41 comments

» MEDIA (10) Optimized Driver Performance in 3dMark05 Benchmark Optimized Driver Performance in AquaMark03 Optimized Driver Performance in Half-Life 2 Stress Test Optimized Driver Performance in Titan Quest Optimized Driver Performance in Supreme Commander (Demo version) Catalyst HL2 Stress Test DNA HL2 Stress Test NGO HL2 Stress Test Omega HL2 Stress Test DHzer0point HL2 Stress Test

Picture this: Late night; it’s almost midnight, and you have to wake up early for work tomorrow, but, not far from you, there’s a mean looking terrorist with a bomb. And you only have one round left in your AWP. It’s up to you to win one for your team – in your final round of Counterstrike for the night.
Noobsauce[DLAT]*, the terrorist (14 kills, 3 deaths), hasn’t seen you yet. And now your half-way across de_dust2, and you’ve been waiting, and you’re ready. No time for a reload. Finger on the trigger. Sweat on your mouse. Your hearts beating a little quicker now. You try to line him up in your crosshairs -- but something’s wrong. Terrorist Noobsauce[DLAT] spots you. He jumps; he bunny-hops toward you, and you got one shot to take and you have to make it count.
You take the shot.
Boom from your speakers.
You miss the shot and the bugger caps you with a Desert Eagle and it’s all over.
You should of have had him.
Your ping is 28 – no problem there -- but there was a couple of smoke grenades belching smoke between you and Noobsauce[DLAT] and you know, from experience, that when there’s a lot of smoke on screen, that is exactly when your non-cutting-edge Radeon begins to choke.
You let out a big sigh. Time to call it a night, you’re thinking.
But in the back of your mind you know that a couple of more frames-per-second could have made all the difference.
If you are a casual gamer, maybe this article will not interest you. But: if you have an ATI card, and are always looking for a way – especially a free way – to squeeze a couple more FPS out of your system, then read on.
Apparently, you’re a hardcore gamer, and there might be a performance increase in store for you.


THE FIVE CONTESTANTS


This article will compare performance between 5 ATI graphics drivers, in 5 different benchmarks. Here’s a rundown of the contestants:

Driver #1: ATI Catalyst 7.2 (Released Feb.21, 2007) [http://ati.de/support/driver.html]

As the official ATI drivers, it’s a safe bet to say that the Catalyst drivers are what the vast, overwhelming majority of ATI users currently utilize. They are as a stable a driver as you are likely to find. Versions are also are available for every conceivable OS, including Linux. AMD/ATI employs a team of software engineers to continually optimize and improve the Catalyst drivers, with new releases coming out fairly regularly.
The most notable improvement in the new version 7.2 of Catalyst is an optimized Catalyst Control Center for Windows XP. Overall, the Catalyst Control Center has quicker start-up times in comparison to previous versions, and demands less system resources. Additionally, a new 3D preview option in the CCC (Catalyst Control Center) enables the user to see immediately what differences different graphic accelerator settings have on performance and image quality.
As these are the official ATI graphic drivers, Catalyst 7.2 will be the baseline performance benchmark that the other four drivers in this roundup will be judged against. The next four drivers in this roundup are all optimized and tweaked versions of the Catalyst drivers.
In the release notes for Catalyst 7.2, ATI/AMD states that the “AMD Catalyst™ software suite is the ultimate in performance and stability” – but let’s see what the benchmarks show us.

Driver #2: Omega 3.8.330 (Released Jan.14, 2007) [http://www.omegadrivers.net]

The Omega Drivers are a labor-of-love project, and the man behind them is Angel “Omegadriver” Trinidad. If you are one of the few that are aware that optimized, 3rd party ATI drivers are out there, there is a good chance that you know of the Omega Drivers -- they have been around for a few years now, stretching back as far as the 3dFX days. While they are unofficial ATI drivers, the Omega Drivers could boast that they are the most ‘official’ of the ‘unofficial’ 3rd party ATI drivers. Terry Makedon, from ATI, has stated in an interview that “Omega Drivers are in fact a good example of ATI’s user community at its best…our relationship with the creator of the Omega Drivers is of the highest working standard…in fact, we would go so far as to say that if a user chooses to go the mod driver route, they go with the Omega Drivers.”
While the latest Omega release, 3.8.330, is based on Catalyst 7.1, the Omega drivers still offer a greater number of resolutions then are available with vanilla Catalyst 7.2. In place of the Catalyst Control Center, Omega comes bundled with Ray Adams’ ATI Tray Tools. Tray Tools is a leaner, less resource-intensive spin on the Catalyst Control Center -- and in addition to the functionality of the Catalyst Control Center, ATI Tray Tools offers further tweaks and an easy to use overclocking utility. Omega also comes with Multires 1.58, which is a program that allows you to quickly change screen resolutions on-the-fly. The Omega drivers website also offers a few great tips on improving the performance of your graphic cards, such as listing recommended motherboard BIOS settings.
Omega drivers are available for Windows XP, 2000, and Windows 9x/Me, but unfortunately, are not Vista ready yet.

Driver #3: DNA-ATi 5.0.7.1 (Released February 8, 2007) [http://www.dna-drivers.com/]

Developed by a small German team from the Netherlands, (perhaps composed only of a man who goes by the name of DrHouse, and his webmaster assistant, KillerSneak), DNA drivers are self-billed as “The Only Drivers that Matter.”
Although they have been around for roughly 3 years, they have flown pretty low under the radar. As of March 6th, new versions of DNA “pre-final” public betas have been released (based on Catalyst 7.2), but for this ATI Driver Roundup, the last stable release was used, which is based on Catalyst 7.1.
If the DNA website’s download page hit-counter is accurate, DNA-ATi 5.0.7.1 has been downloaded somewhere in the neighborhood of 13,000 times – so, compared to Catalyst, and even Omega, these drivers definitely reach a much smaller audience. To their benefit, the DNA team seems very helpful and responsive to individual problems, as evidenced by forum activity, and blog postings that relate personal help given from KillerSneak. Previously, DNA has requested a donation of 1 Euro for their efforts – but this seems to no longer be the case, although, presumably, donations are still very-much welcome.
Like the Omega drivers, the DNA team has opted to use Tray Tools in place of the Catalyst Control Center.

Driver #4: NGO 1.6.12 [http://www.ngohq.com]

Like DNA and Omega drivers, NGO is a modified version of Catalyst 7.2 which strives for “better performance, image quality, and compatibility” and is best “suited for enthusiasts and gamers that want to have the best gaming experience.”
NGO is a project of the aforementioned website, which primarily serves as a source for hardware, graphics, and software news. As far as the more uncommon 3rd party optimized drivers go, NGO is reputed to have strived for a balance between image quality and performance.
NGO has also gone the ATI Tray Tools route, over the Catalyst Control Center. With this most recent release, NGO has removed some bundled utilities, and now the entire driver package weighs in at just over 19 megabytes – the smallest of the five installation packages.
Windows Vista is not currently supported by NGO.

Driver #5: DHzer0point 0.69 (Released December, 2006) [www.driverheavendownloads.net]

The ‘DH’ of DHzer0point is an acronym for the website that currently supports the development of the zer0point driver: Driver Heaven. Driver Heaven is primarily a hardware review website; but also features and develops a handful of useful programs, such as DriverCleaner (a tool to clean out your old graphics drivers), and Spike’s Tools (a set of handy utilities), in addition to putting together there own graphic drivers for the enthusiast public.
The latest DHzer0point release is 0.69, which is a tweaked version of, (unsurprisingly), Catalyst 6.9.
Of all the drivers in this roundup, the least amount of information is available concerning the DHzer0point drivers. The website link to the DHZer0point Team is non-functioning, and the drivers come with little documentation. However, rumors on various forums suggest that the drivers are inclined towards performance over graphic quality and stability -- but let’s see what the benchmarks tell us.


Testing platform


The following benchmarks were run on this computer:

OS: Windows XP, SP-2
Processor: P4 3ghz HT Northwood CPU
Motherboard: ASUS P4PE (Socket 478)
Memory: 512MB of PC2700 Infineon RAM, @ DDR400
Hard Disk: 60GB 5,400 RPM Maxtor ATA
Video Card: ATI Radeon X800 XT, 256 MB

This system should offer middle-of-the-road performance in all five of the benchmarks. Although it is the system components that determine the ultimate performance in benchmark tests, for this article, it will be presumed that differences in graphic driver performance will carryover from one system to the next – although, as these drivers have only been tested on this one system, this can not be guaranteed.


Brief installation/uninstallation comparison


Of all the installation programs, the Catalyst driver is the most polished and straightforward – and the one least likely to run into problems.
The Omega installation is also top-quality and polished, and has the extra feature of unique, soft-mods for Radeon 9500 and 9800 SE. Although, in the installation, there are a few YES/NO near-esoteric questions that may have a few users scratching their heads, such as, “Enable old IC control system for overclocking?”
A notable feature of the DNA drivers is that they include many different language-sets in them, so if you speak Albanian or Turkish (or many languages in between), you are in luck. However, the DNA uninstall utility seemed lacking, somewhat: it does not seem to remove all the files properly; on the benchmark computer, DNA did not remove the memory-using ‘CLISTART’ or ‘MOM.EXE” [a new Catalyst monitoring program, NOT the infamous key-logger program] from the Windows XP startup que.
Both the NGO and DHZer0point install and uninstall utility are effective, and straight-forward. However, the DHZer0point install stands out because it contains (probably) the funniest Licensing Agreement of all-time.


Benchmark 1: 3DMark05

FutureMark’s 3DMark series is a well-reputed benchmark utility that gauges system performance. The 2005 (Free-Edition) version was chosen over the 2006 version because the later version requires a gigabyte of RAM, which was unavailable at testing time.
The benchmark runs through a series of three very graphic-intensive routines, followed by two CPU intensive tests, and generates a precise score in ‘3DMarks’ that is an indicator of overall system performance in graphic-processing.
This benchmark was run at resolution of 1024x768, with no Anti-Aliasing.


Benchmark 1 results


[Please consult graph #1 on the right]

The DHZer0point drivers are the clear winner of this first benchmark, racing to the finish with a 3DMark score of 5381. Not too far behind, in second place at 5319 3DMarks, came the ‘vanilla’ Catalyst 7.2 drivers. Third place goes to NGO. Fourth place is taken by the Omega Drivers, and it is somewhat surprising that they are almost 100 3DMarks behind the Catalyst driver. In this first benchmark, the DNA drivers came in last.
Between all the drivers, there was no noticeable qualitative difference in graphic quality throughout the near five-minute long 3DMark benchmark test.


Benchmark 2: AquaMark3


AquaMark3 is another popular graphics benchmarking utility. As opposed to 3DMark, AquaMark3 utilizes an actual game engine -- (the krass engine) -- to generate a benchmark score. The makers of AquaMark3 would probably argue that this makes for a better benchmarking utility, as the AquaMark3 test measures system performance in conditions mirroring those running an actual game, as opposed to many other benchmarking utilities, which generate scores from a more synthetic environment, based on focused routines and tests.
The settings for this benchmark were: 1024x768 resolution, 32 bit color, no FSAA, 4xAnisotropy (AF), and the details set to maximum.


Benchmark 2 results


[Please consult graph #2 on the right]

Again, the DHZer0point drivers take the top spot for performance – but this time, the margin of victory is huge. The test was run twice to confirm the results, but they held true: DHZer0point proved to be the superior driver, by far, in this benchmark. At a significant speed increase of 11% over the second place Omega driver, it is plausible that the DHZer0point driver utilizes some tweak that the other drivers do not – a certain tweak that particularly impacts performance in some of the graphic-routines utilized in the AquaMark3 test. Behind DHZer0point, there was not much difference in the performance between the other four drivers; although, the NGO driver fell slightly behind the back, to round out 5th place in this benchmark.
And, as same as the last benchmark, there was no noticeable image quality difference between any of the drivers in AquaMark3.


Benchmark 3: Half-Life 2 Stress Test


It is time to move from the benchmark utilities to where the performance counts: in the games.
For the five drivers in this roundup, the next test will be the seminal, perennially popular, Half-Life 2. To compare Half-Life 2 performance, the Valve-made Half-Life 2 Stress Test (included in Counterstrike: Source) will be used. This test should accurately show what kind of in-game difference a graphics driver can have.
As in the next two game benchmarks, two graphic settings will be used: a higher quality setting, and a lower quality setting. For the Half-Life 2 Stress test, the high setting used was: a 1280x960 resolution, high detail settings, 4xAA and 8xAF; for the lower quality settings, a 1024x768 resolution was used, with medium quality graphic-quality settings, and 2xAA with 4xAF.
Each driver (on both settings) was run through the stress-test three times. The best average frames-per-second was taken as the result.


Benchmark 3 results


[Please consult graph #3 on the right]

It’s starting to become a trend: the DHZer0point driver offered superior performance –both in the higher-quality, and lower-quality settings. Although it should be mentioned that the performance difference between all the drivers in this benchmark was relatively slight – a 5-point FPS spread in the lower-quality settings, and a 3-point spread in the higher-quality settings -- there was still a measurable difference in performance, nonetheless.
The Catalyst and Omega drivers veritably tied for second place, with Catalyst performing better on the lower-quality settings and the Omega drivers performing better on the higher-quality settings.
Although it could not be measured with numbers, qualitatively, the NGO driver seemed to offer superior image quality in the Half-Life 2 Stress Test. In comparison with the other drivers, the graphics during the NGO stress-test seemed noticeably better: a little crispier and with increased definition; specifically, with stone textures looking somewhat more realistic in comparison to the images offered by the other drivers.


Benchmark 4: Titan Quest


Titan Quest, a Diablo-esque, overhead, 3D action-RPG, was used for the second game benchmark. Titan Quest is good contrast to use against Half-Life 2, because the two games have significantly different graphic engines.
For this benchmark, the FRAPS program was used. FRAPS has the ability to record segments of gameplay, and then give you a detailed frame-per-second analysis. For this benchmark, a saved game was loaded, then recorded for thirty seconds, with the same actions being performed (the casting of three fireballs at the same spot). Note: because this is not a ‘closed-system’ test, results may be less reliable than a benchmark utility, or in-game stress-test, because results are user-dependent – however, great care was taken to make each 30 second FRAPS as similar as possible.
The Titan Quest benchmark was run at a resolution of 1024x768 with medium-quality settings enabled, and at a 1280x960 resolution with high-quality settings enabled.


Benchmark 4 results


[Please consult graph #4 on the right]

DHZer0point did not take the 1st spot in the Titan Quest benchmark, although it came close: the DHZer0point drivers offered the best performance with the medium-quality settings – however, they faltered on the high-quality settings, coming in last place. Catalyst 7.2 arguably finished first in this benchmark, with the best performance out of the five when the graphic quality was to high, and 2nd best performance overall in the medium-quality setting.
DNA, Omega, and NGO had virtually the same performance, in both the high and medium settings.


Benchmark 5: Supreme Commander


For the final benchmark, the recent RTS game Supreme Commander, was used. This game will push your hardware to the limit -- even if you have a cutting-edge system -- so it makes for an excellent choice to benchmark the five drivers with. For this benchmark, FRAPS was again used, but this time, to arrive at more precise measurement of FPS, a 90 second recording-time was used. For the FRAPS recorded segment, a saved game was reloaded at the exact same point, in which the zoom was set at medium, while many units and buildings were on-screen (with two buildings being constructed simultaneously – which is a graphic intensive process in Supreme Commander). All these conditions lend themselves to a graphic-intensive benchmark; and with no mouse movements or actions performed by the user, there is very little chance of testing error.
The medium-detail settings: 1024x768, 2xAA, ‘Medium’ for all settings, background image off.
The high-detail settings: 1280x768, 4xAA, ‘High’ for all settings, background image on.


Benchmark 5 results


[Please consult graph #5 on the right]

Perhaps worthy of the title “Supreme Commander Driver”, NGO showed the best performance in this benchmark, in both quality settings. While a two point FPS lead in the medium quality settings may not seem too significant, in a game as demanding as Supreme Commander -- where a computer that almost meets the box-printed ‘recommended system requirements’ only gets 20 FPS -- 2 FPS for a change of drivers isn’t that bad at all.
For the rest of the pack, the Catalyst, DNA and NGO drivers all had similar performance, on both quality settings for this game. The DHZer0point driver performed negligibly different at the medium-quality settings, but lagged significantly behind on the high-quality setting.


The fastest driver..?


To offer a relative comparison of the driver’s performance, in each benchmark, 3 points were given for a 1st place showing, 2 points for 2nd, and 1 point for 3rd. If drivers were tied for a place-finish by less than one-quarter of a FPS, they were given the same amount of points. No points for ties were awarded in the AquaMark3 or 3DMark05 benchmarks. Also: points were given for both the higher and lower settings.

Using this system, the results of ATI Driver Roundup are as follows:

1st Place: DHZer0point 0.69 [19 points overall]
Can’t even tell you much about these drivers, but one thing is clear: they are fast. The clear winner of 3 of the 5 benchmarks – including a massive lead in the AquaMark3 benchmark – the DHZer0point drivers came up champions in this roundup. Being the only drivers based on Catalyst 6.9 didn’t seem to hurt their performance at all.

2nd Place: Catalyst 7.2 [17 points overall]
The Catalyst drivers are exactly what official drivers should be: reliable performers. They show that to achieve good performance you do not have to suffer image quality or stability.

3rd Place: Omega 3.8.330 [12 points overall]
The Omega drivers are the ‘grand-daddy’ of 3rd party optimized drivers, and for good reason. Release after release, they have offered great performance. Possibly, these drivers may be especially well-suited for first-person-shooters.

Honorable Mention: NGO 1.6.12 [10 points overall]
The Supreme Commander/NGO performance was nothing to be scoffed at. You may consider trying these drivers if Supreme Commander is in your current game rotation. Also, although it can not be easily verified, they seemed to offer superior image quality, especially noticeable in the Half-Life 2 Stress Test.

Well, that about wraps up the ATI 3rd Party Optimized Driver Roundup. As stated before, mileage may vary with these drivers; but the benchmarks do show two things conclusively: the first being that the drivers have a measurable effect on video card performance; and the second: that some drivers may be better suited for some games over others.
And they just might offer you the edge you need, in any late night run-ins with competitors; be them terrorists, supreme commanders, or titans.



















*name and clan of this horrible CS player who shot me last week changed to protect his cowardly, bunny-hopping anonymity.