RSS Feed

HCW Tech Blog

For the latest info on computer hardware, tech, news, video games, software tips, and Linux, check out our new improved front page: HCW Tech Blog

Reviewed by: Bryan Pizzuti [01.10.01]
Manufactured by: Hercules
MSRP: $59.99 (!)


3D Mark 2000

As expected, the Kyro can't really beat either of the other cards here. No surprise, since it is based on a predecessor chip. The performance is fairly respectable, and scales well as resolutions go up. The original Kyro chip did validate the architecture, but it gets left behind by more current video cores. The key selling point of the Kyro architecture is that performance doesn't scale down as fast when resolution goes up as it does with other cards.

Scalability is important in a card, since resolutions always end up going up with time. Both Kyro chips start off lower than the GF2MX core, true, but they lose much less performance when you scale the resolution up, while the MX core seems to have a much steeper drop.

3D Mark 2001

Since NONE of these cards have hardware support for DirectX8, these low scores are all you should expect. Here they Kyro's lower performance doesn't really make it stick out as much, since full DIrectX8 performance is pretty pitiful on all of the tested cards. Note that the only TRULY DirectX8-compliant cards out there are based on the Radeon 8500 and up, and the GeForce3 series. This benchmark uses an UNBELIEVABLE number of polygons, and doesn't really hide any of them behind other polygons, so raw GPU performance and a strong T&L engine is a must. The Asus card, while it has the T&L engine, it only has about half the raw performance of a GeForce2GTS, so it can't measure up well either. It still performs COMPARATIVELY well, since this mark focuses on strong T&L, but it's not at what would be called a playable level (and it failed to complete the high-res tests). The Kyro2 performs strongly considering its limitations, but it does show that the Kyro architecture is going to need T&L soon to stay competitive.

The 4000XT just can't keep up at all. In many of the individual tests, it averaged 1 FPS, and had a lot of trouble displaying textures in the high-detail tests. It even forgot the sky in the dragon demo a couple of times, because it just couldn't keep up. The 4000XT probably wouldn't be a good idea for games that ultra-stress the graphics card like this.


This benchmark was actually developed by PowerVR to show off the performance of their Kyro chips. Like 3DMark2001, it generates a LARGE amount of polygons. Unlike 3DMark2001, in most places many of said polygons are hidden from view. This is a good test of GPU and graphics memory EFFICIENCY, since something that tries to brute-force all of these polygons will bog down fast due to the GPU, or a lack of memory bandwidth. A GeForce3 with its crossbar memory bus could probably do well (someone want to provide us with a score?) but a GeForce2MX, as we see here, drowns. The Kyro does well again here, though not as well as the Kryo2. It does come close in 800x600 resolution, but as the resolution scales up, the Kyro's 115 MHz GPU just can't keep up as well. It just doesn't seem to be able to scale up as well as its big brother, the 4500. By the way, this benchmark DOES support a hardware T&L engine, even though Kyro chips currently don't have one. So no one can really say that the deck is stacked against other chips here. This sort of benchmark is what many FPS style games are like in indoor areas, where you're clearing passages and rooms.

Next Page: (4)