Ti4400 and Radeon 8500

First, this is not a review of the MSI Ti4400 or the retail version ATI Radeon 8500 64mb. If you want pictures and specs on these cards, there are plenty of them all over the internet. Instead, this is a performance related article showing what you can expect in different racing sims, some old and some new, with these video cards. My computer is running an Intel P4 1.6a Northwood at just over 2.5 gHz and DDR333 memory. You can see an outline of my unfinished article on that here. If you're not running at this speed, your resulting Frames Per Second in games will not be the same as mine. I do have some pictures of FSAA (Full Screen Antialiasing), but if this is important to you to make some of your favorite, old racing games look like new ones, neither of these cards can compare to the 3DFX V 5500 in that department. So, let's begin .......

I had the opportunity to try out an ATI Radeon 8500 64mb for the past of couple months. I bought it to use in my CAD system at work, but things were so busy there that I didn't have time to do all the upgrades that I had planned until now. So, I had to make a choice of either buying a new video card for myself, or keep using my 3DFX V 5500. Since I just upgraded my home computer and my kids' computer, it made sense to buy a new card for myself and give the 5500 to the kids.

I had thought I was going to buy a Ti 4200 because of the price/performance ratio, but the prices didn't come down the way they were supposed to and the Ti4400 seemed like a better deal. I chose the MSI brand because it offered TV-in as well as TV-out and most of the others didn't, it comes with everything including the cables and the DVI to CRT dongle, is equipped for both S-video and composite input and output, and because I read in other reviews that it ran cool, was stable and overclocked well. It doesn't have a fancy red PC Board and the HSF (Heat Sink/Fan) is a standard design. Frankly, I don't care if a card has chrome woofers and pink metal-flake tweeters, I just want it to work well.

Just before I installed the MSI card, I downloaded the newest Win98 drivers for the Radeon, version 9043, and gave them a test run on 3DMark2001se. My score with the older 9017 drivers was 9632. The new drivers gave me 10000. Although this seems like a lot, in perspective it's a 4% increase. However, the new Catalyst drivers are also very stable and give a better looking image. Overall, a nice improvement. My experience with nVidia, which goes way back, is that they produce fast drivers at first, then subsequent releases get slower and more stable. Such is almost the case with my MSI card. It came with version 2760 and I upgraded on their web site (MSI Live Update) to version 2890. Both versions are stable. The new ones are only about 2% slower, but I was able to get more overclocking out of the card to make up for it. My first Ti4400 3DMark was running everything stock and I expected a little more out of it. If I hadn't installed the Radeon's new drivers, the 4400's score may have seemed more impressive. The second 3DMark was overclocked.

I must say, I was impressed with how much the MSI Ti4400 would overclock. First, they give a couple nifty little utilities to do the job. One is call 3D Turbo Experience and it sure is nice to look at, but I used the one built into the Display Properties dialog box. At 662 memory and 320 core nothing crashed, but I could see a few sparkling sprites while running 3DMark, so I turned the memory down to 648 (that was the next step down, the core moves one point at a time) and left the core at 320. That means my OC'd Ti4400 was running as fast as an OC'd 4600. Not bad for the price, and I haven't even tried the bundled games yet. I have tried the TV, however, and GPL looks good on my 50" HDTV.

A bit about 3DMark before I continue. I find it a useful tool for gauging overall performance and differences in overclocking speeds, but it's not the final answer for game performance. Many people get confused about the size of the scores and how meaningful a few hundred points are. It's generally a good idea to divide of all the scores by 100 to get understandable results that are more comparable to actual game frame rates (remember, you heard it here first). Let's say I can get a score of 10500 with a certain stable, overclocked system, but if I really push up the voltage (and heat), I can get 11000. That's a gain of a whopping 500 3DMarks. But, is it worth the strain it's putting on your cpu or video card? Let's look at what it will really get you. If you can get 105 fps the first way, you'll get 110 fps the second way. That's less than a 5% gain (11000 - 10500 / 10500 = 0.0476 or 4.76%). If you were getting 50 fps the first way, you'd get 52 fps the second. Is it worth killing your computer and putting up with stability hassles for 2 fps that you'll never notice anyway? I've found changes in the Radeon's 3DMark scales very well to actual game results. A 4% increase in 3DMarks generally means a 4% increase in game frame rates. With the Ti4400, it's not quite so accurate, but still a close indication. However, comparing the two cards together in 3DMark is not a good indication of how they'll compare when actually playing a game. The OC'd results of both cards in 3DMark shows the Ti4400 as 11% faster than the Radeon, but in F1 2001 its almost 50% faster and in other games there is hardy any difference at all. This may be because some real games are more optimized for the nVidia chip and the suite of games in 3DMark evens the score amongst the games, but I'm not sure. In the Nature scene in 3DMark, the Radeon always outperforms the 4400 by quite a bit, so if a real game was based on the Nature scene, the ATI would perform better.

Now let's look how some of the games performed with each card. I didn't do any Antialiasing tests with the Radeon because it affected frame rates too much, but I did so some with the Ti4400. In most of the tests I used the Act Labs Force RS wheel, but I did use the Thrustmaster Ferrari Force Feedback in a couple of them. All tests used the Act Labs Performance Pedals. There is no difference in frame rates caused by this equipment. The Ti4400 setting of 4xS AA (mentioned in some of the tests) is the highest AA setting available.

Click on the thumbnails to see full size pictures. Ti 4400 tests were done with High Quality video card settings unless noted otherwise. Radeon cards were tested with medium settings (sorry). In general, the picture quality was about the same with the settings this way.

	Description	FPS
	Radeon 8500 F12001 1600x1200x32, full eye candy, full field of cars	37
	MSI Ti4400 F12001 1600x1200x32, full eye candy, full field of cars	55
	Radeon on GPL, 1600x1200, full eye candy, full field of cars	36 (maxxed)
	Ti4400 on GPL, 1600x1200, 4xS AA, full eye candy, full field of cars	36 (maxxed)
	Ti4400 on GPL, 1600x1200. Look at the suspension parts.	N/A
	Ti4400 on GPL, 1600x1200, 4xS AA. Look at the suspension parts.	N/A
	Radeon, Nascar Heat Benchmark	59 (maxxed)
	Ti4400, Nascar Heat Benchmark, 4xS AA	59 (maxxed)
	Radeon Nascar Heat 1280x1024, full eye candy, full field of cars	55
	Ti4400 Nascar Heat 1280x1024, 4xS AA, full eye candy, full field of cars	63 (maxxed)
	Ti4400 Nascar Heat Quality shot showing 4xS AA at 1280x1024 (1600 is not offered in this game) Look at the edges.	N/A
	Ti4400 Nascar Heat Quality shot without 4xS AA at 1280x1024. Look at the frame and the wall.	N/A
	Ti4400 Nascar Heat Quality shot showing 4xS AA at 1280x1024. Look at the frame and the wall.	N/A
	Radeon NR2002 1600x1200x32, full eye candy-mirrors at half, full field	30.21
	Ti4400 NR2002 1600x1200x32, full eye candy-mirrors at half, full field	35.79
	Ti4400 NR2002 1600x1200x32, most eye candy, no reflections or anisotropic, 1/2 mirrors, full field	69.47
	Radeon Rally Championship Xtreme 1600x1200x32, full eye candy, HQ card settings	58
	Ti4400 Rally Championship Xtreme 1600x1200x32, full eye candy, HQ card settings	87
	Ti4400 Swedish Touring Car Championship 2 (demo), 1600x1200x32, HQ settings, full eye candy	63
	Ti 4400 4x4 EVO (demo), 1600x1200x32, 4xS AA, full eye candy, HQ settings. Let's you know what the older games will do.	102
	Ti 4400 4x4 EVO (demo), 1600x1200x32, 4xS AA, full eye candy, HQ settings. Let's you know what the older games will do.	104

Are you surprised at the results? I was. I thought the scores would be more in line with all the tests we've seen that are based on First Person Shooters. If I had to get a new video card now (and I did), I'd make the same choice. The MSI Ti4400 is a good deal and comes with everything. However, if I still had the Radeon 8500 and didn't have to change, I wouldn't. Even after seeing these scores. In F1 2001, 37 fps was the lowest it got. If you've played GPL which is locked so it won't exceed 36 fps, you'll know that F1 2001 is very playable with the Radeon. And of course, you could always lower a couple graphics options or play at some other resolution if you really wanted to increase the frame rates. Nascar Racing 2002 is about 15% faster with the Ti4400, but again, turning off the Reflections or running a lower resolution makes the game perfectly playable with the Radeon. Rally Championship Xtreme, Nascar Heat and other games run fast enough with the Radeon anyway. Once you're over 50 fps, why bother with more?

My recommendation is that if you need a good video card now, the MSI Ti4400 is the one to get. But, if you are just thinking about upgrading, you might be wise to wait 3 or 4 months to see the new R300's from ATI and NV30's from nVidia. If they are as fast as they claim, and that's a big IF, we'll all be looking for graphics quality as opposed to speed very soon. Now that's the kind of graphics card war I could really enjoy. Who knows, maybe they'll even figure out a way to put Reflections on a race car and not make it look like a mirror.