DarkHardware

Hello and welcome to DarkHardware. It is quite possible you may never heard of us before, as we are based in Turkey and since 1997 we are providing the Turkish speaking hardware community with quality reviews and articles. Beginning from 2004, we decided to publish some of our selected reviews also in English. We are aiming to provide hardware enthuasist all over the world another source for reliable information.

Last October, our benchmark report of 33 video cards proved to be quite a success. Although that report  was only available in Turkish and contained benchmark results from just 2 tests, we had many international visitors. That is why we decided to publish the new revision of DarkHardware VGA Charts  both in English and Turkish.

Enough with the introduction. Let's begin with our little report...

What is this all about?

Just like the last time, we have gathered together a bunch of video cards and measured their gaming performance using the latest drivers available at the time. We used Catalyst 4.2 for all the Radeon boards, and Forceware 53.03 for the NVIDIA ones. For all cards, MIPMapping and other driver settings were set to the maximum visual quality levels. Anisotropic filtering and AntiAliasing were off, unless stated otherwise. In-game visual quality settings were also set to the maximum levels. Refresh rate was fixed at 75Hz, and no sound card is used except Call of Duty, as it requires a sound card to run.

It is clear that this charts are far from being the ultimate reference guide for video cards, but we think they are able to answer many of the performance related questions. We will present the charts together with some short comments about the tests and overall performance. We hope you'll enjoy the show...

Cards we used

Here is the list of cards included in this performance review. As this article is only aiming to present benchmark results, you'll find no card reviews here. On the other hand, many of the cards we used here are already reviewed in our website, and you can reach the reviews by clicking on the card name. Unfortunately, our reviews are in Turkish, as we stated at the begining.

Unfortunately, not every board available at the market is presented in the charts. We used what we have at hand, and many of the boards were loans from local companies. Some were borrowed from friends, to whom we are grateful.

Radeon 9700 and 9700 Pro are the most notable absences, but both are missing from the market since a long time, and we were unable to find a sample for our test. We considered adding a Radeon 9800SE, but as Radeon 9800SE boards vary greatly both in memory speed, GPU speed and also memory bandwidth, we decided not to confuse everyone.

And the results...

3DMark2001 SE

Although can be considered quite old, 3DMark2001 SE is still highly regarded by the hardware enthuasist all over the world. Many of our forum readers also use 3DMark2001 SE for performance comparisions, so it is necessary for us to include 3DMark2001 in our tests.

Radeon 9800XT is leading the chart, with 9800 Pro following close. 9800XT's score of 17048 is quite impressive, and it is clear that with some serious overclocking, 20000 points mark could be tried. NVIDIA's FX 5950 Ultra is not bad either. What is interesting is how close FX 59xx series score, compared to each other.

One notable point is the big performance difference between Intel's own integrated graphics core, and ATI's new Radeon 9100IGP chipset. We are really wondering when Intel will realise they can not go on forever with the ancient i740 based graphics. We really hope things will change with the oncoming i915/i925 chipsets.

When 4x AntiAliasing is enabled, the higher ranks of the charts change little, but there are some notable changes in the middle. While older NVIDIA chipsets, GeForce 4 and GeForce 3 series were performing really good without antialiasing, their performance drops considerably with antialiasing enabled. Another point to note is the performance drop of FX 5700 Ultra. Without antialiasing, FX 5900XT and 5700 Ultra are very close, but when AA is enabled, 5700 Ultra'a 128-bit memory bus shows its limitations.

3DMark03

This is the first time we included 3DMark03 in our quaterly performance round-ups. As we want to stay away from the endless debate about optimization/cheating, we decided to present 3DMark03 results for ATI and NVIDIA cards in seperate graphs. In the past, we have used this approach during the AMD-Intel-SysMark2002 mess by not using the SysMark2002 to compare CPU's from different brands. We are planning to keep having a cautious stance towards 3DMark03, until someone manages to bring some sanity into benchmarking scene.

Only DirectX 9 class charts from both companies are included in these charts, to make the comparision a little bit more meaningful.

ATI Radeon series display a fairly regular performance increase in accordance to the card price. It would be great to see how would Radeon 9500 Pro and 9700 Pro fit in that chart, but unfortunately wo don't have any samples of these two at the moment.

There is not much to say here, other than the really low performance of FX 5200, and the almost similar performances of plain FX 5600 and FX 5700 cards. Again, the difference between 5700 Ultra and 5900XT is appearent.

Aquamark 3

Aquamark is a very useful tool for video benchmarking,. Although it is able to run and produce a benchmark result even on the entry level card, we left out DirectX 7 level cards from this graph. The benchmark settings were unchanged and we only used the "Standart Measurement" option.

There is something funny in that results. The arch-rivals, Radeon 9800XT - GeForce FX 5950 Ultra, and Radeon 9800 Pro - GeForce FX 5900 Ultra get almost similar results, creating a head-to-head situation. The plain version of FX 5700 is scoring better than 5600 Ultra, probably due to the better shading abilities of the improved NV36 core. GeForce 4 series are scoring well, as the benchmark uses Pixel Shader 1.3 effects while testing these. That is why the lowly FX 5200 is scoring  worse than integrated Radeon 9100IGP, as the 5200 have to struggle with Pixel Shader 2.0 shaders, while 9100IGP is enjoying easier to handle Pixel Shader 1.4.

These all were the synthetic benchmark results. In the following pages, we'll see the real game benchmarks...