Some of the most dramatic speed improvements in recent years have come in the video card arena. Just two years ago, running a first-person shooter at 1600x1200 would make for more of a slide show than a game on any video card of the day, and even cards a year old have trouble getting remotely playable framerates at that resolution.

Yet today's most powerful video boards can run modern games at such resolutions at playable framerates; cards based on nVidia's GeForce2 GTS chipset can turn in framerates of over 60 frames per second at 1600x1200 at Normal quality settings, which is more than adequate by most standards. But now even video cards may start suffering from the same memory speed problems that processors have endured for many years. Not only has the speed of the video chips themselves increased, but the demands of games and those that play them have increased as well. New games demand more of video cards and their memory, needing more and bigger textures, more triangles, and various other new means of hardware acceleration to look good and still run at playable speeds.

 Also, fewer and fewer gamers are willing to content themselves with lower resolutions like 640x480 and 800x600. 1024x768 now seems to be the standard resolution, and new video cards can even deliver playable, though not spectacular, 45+ FPS framerates at 1280x1024 with 32-bit color. Still, the current Holy Grail of display resolutions is 1600x1200 with 32-bit color, and even the most advanced video cards can't deliver good framerates at that setting in most games. Some can offer framerates of over 30 frames per second, and while that's playable, it's not very smooth.

Before recently, lack of video chip power prevented faster framerates, but on the newest cards and probably on future cards, memory speed is the performance bottleneck. It doesn't matter how fast the video chip itself is if the memory isn't fast enough to feed it all the textures and triangle information it needs. Workarounds like texture compression help, but they aren't enough.

This problem is most evident with GeForce2 GTS cards. Although it has a standard core speed of 200MHz compared to the 120MHz core of the older GeForce 256, it's not all that much faster. Of course, it would be ignorant to assume that simply because the core frequency increased by 66% that overall performance would increase by 66%, since the core frequency alone isn't all that matters.

Not only has the core frequency been increased, but the GeForce2 GTS has several new hardware features the GeForce 256 lacks, so these should provide an additional speed boost over the older GeForce 256. But have a look at one of the many GeForce2 GTS card reviews and compare the new chip's framerates to the GeForce 256's. You'll see that, while they are significantly higher in most cases, they aren't anywhere near 66% higher. That's because the memory of the GeForce2 GTS isn't much faster than that on the GeForce 256, or at least the DDR GeForce 256 cards. Most GeForce2 GTS cards have 166MHz DDR memory, which is the practical equivalent of 333MHz standard memory. GeForce 256 DDR cards use 150MHz DDR memory, equivalent to 300MHz standard. There are also the cheaper GeForce 256 SDR boards, which use 166MHz SDR memory, and the speed differential between these and the DDR-based boards at high resolutions is plain.

© 2023 PCSTATS.com

Next Page >