Choosing a Next Generation Radeon Preview
Aug 14 '01
The Bottom Line Like the original Radeon, the 8500 has potential. The drivers just can't harness its true power.
When ATI introduced their Rage6C/R100 core, today known as the Radeon, it boasted a feature-set that far surpassed the leader in its class - the GeForce2 GTS. With its Charisma Engine, HyperZ, iDCT, 183mhz DDR memory, three texture units per rendering pipeline, and semi-DirectX 8.0 functions the Radeon was considered a marvel. And at the time, it was true... at least on paper. But in reality, it still couldn't compete with a GeForce2 GTS. With ATI's drivers unable to take advantage of the technology, Radeon was forced to take second place. Radeon turned into a product with potential that could never be harnessed. Today ATI previews its sucessor to the Rage6C/R100 core: the highly anticipated R200, now named Radeon 8500. On paper, the GeForce3 has no chance against the Radeon 8500. But history has already proven that if ATI's drivers aren't able to take advantage of the technology, the Radeon 8500 will turn from a winning product into one that merely has potential. Can ATI's drivers take advantage of the technology and should you choose an ATI Radeon 8500 over an nVIDIA GeForce3? Read on and find out.
Like I said, the original Radeon boasted a lot of GeForce2 killing features. The Radeon 8500 carries on tradition, but with GeForce3 killing features instead. One of these features is core and memory clock. The orginal Radeon fell 17mhz short of GeForce2 GTS's 200mhz core clock. This time ATI not only matched the GeForce3's core clock, but also exceeded it. GeForce3, like GeForce2 GTS, is clocked at 200mhz. The Radeon 8500's core is clocked at 250mhz, that's about 37% faster than the original Radeon and 25% faster than the GeForce3. But if you're into graphics cards, you'll know that core clock is not the factor holding the video chipset back, it's the memory speed. The original Radeon beat the GTS's 166mhz DDR memory with 183mhz DDR memory. The Radeon 8500 continues this tradition by beating GeForce3's 230mhz DDR memory with an amazing 275mhz DDR memory, giving it about 20% more raw memory bandwidth than the GeForce3! This is a brute force technique we've seen with the GeForce2 Ultra, clocked at 250mhz and coupled with 230mhz DDR memory. Even with HyperZ-like memory management features, the GeForce3 still can't compete with GeForce2 Ultra's brute force approach on certain non-DirectX 8.0 games. Now ATI is using this approach along with a very effecient HyperZ II memory management system. With the HyperZ II, the Radeon 8500 doesn't use all it's raw power stupidly like the GeForce2 Ultra, but rather it uses it effeciently. With both brains and brawn, the Radeon 8500 is sure to defeat the GeForce3... if it's drivers can take advantage of the technology.
I've already touched on HyperZ II talking about memory speed, so the second GeForce3 killing feature is HyperZ II. The original HyperZ is a technique used to determine what will not be seen in the scene (pun intended). In real life 3D, you can't see behind the monitor you're looking at now because my commentary is blocking it. Traditional 3D rendering techniques would render the back of your monitor whether you see it or not anyway, wasting valuable memory bandwidth. HyperZ will determine you're not going to see the back of your monitor and not waste priceless resources rendering it. The original HyperZ is comprized of three parts: Hierarchical Z, Z-Compression, and FastZ Clear. Hierarchial Z determines what pixels will and what will not be seen and disposes of the useless ones. Z-Compression uses a lossless compression algorithm to compress data and, in turn, conserve memory bandwidth. FastZ Clear is a speedy z-buffer clearing method. HyperZ II improves upon Hierarchial Z by using 4x4 blocks instead of 8x8 blocks, making it more efficient since smaller blocks means you have a finer brush. It's much like drawing with a bulky crayon and drawing with a fine pencil - the pencil is more accurate. The new Hierarchial Z can also discard 64 pixels per clock instead of 8 like the original. The last improvement is to the Z-Compression algorithm. ATI claims a 20% improvement over the original Z-Compression algorithm. Being able to discard 64 pixels per clock beats GeForce3's 16 pixels per clock and nVIDIA has nothing that matches Fast Z Clear, but it's still unclear whether HyperZ II is better than nVIDIA's Light Speed Memory architecture along with their powerful memory crossbar controller. We'll just have to wait until the Radeon 8500 is really released and hope ATI can take advantage of the technology.
ATI didn't stop the improvements with HyperZ, they also improved upon their Charisma Engine. And like HyperZ II, ATI call's the improved T&L engine Charisma Engine II. Charisma Engine II adds ATI's response to nVIDIA's nFiniteFX Engine, SMARTSHADER technology. There may be a little ambiguity in HyperZ II's ability to counter GeForce3's memory management features, but SMARTSHADER takes the crown in programmable vertex and pixel shaders. nFiniteFX is compliant with DirectX 8.0 shaders, while SMARTSHADER complies with DirectX 8.1 shaders. This makes Radeon 8500 able to apply six textures in a single pass, beating GeForce3's 4 textures per pass. This may sound nice, but there is no game that takes advantage of even GeForce3's nFiniteFX engine. We're now seeing games that finally take advantage of GeForce2's non-programmable T&L engine. It's going to be a while until we see anything take advantage of either ATI's or nVIDIA's programmable pixel and vertex shaders but when they do, the Radeon 8500 has the upper-hand. Now they just need to harness Radeon 8500's power with well written drivers.
While improving upon existing technology, ATI also managed to bring in new visual quality enhancing features. Two of these are TRUFORM and SMOOTHVISION. TRUFORM technology divides existing triangles into many smaller ones withing the existing one, increasing the poly count. As any gamer knows, high poly graphics look more realstic. The higher the poly count, the better looking the graphics. Since TRUFORM increases the poly account, without sacrificing performance, Gamers will enjoy better looking graphics without a performance penalty. 3dfx introduced anti-aliasing, a technique that makes lines smoother. However, using anti-aliasing usually brought a huge performance penalty. GeForce3's quincux method allows good anti-aliasing without a huge performance hit. ATI's SMOOTHVISION uses a slightly different technique, but looks better. So gamers can experience awesome anti-aliasing without a huge performance penalty.
Now with all of these GeForce3 killing features, can Radeon 8500 take the crown from nVIDIA or will history repeat itself? Like it copied nVIDIA's 6-month product cycle (or at least tried) and like it copied nVIDIA's manufacturing process by selling chips to card manufacturers instead of making the cards on their own, ATI is copying nVIDIA's unified driver architecture so one driver will work with all Radeon products (and hopefully Rage128s). But do the nVIDIA-like drivers perform like nVIDIA drivers? Looking at the early Beta versions, not even close. The Radeon 8500 still can't even keep up with a last generation GeForce2 Pro, despite its higher memory and core clock speed. Those, alone, should completely obliterate the GeForce2 Pro like the GeForce2 Ultra's extremely high core and memory speed. What makes it worse is the Radeon 8500 uses HyperZ II to efficiently use its extreme power. With ATI's next generation graphics chipset unable to compete with a last generation chipset, I'm not going to even comment on Radeon 8500's gross inability to keep up with the chipsets in its class, the GeForce3. History is looking like its going to repeat itself. In its current state, the Radeon 8500 is just potential, like the original Radeon, but not a winner.
** Update **
I didn't look at the Tom's Hardware benchmarks against a GeForce3, I looked at AnandTech's. After looking at the benchmarks, I can say the Radeon 8500 performance is very questionable now. Tom's Hardware used a Pentium 4 set up, while AnandTech used a Palomino Athlon set up. The 8500 performed horribly on the Athlon, but not too badly on the Pentium 4. Still... with a higher core and memory clock, the 8500 should have performed better on the Pentium 4. But now there is much peformance ambiguity with the Radeon 8500.
So should you buy a Radeon 8500 over a GeForce3? That's a hard one to call. The Radeon 8500 has a bigger feature-set and has the potential to take the crown. On the other hand, it's drivers can't take advantage of the technology, despite the fact that they copied the leader in 3D graphics. The Radeon 8500's official release date is in mid-September. ATI has a couple of weeks to get those drivers up to par with nVIDIA's. Can they do it? I guess we'll just have to see. But history is already looking like it's going to repeat itself and as an computer science major, I know how long it takes to develop algorithms. Heck, it took 6 months for nVIDIA to take advantage of GeForce3's crossbar memory controller. A few weeks may not be enough time to tune those drivers. ATI fans, just hope for the best.
References
-------------------------------------
http://www.anandtech.com/video/showdoc.html?i=1517&p=1
http://www.tomshardware.com/graphic/01q3/010814/index.html
 |
|
|
|
|
|
|
Epinions.com ID:
cptsulu
|
|
Member: Ryan Samiley
Location: University of California, Irvine
Reviews written: 33
Trusted by: 15 members
About Me:
Third year Information and Computer Science major at UCI.
|
|
|
© 1999-2009 Shopping.com, Inc.