Introducing the RADEON X1600

X1600 XT poses with reference GeForce 6600 GT card

A PowerColor X800 XL and the X1600 XT

While the RADEON X1600 and X1300 share the same manufacturing process and are built on the same fundamental technology as ATI’s flagship RADEON X1800, both chips are quite different than their bigger brother. Concessions had to be made in order to make both VPUs cheaper for ATI to produce. This is quite a different situation than the RADEON 9500 PRO, where ATI was essentially taking R300 chips that normally would go into RADEON 9700 cards and instead stuffing them into cheaper 9500 boards.

ATI starts by cutting the number of pixel shaders, down from 16 in RADEON X1800 to 12 in RADEON X1600. The pixel shading units are arranged into groups of four and are known as “quads”, for a total of three quads (4 pixel shaders per quad x 3 pixel quads equals 12 pixel shaders total) Each of the RADEON X1600’s three pixel quads is autonomous and contains its own dedicated branch unit, just like the X1800. Like the GeForce 6600 family, the X1600 features four texture units.

On the vertex shading side, the number of vertex units are also down, with the RADEON X1600 sporting five vertex shaders (versus eight in RADEON X1800). Interestingly enough, this is one vertex unit fewer than ATI’s older RADEON X700 family.



Like ATI’s older mainstream offering, the RADEON X700, the X1600 features a 128-bit external memory interface. We use the word “external” because like the RADEON X1800, ATI uses their new ring bus memory architecture on the RADEON X1600; only it has been scaled down in order to reduce manufacturing costs. The RADEON X1600 features a 256-bit internal ring bus, with a 128-bit external interface.

The biggest addition to RADEON X1600 is its support for 3.0 pixel and vertex shaders, a first for ATI in the mainstream segment. Like the RADEON X1800, ATI stresses threads for the X1600. Its RV530 chip breaks down the pixel processing workload into a large number of small threads. The X1600 features an ultra-threading dispatch processor, just like the RADEON X1800, only the X1600 can only handle up to 128 threads at once, while the X1800 can track and distribute up to 512 threads, four times as many as the X1600.

As a result of these cuts, transistor count in RADEON X1600 is down to about 157 million transistors, roughly half that of the RADEON X1800. As an aside, ATI also provides dual-link DVI support for the primary DVI connection on the RADEON X1600 (both DVI connectors support dual-link on the X1800). The RADEON X1600 also supports ATI’s temporal and adaptive anti-aliasing modes, as well as 3Dc and multisample AA with HDR.

The X700 PRO meets the X1600 XT

Back of the X1600 XT board

X1600 SKUs

For now at least, ATI’s announced two SKUs for the RADEON X1600: the RADEON X1600 XT, and the X1600 PRO. Both cards ship with the same number of pipelines and the same basic memory architecture, the only difference is the clocks. The X1600 XT ships with a 590MHz graphics core, while the X1600 PRO runs at 500MHz.

On the memory side, ATI clocks the X1600 XT at a blazing 690MHz, while the memory on the PRO board runs at 390MHz. Both cards are available in 128MB or 256MB memory configurations (although ATI’s RV530 chip supports up to 512MB of memory max), with prices ranging from $149 to $249. (The X1600 PRO 128MB and 256MB sell for $149 and $199 respectively, while the X1600 XT 128MB and 256MB sell for $199 and $249 respectively.)