Navigate:

Advice
Beginners
BIOS Guide
CPUs
Links
Mainboards
Memory
Network
Storage
Video/Sound Cards

Contact
Forum
SiteMap
Sponsors
WebNews
Home
. .

Prices:

Mainboards

ABIT
ASUS
Chaintech
Shuttle
Soyo
Tyan

CPU
Intel
P4 2.4C-800
P4 2.6C-800
P4 2.8C-800
P4 3.0-800
P4 3.2-800

AMD
AthlonXP
XP 1700+
XP 2000+
XP 2400+
XP 2500+
XP 2700+
XP 3000+
XP 3200+

Athlon64
Athlon64 3200+
Athlon64 FX-51

Opteron
Opteron 240
Opteron 242
Opteron 244
Opteron 246

Memory

Corsair
Crucial
Kingston
Mushkin
OCZ

What are you
shopping for?



































































































































































LOSTCIRCUITS

SHORTCUTS:
The Way to High Definition Gaming
Pixel Pipelines and GDDR III
3Dc Technology
Sparse Sample Antialiasing
AGP8X vs. PCI Express

Give us some feedback

 ATI RADEON X800 Preview
(Review by MS, May 4, 2004)
ASUS RADEON
9600XT At:

3Dc-Compression Technology

The key feature of the new X800 family is the so–called 3Dc compression technology where the acronym 3Dc stands for 3D compression. What this comes down to is the same technology that was used in ATI’s car demo a few years ago, meaning that instead of generating very complicated geometry models, a simple model can be used and the geometry level of detail is added by means of a Normal Map that is then used to generate the 3Dimensional bumps.


Normal maps are a very interesting technology in that the geometry setup can be simplified and the different surface structures can be added to add high structural detail to a rather bleak surface. The only issue with normal maps is that they are a huge hog of memory and, by extension, bandwidth. This is where the compression algorithms come in for a lossless 4:1 compression of normal data.

Even with 256 MB of local frame memory on board, without compression, the normal maps would basically take over. Of course, there is the possibility to increase the frame buffer to the next level but aside from the price issues associated with that, there is also the problem that frame buffer needs to be represented in the virtual address space of the system memory. In other words, more graphics memory eats away a larger chunk of the system memory. Keep in mind that we are talking about virtual address space here rather than physical addresses and, therefore, it is not a 1:1 translation but a dynamic allocation of resources but nonetheless, above a certain size of frame buffer, this can become a problem.

One other interesting feature of the RADEON X800 architecture concerns the Antialiasing techniques used. Among the techniques used are Centroid sampling to avoid bleeding of surrounding colors into the anti-aliased pixel blocks. More interesting is the so-called temporal sampling, which includes programmable sparse sampling patterns that are then alternated. In other words, instead of taking e.g. 8 samples, only four samples out of these eight are taken per frame, however, the sample points will vary in a preprogrammed pattern between frames. That means, that either four sparse samples will alternate with four different sparse samples or else, there will be overlap between the two samples in any possible pattern and degree.

Sparse sample patterns are a very elegant way of doing antialiasing, since they greatly reduce the bandwidth requirements. On the other hand, this technique requires frame rates that stay above the critical flicker frequency of the human eye, which varies with lighting of the environment as well as with retinal eccentricity (from the fovea centralis). As a rule of thumb, for this type of antialiasing, the frame rates should not drop below 45 frames per second.

At launch the RADEON X800 family will feature an AGP 8X interface, but we will see PCI Express versions (AGP 16x equivalent) as early as the third week of may, given support of the interface by the new Intel platforms. This, again, raises the old question about the sense or nonsense of the new interface. At this point, it is clear that there is no benefit for any gaming application in moving beyond AGP 4X, as a matter of fact, we have not been able to see any performance benefit even going from AGP 2X to AGP 4X. The reasons are very obvious, the only data that really need to be transferred through the AGP interface are the geometry data that come from the CPU. Everything else, especially the always misquoted texture bandwidth requirements is no longer valid, since on-board graphics memory has increased beyond the 4-8 MB we used to be stuck with and compression algorithms add their share in removing traffic from the AGP interface.

Looking back at the feature set we just covered, we see an even greater simplification in geometry (at least on a relative scale) with normal maps being substituted for high vertex counts and all of this translates again in a huge redundancy of the AGP path.

Of course, this only holds for gaming, which, coincidentally is the prime target market for the RADEON X800 but there are other applications as well, for example video editing, especially real-time rendering and that is where one may potentially see some benefits of greater AGP bandwidth.

Next Page:    => More =>

If you enjoyed reading this article and found it useful, please consider making a small donation to LostCircuits.
Thank you!

General disclaimer: This page only reflects the author's personal opinion and assumes no responsibility whatsoever regarding any of the contents or any damages that may occur explicitly or implicitly from reading the contents of this site. All names and trademarks mentioned in this review are the exclusive property of the respective parent companies.
All contents of this site are protected by international copyright laws. Reproduction of the contents even in parts is not allowed except after written permission by the author and referral to this site.
Copyright 1998 - 2008 LostCircuits