Here we go!
On my mainboard shown above, I was able to get a NSC Geode GXm 266 to work stable at 300 MHz.
They are not many things to adjust onboard about this - it's only possible to adjust the multiplier (so those CPUs seems to be not locked). I assume, the board adjusts the CPU-voltage automatically.
About the board itself:
- the FSB is 33 MHz (online there is different info about 33 or 66 MHz for those kind of CPUs)
- it accepts AT or ATX-power
- it has at least PCI version 2.0 - but maybe this is managed by the MediaGX-chip itself, as the PCI-controller is integrated in it
- I think, there is USB 2.0 onboard
- there is an onboard-soundchip Analog Devices AD1819 (disabled because of the using of my Creative Soundblaster PCI 128)
- I was only able to install 256 MB SD-RAM (2x 128 MB), as all tested modules of 256 MB are recognized at half capacity (regardless if single-sided or double-sided) -> I think, this is caused by the MediaGX-chip itself, as the RAM-controller is integrated in it, too. The multiplier for the RAM-frequency depends of the CPU-freqency - in BIOS the highest possible RAM-frequency is 150 MHz or 120 MHz (CPU-Ratio of 1/2 or 1/2,5). Sadly, even using PC-133-RAM, 120 MHz was still not stable (BSOD in Windows), so I had to choose a ratio of 1/3 to reach 100 MHz (again, this is maybe caused by the RAM-controller in the CPU)
- it's not possible to use HDDs of 80 GB (no recognition); I had success with a 20-GB-HDD, so the limit is possibly at 32 GB
Used VGA: a GeForce 4 MX 440SE
Used Soundadapter: Creative Soundblaster PCI 128
To say it directly: the overall-performance in 3D of the system is not great - it's (more or less) similar to a very fast 486-system of 200 MHz, or to a midranged Pentium 1 MMX of about 166-200 MHz. For me, the main reason for this are the integrated RAM- and PCI-controllers in the CPU -> possibly they build the "bottle neck" of the performance, as the feeling with e.g. the overclocked Cyrix M1 (shown above) is many faster in Windows98/ME (in 2D and 3D). Of course, the M1 as well the M II use both the onboard-controller for RAM and PCI. Besides that, there is absolutely no L2-cache present, so this is the next big disadvantage of the system.
About the VGA:
- a GF 2 MX-400 was tested (driver versions 6.31, 8.05, 12.41)
- the GeForce 4 MX 440SE was tested (driver versions 44.04 and 53.04)
For both VGAs and for all driver-versions, the 3D-results are very similar (+/- 1-2 FPS, very close), so the performance of the system is slowed down by the MediaGX-chip itself, and/or by the missing L2-cache, I assume.
About the used software:
- the installation of WinME was perfectly easy
- I had to use the APIC- and IDE-drivers from Phil's Computerlab from this link: https://www.philscomputerlab.com/cyrix-mediagx-cx5530.html
- using 3DMark 2001SE is a joke: all tests show about 1-2 FPS, high-resolution-tests are not possible
- in windows, the driver for the build-in-VGA of the MediaGX is not installed (as I think, its performance will be very weak)
- interestingly, the benchmark in CPU-Z shows a difference of the FPU-performance between this GXm-CPU@300 MHz and a GX1-CPU@300 MHz of about 17 %
The Geode-GX1-chip, shown in the pictures, is until now not tested (crooked pins). So I will try to make it good for use and test if and how far it can go on this board.
BTW, once again, I really doubt about a CPU-converter for the MediaGX for other sockets, as there will be a hardware-collision of the PCI- and RAM-controllers inside of the chip with the onboard-controllers (which can not be disabled).
Anyway, we stay optimistic.
I LOVE CPUs RUNNING IN [GonzoHz]
