kixs wrote:

Finally got my GA486AM/S working with AMD 5x86-P75 in WB mode. […]
Show full quote

Finally got my GA486AM/S working with AMD 5x86-P75 in WB mode.

Config:
GA486AM/S v2.21, 256KB 15ns cache
AMD 5x86-P75 133MHz
2x4MB FPM RAM
Tseng ET6000 2MB PCI
BIOS set to fastest settings

With 512KB cache (AMD@160MHz-WB/133MHz-WB):

3DBench...98.8/84.7
PCPlayer...25.6/21.0
Doom...1232/1413
Quake...15.5/12.7

Finally got 1MB cache 😀

Upgraded cache to 1024KB (AMD@160MHz-WB/133MHz-WB):

3DBench...98.6/82.1
PCPlayer...27.8/23.1
Doom...1217/1450
Quake...16.1/13.4

3Dbench has the best results with 512KB cache, all the rest are faster with 1024KB.

I'll do some Windows benchmarks 256/512/1024kb when I get the time.

Added results for DOSBox daum 1/25/15 build.

1      3db pcpb ticks doom  quake
2      560 324.9 294 254.05 168

tldr; it's faster than 0.74, speed comes from using d3d output instead of opengl.

Bad that Speed Test is missing here. It's useful as CPU benchmark for apps without intensive FPU work and seems gives data close to correct inside DOSBox and maybe other emulators too.

awgamer wrote:

Added results for DOSBox daum 1/25/15 build

Daum's build 2014-01-27 is faster and most prefer it on today.

I've added a few results recently Phil including a IDT Winchip & Cyrix MII, any chance you could update/resort?

Cheers,
Chris

BSA Starfire wrote:

I've added a few results recently Phil including a IDT Winchip & Cyrix MII, any chance you could update/resort?

Cheers,
Chris

Sure, will do. Thanks for still submitting results 😀

kixs wrote:

With 512KB cache (AMD@160MHz-WB/133MHz-WB): […]
Show full quote

kixs wrote:

Finally got my GA486AM/S working with AMD 5x86-P75 in WB mode. […]
Show full quote

Finally got my GA486AM/S working with AMD 5x86-P75 in WB mode.

Config:
GA486AM/S v2.21, 256KB 15ns cache
AMD 5x86-P75 133MHz
2x4MB FPM RAM
Tseng ET6000 2MB PCI
BIOS set to fastest settings

With 512KB cache (AMD@160MHz-WB/133MHz-WB):

3DBench...98.8/84.7
PCPlayer...25.6/21.0
Doom...1232/1413
Quake...15.5/12.7

Finally got 1MB cache 😀

Upgraded cache to 1024KB (AMD@160MHz-WB/133MHz-WB):

3DBench...98.6/82.1
PCPlayer...27.8/23.1
Doom...1217/1450
Quake...16.1/13.4

3Dbench has the best results with 512KB cache, all the rest are faster with 1024KB.

I'll do some Windows benchmarks 256/512/1024kb when I get the time.

Very interesting. I normally don't witness any benchmark score improvements when increaing cache from 256 to 1024K. Is your performance increase due to the small size of your installed memory (8 MB)? If you, for example, install 64 MB of RAM with L2 in WT mode, and compare the results at with 256 KB vs. 1024 KB, or 512 KB vs. 1024 KB, do you still witness the same amount of performance gain?

I'd say it's more unusual the performance wouldn't at least somewhat increase with more cache as more code fits in a very fast L2. But it also depends on applications how large is the essential code and 256KB might be enough already. 3DBench is interesting in this case.

About the installed memory... I just test with what is at hand. I think when I tested 1024kb cache it was 1x16MB installed. Will sure check when I get to do the Windows benchmarks. I'm also hoping to get it running at 180MHz (3x60).

Looks like I've come to this party a little late but I have added about 9 different ISA, PCI and AGP cards running on an SS7 board with a K6-2 (that's how I roll on a Friday night). Interesting results, mostly because I've got a pile of AGP cards that refused to benchmark altogether and there seems to be no rhyme or reason to it. Scratch that, I just looked again and it seems they're all ATi cards - a Rage 128, Radeon 7000, 8500 and 9800. The latter nVidia's looked pointless next to the Matrox G100 and Voodoo3, which were almost identical. Will try some more when I get the chance. Most impressive of all was a highly respectable Quake site with one of my ISA cards!

Lots more results added, not all mine either! 🤣
could you resort again please Phil?
Best,
Chris

Sure

Be nice to see some of the other more obscure CPU's added, later 2.2V Cyriz MII's 100mhz FSB models particularly, VIA C series chips, RiSE MP6, Nextgen 686, hard to believe after all this time I am the only one to add VIA c3, C7 & IDT Winchip results.

Someone already had a deal with the problem of "no place for digits". Their decision instead of creating universal algorithm or reserving much space was to add only a single digit place, 🤣. Someone, for example, would may disassemble the code and direct the bench number as plain text when the program closes or write this number to a file.

Yes, version 1.0c only goes up to 999. The thing is some of the faster machines might be pumping out over 2000 fps, so it's hard to tell. It would be great if someone could modify it so it spits out the fps when you exit for example.

PhilsComputerLab wrote:

Yes, version 1.0c only goes up to 999. The thing is some of the faster machines might be pumping out over 2000 fps, so it's hard to tell.

According to the current charts: 1000 in 3DBench ~ 150 fps in Doom. If a machine gets 150-300 fps in Doom and strange low 3DBench, has >=P3 700 and normal video, may be added +1000 to results [with a mark that it's an interpretation like "1200 (+1000)"]. It's bad that some people wrote "Too fast" instead of numbers.
You may fix some numbers in the chart for 3DBench based on Doom fps.
For example:
Skyscraper's P3 850 - Doom 179.11 fp, 3dbench 170.7. I'm sure there should be 1170.7 in 3dbench.

I put in the too fast part. Just use the other benchmarks to compare faster machines it's not a big deal. I'll fix up the rest at some point.

I agree, a modded version of 3DBench to correct for the overflow in some way would be really cool. 😀

kixs wrote:

I'd say it's more unusual the performance wouldn't at least somewhat increase with more cache as more code fits in a very fast L2. But it also depends on applications how large is the essential code and 256KB might be enough already. 3DBench is interesting in this case.

About the installed memory... I just test with what is at hand. I think when I tested 1024kb cache it was 1x16MB installed. Will sure check when I get to do the Windows benchmarks. I'm also hoping to get it running at 180MHz (3x60).

When do you think you'll have an update on this?

feipoa wrote:

kixs wrote:

I'd say it's more unusual the performance wouldn't at least somewhat increase with more cache as more code fits in a very fast L2. But it also depends on applications how large is the essential code and 256KB might be enough already. 3DBench is interesting in this case.

About the installed memory... I just test with what is at hand. I think when I tested 1024kb cache it was 1x16MB installed. Will sure check when I get to do the Windows benchmarks. I'm also hoping to get it running at 180MHz (3x60).

When do you think you'll have an update on this?

I guess I missed your reply... For cache/mem size I did the tests in another thread.

Simple copy-paste to include it here too.

For Windows tests I'll need more time... maybe this spring/summer.

---

I've done some quick test with my Gigabyte 486AM/S UMC 8881 board, Intel 486DX4-100 SK096 WB, 1024KB WB cache, Tseng ET4000/w32p PCI. I've only tested WriteBack L1 & L2. Not sure if WT would be any different in this comparison.

Tested with 3DBench 1.0c, PCPBench, Doom, Quake (Phil's pack) and following memory configurations (1x16MB, 1x4MB, 2x4MB, 2x16MB, 2x4MB+2x16MB)

Every memory configuration was the same. The only variation was in Doom with 1x4MB - result was 1 point faster (1641). Also Quake wasn't run with 4MB as minimum is 8MB.

3DBench... 71.1
PCPBench... 24.5
Doom... 1642
Quake... 11.1

The way I see it, cache is just a super fast mini memory. Larger it is, more data it can hold. Benefits largely depend on the applications itself - how large is it's most used code/data. Cache implementations hold the key only to the efficiency of cache - directly mapped cache has most cache misses and isn't as effective as n-way associative. More cache usually means better performance. But it's true some motherboards (maybe better said: chipsets) don't like large cache as you have to increase wait states and this usually means benefits of a larger cache are diminished.

I just remembered I've done this cache size comparison a few years ago on 386 VLB board with 486DLC-40 from 128KB to 256KB and the same memory configuration - 8x1MB. Here are results from my benchmark notes:

Test........128KB.......256KB
3DBench... .26.3.........26.6
PCPBench.....5.5...........5.8
Doom.........6191........5905

Not much, but it's faster overall.

And now a Samuel 2 is in the spreadsheet.

Looking through here, I'm wondering how much of the suck is the PLE133, and how much is the CPU. There's a P3 Coppermine 850 with a PLE133 that's hanging out with P2 333s...

Added results for a AMD DURON "Spitfire" 650 with nvidia RIVA 128, first "spitfire" in the database 😀