These are the tests that their result wasnt changed with Cyrix optimization software.

Landmark and Sysinfo test results after the Cyrix optimization sw, there is significant increase in results.

PC Player benchmark increased from 3.9 to 6.0. A huge improvement against AMD 386DX-40.

Not 100% sure what "Cyrix optimization sw" is, I assume the internal cache control program? You are still being held back a bit by lack of Write-Back mode in mobo cache.
In Lanmark 6.0 I get 168 MHz AT, 164 MHz 80287, 5146.81 chr/ms.
65.7 in Norton SI 8.0, 6.2 in PC Player Benchmark 320x200 mode.
This is with Trident 8900D in 32-bit/0WS mode (standard ISA 8MHz clock).

Deunan wrote on 2024-05-28, 15:10:

Not 100% sure what "Cyrix optimization sw" is, I assume the internal cache control program? You are still being held back a bit […]
Show full quote

Not 100% sure what "Cyrix optimization sw" is, I assume the internal cache control program? You are still being held back a bit by lack of Write-Back mode in mobo cache.
In Lanmark 6.0 I get 168 MHz AT, 164 MHz 80287, 5146.81 chr/ms.
65.7 in Norton SI 8.0, 6.2 in PC Player Benchmark 320x200 mode.
This is with Trident 8900D in 32-bit/0WS mode (standard ISA 8MHz clock).

Cyrix’s own application. I provided the link you can download and look.

It's 1KB L1 cache enabler. Most BIOSes support Cyrix cpus and have an option to enable it. So you don't need to run any other software.

Also note the PCPlayer benchmark uses FPU and will run faster with one installed.

Deunan wrote on 2024-05-28, 15:10:

Not 100% sure what "Cyrix optimization sw" is, I assume the internal cache control program? You are still being held back a bit […]
Show full quote

Not 100% sure what "Cyrix optimization sw" is, I assume the internal cache control program? You are still being held back a bit by lack of Write-Back mode in mobo cache.
In Lanmark 6.0 I get 168 MHz AT, 164 MHz 80287, 5146.81 chr/ms.
65.7 in Norton SI 8.0, 6.2 in PC Player Benchmark 320x200 mode.
This is with Trident 8900D in 32-bit/0WS mode (standard ISA 8MHz clock).

Added U32 additional cache module for wb cache. Bios was already enabled without u32.

Sysinfo or Landmark didn't change. Difference was observed with cachechk, topbench etc.I didnt make 3d tests. You can compare with previous tests.

I get 33 / 55 us in CACHECHK V7 so I think you're now running the system as fast as it can go (not counting any potential OC). Maybe x87 performance is down to extra wait state cycle on your mobo after all. This is perhaps beneficial if you decide to overclock the CPU. I think it's much easier to find a DLC (or better yet SXL) CPU that will take, say, 50MHz than a coprocessor that will not glitch or hang above 40MHz.

I grabbed one of those ULSI that were going cheaper on eBay a few months back. Some youtuber(s?) have tested them at over 50, I wanna say good to 55. They lag a couple of percent vs a fasmath at same clock in some things, but can go a few percent ahead in others. Just being able to do a higher clock makes up the diff though. Anyway, when I get back to basement dwelling in the fall, I will be working up my BL3 system trying to get 2x50 or 3x40 see which actually goes faster on game demo benches.

Deunan wrote on 2024-07-09, 18:18:

I get 33 / 55 us in CACHECHK V7 so I think you're now running the system as fast as it can go (not counting any potential OC). Maybe x87 performance is down to extra wait state cycle on your mobo after all. This is perhaps beneficial if you decide to overclock the CPU. I think it's much easier to find a DLC (or better yet SXL) CPU that will take, say, 50MHz than a coprocessor that will not glitch or hang above 40MHz.

Not overclocking but maxed out bios settings. Totaly stable system as it is now.

I’m going to have a go with the DLC-40. I’ve mostly ran SXL-40s, but recently I finally got a real Cyrix branded DLC. I always thought a DLC-40 should give about 486-33 performance, but your benchmarks seem more like 486-25 level. I wonder if that extra chip you installed to enable WB L2 is actually doing anything…

From what I've seen, where i486 is approx 2x IPC of i386 is that these SLC/DLC/BL class CPU seem to do 1.5x IPC of a 386. 0.75 IPC of an i486, with about plus or minus 10 percent depending on how much cache can be used, how well the board supports it, and how much the benchmark/application responds to cache.

Therefore I'd expect in optimum circumstances a DLC-40 would split the difference between an i486-25 and an i486-33

Anonymous Coward wrote on 2024-07-10, 11:25:

I’m going to have a go with the DLC-40. I’ve mostly ran SXL-40s, but recently I finally got a real Cyrix branded DLC. I always thought a DLC-40 should give about 486-33 performance, but your benchmarks seem more like 486-25 level. I wonder if that extra chip you installed to enable WB L2 is actually doing anything…

It improved the cache and memory timings.DLC40 is not equal to 486-DX-33. In between DX-25 and DX-33.