VOGONS


First post, by ruthan

User metadata
Rank Oldbie
Rank
Oldbie

Hello,
maybe it could help someone help with building ultimate multi boot machine, as trying, i really dont like waste space with zillions machines and builds.. its not economical, its not ecological and hoarding is stupid, if you can bypass it..

Document: // For your is important only first tab - OS-CompaMatrixBestMultiBoard, less you could ignore its mainly about Windows 98 stuff and main Windows 98 info is already in first tab
https://docs.zoho.com/sheet/open/qu3o108dc7e9 … oard/ranges/C36

Last edited by ruthan on 2019-02-23, 18:24. Edited 2 times in total.

Im old goal oriented goatman, i care about facts and freedom, not about egos+prejudices. Hoarding=sickness. If you want respect, gain it by your behavior. I hate stupid SW limits, SW=virtual world, everything should be possible if you have enough raw HW.

Reply 1 of 58, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++

All in one really isn't possible, unless you want to use emulation for earlier DOS stuff.

All hail the Great Capacitor Brand Finder

Reply 2 of 58, by oeuvre

User metadata
Rank l33t
Rank
l33t

You could find a machine that is Hackintosh compatible, dual boot it with 10 and Linux on 3 or more separate drives, and run 9x/XP in VMs. For DOS, DOSBOX and/or 86box/PCem.

HP Z420 Workstation Intel Xeon E5-1620, 32GB, RADEON HD7850 2GB, SSD + HD, XP/7
ws90Ts2.gif

Reply 4 of 58, by ruthan

User metadata
Rank Oldbie
Rank
Oldbie

Its still better have few (2,3) machines than lots - reconnect them, have cables, more peripherals,monitors etc.

If check my table:
- On Core 2 Duo, you probably could run everything from Pure Dos to Windows 10, but its too slow.
- Z97/X99 boards are quite new - good enough for new games and you could run almost everything - only Pure Dos is not working (sound cards are not working), unless someone will rewrite Sound Blaster emulator driver for new boards, but even without we are pretty successful.

I like emulation too, but i native is usually better and on same quick new machine with Windows 7 or 10 64bit you could run almost every emulator, its good bonus of this project.

Im old goal oriented goatman, i care about facts and freedom, not about egos+prejudices. Hoarding=sickness. If you want respect, gain it by your behavior. I hate stupid SW limits, SW=virtual world, everything should be possible if you have enough raw HW.

Reply 5 of 58, by PARUS

User metadata
Rank Oldbie
Rank
Oldbie
gdjacobs wrote:

All in one really isn't possible, unless you want to use emulation for earlier DOS stuff.

It is not quite true. I have system on Core 2 with ISA bus. This system covers 1983-2009 easy without DOSBOX. A problem is my motherboard is very rare, almost like Roland LAPC-I, and very expensive (different sellers sold them from $200 to $2000 as I saw). Performance can be XT --- Core2. By default it supports only Pentium D and I spent several years for proper Core 2 support. Here is:
9hq
i865. AGP8X, PCI, ISA, SATA1, USB2.0, 4xCOM (if needed).

Reply 6 of 58, by gdjacobs

User metadata
Rank l33t++
Rank
l33t++

Core 2 with unlocked multiplier? What's the minimum multiplier? How does performance scale with multiplier/FSB/cache manipulation? What about drivers for Win98?

All hail the Great Capacitor Brand Finder

Reply 7 of 58, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
PARUS wrote:
It is not quite true. I have system on Core 2 with ISA bus. This system covers 1983-2009 easy without DOSBOX. A problem is my mo […]
Show full quote
gdjacobs wrote:

All in one really isn't possible, unless you want to use emulation for earlier DOS stuff.

It is not quite true. I have system on Core 2 with ISA bus. This system covers 1983-2009 easy without DOSBOX. A problem is my motherboard is very rare, almost like Roland LAPC-I, and very expensive (different sellers sold them from $200 to $2000 as I saw). Performance can be XT --- Core2. By default it supports only Pentium D and I spent several years for proper Core 2 support. Here is:
9hq
i865. AGP8X, PCI, ISA, SATA1, USB2.0, 4xCOM (if needed).

I passed this board up a while ago - preferring backplanes instead. Can you please post the lowest Speedsys benchmark for reference using a Core2?

Reply 8 of 58, by agent_x007

User metadata
Rank Oldbie
Rank
Oldbie

Lowest you can go on Core 2 is ~600MHz : LINK
6a4a9i.png
You need BIOS that can go to 100MHz real (400MHz effective) frequency.
All Core 2 class CPUs have x6 multiplier available (you have to disable C1E/SpeedStep for multiplier control to show up, at least that's how it works on ASRock board I used).

157143230295.png

Reply 9 of 58, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie

^^ Thanks for that. I've managed the same on an Asrock board using an unlocked X6800 Core 2. What I was specifically after was a Speedsys score to validate what Parus mentioned regarding this: "Performance can be XT".

Reply 10 of 58, by PARUS

User metadata
Rank Oldbie
Rank
Oldbie

Don't hurry please, you don't know something. Here PIAGP (Pentium 4 AGP/PCI/ISA) build for DOS/WIN98SE I tell about MSR tuning in DOS. This tool can work with Core 2 too. It can change multiplier alive (like in Windows) - registers 194h, 199h - and activate CPU throttling - register 19Ah. Later I'll find link (I'm not at home now) and make a topic about this. Also I'll tell you about tool for P-2/3/Core which can disable separately L2 and L1 (and enable of course) without entering BIOS setup.

So, if we do follow:
1) FSB 100-133-166 via BSEL mod or via BIOS if supported
2) multiplier 6-8 via DOS MSR Editor
3) CPU throttling 12,5% via DOS MSR Editor
4) L2+L1 both disabled via DOS Cache Control
5) throttle 4 -c -- no comments, 50% chipset throttling
- steps 2-5 in one batch file - and get about XT performance. For the first time you can tune parameters finely, get wished speed and save it to batch files named XT.BAT, 386DX.BAT, P2-400.BAT etc. At home I got 1,04 XT (CHECKIT) or higher or even lower.

Be a little patient, I'll find it soon and show in individual topic.

Reply 11 of 58, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie

^^ Ah, I see - you are using two types of CPU throttling, one directly manipulating CPU registers and the other, THROTTLE.EXE, using ACPI (which is chipset dependent, and I'm aware of this one).

I've managed to get XT speeds using THROTTLE.EXE previously. However, when I did some tests, the results aren't equivalent to 'direct' hardware slowdown. I did some tests with demoscene programs, and there were graphical glitches which weren't present with direct slowdown techniques. Also, at XT speeds, I did some screen scrolling tests which weren't completely smooth compared to direct slowdown techniques.

Reply 13 of 58, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
PARUS wrote:

Yes, sure. You ran command THROTTLE 7, it's not the same that I've shown.

You need to use THROTTLE 4, as mentioned in your point 5)?

THROTTLE. EXE is not equivalent to your typical hardware slowdown, as it has issues, as mentioned in my previous tests.

Reply 14 of 58, by PARUS

User metadata
Rank Oldbie
Rank
Oldbie

Yes, usually 4. I mean 50% of chipset throttling without cahe disabling via throttle.

Please forget about your previous tests where you have used only THROTTLE.EXE with cache disabling because a large rate of chipset throttling brings those issues.
THROTTLE must be execute with key less than 6-5! And "-c" is obligatory! All other operations are via MSR Editor and Cache Control. Now understood? You even didn't see results yet and criticize referring your previous tests.

Reply 15 of 58, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
PARUS wrote:

Yes, usually 4. I mean 50% of chipset throttling without cahe disabling via throttle.

Please forget about your previous tests where you have used only THROTTLE.EXE with cache disabling because a large rate of chipset throttling brings those issues.
THROTTLE must be execute with key less than 6-5! And "-c" is obligatory! All other operations are via MSR Editor and Cache Control. Now understood? You even didn't see results yet and criticize referring your previous tests.

I've used THROTTLE. EXE with 50% slowdown using VIA chipsets that have more granularity compared to Intel chipsets - 8 vs 16 steps, and tested in the XT speed range with a text scrolling program. I saw scrolling that wasn't smooth when compared to direct slowdown. The scrolling had a periodic judder. So I have tested without a large slowdown rate using THROTTLE.EXE.

BTW, I'm not here to criticise, just having a discussion with my experiences. And I look forward to testing this MSR editor in one of my rigs in the future. Thanks for bringing it to my attention.

Reply 17 of 58, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
PARUS wrote:

OK, thank you for more information. I assume you didn't use -c key, right?

I'm sure I tried messing with all available parameters, but as this was a couple of years ago, I can't be certain. I found better options, where I needed both CPU limited and 'GPU' limited slowdown options, not just CPU slowdown. For example, see Might and Magic 3 animations being too fast, even with the CPU slowed down.

Reply 19 of 58, by ruthan

User metadata
Rank Oldbie
Rank
Oldbie

Phil from YTl did some videos with testing with disabling i cache levels and its work too.. By the day there were lots of slowdown programs for Dos.. But usually is not needed, if game is not badly write to expect only present day HW speed, i would say that this problem is overrated..

Info why it is not working on modern boards:
Legacy emulation under DOS Creative/Ensoniq cards will not work on systems
that . .

a) do not route the PCI SERR# signal to the processor NMI
b) have a capacitor between SERR# and ground or between NMI and ground that
is too large.

Some background is required here. The Creative/Ensoniq PCI cards use the PCI
SERR# signal to indicate that someone has accessed a Legacy device register
(i.e., Sound Blaster, MPU-401, ). This SERR# signal must generate a
processor NMI before the I/O instruction completes so that the Legacy
emulation software can perform proper emulation of the trapped I/O event
before the processor executes the subsequent instructions.

In the event that the SERR# signal is not connected to the NMI input, the
software is never notified and cannot perform any Legacy device emulation.

In the event that there is a capacitor on the line that is too large, the
NMI does not propagate to the processor in time to emulate the I/O before
the next processor instruction(s) execute(s). If the I/O event is emulated
too late, it may have an adverse effect on the system as it may change the
state of processor register al at a time when the processor is not expecting
it.

They have a choice to not wire it to the CPU. They have a choice not to
wire SERR# to the southbridge. They have a choice not to enable SERR#
NMIs in the southbridge's configuration registers. They have a choice
not to provide an NMI handler in the BIOS.

They are all sorts of stupid things they can do break NMIs, and something
a simple as not wiring a pin can make the difference as to whether another
expensive layer on the motherboard is needed. Most motherboards don't
support two floppy drives anymore because it means having one less pin
to wire that way

Last edited by ruthan on 2018-05-31, 01:20. Edited 1 time in total.

Im old goal oriented goatman, i care about facts and freedom, not about egos+prejudices. Hoarding=sickness. If you want respect, gain it by your behavior. I hate stupid SW limits, SW=virtual world, everything should be possible if you have enough raw HW.