Our testing methods
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least twice, and the results were averaged.
Our test systems were configured like so:
|VIA KT333||Intel 850E|
|Processor||AMD Athlon XP 2200+ 1.8GHz||Intel Pentium 4 2.53GHz|
|Front-side bus||266MHz (133MHz double-pumped)||533MHz (133MHz quad-pumped)|
|Motherboard||Epox 8K3A+||Asus P4T533C|
|Chipset||VIA KT333||Intel 850E|
|North bridge||VT8367||82850E MCH|
|South bridge||VT8233A||82801BA ICH2|
|Chipset drivers||VIA 4-in-1|
|Intel Application Accelerator 6.22|
|Memory size||512MB (1 DIMM)||512MB (4 RIMMs)|
|Memory type||Corsair XMS3000 PC2700 DDR SDRAM||Samsung PC1066 Rambus DRAM|
|Sound||Creative SoundBlaster Live!|
|Storage||Maxtor DiamondMax Plus D740X 7200RPM ATA/100 hard drive|
|OS||Microsoft Windows XP Professional|
The test systems' Windows desktops were set at 1024x768 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
We used the very latest graphics drivers available online at each chip makers' website as of 8/17/02. Both systems were configured with AGP 4X enabled and 256MB AGP aperture sizes. Fast writes was enabled on the KT333.
Beyond the test rigs listed above, we managed to get test results from a couple of different systems and cards. One was a workstation-class box with dual Athlons on a 760MPX-based motherboard and a Quadro 750GXL with 128MB of DDR memory. The other was a Radeon 9700-equipped system with an Athlon XP 1800+ and a KT333-based motherboard.
All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Here's how each of the cards we tested fared on the Serious Magic benchmark. We're reporting the results in megabytes per second, because that's the most important number for our purposes.
As you can see, pretty much every card we tested, regardless of chipset or motherboard, turned in an abysmal performance. The benchmark's 720x480 32-bit test images, which are about 1.3MB each, would only download at about 8 frames per second in most cases.
To put these results into perspective, a single frame rendered at 1600x1200 with 32 bits of color weighs in at about 7.5MB. Double that to 64-bit color, and it's 15MB per frame. And a single image at 1600x1200 in 128-bit color is over 30MB. That means even if you can render high-quality images at 30 frames per second, you won't be able to get them out of the graphics card at anything near that rate.
At least, not until graphics card driversand perhaps chipset drivers and whatever elseimprove dramatically at AGP downloads. Right now, even the very latest graphics cards aren't ready to do much more than play games and put pretty pictures onscreen. If graphics companies really want to replace CPUs for professional rendering, they've got a bit more work to do.
116 comments — Last by Anonymous at 11:10 AM on 06/14/03
|Corsair's Hydro GFX GeForce GTX 1080 Ti graphics card reviewedNo assembly required||15|
|The Tech Report System Guide: May 2017 editionRyzen 5 takes the stage||111|
|Aorus' GeForce GTX 1080 Ti Xtreme Edition 11G graphics card reviewedThe eagle has landed||35|
|AMD's Radeon RX 580 and Radeon RX 570 graphics cards reviewedIteration marches on||162|
|Nvidia's GeForce GTX 1080 Ti graphics card reviewedI like big chips and I cannot lie||191|
|Where minimum-FPS figures mislead, frame-time analysis shinesA new way to go Inside the Second||250|
|Aorus' GeForce GTX 1080 Xtreme Edition 8G graphics card reviewedFlying high||29|
|The curtain comes up on AMD's Vega architectureRadeons get ready for the workloads of the future||156|
|AMD plugs Radeon ProRender into Blender||4|
|Camera Day Shortbread||3|
|Micron lets all memories of Lexar fade away||7|
|Ryzen Pro platform brings a dash of Epyc to corporate desktops||17|
|Corsair's Hydro GFX GeForce GTX 1080 Ti graphics card reviewed||15|
|Qualcomm hides a fingerprint scanner under your screen||15|
|Toshiba prepares a 96-layer 3D NAND parfait||16|
|Baidu's DeepBench can now measure inference performance||8|
|Toshiba QLC 3D NAND squeezes a fourth bit into flash cells||26|
|So they're part of a fire sale?||+38|