VOGONS


Reply 20 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member
Andy1979 wrote:

Interesting to see how you've approached this. Starting down a similar path, trying to make best use of the old PCs and spare components I have collected over time, but don't have room for them all. It's like trying to choose the perfect "three car garage".

Space has always been my biggest constraint, too. Although, to a certain extent that's a good thing because it keeps things from really getting out of control 😉 Thankfully, I now have a really decent amount of space in our home office to use for this hobby. I hope you'll post about your project as well. I'm always interested in what other retro gamer do with their parts. Sometimes, those projects provide unexpected inspiration for your own big pile of unused parts or remind you of a specific system or part you have lying around somewhere but had already forgotten about. Good luck with your project 😀

Andy1979 wrote:

Good find with the EDO SIMMs.

Yeah, that really was a lucky coincidence. At first I couldn't really believe it and had to double-check the model numbers on my own modules.

END OF LINE.

Reply 21 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

Finally got some time to install the two new RAM modules. Also added a Noctua NF-A6x25 FLX to remove hot air from the V5's backside and the RAM banks. One of the V5's fans was getting louder, so I replaced it with an original spare but I still found the system too loud. When the V5 sits in a tower case below the desk the fan noise is not that noticeable. But in this desktop case it is kind of annoying. So I decided to replace both fans with Noctua NF-A4x20 FLX fans. Definitely a huge difference! I now can choose between having superior airflow with less noise (NF-A4x20 FLX as-is) or having roughly the same airflow than the original fan with even less noise (NF-A4x20 FLX with Noctua's low-noise adapter). Currently, both fans run at full speed.

The post with the system specifications and pictures has been updated accordingly!

END OF LINE.

Reply 22 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

Free time is pretty rare at the moment, unfortunately. But nevertheless, I could fiddle around a little with Retro 1 over the weekend.

I used Partition Magic 7.0 to partition the SSD and performed preliminary installations of MS-DOS 6.22 and Windows 95c. BootMagic 7.0 is handling the active flag for the primary partitions (and gives me a nice OS selection screen, fully configurable through the Windows application). I'm using both applications because I still have my retail copy from back then. Never had any issues with those and really liked them, especially PartitionMagic.

DOS now has the CD-ROM driver working as well as the mouse driver but I didn't invest much time yet. First priority for me was to check whether Escape from Monkey Island would run on this machine (background: refer to the initial post). And it does. That's great. The only issue I encountered is that V.Control won't work under Win95. It runs but all changes applied won't have any effect. Fellow vogons user retro games 100 encountered the same thing, so I'll just revert back to the 3dfx tools instead of wasting hours trying to find a potential solution. This means I'll have to update the IE4 to SP1, according to the 3dfx installer. Luckily, I've already found the update package, just have to test it.

Next step is checking the other end of the performance scale. Hopefully, I can get all the older DOS adventures running on this machine. Prerequisite is to install the AWE64 Gold and change the MIDI port away from 330 which is used by the MPU-401AT. Then I'll have to work a little with the config.sys and the autoexec.bat to get a decent memory configuration. At first, I'll try my luck with just running all the games, starting with the oldest and see what runs with the machine being as-is. In case of speed issues I'll give SetMul a spin to deactivate L1 and/or L2 cache.

In case that doesn't work or if it is not enough, I'm thinking of making the CPU configuration jumpers being switchable from the backside of the case. I'd modify one of the unused slot brackets to hold three flip switches (SPDT) that would be connected to the respective three pin header groups on the motherboard. The lowest available setting is a 150MHz Pentium Pro with 60MHz host bus, 30MHz PCI bus, 7.5MHz ISA bus. The lowest setting keeping the PCI bus at 33MHz and the ISA bus at 8.33MHz is a 166MHz Pentium Pro with obviously 66MHz host bus. And this could always be combined with SetMul cache deactivation.

Another option might be wait state insertion via BIOS which seems possible with this motherboard, it has an option for boot speed. Per Intel's Technical Product Specification for this board, the motherboard then operates at the speed of an AT system. This however might be already too slow for some of the potentially speed sensitive games.

I'll just have to wait until I get to really test the games. If problems with the high system speed should arise then I'll work my way through the possible solutions 😀

END OF LINE.

Reply 23 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

This weekend I only had a couple of hours free time left for the project because of a nice little family trip. So I tried to figure everything out that's left for the Windows 95 portion of the system. Shouldn't take too much time. Or so I thought.

Since V.Control isn't working I will use the original 3dfx tools. The installer told me that I need IE 4.01 SP1. I already managed to track down a Microsoft file for IE 4.01 SP2 which tested fine. 3dfx tools installed fine too. However, setting 4x FSAA still showed no effect in Escape from Monkey Island. After a while of testing with different settings, I installed Grim Fandango and launched it. Result? FSAA is indeed working. Strange.

So I went on installing all original 3dfx V5 drivers chronologically in order to find one that works. The original release showed the same issue, no FSAA. The next release, 1.01.00, worked. 1.03.00 worked too. The latest beta, 1.04.01b also worked. Then I reinstalled 1.04.00, and now it suddenly did work too (at this point I tried V.Control again but it really doesn't work, independent of the driver "issues"). Very interesting.

There were no error messages, no warnings, nothing, over the entire process of installing, rebooting, testing, uninstalling, rebooting, installing, rebooting, testing... A nice waste of time, but at least I do know that everything works and how to get it working. When I'm going to setup the final installation I'll check whether the initial installation of 1.04.00 works. And if it doesn't I'll just reinstall the driver. And if that still doesn't work I'll install all the drivers chronologically in order to finally get it working.

Next step is refining the MS-DOS installation as detailed in the previous post.

END OF LINE.

Reply 24 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

This weekend I had a little more time available to work on Retro 1.

The first session was focusing on refining the DOS installation, as already mentioned. I had just "slapped" DOS 6.22 onto the SSD back then in order to see whether the drive would boot. Later on I had added partitions and Windows 95c as well as a Boot Manager (PowerQuest BootMagic 7.0 in this case).

The initial DOS installation had been modified to load mouse (ctmouse.exe) and CD-ROM (videcdd.sys, switched to xdvd2.sys later) driver to simplify file transfer. I have no way of getting the IDE SSD connected to any of my modern systems in order to easily dump files on the partitions. I didn't want to search for another adapter but most likely I will in the near future. What I do have and what was of great help in the process of setting this thing up is a Toshiba USB floppy drive (for 3.5" disks). It works flawlessly under Windows 7 64-bit (read/write/format). I used it in a VM with Windows 7 32-bit to create boot disks, and BIOS update disks. So, after the initial installation using floppy disks, the majority of files have been transferred using DVD and CD.

Then I spent quite some time trying to get all devices set up and a decent memory configuration. There were a few minor obstacles in the process but nothing severe. The AWE64G driver installation wasn't working at first. I had to unpack the software packages on the SSD instead of trying to copy the unpacked files over. Then the card refused to get initialized although CTCU claimed that everything is OK and there are no resource conflicts. After changing the resource configurations it worked. One driver wouldn't load, startup errors with the buffers, fcbs and stacks values (all typos in the config files). Lost the keyboard localization (not really an issue but annoying) and had to key it back in again. But I got everything sorted out within those few hours.

The results are quite good, I think. Free conventional memory: 638.928, free upper memory: 42.832, largest executable program size: 638.912 (624K), largest free upper memory block: 30.480 (30K), available space in high memory area: 15.392 (15K). I cannot recall ever having this much memory back then... 😀 Of course, many of the tools and drivers available now weren't existing then... Almost everything is loaded high now, leaving only MSDOS, HIMEM and EMM386 in the conventional memory.

I will make a temporary backup of the current config files in case something happens to the installation or the SSD itself (I don't want to go through all the configuration hassle again 😉). When all games are running, I will make a final backup of those files.

So, next up is testing. But first, some additional pictures.

  • retro1_test_post.JPG
    Filename
    retro1_test_post.JPG
    File size
    1.95 MiB
    Views
    2406 views
    File comment
    POST
    File license
    Fair use/fair dealing exception
    retro1_test_bios.JPG
    Filename
    retro1_test_bios.JPG
    File size
    2.09 MiB
    Views
    2406 views
    File comment
    BIOS (CPU info)
    File license
    Fair use/fair dealing exception
    retro1_test_bootmanager.JPG
    Filename
    retro1_test_bootmanager.JPG
    File size
    2.46 MiB
    Views
    2406 views
    File comment
    Boot Manager
    File license
    Fair use/fair dealing exception
    retro1_test_dos_mem.JPG
    Filename
    retro1_test_dos_mem.JPG
    File size
    1.93 MiB
    Views
    2406 views
    File comment
    Memory Report
    File license
    Fair use/fair dealing exception

END OF LINE.

Reply 25 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

So, testing it is...

In order to be able to quickly test games I copied the files from another PC that runs all these games in DOSBox 0.74 and transferred them on discs.

Test setup:

  • - Pentium Pro @200MHz, all caches on
    - AWE64 Gold @IRQ 7, DMA 1 & 5, MIDI port 300
    - MPU-401AT @standard settings

First thing tested was The Secret Of Monkey Island (EGA, patched with NewRisingSun's great Sound Blaster sound effects + Roland MIDI music patch, thanks a lot for that 😀). Called with "mtadl" (using mtadl.exe from the patch package).

Result: Runs fine, music (CM-32L) and sound effects (AWE64G) working

  • retro1_test_monkey.JPG
    Filename
    retro1_test_monkey.JPG
    File size
    2.12 MiB
    Views
    2402 views
    File comment
    The Secret Of Monkey Island (EGA)
    File license
    Fair use/fair dealing exception

Next test was Monkey Island 2 (with one of NewRisingSun's great Sound Blaster + Roland MIDI patches, thanks again 😀). Called with "monkey2 s".

Result: Runs fine, music (CM-32L) and sound effects (AWE64G) working

  • retro1_test_monkey2.JPG
    Filename
    retro1_test_monkey2.JPG
    File size
    2.29 MiB
    Views
    2402 views
    File comment
    Monkey Island 2
    File license
    Fair use/fair dealing exception

Third test was Day Of The Tentacle. Here I expected some potential trouble as this game is on the list of the CPU speed sensitive games on Vogons Wiki (The Secret Of Monkey Island is too but it is patched by NewRisingSun). Called with "tentacle r[330]".

Result: Runs fine, music and special sound effects (CM-32L) as well as speech and sound effects (AWE64G) working

  • retro1_test_dott.JPG
    Filename
    retro1_test_dott.JPG
    File size
    2.47 MiB
    Views
    2402 views
    File comment
    Day Of The Tentacle
    File license
    Fair use/fair dealing exception

So far, everything seemed good. Although I have to state that I didn't play through any of these games. I just tested briefly using the intro sequences and a few minutes at the beginning of each game. Extensive testing has to follow, especially for cases like Day Of The Tentacle. I wanted to check on General MIDI too, so I tried Sam & Max next.

Fourth test: Sam & Max Hit The Road. At first, I didn't expect any trouble, the game is not on the aforementioned Vogons Wiki list. But using setmuse.exe to configure the sound system I was finally encountering issues 😉 Setting up the AWE64G for speech (as SB16) worked fine, the card was tested OK by setmuse.exe. Setting up music as General MIDI on port 330 for the SC-55 and then testing it lead to an instant freeze of the application and the system. And then I recalled that the exact same thing had happened even in DOSBox. This was the first game I had to use a fixed amount of CPU cycles for. Leaving DOSBox on auto cycles would instantly freeze setmuse.exe on MIDI test or the game application on startup.

Using setmul.exe to deativate caches the situation didn't effectively improve. Results:

  • - L1 off, L2 off: Works, but way too slow, unplayable, unbearable
    - L1 off, L2 on: Still too slow, no change
    - L1 on, L2 off: Instant freeze

So, I know that the system needs to be slower, but deactivating the L1 makes it too slow. Switching off the L2 has no effect. Maybe, the setting is not working correctly on this machine.

Next step was to change jumper settings to set other CPU & bus frequencies. Results:

  • - Pentium Pro @166, PCI bus @33, ISA bus @8.33: Instant freeze
    - Pentium Pro @150, PCI bus @30, ISA bus @7.5 (lowest supported setting): Instant freeze

Then I combined those with setmul.exe, but still, to no avail. At least this spares me the trouble of making those jumpers switchable from the outside of the case 😉

Next step is testing a slow down utility that can hopefully decelerate the CPU enough but not too much... I have already downloaded some promising ones that are listed on the respective thread. Does someone have any experiences in this regard? Which ones are reliably working and produce a consistent/reproducible level of performance? From what I've read, I thought about starting with "Bremze" and "WHOA!".

And, at least from my point of view, Sam & Max should definitely be on the CPU speed sensitive games list 😢

END OF LINE.

Reply 26 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member
voodoo5_6k wrote:

Using setmul.exe to deativate caches the situation didn't effectively improve.
[...]

So, I know that the system needs to be slower, but deactivating the L1 makes it too slow. Switching off the L2 has no effect. Maybe, the setting is not working correctly on this machine.

I have checked the BIOS settings now too. Turning off the caches with the BIOS makes the system too slow to even initialize the AWE64G.

voodoo5_6k wrote:

Another option might be wait state insertion via BIOS which seems possible with this motherboard, it has an option for boot speed. Per Intel's Technical Product Specification for this board, the motherboard then operates at the speed of an AT system. This however might be already too slow for some of the potentially speed sensitive games.

Well, I tested it now, and yes, same result as above, the system is even too slow to initialize the AWE64G (at some point I had no more patience and hit reset).

voodoo5_6k wrote:

Next step is testing a slow down utility that can hopefully decelerate the CPU enough but not too much... I have already downloaded some promising ones that are listed on the respective thread. Does someone have any experiences in this regard? Which ones are reliably working and produce a consistent/reproducible level of performance? From what I've read, I thought about starting with "Bremze" and "WHOA!".

So, my first attempts with these two tools failed. With "Bremze" I was not able to find a so called braking coefficient that worked for Sam & Max. I tested the highest and the lowest value, as well as several others somewhere in the middle. Every time I would attempt to test the MIDI music setting setmuse.exe would instantly freeze. With "WHOA!" I had no luck either. There is very little documentation in the package. I will have to search for more information on how to configure it, if possible at all.

My plan for the weekend is now to check out "Mo'Slo". There seem to be plenty of options, including the processor emulation feature which sounds quite interesting (although it of course is more of an approximation than a real emulation). Maybe I have more luck with that tool.

END OF LINE.

Reply 27 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

Today I tested "Mo'Slo", to no avail. No setting worked, everytime I tested the MIDI music or launched the game itself, the system froze. In order to exclude an issue with the DOS installation itself, I transplanted my Asus P/I-XP55T2P4 Rev. 3.00 with a Pentium 233MHz MMX into Retro 1. Booting the unchanged DOS installation didn't change anything (the system still freezes on testing MIDI music or launching the game), but disabling the L1 cache did the trick (on the Pentium Pro the slowdown caused by disabling the L1 cache was too much to run the game). I will look a little deeper into that tomorrow, but maybe I have to change the hardware configuration or Retro 1 again... It would be sad to not be able to use the Pentium Pro, but if the Pentium MMX can do what the Pentium Pro cannot (running all my favorite Lucasfilm/LucasArts adventure games on a single machine), well then... So be it 😀

END OF LINE.

Reply 28 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

Retro 1 is really giving me a headache. Well, not literally, but I didn't really expect this system to be that delicate. I'll first summarize the current situation before I'll elaborate on what can be done with my available parts.

Intended Use:

  • Run Lucasfilm/LucasArts adventure games.

User Requirements:

  • 1. OS
    1.1. MS-DOS 6.22 (older games)
    1.2. Windows 95c (newer games)
    2. CPU
    2.1. CPU >= Intel Pentium class 200MHz (for the newest game)
    2.2. CPU must be able to decelerate "enough" to run the speed sensitive games
    3. MB
    3.1. Intel chipset
    3.1.2. Support at least 64MB RAM, more is preferable
    3.2. ATX format
    3.3. CR-2032 battery
    3.4. At least two ISA slots, three or four are preferable (AWE64G + MPU-401AT)
    3.5. At least two PCI slots, three or four are preferable (Voodoo5 + Intel NIC)

Preferred Solution:

  • Intel Pentium Pro 200MHz 1M & Intel VS440FX

Problem:

  • I haven't found a way to slow down this CPU just enough to run all of the older games. It is either too slow or too fast. Nothing worked. Configuring lower system bus speeds and multipliers wasn't enough, the system is still too fast. Turning off the caches or inserting wait states is too much and the system is too slow. All software based slowdown methods didn't work, i.e. they didn't create a performance level that would allow the speed sensitive games to run. Unless I do find the "magic" solution for the speed issue, there is no way I can use this system for what I want Retro 1 to be.

Alternate Solution 1:

  • Intel Pentium 233MHz MMX & Asus P/I-XP55T2P4 Rev. 3.00 (512KB onboard cache and upgraded with TAG SRAM to cache up to 512MB RAM)

Problem:

  • Although this CPU could be slowed down enough to run the speed sensitive games the Asus BIOS is worse than the VS440FX BIOS in terms of HDD detection/limitations. I had my share of trouble finding a (modern) storage solution that would just work on the VS440FX, but the Asus board is worse. It won't even boot from some floppy boot disks with the wiped IDE SSD connected (interestingly it could boot with the DOS installation of the VS440FX on the SSD, although the boot manager wouldn't work). I tried the latest BIOS, and with that it couldn't even detect the IDE SSD anymore (detection looping during POST, no timeout). I went through all BIOS versions between my original one (0202) and the latest (0207). One update (0203) had obviously the same basics, because everything got detected. But later versions (0204 upwards) felt like a major change, the memory test was way shorter and nothing got detected anymore on the IDE controller. If I decide to go with this system as Retro 1, I'll have address the storage system issues (e.g. dynamic drive overlay software, Compact Flash card with up to 8GB, Disk-on-a-Module with up to 8GB etc.pp.).

Alternate Solution 2:

  • Intel Pentium 200MHz & Intel Advanced/ML (256KB onboard cache)

Problem:

  • This system should be able to do almost the same as the MMX system above, it is just slightly less powerful. This board has by far the best HDD detection I have dealt with on older motherboards. It could even work with a 60GB SATA SSD on an IDE to SATA converter, just boot from floppy, format and install, that's it. So, no storage system issues are expected with this one. But it only has 256KB onboard cache (that might not be too bad since I'm not going to play really demanding games on it) and it can only cache 64MB RAM (there is no upgrade possibility like the Asus motherboard with its TAG SRAM socket provides). I'd like to have more than 64MB RAM for the Windows 95 partition, especially since my lucky find a while ago.

So, what do I have here currently? Of course, trying to get an "all-in-one" machine is hard if not impossible in almost all cases. In this case, the intended use is pretty narrow, and so I gave it a shot. Currently, I'd say the goal is still reachable. There are many obstacles, but overall, it seems solvable.

  • Pentium Pro & VS440FX: I really love this system, but it can't be decelerated to the correct level. Performance either over- or undershoots the target; the middle ground seemingly can't be reached.

    Pentium MMX & P/I-XP55T2P4: A great system, too, with some nice aspects like the 512KB onboard cache and the TAG SRAM upgrade for up 512MB cacheable RAM. Performance can be scaled correctly. I will have to deal with the storage system issues though.

    Pentium & Advanced/ML: This is a very reliable and easy to handle system. Performance can be scaled correctly, comparable to the MMX system. Downsides are the halved cache and the cacheable RAM being limited to 64MB; no cache or TAG SRAM upgrade is possible.

My next steps will focus on eliminating the storage system issues of the MMX system. If that'll work out and the system runs as expected, this will most likely be the new Retro 1.

END OF LINE.

Reply 29 of 134, by Andy1979

User metadata
Rank Newbie
Rank
Newbie

A dilemma to be sure, a shame that the P-Pro system won't behave as it's the most interesting combination of the three.

I always thought that the 430HX could cache >64mb of RAM. Is this a limit of the Advanced/ML board in this case (I see that the manual refers to first 64mb) rather than the chipset?

My Retro systems:
1. Pentium 200, 64mb EDO RAM, Matrox Millennium 2mb, 3DFX Voodoo 4mb, DOS6.22 / Win95 / Win98SE
2. Compaq Armada M700 laptop, PIII-450, Win98SE
3. Core2Duo E6600, ATI Radeon 4850, Win XP

Reply 30 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member
Andy1979 wrote:

I always thought that the 430HX could cache >64mb of RAM. Is this a limit of the Advanced/ML board in this case (I see that the manual refers to first 64mb) rather than the chipset?

The 430HX can, but this is dependent on the TAG SRAM on the motherboard. Usually, motherboards are equipped with a TAG SRAM that would cache 64MB RAM. "Better" motherboards would either have a slot for an additional cache module (COAST = cache on a stick; in case the onboard cache wouldn't already be maxed out) which usually contains an additional TAG SRAM or have a socket for an additional TAG SRAM (on motherboards like my Asus with already maxed out onboard cache). Unfortunately, the very reliable Advanced/ML has none...

END OF LINE.

Reply 31 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

While researching I rediscovered a vogons thread dealing with the question "best atx 430hx board?". The Asus motherboard I have is discussed at the beginning, and a certain post by fellow vogons member vetz made me stop and read again. He states that there is a modified BIOS for K6-III+ support (I don't care about that) which includes a hack for IDE HDD support up to 128GB. Maybe, this will give the motherboard a little friendlier attitude towards my IDE SSD. I'll give this modified BIOS a try as soon as I have some time for that (already downloaded). Very interesting indeed.

END OF LINE.

Reply 32 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member
voodoo5_6k wrote:

While researching I rediscovered a vogons thread dealing with the question "best atx 430hx board?". The Asus motherboard I have is discussed at the beginning, and a certain post by fellow vogons member vetz made me stop and read again. He states that there is a modified BIOS for K6-III+ support (I don't care about that) which includes a hack for IDE HDD support up to 128GB. Maybe, this will give the motherboard a little friendlier attitude towards my IDE SSD. I'll give this modified BIOS a try as soon as I have some time for that (already downloaded). Very interesting indeed.

I have just flashed the modified BIOS but still, the IDE SSD isn't working. Next step is testing the motherboard with either a Compact Flash card or a DOM.

END OF LINE.

Reply 33 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member
voodoo5_6k wrote:
Andy1979 wrote:

I always thought that the 430HX could cache >64mb of RAM. Is this a limit of the Advanced/ML board in this case (I see that the manual refers to first 64mb) rather than the chipset?

The 430HX can, but this is dependent on the TAG SRAM on the motherboard. Usually, motherboards are equipped with a TAG SRAM that would cache 64MB RAM. "Better" motherboards would either have a slot for an additional cache module (COAST = cache on a stick; in case the onboard cache wouldn't already be maxed out) which usually contains an additional TAG SRAM or have a socket for an additional TAG SRAM (on motherboards like my Asus with already maxed out onboard cache). Unfortunately, the very reliable Advanced/ML has none...

Here is a picture of my Asus P/I-XP55T2P4 showing the chip I have used as the TAG SRAM upgrade and another picture showing the verification that the cacheable area is large enough to address all the RAM.

"JP15" has been set for a cacheable area of up to 512MB, so more RAM could be used. "JP14" is permanently set to 512KB onboard cache as this board came with 512KB onboard. There were manufacturing options for 0KB and 256KB. These would include a real jumper for "JP14" and the slot for a COAST module (this is of course missing on mine, it would have been placed next to the TAG SRAM socket).

  • retro1_test_tag_sram.JPG
    Filename
    retro1_test_tag_sram.JPG
    File size
    2.19 MiB
    Views
    2237 views
    File comment
    TAG SRAM
    File license
    Fair use/fair dealing exception
    retro1_test_cacheable_area.JPG
    Filename
    retro1_test_cacheable_area.JPG
    File size
    2.26 MiB
    Views
    2237 views
    File comment
    Cacheable Area Verification
    File license
    Fair use/fair dealing exception

END OF LINE.

Reply 34 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member
voodoo5_6k wrote:

My next steps will focus on eliminating the storage system issues of the MMX system. If that'll work out and the system runs as expected, this will most likely be the new Retro 1.

I had some time to work on Retro 1 yesterday. After removing the IDE SSD from the system I installed a 4GB DOM (Transcend TS4GPTM520) together with an IDE 40 pin to IDE 40 pin adapter. This adapter is needed to connect the DOM to a ribbon cable. Per design, the DOM is supposed to be plugged directly onto the IDE header on the motherboard. Therefore, a special adapter is needed to map the pin-out of the ribbon cable connector to the DOM connector. I want to use the ribbon cable because if the DOM solution works I'll be needing more than just one to address the storage needs I have.

With the DOM connected, the system went through the POST without any issues and the DOM got detected (with the latest BIOS 0207). I booted with the PartitionMagic rescue floppies and partitioned the DOM, set the active flags and rebooted with the MS-DOS boot disk. The installation of DOS completed without any issues, and subsequently, the system could boot from the DOM to the DOS command prompt. Yay! 😎

As I now have a working storage solution, I can move on with setting up the system again to go back to testing. The plan is to use two 4GB DOM for storage and a smaller one (512MB) for the Win95 swap file. This way, the constant writes for the swap file will not exhaust the available write cycles of the DOM used for OS & data. When the swap file DOM dies some day it can be easily replaced, all that's lost then is the swap file (and of course it is a lot cheaper than a 4GB DOM). The optical drive will be jumpered as slave on the second IDE channel, the three DOM will then take the remaining positions. As PIO 4 is all the integrated IDE controller will do, there is no performance loss expected from sharing an IDE channel between the optical drive and a DOM because the optical drive is also capable of running in PIO 4 mode.

The partitioning will be something like this:

  • DOM #1 (4GB, IDE 1 Master):
    • 1x 512MB FAT16 for MS-DOS 6.22 + Add-Ons (drivers & tools)
      1x 2GB FAT16 for the DOS adventure games, everything copied to disk
      1x remaining space (1.xGB) FAT32 for Windows 95c + Add-Ons (drivers & tools)
  • DOM #2 (4GB. IDE 1 Slave):
    • 1x all space FAT32 for Windows adventure games, everything copied to disk
  • DOM #3 (512MB, IDE 2 Master):
    • 1x all space FAT32 for Windows 95c swap file

I am hoping to not find any more issues with speed sensitive games. But if I recall everything correctly, Sam & Max was about the only game that had issues when I set up DOSBox on one of my non-retro PC. And with the Pentium MMX, disabling the L1 cache worked for that.

Next step, hopefully completed over the weekend, is completing the hardware setup (connect all DOM and redo the cabling, take a few pictures of the system) and then complete the DOS installation, reinstall Windows 95, reinstall BootMagic, reinstall all games and perhaps start testing (and take even more pictures while doing that). Let's see how much of that I can get done 😀

END OF LINE.

Reply 35 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

So, this is the current iteration of Retro 1. Since the Pentium Pro setup can't be decelerated to a performance level that would allow for all DOS adventure games to run, the Pentium MMX was setup for further evaluation and as of now replaces the Pentium Pro.

Retro 1:

System specifications:

  • CPU: Intel Pentium MMX 233MHz
    Cooler: StarTech Socket 370 cooler (mounted with Socket 7 clamp) + Noctua NF-A6x25 FLX
    Mainboard: Asus P/I-XP55T2P4 Rev. 3.00 (BIOS 0207) + TAG SRAM Upgrade (Cacheable Area: up to 512MB RAM)
    RAM: Kingston 128MB EDO 60ns (4x 32MB)
    PSU: Enermax EG301AX-VE(G) 303W
    Case: Compucase CI-7106W ATX

Add-in cards:

  • PCI1: 3dfx Voodoo5 5500 PCI 64MB + 2x Noctua NF-A4x20 FLX
    PCI4: Intel PRO/100+ Management Adapter
    ISA2: Roland MPU-401AT
    ISA4: Creative Labs Sound Blaster AWE64 Gold

Drives:

  • FDD: Alps Floppy Drive
    DOM: 2x Transcend TS4GPTM520 4GB, 1x Transcend TS512MPTM520 512MB
    ODD: Plextor PX-740A

I/O Devices (via KVM):

  • Display: Eizo FlexScan F520
    Keyboard: Microsoft Natural Keyboard Pro
    Mouse: Logitech S96
    Speaker: Canton Plus Media 3

MIDI Devices:

  • Roland Edirol UM-550 (5 in/5 out MIDI patcher, provides access to the CM-32L and SC-55 for currently three systems)
    Roland CM-32L
    Roland SC-55

Pictures:

  • retro1_mb.JPG
    Filename
    retro1_mb.JPG
    File size
    2.26 MiB
    Views
    2181 views
    File comment
    System Overview
    File license
    Fair use/fair dealing exception
    retro1_cpu+ram.JPG
    Filename
    retro1_cpu+ram.JPG
    File size
    2.16 MiB
    Views
    2181 views
    File comment
    Pentium MMX 233MHz + 4x 32GB Kingston EDO
    File license
    Fair use/fair dealing exception
    retro1_vga.JPG
    Filename
    retro1_vga.JPG
    File size
    2.17 MiB
    Views
    2181 views
    File comment
    Voodoo5 5500 PCI
    File license
    Fair use/fair dealing exception
    retro1_mpu-401.JPG
    Filename
    retro1_mpu-401.JPG
    File size
    2.08 MiB
    Views
    2181 views
    File comment
    MPU-401AT
    File license
    Fair use/fair dealing exception
    retro1_sb.JPG
    Filename
    retro1_sb.JPG
    File size
    2.14 MiB
    Views
    2181 views
    File comment
    Sound Blaster AWE64 Gold
    File license
    Fair use/fair dealing exception

END OF LINE.

Reply 36 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

...Continued from above. More pictures:

  • retro1_dom.JPG
    Filename
    retro1_dom.JPG
    File size
    2.43 MiB
    Views
    2180 views
    File comment
    3x Transcend DOM
    File license
    Fair use/fair dealing exception
    retro1_psu.JPG
    Filename
    retro1_psu.JPG
    File size
    2.01 MiB
    Views
    2180 views
    File comment
    Enermax PSU
    File license
    Fair use/fair dealing exception
    retro1_fdd+odd.JPG
    Filename
    retro1_fdd+odd.JPG
    File size
    2.21 MiB
    Views
    2180 views
    File comment
    Alps Floppy & Plextor DVD
    File license
    Fair use/fair dealing exception
    retro1_detail.JPG
    Filename
    retro1_detail.JPG
    File size
    2.28 MiB
    Views
    2180 views
    File comment
    Case badge
    File license
    Fair use/fair dealing exception

And the MIDI rack...

  • retro1_midi.JPG
    Filename
    retro1_midi.JPG
    File size
    2.08 MiB
    Views
    2180 views
    File comment
    MIDI rack
    File license
    Fair use/fair dealing exception

Edit:
- Corrected formatting
- Task list removed, hardware configuration not working according to requirements

Last edited by voodoo5_6k on 2017-11-20, 08:34. Edited 2 times in total.

END OF LINE.

Reply 37 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member
voodoo5_6k wrote:

Next step, hopefully completed over the weekend, is completing the hardware setup (connect all DOM and redo the cabling, take a few pictures of the system) and then complete the DOS installation, reinstall Windows 95, reinstall BootMagic, reinstall all games and perhaps start testing (and take even more pictures while doing that). Let's see how much of that I can get done 😀

This time, I have some good news and some bad news, as always 😉 Hardware setup is completed (as detailed and pictured in the two previous posts), DOS installation is completed, and some of the DOS adventure games are installed to test. This part of the testing has been started, but I ran into some new problems.

The hardware installation is now finalized, all cabling redone and the DOM installed. They are all detected during POST, and I have partitioned them according to plan using PartitionMagic.

  • retro1_test2_post.JPG
    Filename
    retro1_test2_post.JPG
    File size
    2.03 MiB
    Views
    2179 views
    File comment
    POST (DOM Detection)
    File license
    Fair use/fair dealing exception

I have completed the DOS installation and installed almost all of the DOS adventure games to have something to test. First stop was Sam & Max, and unsurprisingly it still worked with L1 cache deactivated. But now, Day Of The Tentacle started acting up. MIDI and special sound effects (CM-32L) were working, but no Sound Blaster sound effects, and no speech. Calling the game's executable without any switch brings on Sound Blaster music (working) but still no sound effects and no speech. I was already running out of time by then, but I briefly tested with the L1 cache disabled and enabled, calling "tentacle.exe" with the switches "r s", "r", "s", but so far, nothing could be gained. With some settings, I at least get garbled sound effects and speech. I would exclude some sort of resource issue as Sam & Max runs fine with music, sound effects and speech. Currently, I assume, I'll have to decrease the CPU clock rate a little since this game was running OK on the 200MHz Pentium Pro without any tweaks. The 233MHz MMX can't be too far off, I would think. This will be the focus of next weekend.

END OF LINE.

Reply 38 of 134, by voodoo5_6k

User metadata
Rank Member
Rank
Member

There was a change of plans for this weekend. I had to make a few changes to our network infrastructure, and that took most of my time. However, I tweaked Retro 1 a little to allow for easier testing.

It is always quite annoying to change jumpers in this system. I have to disconnect the laser printer that rests on top of it and remove it, pull the case away from the wall so I can reach the screws with a screwdriver, remove the top of the case and use tweezers to move the jumper caps and then, when done with testing, screw the top back on, push the case back to the wall, put the laser printer back and reconnect it. It is OK to do this once in a while, but when I have to do this over and over again, this is really unnerving.

So, I got back to my earlier plan to make the jumpers switchable from the outside of the case. I installed four SPDT switches, repurposing the unused fan position next to the CPU heat sink, and connected these switches to the respective jumpers on the motherboard. Now, I can control the jumperblocks for the system bus frequency and for the CPU internal multiplier without having to open or move the case. This really simplifies this part of the testing, changing the CPU clock rate is now a matter of seconds (toggling a few switches) rather than minutes (clearing and opening the case, moving jumper caps etc.).

  • retro1_test2_jumper.JPG
    Filename
    retro1_test2_jumper.JPG
    File size
    2.06 MiB
    Views
    2135 views
    File comment
    Wired jumpers
    File license
    Fair use/fair dealing exception
    retro1_test2_switch.JPG
    Filename
    retro1_test2_switch.JPG
    File size
    2.36 MiB
    Views
    2135 views
    File comment
    SPDT switches
    File license
    Fair use/fair dealing exception

END OF LINE.