pshipkov wrote on 2022-08-28, 19:19:

of course you can create more partitions.
they can be uup to 2gb. Just the first one is limited to 512mb for maximum compatibility.

I would agree but fdisk days otherwise.

Chadti99 wrote on 2022-08-28, 10:48:

I couldn’t find the exact data sheet for the 8ns SOJ we’ve been using but I think this is similar.
LP61256GS.pdf

Here’s the one for the DIP.

Need more from you here to help you, I need to see the problem

Show me the motherboard, size/clearance constraints,

Sketch your design suggestion, socketed? Smd soldered? Vertical daughter board?

Is the datasheet you linked the same dimensions? Same pinout?

I can’t open the datasheet right now on my phone I will have to look later.

Show me the chip you want to adapt physical pictures

Show me the exact motherboard it goes on. The chip it is replacing, show both on the socket that I san see the size relationship

Do you have a caliper? Show measurements. Make me feel like it is in front of me so I can visualize the problem.

May wanna make a build thread for this, and quote these replies 😀

@chadti99
you should be able to create more partitions from fdisk option 2.

@sphere478
lets split this into its own thread.
it is not clear yet which 1Mb L2 cache mobos can handle reliably 3x60 and 3x66 MHz and may benefit from such adapters.
that needs to be established first. the few inspected already are not quite there. couple of more to go.
but in the meantime we can sort-out the adapters spec.

pshipkov wrote on 2022-08-28, 22:57:

@chadti99 you should be able to create more partitions from fdisk option 2. […]
Show full quote

@chadti99
you should be able to create more partitions from fdisk option 2.

@sphere478
lets split this into its own thread.
it is not clear yet which 1Mb L2 cache mobos can handle reliably 3x60 and 3x66 MHz and may benefit from such adapters.
that needs to be established first. the few inspected already are not quite there. couple of more to go.
but in the meantime we can sort-out the adapters spec.

Sounds good. You go ahead and make it and outline the problem since you are closer to the subject and gather info and compile it and I’ll help where I can. This will be your project, just tell me what you need.

split-out adapter thread: soj-dip adapter

---

CDTek 286M, based on Texas Instrument's TACT82301PB/TACT82302PB/TACT82303PB chipset.
After all this time from the late 80ies early 90ies it is still in great physical condition.

Came with 12MHz rated AMD processor.
Does not go a single megahertz above that.
Tried hard with good quality SIMM memory and 25MHz rated Harris/Intersil CPU - no and no.
BIOS offers options for shadowing, memory relocation and EMS.

Even at 12MHz its clock-to-clock performance is unimpressive.
Didn't bother capturing performance metrics.

Works fine for what it is.

pshipkov wrote on 2022-08-29, 01:07:

Even at 12MHz its clock-to-clock performance is unimpressive.

How much variance do you see in 286 systems, clock to clock, assuming 0WS? My understanding of the platform is that unless there are wait states being inserted there simply isn't that much room for the chipset to hamper the CPU since it accesses everything as a "local bus"

There is a lot of variance between the different motherboards.
That will be evident if you skim through the previous posts in this thread touching on the 286 subject.

Some mobos handle computation tasks really well but their graphics performance is bad.
Some allow for overclocking of up to 35 MHz but perf is still low despite that.
Some handle FPU well and the few apps that can take advantage of that benefit from it.
Some are slow on the IDE side.

There are various factors for that - the quality of the chipset itself, imposed wait states, BIOS microcodes.

Anyone have the jumper settings for POD on the m919? I keep getting HIMEM errors @100MHz on a known 100MHz capable cpu. I think I’m setting voltage correctly. As it does need at least 4v on other boards I’ve used it in. 83MHz is fine.

L1 cache is stuck in WT if that is any indicator.

In this post where i looked at bunch of mobos + POD100 there is a note for M919:

IDE driver version 3.2 and /D0:17 /MM0 /F0 flags.
L1 cache in WT mode since motherboard does not have proper support for P24T. Only for P24D (L1 WT).
Ark1000VL didn't work at all.
S3 Trio64 was the second best VLB video card that produced better results in some of the tests than MGA PCI.

pshipkov wrote on 2022-08-29, 21:21:

In this post where i looked at bunch of mobos + POD100 there is a note for M919: […]
Show full quote

In this post where i looked at bunch of mobos + POD100 there is a note for M919:

IDE driver version 3.2 and /D0:17 /MM0 /F0 flags.
L1 cache in WT mode since motherboard does not have proper support for P24T. Only for P24D (L1 WT).
Ark1000VL didn't work at all.
S3 Trio64 was the second best VLB video card that produced better results in some of the tests than MGA PCI.

Got it, missed that somehow on the L1. Odd this cpu is struggling at 100MHz on this board. Or maybe it isn’t 🤣.

The main reason i started posting my little investigations here was to have a public record and be able to go back and remind myself about stuff that otherwise gets forgotten.

At the beginning of the linked POD100 post a premise is established that unless issues are listed with given motherboard - everything works flawlessly.
The only note for M919 does not indicate instabilities, so things were working as expected with L1 WT.

Yes and I thank you for the index in the first post, I reference it often. The board is listed under the POD100 summary as working well with the P24T so that is probably where my confusion originated.

For whatever reason my experience with the m919 has been a bit different. I am using a different storage solution which might explain why I can’t allocate more than 504MB of disk space. LBA mode won’t work with the AMI bios and my SD adapter. LBA mode will work with the Award Bios, although doubtful it’s meant for this board as It disables VLB graphics.

My POD100 instability could be related to the award BIOS, I’ll try switching back to AMI.

I was also thinking the instability could be related to incorrect jumper settings as there are no specific P24T settings mentioned in the manual or from online sources I’ve found.

Was hoping I could provide some benchmark results with the POD@100MHz using 1024k cache but it’s fighting me.

In better news, my Voodoo4 isn’t dead, fires up on a slot 1 board just fine. Was worried there for a bit.

Your contributions are great addition to the effort, so thank you too. : )
2 > 1

M919 is a prick but you are having more trouble than i experienced.
Did you try a normal CF card + CF-to-IDE adapter + AMI bios.
Attached you the page of the electronic manual for the CPU jumpers as well as the latest AMI BIOS.
If you can provide POD100+1Mb_L2_cache numbers that will be great. Otherwise you reminded me that i didn't update that POD100 post accordingly and will try to eventually find the time to do it.

You guys should kick me out of here, thought I was using the latest AMI bios this whole time and I wasn’t. Thanks for posting this here pshipkov!

So I’m able to use LBA mode with this bios if I manually adjust the cylinder count down.

Also my POD is running stable at the 5 volt setting.

And my Voodoo4 is posting again.

With L1 in WT and L2 in WT I get the following:

POD100, 1024k cache, Voodoo4
3d Bench 1.0c - 89.6
Chris VGA - 74.2
Chris SVGA - 15.5
PCP VGA - 27.3
PCP SVGA - 8.5
Doom Max - 1486
Quake - 25.5

Not too impressive, lack of WB support for L1 def hurts.

Good we sorted this out.
Also, no kicks will be awarded this time. : )

This board with Voodoo4 is becoming a bit special for UM888# implementation . I would say even - unique.
Good finding there.
But still no-go with Voodoo3 ?

Thanks for the results.
It will be really great if can run and 3DS chevy test, LW3D blade runner and WinTune2 ?
Or at least some of them.
The simple DOS tests are kind of meaningless on their own and don't really tell good story.

Yes, M919 does not support P24T for some reason.

Are you sure your linked BIOS for the M919 with file name 9191016S.ROM is the latest? This is what I have recorded. Did I make a clerical error?

9190506.BIN -- 05/06/1996
9191016S.ROM -- 10/16/1996
9190914S.ROM -- 09/15/1998

Chadti99: YOu should always have the motherboard jumpered for 5 V when using a POD because the POD has an onboard 3.x volt regulator.

The m919 cannot do POD in WB mode, so trying to get optimal performance out of this configuration is destined for failure. only POD in WT mode will work properly.

From Chadti99's tests, it sounds like that AWARD BIOS for the m919 is a waste of time. I sort of recall doing some tests to this degree many years back and ultimately deleted the AWARD BIOS.

Just tested the Voodoo3, it kicks the monitor on but never displays anything.

I’ll run more tests later this evening.

@feipoa, this is a voltage modded POD and it’s usually happy at 4v but now I’m remembering someone said the 4v setting on the m919 actually delivers less than 4v, so maybe that’s the issue.

Thanks for the note about POD and L1 WB, a shame really, could of been a performer.

The Award bios seems fine if you don’t want to use VLB graphics or a POD@100. So an AMD@180 with PCI graphics would be a good use case, no messing with jumpers to get full speed on PCI.

Trying the Voodoo4 once more in Windows on the M919.

Checking my posts in the “486 Overkill Graphics” thread I did manage to get this card working on an LSD but was also having similar power related issues. Maybe these socket 3 boards or my PS just can’t deliver enough amps. I mean look at the Tesla Coils hanging off this thang.

Yep, so I ran GLQuake for about a minute and black screen.

Ok, so Voodoos are really not happening on UM888# boards.

---

Spent a moment over the weekend to upgrade an Alaris Cougar II motherboard with better BL3 CPU.
This post talks about running the system at default 75MHz (3x25) and then a narrow overclock to 100MHz (3x33), where the second configuration was very unstable - mostly because of the CPU not able to handle the additional 25MHz well.

Since then i found few more BL3 CPUs. They have the same markings on top but their overclocking capabilities are quite different.
The best one is used in the 386 PC based on DTK's Symphony mobo - blazing fast and stable.
The second best went to the Cougar board in question here and enabled it to handle the 100MHz much more reliably - that is with air cooling.
System is still not 100% stable, but very close to there.
All BIOS settings on max. The hardware can handle it just fine.
Complete stability can be achieved by disabling the L1 cache. Obviously performance tanks in this case.
With L1 enabled several limitations are imposed:
- adding an FPU results in no POST
- things are pretty good with 4Mb RAM only. Can work ok with 8Mb but with even more lowered stability. Further increasing the installed memory significantly lowers stability.
As mentioned in the linked above post - Ark1000VL video cards are problematic with this motherboard. Performance is great but frequently enough they will lock it to a level that requires power-off for some time. S3 Trio64 is the way to go at a slightly lower performance in DOS.

I know that applying active cooling to the CPU using Peltier element will take care of the instabilities, but that won't be as easy as it sounds in this case and needs to be given some thought.
- The CPU is surface mounted which limits attachment options for the Peltier+heatsink+fan contraption.
- There are ISA/VLB/FPU slots and sockets that limit the size, shape and structure of the cooling system.

The improvement is an evolutionary step.
Hope i manage to achieve a fully stable system soon.
If that happens - it will be the best pixel pusher in its class.

Video showing things in action - Doom, PC Player Benchmark, Superscape, Wolf3D, Landmark, SpeedSys (with long and boring memory graph), BIOS settings, physical world.

Interactive graphics DOS tests:
Doom: 27.5 fps
PCPlayer: 10.7 fps
Superscape: 52.6 fps (55.5 fps with Ark1000VL)
Wolf3D: 90.7 (99.8 fps with Ark1000VL)
Quake 1: impossible because of FPU related instability

IDE: ~7 Mb/s

Windows and related tests:
very unstable

Complex compute:
impossible because of FPU related instability

DTK system metrics for reference