Update on the Terminator II with AMD DX4-100, the first impression was quite disappointing, maybe I understood why, entering the BIOS I saw that by default, the RAM times are more conservative, maybe because the RAMs at that time could be slow, and in this way it ensured compatibility with any RAM, so considering that I have a FSB 33 and 60 ns EDO RAM, from the BIOS I selected all the fastest settings (Fastest 2-1-1-1 0WS), I memorized them, and I redid the benches (including those of Phil's).

At startup, I notice that the PC is better, I have the doubt that now the L2 cache works, in fact it doesn't work yet ️, but a higher DOS startup speed is visible.

Even Phil's benches confirm a clear improvement, which was more or less what you could expect, so let's see the best scores, with slow and fast settings, AMD 486DX4-100 CPU:

1) 50.0 > 66.6 2) 49.1 > 64.8

3) 35.7 > 45.0 4) 10.4 > 12.6

5) 11.4 > 15.1 6) 4.6 > 6.0

A) 67.59 > 98.02 b) 28.07 > 36.20

C-e) 8.1 > 10.4 d) 3.8 > 4.8

K) 127.8 > 192.0 m) 180 > 201

N) which would be Speedsys 4.70, with the score of 37.24 obtained only with the fast settings, with the slow ones it could not work.

In addition to these I have done others, and they all show an improvement, but the L2 cache problem always remains to be solved, I think it would make even better scores, we will see later if there is a problem with the chips (failures?), or it is something else.

One of the things I would like to do in the next few days, is to continue experimenting on the SOYO SY-5BT (5BT5), the model I own has an L2 cache chip with missing pin, not long ago I managed to discover, milling the plastic, the inner part of the PIN, so that I can weld a micro wire, and restore the original connection.

This will most likely allow the operation of the L2 cache, with relative performance gain.

But if for some reason it doesn't work, the only way to have an L2 cache, is to use a CPU with an integrated L2 cache, like for example an AMD K6-2+, but this requires a VCORE of 2.0V or less, it depends on which we have, the minimum voltage for these CPUs, should be 1.5V, a little far from the 2.2V of the normal K6-2, so some time ago, I had found all the necessary resistances to make the voltages from 2.3V to 2.7V, which can be added to the position of the jumper for 2.2V, or bypassing the resistance, they can be connected directly to PIN 2 (of the jumpers).

For voltages below 2.2V, so from 2.1V up to about 1.4V, I have to weld a wire on the track that connects the resistors, in order to bypass them, then I will only have to put a guinea pig CPU, and measure the VCORE on the mosfet, it doesn't matter if the CPU with much lower voltages doesn't start, but I think that up to 1.6V any K6-2 can start, I'm interested in checking the VCORE, I don't have to do stability or bench tests.

In short, in a few days I could do the welding of the wire, for the VCORE, regarding the L2 cache chip, I don't think I can weld a micro wire, for the moment I don't have adequate equipment, to do such a job.

Today I dedicate a few minutes to the 486 PCI Terminator II motherboard, the PC currently has an AMD 486DX4-100, 16 MB RAM 72 PIN (2X8), a PCI Miro S3 1 MB video card, an ISA SB16 sound card, a HD of about 540 MB, floppy 3.5 from 1.44 MB, in short, an adequate equipment for using DOS Windows 95, but currently I only use DOS, without Windows.

After running the various benches, I changed VGA to see if it made a certain difference, so from the results of the bench we are the same or very close, a sign that the card I used does not make much difference, for the tests I used the 2 MB PCI SiS, maybe a different card like a Matrox PCI, it could earn a little (better scores?), or they could be limited by the speed of the CPU, which in this case could slow down the flow of data coming from the GPU (?).

I'm not going to report the scores obtained, and given the results, I put the VGA PCI Miro S3 from 1 MB, I don't know if by expanding the memory to 2 MB, you can gain something, in the bench with higher resolution, I don't think so, but most of the benches use low resolutions, so it should be the same to have 1 MB or 2 (?).

A further possibility could be the use of an accelerator board, but I think that even in this case, it would be limited in performance from the CPU, maybe it improves the bench a lot, but maybe it works below the real potential, surely it would be better on an S.7 of about 200-500 MHz, in short the 486 are a little slower than the Pentium equal frequency, but even if it is possible to push them beyond 133 MHz, there is always a certain gap, and for me at that point, you might as well go on the S.5 or the S.7.

Going back for a moment to other PCs, lately I have evaluated the reconfiguration of my PCs, some are quite similar if not even the same, I will see in the future to differentiate them, and not create the same PCs, both for CPU and for equipment.