The only slower crystal I have is a 12,28 MHz one. I will try it later today and report back if the floppy drive still works 😀

While running dos, the time runs too slow (around 10%). In case of the benchmark results, all benchmarks that use the timer will be off for sure. I'll repeat the benchmarks later (I'm not sure if I'll get around to it today) with the stock 14,318 crystal @120Mhz so we have some values to compare.

Speedsys also fails to display the correct CPU speed but displays 150MHz instead. Do you know another tool to measure the CPU speed perhaps? CHKCPU also displays 150, so I guess it uses the same approach (using the timer) as speedsys...

As I recall it, something called "maxspeed.exe" doesn't use the normal routes to determine CPU speed. I've attached it. However, I also recall it not as precise after 100-ish MHz.

feipoa wrote on Yesterday, 06:50:

As I recall it, something called "maxspeed.exe" doesn't use the normal routes to determine CPU speed. I've attached it. However, I also recall it not as precise after 100-ish MHz.

Thank you! I will try this when I'm back home 😀

I tried maxspeed, but sadly it just locks up the entire pc without displaying any value. I also tried skipping config.sys and autoexec.bat, disabling all Cyrix CPU registers except for lsser and I tried it with the stock 14,318MHz crystal on a 40MHz FSB but all to no avail 🙁 I have no idea what is causing this behavior. Are there any optional parameters for the exe? I searched it up online but could not find anything related.

While I had the stock 14.318Mhz crystal installed, I ran a few benchmarks with the following results:

120Mhz / 134Mhz / increase in % / program name

95.3 / 119.3 / 25.2 / Playhouse

24.4 / 30.6 / 25.4 / PC Player benchmark

17.2 / 21.5 / 25.0 / Quake timedemo

Considering that the cache, RAM and video performance also improve with the higher FSB the results may be realistic? Are these results somewhat close to your results with the IBM 5x86c running at 2x 66Mhz?

btw I also tried if the floppy drive works with the 12.28MHz crystal and it does 😀

It will be hard for us to compare our Cyrix systems at 133 MHz because of vastly different conditions, like FSB, SRAM quantity, wait states, Cyrix register settings, and graphics card bus frequency.

My system uses:
MB-8433UUD
2x66 MHz at 1/0 ws read/write EDO
1024K at 3-2-2-2
PCI at 33 MHz

With BTB, FP_FAST, LSSER, MEM_BYP, LINBRST, DTE, and BWRT optimal, I get:

3Dbench2 = 95.2
PCPBench /vgamode = 27.2
Quake 1.06 = 19.8
Doom = 1396 = 53.5 fps
Landmark v6 = 1110 / 1629 (CPU/FPU)
Sysinfo = 422.4

Without BTB (branch prediction), the results drop to:

3Dbench2 = 91.8
PCPBench /vgamode = 26.9
Quake 1.06 = 19.4
Doom = 1434 = 52.1 fps
Landmark v6 = 684 / 1605 (CPU/FPU)
Sysinfo = 352.1

feipoa wrote on Today, 00:20:

It will be hard for us to compare our Cyrix systems at 133 MHz because of vastly different conditions, like FSB, SRAM quantity, […]
Show full quote

It will be hard for us to compare our Cyrix systems at 133 MHz because of vastly different conditions, like FSB, SRAM quantity, wait states, Cyrix register settings, and graphics card bus frequency.

My system uses:
MB-8433UUD
2x66 MHz at 1/0 ws read/write EDO
1024K at 3-2-2-2
PCI at 33 MHz

With BTB, FP_FAST, LSSER, MEM_BYP, LINBRST, DTE, and BWRT optimal, I get:

3Dbench2 = 95.2
PCPBench /vgamode = 27.2
Quake 1.06 = 19.8
Doom = 1396 = 53.5 fps
Landmark v6 = 1110 / 1629 (CPU/FPU)
Sysinfo = 422.4

Without BTB (branch prediction), the results drop to:

3Dbench2 = 91.8
PCPBench /vgamode = 26.9
Quake 1.06 = 19.4
Doom = 1434 = 52.1 fps
Landmark v6 = 684 / 1605 (CPU/FPU)
Sysinfo = 352.1

Thank you for the benchmark results! Yes, comparing our systems is difficult, but at least I now have a few comparative values 😀

I’m after some advice from the experts. I’m going to start doing some benchmarks. I have 3 CPUs and 3 motherboards. What CPU should I match to what motherboard?

AMD 5x86 133
Cyrix 120GP
Intel PODP5V83

PCChips M919 v3.2 (also have a v3.4) / 256Kb cache / 32Mb 60ns FPM
Shuttle HOT-433 v4.0 / 256Kb cache / 32Mb 60ns FPM
Asus VLI-486SV2GX4 v2.0 / 1Mb cache / 32Mb 60ns FPM

The POD, with L1 set to WB mode, works properly on very few motherboards, and the M919 is not one of them. It will work in WT mode, or if it appears to run in WB mode, it is actually in WT mode. You can confirm this by switching between WT and WB mode, and if the benchmark scores don't change, it is likely only working in WT mode. From those 3 motherboards, use whichever one works with the POD in WB mode. Probably the HOT-433, but I do not recall. Note that just because a motherboard's manual has jumper settings for the POD in WB mode, doesn't mean it actually works. Apparently, Intel changed their POD specifications late in the game and many motherboards have the POD implemented incorrectly.

For the VLI board, use the Am5x86. The latest BIOS update lets L1:WB mode work correctly at 4x. You should be able to do 160 MHz without much issue. The Cyrix 5x86, as I recall it, has some issues with WB mode on this board, but this may only be with VLB SCSI controllers. I don't recall the exact issue I had with the Cyrix, but I ran into issues on my system with the Cyrix 5x86. Thus, use the Cyrix 5x86 in the M919.

For the M919, maybe switch to EDO RAM or 1MB cache module.