Reply 60 of 137, by jwt27
- Rank
- Oldbie
wrote:This is an issue with 3dbench. There are too versions. […]
wrote:I just tried slowing down my slot 1 machine. 3dbench runs at about 1 frame per 3 seconds, took 15 minutes to finish, and yet still says 62.1 fps... uh, okay.
This is an issue with 3dbench. There are too versions.
1.0c is part of the 486+ benchmark suite, this is the one used in the VGA Benchmark database project. It works great for fast machines, and ok for fast 386 machines, but slow machines it glitches out.
Version 1.0 glitches out with fast machines and it counts over after 99.9 frames, but it's much more accurate with slow computers: http://www.philscomputerlab.com/386-benchmark-suite.html
15 minutes is a very long time. What processor did you use?
Tried the 3DBench from your link, but it says "Packed file is corrupt". I've seen this before (on perfectly good executables) but don't recall offhand what caused it.
I used a Katmai 550MHz, stock speed, with Throttle at max and cpucache off. And then I still have the option to underclock the whole thing to 133MHz...
wrote:More interesting for benchmarks would be Topbench as it benches the CPU, memory and graphics subsystem independently. Usually this becomes quite unbalanced when slowing down a faster PCI based system compared to a 386 with ISA.
Seems pretty consistent, total speed varies between about 4900-5300µs. Using acpi throttle without disabling cache is much less stable, total time keeps jumping up and down by about two orders of magnitude.