Rage3D.com
Click for Page List Next Page
Print Article Discuss Article
Product : Radeon X1800 XT
Company : ATI Technologies
Author : Ryan "MrB" Ku and Mark "Ratchet" Thorne
Editor : Eric "Ichneumon" Amidon
Date : October 13th, 2005

Radeon X1000 Family

ATi has launched an entire family of products based on the X1000 family architecture. The X1800 line covers the enthusiast segment, the X1600 fills in the middle performance market, and the X1300 rounds off the mainstream consumers. This is an industry first to have launch products that cover the whole spectrum, with the entire line sharing the same basic DNA, and therefore the same feature set.  The name itself I find a little humorous. ATi jumped from 9800 to X800 where X kind of stood for Roman numeral 10. By going to X1000 it’s like 101000, but I digress.

The X1800 XL and the X1300s should be on store shelves as you read this, with the ultra high-end X1800 XTs available in about a month, and the mid-range X1600 XTs and X1600 Pros available at the end of November.

Here's a nice chart outlining the new lineup:

Model Memory Amount Core Speed Memory Speed Interface PS VS SM ETA MSRP ($US)
X1800 XT 512 MB 625 MHz 1.5 GHz 256-bit 16 8 3.0 11/5 $549
X1800 XT 256 MB 625 MHz 1.5 GHz 256-bit 16 8 3.0 11/5 $499
X1800 XL 256 MB 500 MHz 1.0 GHz 256-bit 16 8 3.0 Now $449
X1600 XT 256 MB 590 MHz 1.38 GHz 128-bit 12 5 3.0 11/30 $249
X1600 XT 128 MB 590 MHz 1.38 GHz 128-bit 12 5 3.0 11/30 $199
X1600 Pro 256 MB 500 MHz 780 MHz 64-bit 12 5 3.0 11/30 $199
X1600 Pro 128 MB 500 MHz 780 MHz 64-bit 12 5 3.0 11/30 $149
X1300 Pro 256 MB 600 MHz 800 MHz 128-bit 4 2 3.0 Now $149
X1300 256 MB 450 MHz 500 MHz 64-bit 4 2 3.0 Now $149
X1300 128 MB 450 MHz 500 MHz 64-bit 4 2 3.0 Now $99
X1300 HM 32/128 MB 450 MHz 1.0 GHz 32-bit 4 2 3.0 Now $79
PS = Pixel Shader Processors, VS = Vertex Shader Processors, SM = DirectX Shader Model Version

In designing the X1000 architecture ATi looked at a number of industry trends:

  • Resolution
    Today’s resolutions have not really moved past 1600x1200 for the large majority of users. They actually have dropped due to the popularity of LCDs, with the majority having native resolutions of 1280x1024. Fill rate is now less important, therefore increasing the number of pixel pipelines is an inefficient use of die space as its gains aren't fully utilized.
  • CPUs
    Improvement in pure CPU speed has been minimal with the movement towards multi-core processes. In most scenarios, application performance is CPU-bound and not GPU-bound. Until software takes advantage of multi-core processors, CPUs will be limiting the GPU which means other ways will need to be found to improve IQ per frame on existing CPUs.
  • Shaders
    Next generation games are using more and more shaders that are increasing in complexity. Next generation consoles support SM 3.0 resulting in an even greater number of SM 3.0 games. GPUs are becoming more shader-bound than fill-rate bound and therefore focus should be directed to shader performance
  • HDR
    HDR was the most popular word of the event, mentioned even more than ATi. HDR is the next big thing. It requires full precision FP rendering capability and requires plenty of memory and bandwidth.
  • Scalable Graphics
    GPUs working in parallel with other processors is a hot topic. SLI and CrossFire are on everyone’s wish list.

These trends led to two key priorities in designing the X1000 architecture in my opinion. First, the largest priority was designing a very efficient architecture, allowing the GPU to continually process the right work at the right time.  Second, as shaders are becoming one of the main performance bottlenecks, providing Shader Model 3.0 support and maximizing shader performance is becoming more important.

At the foundation of this new family is the latest manufacturing process. All X1000 products are manufactured at Taiwan Semiconductor Manufacturing Co using a 90 nanometer (nm) fabrication line. This allows for double the number of transistors being packed into the same area, compared to a 130nm process. Ultimately this creates the opportunity for faster clock speeds while using less power.

The X1800 weighs in at an industry leading 321 million transistors, over double the X850 and a bit over the 7800. At 157 million transistors, the Radeon X1600 has the same transistor count as the X850. Finally the X1300 is the first entry level GPU to break 100 million transistors, which is roughly comparable to the Radeon 9700. It therefore can be seen to have taken roughly 3 years for the complexity of a high-end chip to reach the mainstream level at launch, which is fairly impressive in thinking about it. Technological progression is a wonderful thing.

A large majority of the transistors have gone into bringing the shader engine up to 3.0 level, and implementing a new memory controller. ATi focused heavily on efficient dispatch logic with this new product line.





Radeon X1800 Architecture Diagram
Radeon X1800 Architecture Diagram

Radeon X1600 Architecture Diagram
Radeon X1600 Architecture Diagram

Radeon X1300 Architecture Diagram
Radeon X1300 Architecture Diagram


All graphics and content on this page are ©2022 Rage3D.com
Reproduction in whole or in part is strictly prohibited.