A while back I came into possession of an Intel 486DX 50 MHz running on a 50 MHz Front Side Bus (not a 486DX2 50 MHz running on a 25 MHz Front Side Bus).
Although I've seen various posts and some benchmarks on the forums, I decided to see what results I could achieve from my setup.

The CPU came with a Chicony CH-471A (rev.3) motherboard which utilises the SIS 85C471 chipset.
I upgraded the RAM to 16 MB (60ns) FPM RAM and installed an S3 805 VESA Local Bus graphics card with 1 MB of RAM.
The motherboard came with dual banked 256 KB (20 ns) cache SRAM chips but, I installed 128 KB (15 ns) cache SRAM chips for comparative tests with my 486DX 33 MHz (model SX810) that is currently installed on an Asus VL/I-486SV2GX4 motherboard utilising the same SIS 85C471 chipset.

Below are the results with some well known DOS benchmark utilities.

As also noted in other threads, setting the L2 cache to write back, results in quite a performance penalty in terms of memory bandwidth under Speedsys & RAM access (read) time under Cachechk. There is also definitely a slow down in some of the other benchmarks when setting L2 cache to write back.
Both benchmark tests were thus performed with L2 cache set to write thru.

Also of interest, on the Asus VL/I-486SV2GX4 motherboard (which had the 486DX33 MHz CPU installed), both Speedsys & Cachechk wasn't able to "pick up" the L2 cache and therefore unable to perform a benchmark (I tried this with both the 0305 & 0402 BIOS revisions). Speedsys is, however, able to perform a L2 cache benchmark when I set the L2 cache to write back but, as mentioned above, this results in quite a performance penalty in terms of memory bandwidth & RAM access (read) time.
At this point, I'm not sure whether this is a hardware error with the motherboard (could simply be a faulty cache RAM chip). It could also be that I have to install dual bank cache memory on this motherboard and test the results again. I also happen to have another Asus motherboard which I can test this with again but, compared to the results of the Ultimate 486 Benchmark, my results seems to be more or less on par.

Below are some pictures of my CMOS setup and the best settings I could achieve to ensure a stable setup (under DOS) for my 486DX 50 MHz. With the 486DX 33 MHz, I was able to set everything to the fastest settings and shortest wait states but, still keeping Local Bus Ready at "Synchronize".

EDIT: For the Asus VL/I-486SV2GX4 motherboard (which had the 486DX 33 MHz installed), it appears that selecting the "Fastest" setting for DRAM Speed (under the Chipset Features), results in both Speedsys, NSSI & Cachechk not being able to "pick up" the L2 cache. By changing this to "Faster", results in both NSSI & Speedsys being able to correctly "pick up" the L2 cache on the PC and, under Speedsys, display a test result. Cachechk, however, doesn't seem to be able to "pick up" the L2 cache, even if I set it to the lowest value. By lowering the DRAM Speed to "Fastest", results in a reduction in the memory bandwidth from 100 MB/s to around 76 MB/s in the Speedsys results.
However, even if I set the DRAM speed to "Fastest", based on comparative benchmark results from other users, it seems that the L2 cache is being utilized correctly.