Good comments. Unfortunately, I've put the system away for now. I feel like we should be live on IRC when I'm testing things so that I don't have to backtrack later. I don't always take notes when there's a failure.

I do plan on re-evaluating the BL3 hack once I've ordered the ultra low dropout regulator, which I think should only dropout 0.2 to 0.3 Volts max, meaning I can take it to 4.7 - 4.8 V. I don't have plans to connect my variable DC supply. if I'm going to desolder wires/components and put it all back together when I'm finished, I just as well wait for the new VRM and do everything at once.

It is a little intimidating doing these tests on my finalised system BL3 system. At one point, all the system hangs caused some damage to my system.ini and I wasted 30 minutes trying to fix that. I think I'll try it again with my spare ALi Panda board. However, I think it best to try other motherboards first. My Daewoo ALi Panda board is finecky. You don't even have control over the ISA bus speed - its always at 7.1-ish MHz. Thus, VLB is a must for it, and thus the tower of sockets!

I think the 46 C was the temp of the VRM, not the CPU. The CPU wasn't even warm to the touch when that 25 mm fan was on.

I am not using an SD-to-IDE adaptor; I don't own one. I am using a SCSI2SD adaptor, version 6. I have a dedicated unit for testing purposes.

It is possible that the BL3 needs a peltier to get a higher speed. Will have to wait for heatsink removal before I can experiment with this. The fact that 3x33 is way more reliable than 2x40 implies to me that there's more at play than just heat, voltage, and operating frequency. My goal is to figure that out, and to do so, I should find another board in my bin which works with the BL3, at least in DOS.

I'm pretty sure I tried the BL3-100 with the L1 disabled and noted it would boot to DOS and run DOOM. DOOM ran so slow, I didn't let it finished and decided to try more settings with L1 enabled.

Your last line is a strong clue that the CPU is not happy with the 100mhz on its own.
It may need extra V, or Peltier, or both. I bet on 5V Peltier with the CPU V you have now.

I managed to get my SiS Rabbit board working with the BL3 hack. I'm not sure exactly what the issue was before, but I suspect some ISA PnP trouble between the SCSI and LAN cards. Luckily, I was able to get it working at 40x2 MHz easily, so as I guess the Daewoo ALi Panda board just doesn't do well with a BL3 and 33/40 MHz FSB.

At 3.6 V,
75 MHz, 25.0 x 3 , ISA at 10 Mhz, DOOM = 4736
83 MHz, 27.6 x 3, ISA at 11 MHz, DOOM = 4286
90 MHz, 30.0 x 3, ISA at 12 MHz, DOOM = cannot

80 MHz, 40.0 x 2, ISA at 13.3 MHz, DOOM = 3842
90 MHz, 45.5 x 2, ISA at 11.3 MHz, DOOM = cannot

I had hoped that upping the voltage to 3.9 V would let me run 90 MHz, but it didn't. Unfortunately, my 87 MHz oscillator, when installed in this board, heats up and screen doesn't turn on. So until I update the VRM, looks like 83 MHz is the max. Divide by 5 was the least divisible option, so 11 MHz to the ISA is it. I will try a MR BIOS to see if I can get a denominator of 4 to bump the ISA up to 13.8 MHz.

I also plan to put in my Evergreen BL3 module to see ensure that the Evergreen logic isn't slowing down the performance.

Too bad ALI Panda cannot handle well the BL3 CPU beyond the 25MHz.

I like it when things out of a sudden decide to work. Good outcome with the SIS Rabbit motherboard.
And the Doom number for 2x40 is not bad at all - 20 fps. It is getting into the playable zone.
Can you remind me if you can run SXL2 natively at 45/50 MHz on that board ?
That will be a good indication if you the inability to run the BL3 chip at 2x45.5 is the chip itself or the mobo.
Maybe extra voltage later will help it get over the curb.

It would be the cpu, and not necessarily the motherboard, because the CPU also would not run at 30 x 3.

Dont know what is the lesser evil inthis case.
The 3 cpus here are quiet different in OC capability.
One of them cannot go past 80mhz, second one maxes around 90, third one can go up to 105-110.
Feels like lacking QC at the fab for this silicon.

Looks like your sxl2-66 running at 2x40 will be a better deal.
Wish i could lay my hands on such cpu.

I tabulated quite a bit of data on the SiS Rabbit board using the BL3 hack, Evergreen BL3, and SXL2-66. Some observations,

1)
Clock-for-clock, the IBM Blue Lightning, or BL2/BL3, is between 5% and 7% faster than the SXL2 when run on the Chaintech 340SCD (SiS Rabbit). By comparison, the SXL2 was 8.4% faster clock-for-clock than the BL2 on the DTK Symphony board (DTK PEM-4030Y). Can anyone explain this?

2)
In light of the results from 1), I wanted to take some clock-for-clock values of the Evergreen BL3 vs. the BL3 hack on the SiS Rabbit. This is to determine if the onboard logic on the Evergreen BL3 is hindering performance. From the results, we see only a 0.7% drop when using the Evergreen module compared to the logicless BL3 hack.

3)
Unfortunately, my stacked SXL2-66 device wouldn't run at 80 MHz, so 17.33 fps at 75 MHz was the best it would do on the SiS Rabbit, however it went up to 20.52 fps on the Symphony 461 board. Similarly, the BL3's best performance on the SiS Rabbit was 19.44 fps at 80 MHz. On the SiS Rabbit, the BL3 scored 17.57 fps at 75 MHz, however it went up to 18.93 fps on the Symphony board. Thus, when transitioning from the SiS Rabbit board to the Symphony board, and keeping everything else identical, e.g. settings/hardware, the SXL's improvement was 18.4%, while the improvement for the BL3 was only 7.7%. Hopefully, the newly designed SXL2 interposers will let us hit 80 MHz or higher, for which I predict a DOOM score approaching 22 fps.

4)
A strange and repeatable oddity observed was concerning ISA speeds; sometimes an increase in ISA frequency results in lower benchmark results. Why is this? Are frames beings skipped? An example is the SXL2 at 66 MHz - 16.21 fps for ISA at 11.1 MHz and 15.42 fps for 13.3 MHz. Also look at the SXL-45 and BL-45 between 11.3 MHz and 15 MHz and BL2-66 from 11.1 Mhz to 13.3 MHz.

5)
The BL3 was able to do 1x50 MHz on the SiS Rabbit, while the SXL was not.

6)
The Evergreen BL3 wasn't able to cope at 83 MHz, yet the BL3 hack was. I kept pushing for 90 Mhz, but no go. I will have to get my Buffalo unit out to tabulate results from 90-110 MHz.

Good story.

1. Can this be a cache policy thing ? The DTK BIOS does not allow control over that. Cannot think right now of a tool that can do it so we can experiment.

2. Good info.

3. Same as 1 ?

4. This is strange. Will check tomorrow on my side.

5. I also had some issues with SIS Rabbit at 50MHz base frequency. Not surprised it was not ok with some cpus.

6. Less components - smaller change for things to go wrong.

Letss see what the buffalo guy can do next.

From my previous test results with this board, the Buffalo at 100 Mhz, or 3x33, yields 24.34 fps at 11.1 MHz ISA.

I also have SXL2-80 results written down from years back, which is 20.9 fps at 10 MHz ISA. From 3), we saw that transitioning to the Symphony board, an 18.4% improvement was observed for the SXL2 CPU. Adding this 18.4%, we get an SXL2-80 result of 24.75 fps on the Symphony 461 board. Also, if we increase the ISA rate from 10 MHz to 13.3 MHz, we may get yet another 1 to 1.5 fps improvement, so around 26 fps may be possible. Hail to the king baby.

I tested three MR BIOS ROM's for the SiS Rabbit board, with files names 9B303, 9B306, and 9B307. While these ROM's are functional, they would detect a maximum of 8 MB RAM and wouldn't detect any L2 cache. With the BL3 installed, there was an option to enable/disable the internal cache. For L2 cache, there's the option to select 32K, or >=64K. I have 256K, but selecting the latter didn't help. As usual, MR BIOS works with PS/2 mice out of the box.

That's all the testing I plan on doing with MR BIOS on this system.

I wonder if someone couldn’t become a expert on those mr bioses and reconfigure them better for specific motherboards. They may be easier to decode than other bioses 🤔

@Feipoa
The standard implementation for SIS 303 306 307 seems to be 8Mb RAM and 64Kb L2 cache.
Your board is one-off thing. That's why Mr.BIOS is not handling it properly.

pshipkov, I'm not using SiS 303, I'm using SiS 310/320/330. Were those MR BIOS ROMs that you sent me for the 303/6/7? If so, that would explain the naming convention used. Do you have for 310/20/30?

My bad. I meant 310/320/330.

Thermalwrong wrote on 2022-10-19, 23:45:

I'm not going to take the CPU off my Blue Lightning board, but I would like to say that you've done a great job with remodelling that PCB to accomodate the voltage regulator. That's excellent that it works with the AM1117 too, does make me wonder why my board has such a powerful regulator instead.
Thank you for the solder paste tip 😀

I know the feeling as well when things don't work quite right - I've been remodelling one of my laptops to mod from DSTN to TFT and when I think things are working great initially, problems have cropped up and it really kills my energy for working on it.

Hello thermalwrong,
I saw that you bought a Toshiba t4900ct laptop. I also want to buy this model and have one Question.
Is sound working under Dos games?.
Thank you

How hard is it to find qfp-132 bl3 chips?

Are they cheap?

MB17 wrote on 2022-12-27, 19:40:

Hello thermalwrong, I saw that you bought a Toshiba t4900ct laptop. I also want to buy this model and have one Question. Is so […]
Show full quote

Thermalwrong wrote on 2022-10-19, 23:45:

I'm not going to take the CPU off my Blue Lightning board, but I would like to say that you've done a great job with remodelling that PCB to accomodate the voltage regulator. That's excellent that it works with the AM1117 too, does make me wonder why my board has such a powerful regulator instead.
Thank you for the solder paste tip 😀

I know the feeling as well when things don't work quite right - I've been remodelling one of my laptops to mod from DSTN to TFT and when I think things are working great initially, problems have cropped up and it really kills my energy for working on it.

Hello thermalwrong,
I saw that you bought a Toshiba t4900ct laptop. I also want to buy this model and have one Question.
Is sound working under Dos games?.
Thank you

This is the better thread to post this on: Re: The quest for the perfect retro laptop: a saga I think you're new enough to the forum that you can't use PMs yet otherwise I'd send you a private message as reply. The T4900CT has a Windows Sound system so you need WSSXLAT.EXE to have sound blaster functionality in DOS, otherwise you've only got FM / Adlib sound, not digital audio.

I have removed the AMS1117 800 mA low dropout regulator in favour of a 1 A ultra low dropout regulator, the Texas Instruments TPS72501DCQR. If I recall right, the AMS1117 couldn't go above about 3.9 V on this setup. The 72501 should theoreticaly output up 4.7 - 4.8 V.

On the AMS1117, the tab was Vout, while on the 72501, the tab is GND. Thus I have re-positioned the VRM as shown so that I can connect the tab to the ground ring. Maybe that will help with cooling?

I'm using a 124 K-ohm 0805 resistor with a 500 K-ohm 12-turn mini trim pot. I'm using a 10 uF ceramic for Vin and a 22 uF tantalum for Vout. Soldered it up as shown,

The unit is functional. I tested 3.2 V at 83 MHz. I cannot go to higher voltages until the epoxy dries. I decided to add four 100 nF 0603 caps to each side of the QFP package in the middle, as shown on two sides here,

A major drawback to this type of hack job is that I cannot easily swap out the BL3 CPU for another. I would have to re-assemble all the external components again.

I was able multiple times to solder/desolder ICs in close proximity. Takes carefully placed multi-layer aluminum foil screens to protect stuff that shouldn't get hit by the hot gun. Looking at what you have here - it will be fine, given proper preparation.

Nope, won't work here. If you look carefully, you will see the hurdles.

a) 30 AWG wires going under and over some of the QFP leads - the cladding will melt and shorts will evolve.

b) plastic encasing of the trim pot will melt. it might survive, but will look horrible.

c) those 4x 100 nF decoupling caps on the four corners each have a 30 AWG wire lead going to a QFP pin directly. Zoom in on the photo. Those would all need to bre redone.

d) not sure if the kapton insulation layer tape melt or not, haven't tested it. This point is last for good reason.

Ideally, I would want to find another of these interposer and modify it as well. I have one extra trim pot and one extra ultra low dropout reg. I have two NOS BL3 chips, one to fill the void on my IODATA module, the other to experiment with.