Related Articles

Specs

Do the clockspeeds look somewhat like another familiar part? The 9600XT maybe? With the R3xx series, ATi took a route that was a lot less riskier than Nvidia- they banked on proven technology with 0.15 micron parts for their high end and went for the riskier venture with their mainstream parts, the 9600 Pro and 9600XT. The 9600 Pro validated ATi's 0.13 micron architecture, while the 9600XT improved upon the Pro with the inclusion of a low-k dilelectric which reduced signal crosstalk and allowed ATi to ramp up speeds even higher. Fast forward 6 months to the present day - ATi's current high end parts are now using the 0.13 micron low-k dielectric process. ATi claims that low-k has enabled them to improve clock speeds by roughly 10%. Both the X800 Pro and the X800 XT have high clockspeeds comparable to the 9600XT. Besides the increase in raw clockspeed, the X800 XT and the X800 Pro have 8 and 4 more pipelines than the 9800XT respectively. The raw fillrate of the X800XT is more than twofold over the 9800XT, 8.3 Gigapixels compared to 3.3.

Memory Clock

Memory clocks have also increased significantly over the previous generation especially with the XT. Like the NV40, ATi is now using GDDR3 on their graphics cards. (A coincedental sidenote is that ATI is the company that pioneered GDDR3).

One of the many interesting comments made at Technology Days is the growing divide between memory speeds and core clock speeds. Bob Drebin, ATi's director of Hardware Engineering remarked that the X800 Pro is more of a balanced part than the X800XT because of the memory bandwidth limitations- even GDDR3 at 1.12 Ghz is not enough to feed it. This will likely not change with 3rd Party board makers as ATi is encouraging them to stick to stock speeds with regards to both the core and the memory. 3rd parties do however have a free hand in deciding what speed of memory to use so there could possibly be cards with faster memory timings.

GDDR3

ATi spearheaded the effort into bringing GDDR3, the next generation standard for graphics memory, into the market. A brief couple of key points:

- GDDR3 requires lower power than GDDR2, which translates into higher clock speeds
- Simplified clocking - a simplified clocking scheme also allows for higher clock speeds

The end result? Faster memory and lower heat output. Remember we mentioned that the heatsink on the backside of the 9800XTs was dropped from the X800? There is a lack of a heatsink AND the clockspeed of the memory is higher.

One of GDDR3's big improvements over GDDR2 is because it is actually an industry standard. There were several variations of GDDR2 which made production expensive because graphics vendors would have to specially order GDDR2 chips, hence batches were relatively small and costs were relatively high. There is a lot of technical info details about GDDR3 but the relevant parts are that it is standardized meaning all graphics vendors requesting GDDR3 will get the same type of chips, GDDR3 will run faster and cooler than GDDR2 chips . ATI is trying to nail down the specs for GDDR4 by year's end and it will feature more incremental improvements like GDDR3 has on GDDR2.