Meatball wrote on 2023-08-19, 20:16:

fosterwj03 wrote on 2023-08-19, 19:53:

You might be thinking about Windows XP. […]
Show full quote

Meatball wrote on 2023-08-19, 19:26:

Sounds great! […]
Show full quote

Sounds great!

I presume there's a "for fun" angle (of course!), but there are some caveats to be wary (of which you may already be aware):

You don't need Advanced server for multi-core or hyperthreaded machines. It is licensed by socket up to 8 physical sockets. Regular server is licensed up to 4 sockets. Professional is licensed up to 2 sockets. You won't see any advantage performance advantage for server versions. You could have a million threads and cores on a single socketed machine, and Professional is still the requirement.

Windows 2000 has no idea about hyper-threading (or cores for that matter). You should disable hyper-threading for Windows 2000 (and NT 4.0). You are likely hurting performance as the OS thinks they are actual physical processors. You can get away with multi-core, but still it not designed with cores in mind. With that said, to make sure you are getting your money's worth, check this thread: https://msfn.org/board/topic/174127-info-on-w … r-cpu-machines/ Alternatively, upgrade to Windows XP/2003. I know it's not the goal, but these are the first Microsoft OSs to recognize multi-core and hyper-threading.

If it's RAM you're concerned with, Professional and Server support/licensed up to 4GB, and 8GB is the maximum for Advanced Server. If you want more, you'd have to upgrade to Datacenter for support as high as 32GB.

You might be thinking about Windows XP.

While you're generally correct about the licensing, Windows 2000 treats each core (and virtual thread) as though they are separate socketed processors. I need to use the server version (or the server hack) to allow Windows to use more than two cores.

It is true that Windows 2000 was never officially updated to optimize hyperthreading, but it does see all eight virtual processors when enabled on the Xeon. While not optimal, multithreaded apps will use all of the threads.

It is just for fun. I don't really have a use case for Windows 2000, but I love retro rockets.

That Windows 2000 would only use the first two cores, I had no idea, so I learned something there. Yes, it does see eight virtual processors, but it treats them like physical processors as we've both outlined. And because hyperthreading shares many of the same resources with the primary core, they will have resource scheduling issues. It may be what is causing 3DMark issues. Anyway, I'm glad I responded as I learned something new!

I would have thought the same thing about the lack of optimization based on what I read about hyperthreading back in the early 2000's. Lots of articles said it wouldn't help or might hurt older OS's, but those articles never gave any real-world evidence. They probably just figured no one would ever bother to use an old OS on newer hardware.

I decided to run a couple of quick synthetic tests and one real-world test to see what the impact would be. Here are the results:

4 Cores; HT Disabled (4 Threads)

Sandra 2001 ALU - 47938
Sandra 2001 FPU - 19603
CPU-Z Single - 709
CPU-Z Multi - 2802

Crysis Run 0 - 56.28 FPS
Crysis Run 1 - 59.33 FPS
Crysis Run 2 - 59.35 FPS
Crysis Run 3 - 59.52 FPS

4 Cores; HT Enabled (8 Threads)

Sandra 2001 ALU - 58604
Sandra 2001 FPU - 35049
CPU-Z Single - 709
CPU-Z Multi - 3290

Crysis Run 0 - 56.21 FPS
Crysis Run 1 - 59.54 FPS
Crysis Run 2 - 59.33 FPS
Crysis Run 3 - 59.36 FPS

Based on these numbers, Windows 2000 hyperthreading works without detriment in the synthetic tests. These results indicate a 17 - 78% improvement in multi-processor scores with hyperthreading enabled. CPU-Z's single thread performance also seemed unaffected by the presence of hyperthreading.

Crysis also seems unaffected by hyperthreading with results in-line with the disabled results.

I wonder if Intel improved on-die thread scheduling in the Core i-series to the point where it didn't matter so much if the OS was "Hyperthreading Aware". Just a thought.

I am taking apart some old medical equipment and I think I have found the cutest Pentium 3 Mainboard ever. It’s the size of 2 hands. My plan ist to make this board my PC104 test station.

DerBaum wrote on 2023-08-20, 02:14:

I am taking apart some old medical equipment and I think I have found the cutest Pentium 3 Mainboard ever. It’s the size of 2 hands. My plan ist to make this board my PC104 test station.456AE217-A98E-421E-B596-403D4A208D57.jpeg

What did you take that out of? It's always cool to find PC components in unexpected places.

Ensign Nemo wrote on 2023-08-20, 03:17:

What did you take that out of? It's always cool to find PC components in unexpected places.

Its was a box with an integrated speaker and an external scanner to read letters and books to blind people.
It was running Windows 98. The software on it was controlled by some buttons on the device (no keyboard and monitor). Its a OCR program with integrated TTS functionality.
The software is still working. I have no scanner but i can read regular text files in a quite natural voice (for 2003 standards...)

appiah4 wrote on 2023-08-19, 20:49:

Win2K Pro licensenindeed capas out at 2 processors, no limits on Win2K Server though.

So not true, pached you can use 4 core cpu. Phenom II 965

ediflorianUS wrote on 2023-08-20, 08:03:

appiah4 wrote on 2023-08-19, 20:49:

Win2K Pro licensenindeed capas out at 2 processors, no limits on Win2K Server though.

So not true, pached you can use 4 core cpu. Phenom II 965

Patched with what?

It is an unofficial solution and more a hack than a patch.

appiah4 wrote on 2023-08-20, 10:32:

ediflorianUS wrote on 2023-08-20, 08:03:

appiah4 wrote on 2023-08-19, 20:49:

Win2K Pro licensenindeed capas out at 2 processors, no limits on Win2K Server though.

So not true, pached you can use 4 core cpu. Phenom II 965

Patched with what?

It is an unofficial solution and more a hack than a patch.

There is a "patch" of sorts. Someone wrote a program a few years ago that modified the registry in such a way to make Workstation think it was Server.

Edit: It's called "NTSwitch". It modifies several registry entries simultaneously to circumvent MS's license protection features. The registry changes survive reboot, and Windows 2000 Workstation will boot with the Server features (including the boot screen and log-in prompt). NTSwitch can change it back, too.

I've used it as well on my Core i5-3570, and it does work. It's a bit wonky, though, and some programs don't respond well to the Server features. In my case, I needed Advanced Server to use all 8 threads, so NTSwitch wouldn't have worked for me.

Win2k will use the second CPU, be that a Hyperthreaded one or the second core on the same CPU socket, but the licencing sees the physical CPU socket on the motherboard and its a 1-2CPU licence.
I have it installed on a Dell Precision 690 with 2x X5365 Xeons and it only sees 2 of the 8 CPUs cores the system has.

I'm doing a bit of cleanup of my tech/retro room... I need at least space to move around 🤣

Among many things... I also found one unopened package from Oct.2019 - bought something from eBay 🤣 Will check it now 😉 In case some are wondering why you don't receive a feedback... 🤣

kixs wrote on 2023-08-20, 14:16:

I'm doing a bit of cleanup of my tech/retro room... I need at least space to move around 🤣

Among many things... I also found one unopened package from Oct.2019 - bought something from eBay 🤣 Will check it now 😉 In case some are wondering why you don't receive a feedback... 🤣

Opened... it's just an Adaptec AHA-1542CP SCSI ISA controller card with floppy connector 😀

I did a reball on a broken GF256 DDR. Didn't help. It's dead Jim.

Installed Windows on one of our kids Scenic-T 815E. Sure is a nice system, although it's a bit picky on the ram. Dunno. I think it'll need a pass with memtest at some point.

There's three motherboards waiting repair..
- Asrock K7S8X rev3.01 with only half the ram voltage. A schematic would be needed
- Epox P55-VP3 waiting to ID a part that blew after a recap (yes these are installed the right way around 😉)
- K8VM800-RH waiting for a recap. Maybe some evening next week. 20+ caps to do.

Plus a few awaiting bios flash to something recent

How certain were you that it was the GPU and not the memory or some over component? Reballing the GPU is one of the riskiest forms of repair, and should only be considered as a last resort.

What equipment did you use for the reball?

Kahenraz wrote on 2023-08-20, 16:01:

How certain were you that it was the GPU and not the memory or some over component? Reballing the GPU is one of the riskiest forms of repair, and should only be considered as a last resort.

What equipment did you use for the reball?

Not detected properly by windows (Win98).
I used a hot air station to remove the GPU (after preheat), reballed it with proper (0,76mm IIRC) balls using a reballing jig. Mind you, I've reballed TFBG88's and AD6528 cpu's successfully before. This was way too easy (big balls)

Patch was somewhere on win98 patches webpage, don't remember exactly. I can find and upload on request.

Dig up old documentation about the DOS Disk Transfer Area, the FILEINFO structure, and the DOS function FindFirst, FindNext.

Also read the FreeDOS kernel source code.

I didn't know it would be so complex... You can search remote files, character devices, Volume IDs, etc... 😮

DerBaum wrote on 2023-08-20, 03:35:

Its was a box with an integrated speaker and an external scanner to read letters and books to blind people. It was running Windo […]
Show full quote

Ensign Nemo wrote on 2023-08-20, 03:17:

What did you take that out of? It's always cool to find PC components in unexpected places.

Its was a box with an integrated speaker and an external scanner to read letters and books to blind people.
It was running Windows 98. The software on it was controlled by some buttons on the device (no keyboard and monitor). Its a OCR program with integrated TTS functionality.
The software is still working. I have no scanner but i can read regular text files in a quite natural voice (for 2003 standards...)

That's a great find. Now I just need to find a bunch of these for cheap and make a YouTube video describing how to turn them into cool retro gaming machines. Then I can sell them at a marked up price on eBay. Lol. All kidding aside, that's not far from what has happened to old thin clients that can run DOS and Windows 98.

For fun and the hope of an easy fix, I bought this 'broken isa VGA card' off the internet auction site for cheap:

But it was no easy fix, I almost gave up on it.
This card had a bunch of problems that I think the seller tried to fix - it had no BIOS for a start, that's tricky to fix since the original BIOS is a mask ROM rather than an EEPROM. I think the seller either never had it, or removed it when putting it up for sale. Maybe it was removed long ago when disposed of.
Such an odd card, why was some of the 16-bit ISA connector almost cut off? Why remove half the VESA feature connector? They'd tried jumpering links to make up for the removed bits of the PCB but it had way more problems than just a couple of missing traces around the RAMDAC section.

The ROM situation was tough to solve, I learned the hard way that the ROM data that can be read from the OS is not what the EEPROM contains. This is because the address pins are ordered differently, maybe on purpose? Specifically, the ROM's A14 links to ISA A0, then A0 > A1 etc. This is probably an artifact of the switch from 2x 27128 ROMs to 1x 27256. Maybe the data pins are mixed up too, not sure.

There was no complete ROM that I could find that was specifically for the 8900C, there was one for the 8900D that was taken from the ROM itself but most other ones I could find were extracted in the OS. Putting regularly formatted data into the EEPROM gave me the "no video card" beep error, because the different pinout means the computer would then see garbled data.
The TVGA 8900D BIOS did work but the display was a garbled mess because the RAM wasn't fixed yet so I kept looking. Eventually found the ROM but not on the live internet - it was on a russian site which isn't accessible now, possibly for reasons, but the internet archive had the page and thankfully had the ROM data saved too.
This website and the internet archive backup of it, seems to be the best repository for ISA VGA ROMs: https://web.archive.org/web/20230430144018/ht … om/video_en.htm
To fit my 28C256 EEPROM where the card was designed for a 27256 mask ROM, EEPROM Pin1 had to be kept out of the ROM socket, along with pin 27. Run a blue wire from the EEPROM's Pin1 goes into the socket's Pin27, then the EEPROM's Pin27 is soldered to Pin28 (VCC) for normal operation.
And that has to be undone/redone each time to try a different ROM 😀

Even when I got it detecting / running, the card displayed a garbled mess and swapping out the ram didn't immediately resolve that. In the end I had to clean out the original solder from the VIAs in the RAM area, replace the RAM chips (did this twice) but put the chips on two at a time - top & bottom, furthest column first, touch up the solder and then do the next column. The RAM chips were too close together to get a soldering iron in between them otherwise.
Normally SMD soldering like this wouldn't be an issue, but this card has open VIAs practically on the SMD pad, so solder likes to flow down there instead of staying on the pad, meaning the SMD pads with VIAs next to them have far less solder than the surrounding pads, leading to unconnected traces.
The temperature I used was 280C from the hot air station, the solder was bubbling with 320C. There's no ground plane really so it was quick to heat up.

There were damaged traces around where each capacitor had been swapped out, I know it's the seller that did this because he was also selling that particular brand of 100v capacitors recently too. 100V capacitors on a card that might see 5.5v 😀 Mostly those now have 16v capacitors and MLCCs put in in place of the 100nF capacitors, one was missing and another crumbled to pieces when heated, maybe that even caused the RAM fault, I'm not sure.

The RAM chips were swapped out but I don't think they were at fault, probably a bad solder joint because of the VIA situation. Also all the main chip legs got reflowed because that's the easiest way to fix most VGA cards from this era.

There's a couple of things I really wish I had not bothered with:
1. Removed all the 74LS logic components to test them out of circuit - pointless
2. Removed the ROM socket and replaced with a new one, the socket was loose but it was working. If it's posting and detecting the card, it's not a ROM fault. Sadly in doing this I took some of a via/trace out and had to run a bodge wire for the ROM to work again.

I got another ISA VGA card too, also cheap and had artifacting on the display, but that was fixed in less than 10 minutes of troubleshooting because it was a damaged trace from a bash that's visible in the seller's pictures.

Ensign Nemo wrote on 2023-08-20, 19:52:

DerBaum wrote on 2023-08-20, 03:35:

Its was a box with an integrated speaker and an external scanner to read letters and books to blind people. It was running Windo […]
Show full quote

Ensign Nemo wrote on 2023-08-20, 03:17:

What did you take that out of? It's always cool to find PC components in unexpected places.

Its was a box with an integrated speaker and an external scanner to read letters and books to blind people.
It was running Windows 98. The software on it was controlled by some buttons on the device (no keyboard and monitor). Its a OCR program with integrated TTS functionality.
The software is still working. I have no scanner but i can read regular text files in a quite natural voice (for 2003 standards...)

That's a great find. Now I just need to find a bunch of these for cheap and make a YouTube video describing how to turn them into cool retro gaming machines. Then I can sell them at a marked up price on eBay. Lol. All kidding aside, that's not far from what has happened to old thin clients that can run DOS and Windows 98.

It's a good use for them though, all that thing needs is something to convert the PC104 to ISA and it'd be quite a novel little system.

Ensign Nemo wrote on 2023-08-20, 19:52:

That's a great find. Now I just need to find a bunch of these for cheap and make a YouTube video describing how to turn them into cool retro gaming machines. Then I can sell them at a marked up price on eBay. Lol. All kidding aside, that's not far from what has happened to old thin clients that can run DOS and Windows 98.

I think that will be a little difficult...
In the 10 years im activly searching for them i only found 3.
One of them came with the original paperwork and i could see that one of these things cost as much as a small used car... Super high end medical market...
Only pretty wealthy people would have got one of these things. They are made in germany and probably just sold here... If they sold 500 of these things in the 10 years it existed that would be much...
AND they come with different hardware inside from 486 to pentium 3 ... you will never know what you will get...

Broke out the Versas last night, M/75 survived the move just fine, was BBQing on the patio while playing Postal on it. The P/75 however, is cracking to pieces again. I"m almost tempted to just fix up the plastic pieces for the Versas and make molds for the plastic on them. I also may have found a way to get the M/75 Adlib and SoundBlaster support using the KXLd/20 card. I need to back up the M/75 so I cna xfer some of the stuff on it over to my FMAK9200D.

I tried getting into the HDD in the FMA3500 via OSX, but forgot that it has STACKER on it - oh the joys of File Compression. Probably time to put WFW and that Serdashop WiFi Modem thing on it instead to back up the HDD. I want to slap a big 8GB in it. I did find the battery though, finally.

Replacing the chips on my MS-6168 V2 board to get 16MB of video memory.

First boot:

Oops. Somewhere a loose contact, when I press a chip, the output changes.