| | Kyro II Processor |
| Transistors | 15 million |
| Die Process | 0.18 micron (180nm) |
| Power Dissipation | Approx 4 Watts |
| Rendering Pipelines | 2 |
| Texture Units per Pipeline | 1 |
| Memory Interface | 128-bit SDR |
| Memory Configurations | 16MB, 32MB, 64MB |
| RAMDAC | 300 MHz |
Looking at the specifications, the Kyro II appears unimpressive. In fact, its specifications look more like those of a TNT2 or Voodoo3 than a card designed to compete with today’s GeForces and Radeons. But appearances can be deceiving, particularly when it comes to Tile renderers. We’ll see just how deceiving when we get to the benchmarks.
Fig 3. The Kyro II processor, seen here on Hercules' 3D Prophet 4500.
One of the Kyro’s biggest strengths is something Imagination and ST Microelectronics like to call ‘Internal True Color’. In essence, unlike most other video cards, all z-buffering (depth-buffering) and texturing are performed on-chip, in full 32-bit color. This has two primary benefits.
Firstly, since all rendering is actually done in 32-bit color, images displayed in 16-bit color only look much better, as they’re kept in high-quality 32-bit color internally, and then dithered down to 16-bit for display. In other words, because the Kyro II renders everything at 32-bit internally, 16-bit color should look much more vibrant than other cards.
The second major benefit is that, again, because all rendering is in 32-bit anyway, there should be little to no performance difference when switching from 16-bit to 32-bit color. There will be some decrease in performance, as 32-bit textures still take twice as much memory as 16-bit textures, so if the card runs out of texture memory, it may be forced to swap textures over the slow AGP bus. Since most scenes common to today’s software don’t generally contain more than 64MB of textures, we suspect that this shouldn’t prove to be a big issue. (This is the same reason that switching from AGP 2X to AGP 4X makes virtually no difference most of the time -- texture swapping isn’t very common)
It’s also interesting to note that because of its small size and advanced process, the Kyro II dissipates only about 4 Watts of power, meaning it could be integrated into notebooks and other devices fairly easily. It will be interesting to see if any manufacturers jump on that one.
