Alright here we go.

AMD 5x86 200MHz 66*3 @5volts
LS486e-RevD w/8ns Cache mod
16MB 50ns EDO
Voodoo 3 2000

1:1 66MHz 313 Fastest
Doom 996
PCP 480 13.9
PCP 240 31.2
Chris SVGA 26.6
Chris VGA 84.4
3D Bench 1.0c 115.6
Quake 21.8
VESA Memory speed 44,588 KB/sec

Some issues at 1:1 66MHz
Can’t ctrl alt del
Blank screen exiting bios
Lose bios settings on power cycle sometimes

1:1/2 66MHz 313 Fastest *Windows Install Stable
Doom 1125
PCP 480 13.2
PCP 240 31
Chris SVGA 25.5
Chris VGA 82.4
3DBench 1.0c 110.8
Quake 21.8
VESA Memory Speed 37,141 KB/sec

I really wish the Biostar could compete with these numbers. I really like the mounting solution I Macgrubered. Not that it’s that far off and this CPU may still settle there. As it is I can’t case this LSD/5x86/370 cooler setup without a horizontal desktop style case.

Chadti99 wrote on 2022-02-14, 00:31:

Some issues at 1:1 66MHz Can’t ctrl alt del Blank screen exiting bios Lose bios settings on power cycle sometimes […]
Show full quote

Some issues at 1:1 66MHz
Can’t ctrl alt del
Blank screen exiting bios
Lose bios settings on power cycle sometimes

Sounds pseudo-stable. I'd stick with 1:1/2 for peace of time. Might also increase lifetime of the hardware?

feipoa wrote on 2022-02-14, 02:17:

Chadti99 wrote on 2022-02-14, 00:31:

Some issues at 1:1 66MHz Can’t ctrl alt del Blank screen exiting bios Lose bios settings on power cycle sometimes […]
Show full quote

Some issues at 1:1 66MHz
Can’t ctrl alt del
Blank screen exiting bios
Lose bios settings on power cycle sometimes

Sounds pseudo-stable. I'd stick with 1:1/2 for peace of time. Might also increase lifetime of the hardware?

Agreed everything is running really well at 1:1/2.

Really puzzling - the 1:1 at 3x66MHz issue you are facing.
Is it the same if you hard-freeze the CPU for example ?

I would go with a copper heatsink - it is better than aluminum.
Maybe consider some 3D printed contraption to hold the thing.

Btw, i have been slowly curating from time to time the 8ns L2 cache chips on one of the LSD motherboards (really asking for trouble here, hope i don't damage too much the solder pads) - but i think stable 3-1-2 at 3x66MHz is getting within reach.

I have this PCB for an LED MHz display that I had to keep soldering and desoldering the pins on. Perhaps about 6 times. After each 3 times of solder on/off, I would ultimately damage a via. On one such case, the connection between the via and the trace got severed. On another instance, the via itself was sucked into the desolder pump. One has to be really careful with repeated heating/cooling. However, the manufacturing of some PCB's are much better than others in this sense.

I got a load of 8 ns chips from China. They look legit, along with all my other parts. I was going to put 8 ns for the TAG on my LSD board. Not sure what it buys me, but why not 'eh.

I know very well what you mean.
I have 3 lsd mobos. 2 are of great pcb quality - one of them is getting the treatment.
The third one - not so great quality. It literally melts under the heatgun with pads and rails all falling like mushrooms growing after rain.
Brought it up in previous posts.

In hindsight, when swapping the cache modules, I wouldn’t have added solder back to the pads. I should have set the chips direct on the bare pads so they would have been level. Was a huge pain checking and rechecking continuity for every pin. Fortunately I didn’t get one of the really fragile boards that pshipkov mentioned. I swapped my tag chip as well, can’t hurt.

pshipkov wrote on 2022-02-14, 08:51:

Really puzzling - the 1:1 at 3x66MHz issue you are facing. Is it the same if you hard-freeze the CPU for example ? […]
Show full quote

Really puzzling - the 1:1 at 3x66MHz issue you are facing.
Is it the same if you hard-freeze the CPU for example ?

I would go with a copper heatsink - it is better than aluminum.
Maybe consider some 3D printed contraption to hold the thing.

Btw, i have been slowly curating from time to time the 8ns L2 cache chips on one of the LSD motherboards (really asking for trouble here, hope i don't damage too much the solder pads) - but i think stable 3-1-2 at 3x66MHz is getting within reach.

It’s possibly the stability issues at 1:1 would be resolved with a peltier, but I want to keep this build air cooled, although I might try it down the line. The other CPU I have that requires a peltier at 200 runs better at 1:1.

Stable 3-1-2 at 66 would be sweet, keep us posted!

A 3d printed part might work if I can find a lip or edge around or under the socket to grab. Was also thinking about something that would plug into PCI bracket and help hold it. Someone with better skills than me could probably swap out the socket.

Chadti99 wrote on 2022-02-14, 11:06:

In hindsight, when swapping the cache modules, I wouldn’t have added solder back to the pads. I should have set the chips direct on the bare pads so they would have been level.

That was certainly my finding as well, that is, to get absolutely all the old solder off the pads and off the RAM modules (if reusing). It can be difficult to make a sold bridge in air.

feipoa wrote on 2022-02-10, 14:15:

There is no size trim pot which can be used with the onboard regulator to get 4.3 - 5.0 V regulation. The low drop-out voltage regulator drops a minimum of 0.7 V. You'd have to switch to an ultra low drop-out regulator, which might drop 0.3 V.

Is there a part you might recommend that would work in this application?

Not really. You'd have to hunt down some ultra low drop-out regulators on digikey and see if the pinouts match.

feipoa wrote on 2022-02-14, 12:09:

Not really. You'd have to hunt down some ultra low drop-out regulators on digikey and see if the pinouts match.

Gotcha, already see one with a 0.5 volt drop out, I’ll keep digging.

Chadti, for the printed contraption - the support should not be tied to the motherboard itself, but some support beams or other elements that are connected to the PC case frame and firmly presses the fan + heatsink to the CPU. I can think of at least 2 designs for that.

@pshipkov, maybe you could do a couple quick sketches of the designs, I’m failing to envision it.

Mocked up something quick using the models i had from previous print jobs.
I think it can be as simple as that.
The "?" suggest that the element connecting to the heatsink and the main bar can be bolted, but i can think of a snap-kit like construction that will allow easy insertion/removal.
The exact connection between the bars pushing on the heatsink and the heatsink itself is TBD - depends on its exact design.

This is from the top of my head, but there are other designs that can be utilized.

Something different.

The whole computing thing started for me in the second half of the 8-bit and the beginning of 16-bit periods. So naturally everything 8 / 16 bits is of significant interest here.
For long time i was able to resist the urge to get back to it given the ongoing activities around 286, 386, 486 and Pentium 3 class hardware, but i was only lying to myself for as long as i can before the time comes and i cannot keep resisting.
Reached that breaking point last year.
Here is what happened next.

I wanted to setup an 8086 machine since never touched something like that back in the day. Only 8088 stuff.
Unfortunately there are not a lot of options to pick from - a handful of branded PCs and very few (1?) standard XT motherboards.
Never owned branded computers - too rigid and constrained ... but at a premium price, so there was a strong resistance to break the "rule" and instead kept going for and back about the different options.
Discarded portable PCs.
Discarded the IBM, Amstrad, Siemens stuff for obvious reasons.
Discarded other highly proprietary implementations and designs with low-level of integration.
Only 3 real candidates kind of passed the filters:
Juko Nest XT - XT standard through and through, looks like potent assembly ... but unobtanium.
Olivetti PCS86 - has a proprietary PSU interface, soldered on-board CPU and RAM. Not ideal for my peak performance needs (overclocking). Ignored it at the end.
Tandy 1000 SL

Tandy 1000 SL PCs are still available online from time to time.
Conform to the XT standards - basically, normal proper computers.
Have few minor customizations that can be turned-off easily not to get in the way.
8086 based.
Socketed CPU, FPU, RAM - good for overclocking activities.
An obvious choice.

Managed to obtain the full package not far ago.

On the outside (before whitening - covered in following posts):

Monitor damage during transit to here. 🙁
(Fixed it later with custom 3D printed pieces and matched paint color.)

There is a lot of charm in CGA graphics, but let's be real - this will be VGA at the end.
Latest model Diamond SpeedSTAR24 - with those surface mounted components. This one is with 8-bpp RAMDAC.
Grossly overpowered for the host machine. May downgrade to something more appropriate.
Have several pre-90 ISAs. Will see which one fits best.

On the inside - barebone - leaves a lot of room for improvements, which is great since journey is the mission.
All PCBs have this factory fix with the blue wire.
Needs thorough cleaning. Did it later.

--- Upgrades ready to go in.

CPU: 16MHz rated NEC V30 HL:

FPU: 10MHz rated Intel 8087:

RAM: 80ns rated 640Kb:

CLOCK: Smart watch module. Because i really care about correct date on this machine:

EMS:
Acculogic RAMpAT! memory board for up to 16Mb of RAM.
High chance to be discarded - there is no software to take advantage of it. Most likely will end-up in the fast 286 VLSI PC.
Its place can be taken by something more modest like Lo-tech 1Mb RAM board. We will see.

LOCAL STORAGE:
HDDs are inconvenient now-a-days, especially "standard" stuff like Tandy's hard card.
Lo-tech XT-CF-lite rev.2 takes the lead.
Not sure about the red color ...

Not convinced about adding dedicated sound card and modem.
There is a built-in audio system - more of a buzzer, but it goes well with the whole XT thing.
Modem - feels like another piece of baggage in the trunk.
Thoughts ?

Most likely won't bother with a second floppy drive.
Not sure if want to splurge for Tandy mouse, so standard one will do for now.

While the PC case is sustained its color, the monitor and floppy have this Tabaco shade that may need treatment.

---

Tried briefly last night the XT-IDE card - BIOS menu does not show up on screen.
Verified the card is functional in 486 PC, so some issue with memory addresses.
Will be looking at this over the weekend.
Probably need to switch the boot from DISKMATE to MS-DOS.
Any ideas ?

---

Objective is to setup a well rounded XT and see how far it can climb on the performance scale - the usual.
About to get to it in the coming days.
Any feedback welcome.

(EDIT: story continues here and here)

Thanks for the drawing pshipkov, this really helps.

In other news, thought I got a decent deal on a Promise EIDE2300+ VLB controller that looked brand new. When I fired it up I got a spark and magic smoke from component marked TC3 on the board. I don’t think I’ve damaged my motherboard thankfully. Is there any saving this guy or am I asking for trouble?

@Chadti99

Very recently i had the same close encounter with Hornet Technology's VL-300 Multi I/O controller.
Long story short - plugged it into Asus VLI - power supply didn't start.
Plugged it into PC-Chips -> explosion. Funnily, after that capacitor exploded the controller actually started working.
Lesson learned - going forward new peripherals will be plugged for a first time in a low value motherboards.

So, is your controller working or not ?
Hi change it does.

---

Spent some time with the XT.
Gutted it.
Cleaning pass, but more is needed.
Soldered Crystal Oscillator sockets.
Swapped the memory chips, CPU and FPU.
FPU runs really hot. Decided to put heatsinks. Most likely will add a fan too. After the ADZ fiasco cooling is high-on my list.
Disabled Deskmate. Switched BIOS to boot from HDD into MS-DOS. XT-IDE adapter worked right away.

Few minutes later the system was ticking at 16.7 MHz - "brutally fast". I'll take it.

Can boot with 55MHz oscillator but with 320Kb RAM only. Need to curate the chips as well.
Also, the system does not boot every time.

ISA bus can be overclocked independently from CPU. That will be next on the list.

Power supply location and placement really benefit from the professional test bench setup.

Captured couple of performance metrics with the usual synthetic benchmarks.

It is not like there is a lot of "competition" in the 8-bit department, so this can easily become one of the fastest XTs around.

(EDIT: story continues here)

pshipkov wrote on 2022-02-19, 23:40:

@Chadti99 […]
Show full quote

@Chadti99

Very recently i had the same close encounter with Hornet Technology's VL-300 Multi I/O controller.
Long story short - plugged it into Asus VLI - power supply didn't start.
Plugged it into PC-Chips -> explosion. Funnily, after that capacitor exploded the controller actually started working.
Lesson learned - going forward new peripherals will be plugged for a first time in a low value motherboards.

So, is your controller working or not ?
Hi change it does.

I’ve not tried firing it back up. I probably at least need to change the caps and find a throw away vlb board to test it with.

i would remove the capacitor and try it that way, but in a cheapo mobo.
still, it is your call really.