Inno3D Tornado GeForce2 MX400 in the test: 32 and 64 MB VRAM in comparison
Now we come to one of the most interesting topics with graphics cards, overclocking. How well you can overclock a graphics card depends, among other things, on the cooling and the access time of the memory. So the cards are not that bad for us. After all, both graphics cards from Inno3D have an active fan on the graphics chip. Furthermore, memory with an access time of 5ns is built into the 32MB variant, so you can probably get a lot out of it. The 64MB version has 6ns memory, but you should also be able to tease out a bit more performance here. From the memory of the Inno3D GeForce2 MX400 32MB, which is already operated at 184MHz a bit above the specifications, 16MHz more, i.e. a total of 200MHz, can be extracted. When we wanted to increase the memory clock by just 1 more megahertz, the computer immediately crashed on the Windows desktop and the image froze. So here the limit was reached. By cooling the memory, you might have gotten more out of this, but we're finally testing the graphics cardin the delivery condition and not after some changes. So it went on with the chip clock. This could initially be set to 235MHz without any problems, while Quake 3 still ran flawlessly. In Unreal Tournament and 3DMark 2001, however, we quickly noticed that we had reached a little too high here. After reducing the graphics chip clock to 230MHz, the graphics card ran stably again and we were able to run the benchmarks.
With the Inno3D GeForce2 MX400 64MB things turned out to be a bit more difficult. The memory could be increased in Windows to up to 190MHz, but above this a number of image errors occurred on the desktop. In the end, the graphics card only passed all tests with a maximum of 180 MHz memory clock without crashes. As with the 32MB version, you could also set the clock of the graphics chip to a maximum of 230MHz, otherwise there were freezes in pretty much all tests. With 180Mhz memory and 230MHz chip clock, we carried out our benchmarks here. It has to be said that there were some unsightly image errors in 3DMark 2000. The system ran stably, however, and no image errors occurred in other benchmarks/games. Maybe you should clock the graphics card down a bit here.
The following table clearly lists the clock rates of the two graphics cards with which they could be operated stably.
on the next page : Overclocking Benchmarks (Part 1)