Hands-On Review of the nVIDIA GeForce2 MX vs the ATI Radeon
|
Hands-On Review:
GeForce2 MX vs Radeon Revisited By Bill Fox In last week's review of the new Power Mac G4 533DP, we briefly compared its nVIDIA GeForce2 MX graphics card with the retail version of an ATI Radeon graphics card, a BTO option. We followed up with a comparison using ThroughPut 1.5. on Friday. Our conclusion from these comparisons was that the GeForce2 MX excelled in 3D graphics but the Radeon excelled in 2D graphics, at least on our Power Mac 533DP. A couple of other web sites, Bare Feats and Accelerate Your Mac!, have also posted various comparisons of the two cards but with different computers. Some of their results seemed to conflict with ours. For example, AYM found similar 2D performance to be similar using the MacBench 5.0 "Publishing Graphics" test rather than the "Graphics" test that we used. Our test came out with the Radeon ahead by 10 percent. Bare Feats found in several runs that the Radeon was faster in 3D graphics. Different computers may be the reason or perhaps different Radeon's. We used the retail version and AYM the BTO version. It is not clear which version Bare Feats used but is likely the BTO. The principal differences between the Radeon's BTO and retail version are the S-video port, present on the retail version but absent on the BTO card, and the digital port. The retail card uses a DVI connector and the BTO card, a bit smaller, uses Apple's new ADC connector as well as supplying 28 volts to it. We decided to run a more complete and controlled comparison. We also threw in some results from 3dfx's Voodoo5 set at 2xFSAA, 3dfx's last effort before being bought out by nVIDIA. 3D Game Comparisons In our Power Mac G4 533DP review we conducted only two comparisons, one run of Demo1 with Quake III Arena v1.17 and one City Intro run with Unreal Tournament v436. The settings for both games were as we normally play, at 800x600 for Q3A and 1024x768 for UT with a variety of texture and sound settings. The video was set to play full screen on a 17" Samsung flat panel LCD monitor. Quake III Arena: We re-ran the Q3A Demo 1 tests with version 1.17 and the newer 1.27beta using the same high end settings (Lightmap, Geometric Detail High, Texture Detail on 3rd click, Texture Quality 32, Texture Filter trilinear, Sound Quality Low) for three resolutions (640x480, 800x600, 1024x768) and 32 bit color. We also conducted some 16 bit color runs for comparison. Here are the results:
The GeForce2 MX out performs both other cards over the test range that covers the resolutions that most people are likely to use. Only the Voodoo5 is not particularly playable at the highest resolution and color depth. Note that both the GeForce2 and Radeon display little difference between 32-bit and 16-bit color. Now let's take a look at the results for Q3A v1.27. We only repeated the 32-bit runs and left out the Voodoo5. We did, however, use the fastest possible settings (16-bit color, Vertex lighting, Geometric Detail low, Texture Detail 1st click, Texture Quality 16, Texture Filter bilinear, Sound Quality low) for one run at 640x480 to see the maximum attainable frame rate.
The GeForce2 MX consistently beat the Radeon but by less than compared with the results for Q3A v1.17. The maximum frame rate attained by degrading the quality of the visual experience is not remarkable. I don't know exactly why the frame rate is less for v1.27beta nor why the performance gap closed. However, the former is likely to be because Demo127 uses a different set of scenes from a different map which may be more complex than those for Demo 1. The Q3A tests are the best 3D game comparison because both cards have drivers for and use Q3A's OpenGL API. Unreal Tournament v436: We ran the City Intro demo with everything set on high. We would have tried the Wicked400 and Thunder demos but we have never been able to get either to run. UT supports the excellent 3dfx Glide API and the far less than excellent RAVE API. There is a "beta" OpenGL version of UT but it's lousy and virtually no one uses it. No nVIDIA RAVE-specific or Glide-specific driver came with the Power Mac G4 533DP, however, UT detects a RAVE driver. The Radeon, like most of the ATI cards for the Mac, supports RAVE and runs fast on RAVE if not prettily. Here are the results set on RAVE for the Macintosh which also include the 3dfx Voodoo5 at 2xFSAA (full scene anti-aliasing) using its fabulous Glide API:
The Radeon bested the other two cards. The Voodoo5 even beat the GeForce2. However, the visual experience with RAVE on the Radeon and GeForce2 is not good. It's very dark and almost unplayably dark on the GeForce2 despite all monitor adjustments. The visual experience on the Voodoo5 is terrific with 2xFSAA, even with 16-bit color. While we do hope nVIDIA will produce a better RAVE driver for RAVE-only Mac games, we mostly hope that nVIDIA will ask its ex-3dfx engineers to produce Glide drivers for the GeForce2 and upcoming GeForce3. We all know that Apple has embraced OpenGL as the future 3D game API but Glide has always been the best visually in our opinion, especially with Unreal, Unreal Tournament and games based on the Unreal engine. 2D Comparison Since we don't process many large images, 2D graphics performance on Macs has been okay for us and, we suspect, most Mac users for a very long time. In our Power Mac G4 533DP review we noted that based on MacBench 5.0's "Graphics" test that the Radeon seemed better at 2D, scoring some 10 per cent better. We also noted that the CopyBits score from ThroughPut 1.5 was much higher for the Radeon. However, we ran the "Publishing Graphics" test, a series of image manipulation routines, and one "real world" test to look further at 2D performance. The only 2D graphics operation that we do frequently that is fairly slow is to page through PDF files, especially those with graphics added to the text. So we opened the Virtual PC 4.0 PDF manual in Acrobat Reader v4.0.5, set it to "fit visible", clicked and held on the down arrow and timed how long it took to scroll through to the end of the 136 page document loaded with graphics. It took 113 seconds on our 400 MHz PowerBook 2000. Here is how long it took on our 533DP with the GeForce2 and Radeon cards:
We repeated the scrolling three times each and got the same result each time. Surprisingly the nVIDIA GeForce2 MX was 10 per cent faster than the Radeon. You are unlikely to notice that kind of difference in actual use. What about the index tests?
The two MacBench 5.0 tests indicate that the Radeon should be slightly better at 2D graphics, as does the CopyBits test of ThroughPut 1.5. So why was the GeForce2 faster on the PDF file scrolling test than the Radeon? I don't have an answer. Conclusions The GeForce2 MX is faster than the Radeon for the 3D game that uses OpenGL drivers. The Radeon is faster on the 3D game that uses RAVE drivers. On the whole, however, RAVE is not a very good 3D gaming API. OpenGL is the future so we recommend the GeForce2 MX for gamers. The GeForce2 MX is somewhat slower on tests of 2D but not a lot. On the one "real world" test we performed, the PDF file scrolling test, the GeForce2 MX was actually faster than the Radeon. Based on our observations, we can not conclude that the Radeon is significantly faster, if at all, for 2D graphics. nVIDIA is likely to release updated drivers through Apple that will improve the GeForce2's performance. So it's a toss up from where we stand at this time unless you are a serious gamer, then we clearly recommend the nVIDIA GeForce2. Mac OS X secrets revealed right here! |
Click On Images Above!
Saw something? Send a tip or a correctionThe archive ran on reader tips. What did you see, where, and do you want the credit? Something wrong on this page? Say so and it gets fixed. |
