Currently I have a Radeon 9000 All-in-Wonder (64M DDR), and a Radeon VE/7200. I have only recently switched from a Voodoo5 5500. The main game I'm interested in is Grand Prix Legends, which was written for the Voodoo and Redition technologies. The Voodoo held up very well from a Voodoo3 2000 through a Voodoo3 3500, a Voodoo4 4500, and on to the Voodoo5. When I originally acquired (Thanks, Chris) the Radeon VE I did a great deal fo testing against the Voodoo5 to see if it was worth the switch. It was not really worth the switch. The Radeon VE performed pretty much the same, from a frame rate point of view, as the Voodoo5 in D3D mode. The Voodoo, on the other hand, could also do Glide mode, and GPL looked so much better this way. The key componenet in driving sims, all performance issues aside, is depth perception. If you can't judge the simulated distance well at a simulated 200mph, or so, you're not getting all you can out of the simulator. The Voodoo beat the Radeon VE hands down in the simulated distance rendering department. The Radeon 9000, running D3D, on the other hand, is light years ahead of the Voodoo running Glide.
This is a P4 2G, 512M RAM, Radeon 9000 All in Wonder 64M DDR RAM, Soundblaster Audigy. I'm running GEM+ with the bandwidth patch, carsound patch, 1.2.0.2 patch, a 21" monitor at 1280x1024/GPL at 1600x1200, and D3D rasterizer. Z-buffer is on (not disabled), all other graphics enhancements are on, detail is full, dynamic lighting is off, but specular highlights are on.
Here's the problem I was trying to solve:
Next day I read all of the above, and ran a couple of iGOR, and a couple of VROC races, including Monza, and all seemed fine. I think increasing the AGP aperture did it. This is the only thing that makes any sense. The AGP window had been set at 32M, but I have 512M of RAM. Increasing my monitor size, and hence the size of each, and every frame, increases the requirement for buffers. I would have thought the software buffers of the ATI drivers, Windows, and D3D should be at least somewhat dynamic, so which part of the system could be filling up, and causing a crash? These thoughts lead me to believe hardware issues are the most likely suspects, so I increased the AGP window to 256M, and turned off the BIOS's direct draw capability (see the first article above). I also changed all the GPL core.ini settings as per information gained from the GPL FAQ, and the included links regarding the core.ini file settings. I'm now running more detail, and more graphics features than before, and I have better graphics, a better frame rate, and smoother transitions. For example: I'm practicing Clermont-Ferrand/Charad 1970, and there's a high resolution graphics patch. Previously I had installed this patch, but after a few laps GPL would freeze solid. With the current settings I have no problem. I can run Charad forevery without any trouble, and, as I said before, I can run iGOR, and VROC races without any trouble.
A couple of months later I have now tweaked things out, and am running 1280x1024 with full car sets, and full track graphics, and can get 24 fps on a full AI grid. This picks up to the max 36 fps almost immediately after the flag drops. I dropped the screen resolution to 1280x1024 from 1600x1200 when I discovered a setting/tweak that refused to do FSAA above 1280x1024. The Radeons seem to be a bit odd in certain regards. I don't really mind this, but it would be nice to have the information up front from ATI.
I recently noticed that my box was making a funny noise, and a team mate suggested I should check the temperature of my video card. I did not have a temperature sensor on my Radeon, but I do have an infrared pyrometer (Raytek MiniTemp FS), something I now think no gamer, or overclocker should be without. Anyway... Finding the temperature specs on the Radeon, or its chip set was impossible, but I found some information on a similar card, and guessed mine was in a similar range. Short answer: I was well within the range. However, as the result of examining the card, I noticed at one point that the fan was stationary. But spun up when I helped it. I removed the card, and examined the fan. The bearings were extremely sloppy, so I installed a cheap ($10) VGA cooler from Evercool. This resulted in a 10° Farenheit, or 7° Celsius drop. What I think is a secondary benefit of this kind of kit is that the power for the fan is no longer being taken from the Radeon itself, but comes directly from the power supply. I imagine this helps the Radeon a bit.
Because the pyrometer was running in Farenheit at the time, and I was only interested in the difference, I measured in Farenheit. Also, because I was only interested in the difference, these tests were all done with one side of the case removed. At rest (Just the Windows Desktop) I used to get 106° F. Under load (GPL) for an hour, or so, I managed to push it up to 124° F. Now, with the Evercool fan, and heat sink, I get 99° F, and 104° F. These measurements were taken from the 'back' of the card behind the centre of the main chip. When the card is installed, the chip, and fan are facing down. I believe this contributes to the failure of the fan. The bearings don't seem to be designed to run that way up. I will be monitoring its health closely.
NOTE: This product is no longer listed on the Evercool Web site, but you can still find similar products, or old stock, so this guide is still relevant.
This bit is here because I can't find any information anywhere about installing this kit, and there are no instructions with the kit.
Notice that the kit is referred to as VC-RB, but one of the images is labelled RC-RV. I have found referrence to both numbers, but here is my experience:
The heat sink on the fan has two ears sticking out from two opposite corners with holes in them to allow mounting this 'universal' kit with the nylon press-in posts provided. The Radeon has two holes on the board aligned with the corners of the chip to allow mounting of a post/screw-mounted heat sink. The board, and the heat sink don't match.
The kit comes with thermal paste, and any of the overclockers' Web sites always talk about using the paste when installing aftermarket coolers to your video card. Thermal paste is not glue! If you use the paste you must have a mechanical method of securing the heat sink to the board, or chip, and since we can't do that with this board/cooler combination, we have to use one of the adhesive pads provided. I tried to use the thermal paste, but after the card sat on the shelf for eight hours, the paste was still too wet, and maleable, to consider reinstalling it, so I cleaned off the paste, and applied the pad.
To remove the original heat sink from the chip, I inserted a medium sized flat screwdriver head between the edge of the heat sink, and an empty area of the board, and twisted firmly but slowly.
To clean the old adheasive off the chip I used a product called
Goof
Off. This is primarily isopropyl/rubbing alcohol, which is what the overclockers
recommend for cleaning chips, but I was out of alchohol. I did buy some alcohol
afterward, and used that to remove the thermal paste. Both of these products
work fine. When you buy the alcohol, make sure there's nothing else in it. Sometimes,
if it's sold as rubbing alcohol, it will contain a perfume.
HTH