RADEON X300

Besides the X600, ATI is also releasing a new PCI Express product for the value segment, X300. Fortunately, X300 isn’t just a PCI Express version of RADEON 9200; ATI has begun with a clean slate for X300.

For starters, X300 is a true DX9 product, taking advantage of ATI’s SMARTSHADER 2.0 and SMOOTHVISION 2.1 technologies, as well as HYPERZ III. One of NVIDIA’s key advantages with GeForce FX 5200 in the eyes of OEMs is its DX9 capability, even though it really doesn’t have the horsepower to run DX9 titles in full precision with adequate frame rates. Now ATI’s value part matches NVIDIA in this respect.

Like the X600 line, X300 features a four-pixel pipeline architecture, with one texture unit per pixel pipe, and dual vertex pipelines. The core itself is clocked at 325MHz, the same clock speed as ATI’s previous DX9 value offering, the RADEON 9600 SE and NVIDIA’s GeForce FX 5200 Ultra. This puts all three cards at the same peak fill rate. Unlike the RADEON 9600 SE however, X300’s 200MHz memory (400MHz effective) sits on a 128-bit wide memory interface, giving it up to 6.4GB/sec of peak memory bandwidth, twice that of the RADEON 9600 SE.

X300 in active and passive cooling

The wider memory interface should make X300 more competitive with GeForce FX 5200 Ultra once AA or AF is applied, but still falls short of the 5200 Ultra by a factor of 4GB/sec. At the same time however, the 5200 Ultra’s NV34 graphics core isn’t the best shader model 2.0 performer, so X300 could prove to be very competitive in 2.0 shader titles such as Half-Life 2: at one point Valve was going to treat the 5200 line as DX8 hardware in HL2, meaning you wouldn’t get eye candy such as high dynamic range lighting.

ATI will also be producing a lower cost variant of X300, the X300 SE. The X300 SE boasts the same clock speeds as the X300, 325MHz core/400MHz effective memory, but relies on a 64-bit memory interface rather than the 128-bit interface found in X300. This cuts memory bandwidth in half, essentially matching the specs of ATI’s RADEON 9600 SE from last year. A low-profile, half-height card X300 SE option is also available for OEMs and card manufacturers that want to conserve space, and since power consumption and heat generation isn’t any greater than RADEON 9200, passive cooling is available as an option on low profile and full height X300 cards.

One first for X300 is its new manufacturing process: X300 is the first graphics product to be built on 0.11-micron. According to ATI, 0.11-micron provides 40% more transistors per unit area than 0.13-micron, freeing ATI to incorporate more features into the chip without needing a larger die. This also allows ATI to produce more X300 chips per silicon wafer, lowering their production costs (assuming good yields). To further reduce manufacturing costs, low-k dielectric material isn’t used at 0.11-micron.

ATI claims that the X300 board delivers up to twice the performance of RADEON 9200, although this figure admittedly comes from 3DMark 03, where the X300 gets extra points for being able to complete the Mother Nature test. In most benchmarks ATI expects the X300 to deliver just shy of 1.5 times the performance of RADEON 9200. We’ll of course provide full benchmarks once we get our hands on a card.

Like X600, the X300 series will be sold in PCI Express configurations only. ATI’s RADEON 9200 family will continue to serve the value segment in the AGP space.