Yea there are two versions of 3dbench for a reason 😀

Version 1.0 is recommended for slow machines. 286, 386 and will work reliably. But it doesn't work precisely with faster machines, having huge steps. Here 1.0c is what you should use.

melbar wrote:

You forgot your K6-2 system. Even with Sis530 board you can control L1 and L2 separately, or not?

No. 🙁 This motherboard has no option to disable the L2 cache:
http://www.motherboards.org/files/manuals/1/p5svm-103.pdf

Thumbs down.

Well, my hope is that people would like to add their own results for a wide variety of CPUs, so that someone with the ability to disable L2 can add their result to the list.

On a positive note, I have a POD 200MMX on the way that will go into the Triton 430FX board. It will be nice to add its L1D results to the chart.

PhilsComputerLab wrote:

Yea there are two versions of 3dbench for a reason 😀

Version 1.0 is recommended for slow machines. 286, 386 and will work reliably. But it doesn't work precisely with faster machines, having huge steps. Here 1.0c is what you should use.

Good to know. Thanks! I will have to re-do the PII and Celeron results with L1D and if this ever gets off the ground, I'll note it in the opening thread.

clueless1 wrote:

No. 🙁 This motherboard has no option to disable the L2 cache:

I see, didn't know that there are also L2 less cache memory outside. It's a micro-ATX board with 3PCI & 1 ISA and onboard VGA.
When i look at the pdf you linked:
That means the L2 chips + Tag ram must be desoldered from board...
At p. 8: "Features 512KB/1MB (when chip is available) pipelined-burst SRAM/L2 memory cache and integrated Tag RAM"
At p.11: You see the board with L2 cache.
It seems, there are SS7 boards with 512kB, 1MB, 2MB or nothing.

I think someone needs to have a talk with that motherboard. It doesn't sound like it's interested in being a team player.

This should be a joke right 😀 Well, maybe it is the revenge of it due to mishandling of the previous owner... (Possibly PS/2 port defect?)

My head is spinning with all the combinations to run on this board WITHOUT the option to disable L2. Between 100Mhz, 95Mhz and 66Mhz, then multipliers between 2.5 and 5.5 for each...

melbar wrote:

Since you have also DOOM values, i've made a re-bench of my K6-2 171Mhz, now with another VGA (Geforce2-MX400 AGP), including DO […]
Show full quote

Since you have also DOOM values, i've made a re-bench of my K6-2 171Mhz, now with another VGA (Geforce2-MX400 AGP), including DOOM and also a investigation in SETK6.
I have included results with "full caches" and "L1 on, L2 off" and 3 different runs:
- standard setting, write allocation on, write combining off
- write allocation off, write combining off
- write allocation on, write combining on
These setting have only a effect when you have L1 enabled.

At the end, I've got a strange result for DOOM: with write allocation/ write combining Off, i'll get the best results. I don't know why....?!

melbar, would you be willing to add speedsys results to your chart? And FSB, multi. and graphics bus? I could add them to my chart so we can have an example entry with L2 and L1+L2 disabled.

Ok. I've have benched now my K6-2 with all the benchmarks in your list. Because it's a lot of work and i'm mainly interested in the extreme values of this CPU (except overclocking),
so i've clocked my CPU only with the minimum (170MHz) and maximum (500MHz). And then i've run all the tests with and without caches.

Due to my last tests with write allocation/write combining with SETK6 tool and the abnormal results, i've decided to do all test with standard configurations.
The memory is set to CL3, and all special memory features in BIOS are disabled. A short test with AGP speed 1x vs 2x shows not a real impact, so i've decided to let it at AGP 1x.

Here are the properties of WCPUID and CPU-Z:

Here are my results.
The most interesting thing is, that my values of L1D (@500MHz and @170MHz) are really next to your values (@550/400MHz and @266MHz), but i have with these values L2 enabled, and you have said that your board has no L2. When i disable all caches, the speed is falling to 386SX33...

Also the Speedsys Benchmark has no effect in disable any L2 cache.

melbar wrote:

Ok. I've have benched now my K6-2 with all the benchmarks in your list. Because it's a lot of work and i'm mainly interested in the extreme values of this CPU (except overclocking),
so i've clocked my CPU only with the minimum (170MHz) and maximum (500MHz). And then i've run all the tests with and without caches.

I know. I was a bit dismayed when someone made a comment that sounded to me like I should be testing every speed possible. In the end, I did do that, but then I cut out the results that were very similar to one another. I just tried to leave results that landed in key areas with L1D. So next chance I get I will add your values and post another screengrab. Thank you very much. I do appreciate the time you took to help out! 😀

melbar wrote:

Here are my results.
The most interesting thing is, that my values of L1D (@500MHz and @170MHz) are really next to your values (@550/400MHz and @266MHz), but i have with these values L2 enabled, and you have said that your board has no L2.

No, the BIOS tells me that there is 512KB of L2 on the mainboard. What I meant was that there is no way to disable the L2.

I should have some time this evening to add your results, melbar. I've also got another member working on results for me. I'd like to have a couple of different users' results in the database before I start a new thread. I think having a few different examples besides my own will make things better/easier for people to contribute.

Also, I got my POD 200MMX in, so I will benchmark it tonight and add it to the results as well. I'm very hopeful that it will be in the 386dx-40 range with L1 disabled. *fingers crossed*

I'm a little excited...I had time at lunch to install the POD 200MMX and run a couple of tests. Doom results at full speed: just a hair slower than the Pentium II 233. Doom results with L1D: just a hair slower than the Cyrix 486SLC2-40. I will have to test with Wing Commander 1 and 2 tonight, but this might be as close to ideal as i'm going to get on an Intel platform.

For better compatibility with a few DOS games, clock speed can matter. Some games have issues at high clock speeds, even with cache disabled. My main DOS PC is basically a Pentium 100 MHz (200 MHz model clocked lower) for maximum compatibility.

PhilsComputerLab wrote:

For better compatibility with a few DOS games, clock speed can matter. Some games have issues at high clock speeds, even with cache disabled. My main DOS PC is basically a Pentium 100 MHz (200 MHz model clocked lower) for maximum compatibility.

Thanks. I hope this processor works well when I check the rest of my games, but after a quick check, WC1 and WC2 ran great with L1D and System Shock SVGA, Duke 800x600, and Descent 640x480 ran great at full speed.

Do you play any SVGA modes on the P100? It seems like a P100 would struggle with most SVGA stuff.

This will probably be the last screengrab I post. I will soon start a new thread with a Google spreadsheet for people to add their own.

This one has melbar's K6-2 results and my POD 200 results. melbar has a proper motherboard (unlike mine) that allows disabling L2 cache.

Here's the scores for the AMD K5 PR 166(real speed is 117 MHz, 66MHz x 1.75), DFi Intel 430TX, 32 MB SDRAM, Diamond stealth 64 DRAM(S3 TRIO 64 1MB), DOS 6.22.

All caches on:
3D Bench2 143.9
PCP Bench 40.5
Doom Realtiks 1112
Speedsys 122.9

Now with Level 2 cache off:

3D bench2 132.5
PCP Bench 33.8
Doom realtiks 1210
Speedsys 122.9

Now with both level 1 & level 2 off:

3D bench2 12.8
PCP Bench 3.1
Doom Realtiks 16633
Speedsys 10.14

AMD K5 really needs it's level 1 cache!!!

I'll do some flavour of Cyrix 6x86 next. After that IDT Winchip.