ATI Radeon X850 XT PE and X800 XL

Introduction

Earlier this autumn, ATi revealed details about their new PCI-Express video cards, the X850 XT Platinum Edition (PE) and the X800 XL. While the X850 XT PE of course is the new high-end video card, the X800 XL is ATI’s answer to NVIDIA’s GeForce 6800 GT. ATI was kind enough to lend us two reference cards for a quick spin in some of our benchmarks.

Technical Information

Both the X800 XL and the X850 XT PE use the new PCI-Express slot.

The main difference between the R480 chip in the X850 XT PE and the R423 chip in X800 XT’s is ‘process enhancements’. These have made it possible for ATI to clock it slightly higher than the X800 XT PE. There’s a bigger difference between the R430 than can be found in the X800 XL and the R423. While the R423 is fabbed with a 0.13 um low-K process, the R430 is fabbed with a 0.11 um process. It cannot reach the clockspeeds of the R480 but runs cooler. In the end, it is the perfect chip to power the cards that ATI wants to rival the successful GeForce 6800 GT.

The Cards

The X850 XT PE

As you can see, this is the first Radeon card that is dual-slot from the start.

The fan is big and makes the X850 XT PE a 2-slot solution

The fan is big and pretty noisy for a few seconds when you start. It never reached any high volumes while using it though.

This card needs more power than the PCI-E slot can provide and thus has a power connector. This connector is of the new type that you see on PCI-E cards, and thus you cannot just connect it to your old power supply unless it is of the new type that has that special connector type. Luckily, ATI realizes this and supplies an adapter that lets you use a normal 4-pin Molex-connector.

The memorychips are covered by the fan/heatsink

The X800 XL

In contrast, the X800 XL really looks like the older ATI cards.

Even though it has the same amount of pipelines like the older X800 XT cards as well as just being slightly lower clocked, it still does not need any additional power thanks to the more power-efficient 0.11 um process.