First post, by Marvin
I managed to get my hands on these rare 64MB 32 chip DIMMs a few years ago:
These are rare because they can max out 430VX 64MB cacheable RAM limit with a single DIMM. On some threads their existence has even been questioned.
According to the manufacturer datasheets they are rated as 100MHz CL3 but as the bus speed on 430VX is 66MHz they are specced to run at CL2.
I looked around and never found any real numbers how these compare to SIMMs, so I ran a few benchmarks on my QDI P5I430VX-250DM. Processor was Intel Pentium MMX 233MHz.
By default the Award BIOS on this motherboard has no SDRAM CAS latency option, but that can be unhidden with modbin. So the manufacturer only half-implemented it, with timings 3/3 for DIMM by default. After unhiding the option one can select the usual 3/3, 3/2 or 2/2 SDRAM(CAS Lat/RAS-to-CAS) in BIOS. I tested the machine heavily on the fastest 2/2 setting and encountered no instabilities.
As a comparison I used 2x32MB 60ns EDO SIMMs. To make things interesting I tested both 60 and 70ns speeds in BIOS.
I used PhilsCompLab DOS benchmark kit, but only included benchmarks that showed real difference:
The most interesting is of course the raw memory speed that SpeedSys shows. Here are the screens for that:
EDO 70ns:
EDO 60ns:
SDRAM 3/3:
SDRAM 3/2:
SDRAM 2/2:
SpeedSys uses same strange logic for the upper left "Memory Bandwidth" calculation that I would ignore. It is especially strange it showing less for SDRAM 3/3 and 3/2 that are clearly faster than the EDO SIMMs. The real speed is on the graph at the bottom.
To conclude: going from 60ns EDO SIMM to SDRAM DIMM on 430VX is clearly measurable for the memory throughput, especially if the BIOS allows one to select 2/2 as SDRAM(CAS Lat/RAS-to-CAS). But for real DOS games it shows minimum gains.