| ||||||
| ||||||
|  |
 
INNO3D Tornado FX5700 Ultra with DDR-IIReview Date: April 19, 2004 Reviewed by: Ashley E. Glenn Sponsored by: InnoVISION
The following computer was used for testing:
The benchmarks used for this review include 3DMark 2001 SE and 3DMark03, both with their most recent patch; Wolfenstein: Enemy Territory; Serious Sam: The Second Encounter; Unreal Tournament 2003; and GunMetal. Enemy Territory makes use of 3DCenter’s Battery Demo. Serious Sam: SE uses the built-in performance diagnostic and the Grand Cathedral demo. Unreal Tournament 2003 was benchmarked with the UT2k3 Bench tool; the BCC Botmatch and [H]ardOCP Flyby benchmarks were used. There has been question in the past concerning the use of Futuremark’s benchmarking tools with certain drivers. The 53.03 drivers were WHQL approved and readily available from NVIDIA’s website at the time of the benchmarking. Furthermore, as this card is benchmarked against itself and not against another brand’s card, there is no driver bias and the results stand by themselves. These benchmarks are provided solely to show the performance of the card in different situations and not to create bias toward or against another card or brand. We’ll be starting off with Futuremark’s 3DMark 2001 SE. While not a young benchmark by any means, it still accurately represents a majority of current games as the DX9 standard is still slowly creeping into the mainstream of distributed games.
Starting off at 1024x768, the card performs nicely. It pulls in a score of over 14,000 3DMarks; this breaks down to performance of over 200 frames per second in each of the lower-detailed game tests and some rather scattered performance at the higher detail levels. The card reaches 73.5 frames per second in the nature demo.
Moving up to 1280x1024, the card takes a slight performance hit but still holds itself to more-than-playable speeds. The total score of 11,389 breaks down to an average of around 170 frames per second in the lower-detailed game tests an average of around 80 frames per second in the higher detailed tests. The Nature scene shows a significant performance drop, dipping to 47.6 frames per second. This was still smooth and without jumps or stutters, and is definitely at an acceptably playable speed.
At 1600x1200, the card begins to show its limits. The lower detailed game demos average around 130 frames per second, while the higher detailed game demos are all slightly above 70 frames per second. Nature reaches 33.6 frames per second. This last score is approaching what too many consider the “playable” cutoff zone. There were a couple pauses and jitters, but none that severely affected the flow of the demo. These scores combine for a score of 9012 3DMarks at this resolution.
As these benchmarks are in theory supposed to represent how a card should perform in real situations, we can make a couple assumptions at this point. First, DX8 games shouldn’t phase this card much at 1024x768, and even 1280x1024 should be fine for many older games. 1600x1200 might be a stretch for some titles, but should be good for others. We’re going to take a look at what happens when we crank up the eye candy in a little bit, but for now it’s time to move on to 3DMark03. 3DMark03 is built to test how well a card performs in next-generation games. I have to use the qualifier “next generation” as I currently haven’t seen any games on the common market that make use of this level of detail throughout the entirety of the game. They are starting to appear, and therefore this benchmark tool is becoming a little more relevant in deciding if a card is up to snuff or not.
The first result we’re greeted with at the end of the 1024x768 test is the aggregate score – 3915. This breaks down as follows between the following game tests: Wings of Fury, 154.6; Battle of Proxycon, 25.8; Troll’s Lair, 21.2; Mother Nature, 21.6. Wings of Fury was an eye-catching flight, but that’s about it for this benchmark. Even at this resolution, the final three game tests barely broke 25 frames per second. At this resolution, Battle of Proxycon and Troll’s Lair were full of jagged edges. These should diminish at higher resolutions, but at somewhat of a performance cost.
As I mentioned in the last paragraph, bumping up the resolution to 1280x1024 causes a performance hit of close to 40% in most cases. Wings of Fury drops from 154.6 to 113.1 frames per second. Battle of Proxycon looks a lot better at this resolution, but at a penalty of over 8 frames per second; this leaves the demo as an occasionally choppy experience. The same can be said for both Troll’s Lair, which loses 6.5 frames per second, and Mother Nature, which loses 5 frames per second.
At 1600x1200, the card has hit the performance ceiling. Not much can be done past Wings of Fury, which turns in a still-respectable 86.9 frames per second average. Unfortunately, the performance for the remaining three game demos is abysmal. I won’t ponder here what can be done with further driver releases, as that seems to be taboo when it comes to Futuremark’s benchmarks, but there is hopefully some room left to up this card’s performance a bit in the higher end of things.
|
 |
|
| |
| |
|
