I guess I just like the hard numbers 😉

Thanks again

Phil, the main (stock) table has evolved since it was first introduced. Perhaps colomn AB, that was going to be deleted eventually, should finally be removed? Plus, the overclocked table is still as it was when first introduced. Don't you think it should be brought to uniformity with the stock table? Not only does it look different, it also represents a different classification of systems (i.e. Socket 4, 7, 8 in the same class, etc.)

There are also a couple of small things that need fixing:

1) The second line of the Stock table is blank.
2) Colomn C of stock list is protected only as far as row 309.
3) The second line of the Overclocked table is also blank.
4) Lines 92,93 of the overclocked list feature systems that did not pass more than one test, should they go?

@jwt27

- Doesn't 3dbench only have 3 digits? Your results are 1000+
- Missing benchmark entries

@RacoonRider
- Removed the second lines in stock and OC
- Column AB stock stay
- OC section not focus of project, good enough
- Protection Column C fixed

@Easy John
-check spelling please Am486DX2 not am486DX2

easy_john wrote:

This tests is more lemitrd by cpu, chipset and bus speed on PCI systems and more modern. Video card and his bus is not a bottlen […]
Show full quote

This tests is more lemitrd by cpu, chipset and bus speed on PCI systems and more modern.
Video card and his bus is not a bottleneck from about 486 systems.

I've started my test on 486 DX2-66 with bios default.
Tested few videocards and found similar benchmarks.
Then, I've statret bios optimisation, and on absolutely same system got 10 and 20% higher bench: (two bottom lines)

<snip>

So, cache and ram timings, bus speed - is highly sensivity for dos gaming system.
I'm upload full results later, after finish my tests.

Try an ATI PCI card, on the RAGE chipset. I've done a lot of testings in the past myself in a Pentium class system and found that quite a lot of PCI cards similar to what you tested were all the same in DOS gaming. Then I tried a Rage 2 PCI card and it was around +40% faster than everything else. If you can find one.. try it in your 486. Also of course though, a voodoo3 PCI card would probably be the fastest you're going to get for dos gaming.

My testing show that Rage is about 20-30% slower then some S3. I find that the fastest PCI ones are Tseng ET6000 and Nvidia TNT, followed by a 3dfx Voodoo3 and Matrox Millennium.

philscomputerlab wrote:

@jwt27

- Doesn't 3dbench only have 3 digits?

Yes, but the 1st number is actually calculated even on the fastest systems, but there is no space on the program to display it 😜

GeorgeMan wrote:

philscomputerlab wrote:

@jwt27

- Doesn't 3dbench only have 3 digits?

Yes, but the 1st number is actually calculated even on the fastest systems, but there is no space on the program to display it 😜

But how do we know? Sounds like an assumption. The program is clearly not designed for scores higher than 1000 and I think these results will have to go.

Are you sure you meant to address that to just jwt27? There's a few other people who have results like that in there. I'll get the tests run on a couple more laptops (Toshiba Satellite 200CDT and Dell Inspiron 8000, probably) soon.

Lukeno94 wrote:

Are you sure you meant to address that to just jwt27? There's a few other people who have results like that in there. I'll get the tests run on a couple more laptops (Toshiba Satellite 200CDT and Dell Inspiron 8000, probably) soon.

Put them all to "too fast". Now we have the problem some fast machines with low 3 digit results. Might just kick out 3dbench2 for fast machines as it might confuse people.

philscomputerlab wrote:

GeorgeMan wrote:

philscomputerlab wrote:

@jwt27

- Doesn't 3dbench only have 3 digits?

Yes, but the 1st number is actually calculated even on the fastest systems, but there is no space on the program to display it 😜

But how do we know? Sounds like an assumption. The program is clearly not designed for scores higher than 1000 and I think these results will have to go.

Not go. Just as is now, not calculated. 😀

kithylin wrote:

Try an ATI PCI card, on the RAGE chipset.

One of local forum members perform that tests, rage is not in a first place.

kixs wrote:

I find that the fastest PCI ones are Tseng ET6000 and Nvidia TNT, followed by a 3dfx Voodoo3 and Matrox Millennium.

My own top list for classic PCI 2d cards almost the same: 1) Tseng et6000, 2) Trio64 v+, 3) Virge, Matrox (don't remember which i've had), Voodoo3, ... 4) Trio3d/3d2x, CL5430-5436, ...
But fo all this cards difference in lowres is about 1-3%, only in svga modes is about 5-15%.
Don't have pci version of Riva TNT.

philscomputerlab wrote:

@jwt27

- Doesn't 3dbench only have 3 digits? Your results are 1000+
- Missing benchmark entries

I was overclocking a Katmai 550 at the time, and as I went from 660 to 682MHz the 3DBench2 result rolled over from 975.6 to 010.5, so I'm fairly confident the actual score would be 1010.5.

For the other CPUs I've submitted I'm not too sure since the lowest results I got were already >1000. You're right, saying "too fast" is probably better here (although it takes the total score down)

The missing entries mean... it crashed there 😀

JFYI.
I've try to find difference in benchmark, depends on vga ram size.
Found nothing in PC Player Benchmark in any mode, and only drop-down on 1mb in vesatest (640x480x32):
.
Platform: Am5x86-33 (33x3), atc-1415, 32mb ram.

Second that, my results say the same. Results are exactly the same for what concerns dos, but again this is also the expected behavior.

It's a shame that the Quake demo doesn't appear to run under Win2K, as everything else did. Ah well, I've got the results from my Toshiba Satellite 200CDT and from both Speedstep settings of the Dell Inspiron 8000 added, at least. It's quite interesting to see how much difference the extra 300 MHz made to Quake, whilst doing barely anything at all for Doom.

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 2.25MB PCI
BIOS set to fastest settings

In the process of adding the Dell Inspiron 8100 to the database... interesting how much difference using a Tulatin PIII in place of the Coppermine unit makes at the same clock rate (although an extra 128MB of RAM may have a small effect, I would imagine most of it is the faster FSB) in everything bar 3DBENCH. I'd be intrigued to see how much faster the Radeon Mobility-equipped models are!

Memory timing can make a huge difference. Especially on BX440 chipset machines with 100 MHz FSB, most PC133 sticks are rated for CL2 at 100 MHz 😀

philscomputerlab wrote:

Memory timing can make a huge difference. Especially on BX440 chipset machines with 100 MHz FSB, most PC133 sticks are rated for CL2 at 100 MHz 😀

Both use the 815E chipset, and I think at least one of the PC133 sticks in the Inspiron 8100 is actually rated for CL2 at 133 MHz, not sure though. The Inspiron 8000 has 256 MB of PC100 RAM, no PC133 sticks.

Double post, but switching to Low Power mode on the Toshiba Satellite 200CDT makes an enormous difference - even though the CPU clock stays the same, framerates are consistently halved across the benchmarks.