ultra be able to play this game

23 - Posted by Guest on December 3, 2004 - 1:06 pm

i'm sure it could... but you might have to turn down some of the graphics for a better framerate.

24 - Posted by Guest on December 20, 2004 - 5:20 am

I have a 5700 ultra, and it works great, the graphics are nice too...

25 - Posted by Irrguy on January 10, 2005 - 3:42 pm

I can't figure out how to run these "timedemos" and get a reading. I've double-clicked the demo file but that just loads the game. I got my HL2 via ATI's Steam offer, is that a problem? Any help would be great.

26 - Posted by Kurtis on January 10, 2005 - 5:45 pm

You are doing it incorrectly, that is the problem (not how you acquired the game).

You have to place the timedemo file in a the HL2/HL2 directory (maybe brian can reply or edit this post and elaborate b/c i don't have access to my computer at the moment to tell you the exact location)

Oen HL2, go to options>advanced>enable developer console.

open the console with the ` key

type "timedemo DEMONAME"
- DEMONAME = the filename

It will run and spit out the results in the console when it is finished.

27 - Posted by Irrguy on January 10, 2005 - 6:04 pm

Nice, I'll give that a try.

OK so I get...

ERROR: demo network protocol 6 outdated, engine version is 7
Failed to read demo header.

Any thoughts?

28 - Posted by Brian on January 10, 2005 - 6:40 pm

The demo was recorded on an earlier version of the game. Valve has been apparently changing a lot with their updates and is constantly outdating the demo recording and playback methods. These won't work anymore sadly :-/

29 - Posted by Kurtis on January 10, 2005 - 6:50 pm

damn its on engine version 7 already? it was on 6 when i did that review... valve is really annoying me with that because i have to create a new timedemo every time they do that. maybe i'll call them to complain, even though it won't accomplish anything. :P

30 - Posted by Irrguy on January 10, 2005 - 7:20 pm

No kidding about annoying, now we can't even make a comparison until the new demos are out, which means some serious bench-time by you guys and wait time from us ;)

31 - Posted by Kurtis on January 10, 2005 - 7:39 pm

well, i won't be doing a new bench until we have more video card reviews to do. it really is a pain in the butt, i wish valve wouldn't update so much for such small stuff, or at least not make stuff like this incompatible when it shouldn't be for these minor updates.

Add Comment

To add a comment without being a member, you may omit the password field, but you must enter your name (or nickname) along with your comment. * Denotes required fields.