facebook rss twitter

Review: ATI Radeon 64 Meg DDR OEM

by David Ross on 19 August 2000, 00:00

Tags: ATi Technologies (NYSE:AMD)

Quick Link: HEXUS.net/qaz

Add to My Vault: x

Results

The following review compares my old 32 meg TNT2 (Oced to 147/205) with the Radeon, which is of course not a fair comparison. The point of this comparison is not as a head-to-head, more an idea what kind of improvement upgraders will get.

MDK2

The test: This is the standard ‘Test settings’ timedemo that comes with MDK2, which uses OpenGL as the API.

Settings: In all cases below I have used 32 bit colour, bilinear filtering. With the Radeon, I have enabled Hardware T&L. EAX sound is ON and texture quality is at MAX (i.e. far right on slider)

800x600 (TNT2 - 64 Radeon - 76)

1024x768 (TNT2 - 44 Radeon - 71)

1280x1024 (TNT2 - 15* Radeon - 54)

* Note: At 1280x1024 ithe TNT2 seemed to either run at 32 fps or 7 at any given point in the demo. Texture memory limitations seem to be a problem at this high resolution.

MDK did exhibit some strange graphical crap on the screen for a few seconds after it closed down each time on the Radeon, but it cleared itself pretty quickly. I have read on some messageboards that some people have graphical artifacts in-game on MDK2, but none were apparent on my system.

Unreal Tournament (ver 4.28)

The test: This is the Wicked405 demo that comes from The Pulpit

Settings: D3D rendering, max world detail, medium model detail, 32-bit colour, full HUD, 3d sound on with high quality sounds. No tinkering with .ini files other than that.

TNT2 Radeon

Min Max Ave Min Max Ave

800x600 24 49 37 24 53 36

1024x768 19 47 33 23 37 35

1280x1024 15 33 22 22 51 33

Going by figures alone, not massively impressive really. But UT benchmarks are notorious for being limited by the processor speed and by the flakey d3d.dll used by UT. What is very worrying is the persistently low min FPS scores in UT. I have played for a while with UT and Deus Ex (UT engine) and the actual gameplay does seem a lot better with the Radeon, even though the benches don’t show it. The biggest improvement is in the handling of models, where UT has a tendency to freeze up for a period when a new player joins or translocates in.

Villagemark

The test: This is the only synthetic benchmark I am going to use as it should show the effectiveness of the Radeon’s HyperZ technology. In short, the demo runs like a first-person shooter with you walking round a village, but the scene has absolutely masses of overdraw. To get a view of some other benchmarks run on it (including a Kyro it was designed for and a Geforce 2), go to the download page at Beyond3D

Using default benchmark 1 settings (1024x768, trilinear filtering on, 16 bit colour)

TNT2 - 15

Radeon - 32

Not having any other cards to compare, it is difficult to tell how good this is. All I can compare with other than the TNT2 is some figures off the link above –

‘An NVIDIA GeForce 256 DDR on an Athlon 700 runs the demo at only 21 fps and an NVIDIA GeForce2 GTS DDR runs this demo at 32 fps on a P3-750 system.’ While still a very good score, I thought the Radeon might have beaten the Geforce at this, not just drawn (although I do have a slower CPU)