VOGONS


Reply 20 of 46, by Scali

User metadata
Rank l33t
Rank
l33t
jesolo wrote:

I recall you actually had to switch the PC off, chose your setting and then switch the PC back on (not that we ever did run the PC at 4.77 MHz).

Haha, that's bad!
I have two 8088 machines with switchable clock speed. Both can be switched on-the-fly either with a keyboard combination or a commandline utility (but no turbo button on the case).
One is a Philips P3105, which does 4.77 and 8 MHz. The other is a Commodore PC20-III, which does 4.77, 7.14 and 9.54 MHz.

http://scalibq.wordpress.com/just-keeping-it- … ro-programming/

Reply 21 of 46, by 5u3

User metadata
Rank Oldbie
Rank
Oldbie

I've got an Asus PVI-486SP3 board and there is an issue similar to the one the OP posted, only reversed: The turbo switch works with write-back CPUs, but not with write-through ones. Also, when I disable the L1-cache, the L2-cache gets disabled as well (i don't know if this is normal, a configuration error, or just a quirk with my board).

I think the main reason why newer 486 boards rely on waitstates instead of actually reducing the clock speed when turbo is off is because of the VLB/PCI buses, which won't work properly at low clocks (VLB: <16 MHz, PCI: <20 MHz).
On old ISA-only boards, a real 8 MHz deturbo was common, after VLB/PCI was introduced, this feature was only found on certain brand-name OEM machines.

Reply 22 of 46, by RacoonRider

User metadata
Rank Oldbie
Rank
Oldbie

My 486 rig (to the left) features Acer AP43 board. Its turbo function effectively halves FSB. It can be tweaked in BIOS (Button up = turbo ON or Button up = turbo OFF). The Turbo button is wired to turbo header, the keylock is wired to multiplier jumper and the two switches in 3.5" bay are wired to FSB jumpers. The overall result is freedom of picking FSB from 25MHz/2 to 50MHz at multipliers of x3 and x4. That's from 37.5MHz DX4 to 150-160MHz 5x86.

P.S. Turbo fucntion works with both WT abd WB CPUs

Reply 23 of 46, by shamino

User metadata
Rank l33t
Rank
l33t

A long, long time ago, I had an ultra cheap PC Chips manufactured 486 board with an Intel 486DX2-66 in it. The board had VLB.
I remember being curious about this, and experimented with the turbo switch in a simple game, where the speed change could be easily timed. The speed with turbo turned off was about 1/3.5 of normal speed. At the time, I assumed it was both halving the bus speed and also turning off the clock doubling, but that might not have been exactly true. It seemed to fit with the amount of speed change I was getting though.

As unstable as that crappy motherboard was, I don't remember the turbo switch ever causing any problems, but I rarely had any reason to use it other than curiosity.

Reply 24 of 46, by jesolo

User metadata
Rank l33t
Rank
l33t
shamino wrote:

As unstable as that crappy motherboard was, I don't remember the turbo switch ever causing any problems, but I rarely had any reason to use it other than curiosity.

I must admit, I also rarely use the turbo switch functionality on my 486 machine.
Most of the DOS games I play are from the late 80's (that's 1980's for the younger audience here) up to the time when DOS games were no longer around and, most actually do run fine on my 486DX4-100 (with the exception of a patch that needs to be applied first in certain cases).
However, there are occasions where a game might start up fine in "turbo" mode but, in game performance might be too fast to allow you to finish the game.
A couple examples:

  • The dog fight sequence in Space Quest 3 (at the end of the game) - without the PC in "de-turbo" mode, it's practically impossible to win.
    The original versions of Wing Commander 1 & 2 (although, I believe that patches have since been released to address the speed related problems).
    Martian Memorandum - runs fine on a 486DX-33 but, I've experienced hanging music notes on my 486DX4-100 (no problems in "de-turbo" mode).

So, quite a useful feature to have and a much safer method than to scratch around in your case trying to adjust jumper settings or, adjust BIOS settings which might inadvertently damage your PC.
Considering that the type of hardware from that era (circa 1995) are becoming more scarce (and more expensive by the day, if you can still find some), I'd rather try and take good care of it for as long as possible.

Reply 25 of 46, by jesolo

User metadata
Rank l33t
Rank
l33t

After my last post, I acquired an AMD 5x86-133 CPU and tested it on my Asus PVI-486SP3 motherboard.
With the motherboard's jumpers set in write back mode for the CPU, the turbo switch does work but, the slow down in CPU speed is negligible (so much that it still out performs a 486 DX4-100).

However, when I set the motherboard jumpers for the CPU in write thru mode, then the turbo switch functionality does actually work much better in the sense that it slows the CPU down to the equivalent speed of a 486DX-40.

Haven't tried this yet in my other 486 motherboards but, the turbo switch functionality does seem to work better with CPU's in write thru mode.

Reply 26 of 46, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie

^^ What kind of speed differences are you getting just by switching between Write-thru and Write-back mode, and without using the Turbo?

Depending on the CPU being used, this type of switching can almost behave like a Turbo switch. For example, I forced a K6-III+ CPU into Write-thru mode using a jumper on a motherboard, and this basically killed the performance by around half. That's quite a significant change in speed...

EDIT:
It looks like that k6-III+ can drop its performace even more dramatically by forcing it to use Write-thru mode. By running it at 480 MHz, and using Write-back, I get a Speedsys score of around 544, and this drops to around 118 using Write-thru! The higher the initial speed, the larger the drop...

Reply 27 of 46, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++
j^aws wrote:

It looks like that k6-III+ can drop its performace even more dramatically by forcing it to use Write-thru mode. By running it at 480 MHz, and using Write-back, I get a Speedsys score of around 544, and this drops to around 118 using Write-thru! The higher the initial speed, the larger the drop...

Could you please share how you do this, meaning what tool, what commands?

This could be of interest for "slow down" projects and helpful in getting old games to work a bit better 😀

YouTube, Facebook, Website

Reply 28 of 46, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
philscomputerlab wrote:

Could you please share how you do this, meaning what tool, what commands?

This could be of interest for "slow down" projects and helpful in getting old games to work a bit better 😀

I don't think I've come across a software tool that does this, but maybe one exists and is lost in time...?

Basically it's done in hardware, more specifically, there is a jumper on my Socket 7 board that switches between L1 cache as Write-back or Write-thru. And the K6-III+ is installed via an adapter, so I'm not using a Super Socket 7 board. I'm running the board with L3 cache (external) disabled, so that I can use 6x multi without issues. There aren't any BIOS options for Write-back or Write-thru modes - only the jumper on the board. Ironically, the board also has a Turbo Switch header that does nothing.

The K6-III+ can run between 50MHz to 480MHz using 2x25 to 6x80, and with both L1/L2 enabled using Write-thru, I'm getting between 23-544 scores in Speedsys (544 in WB). So that's around as low as a 486DX-33, and will go even slower by disabling L1/ L2 caches too.

I'm not sure how this is exactly working - maybe there are some undocumented registers that could be enabled to do this in software?

Reply 30 of 46, by PhilsComputerLab

User metadata
Rank l33t++
Rank
l33t++

Thanks for the answer j^aws.

There are a few tools for the K6, I might check it out. One of them has lots of flags, and I remember some affecting caching.

YouTube, Facebook, Website

Reply 31 of 46, by jesolo

User metadata
Rank l33t
Rank
l33t
j^aws wrote:

^^ What kind of speed differences are you getting just by switching between Write-thru and Write-back mode, and without using the Turbo?

EDIT:
It looks like that k6-III+ can drop its performace even more dramatically by forcing it to use Write-thru mode. By running it at 480 MHz, and using Write-back, I get a Speedsys score of around 544, and this drops to around 118 using Write-thru! The higher the initial speed, the larger the drop...

By using a jumper setting on my motherboard (not in the bios), I can also switch between write back and write thru mode for the L1 cache.
Interestingly, with my AMD 5x86 CPU, when I select write thru mode, the CPU speed is actually slower than my AMD 486 DX4-100 write thru CPU but, not as drastic as with your K6-III.
In write thru mode, the turbo switch does actually work much better and, in de-turbo mode, the CPU slows down to 486-25 speeds.
However, this is not what I was trying to achieve. I was hoping that the 5x86-133 would slow down more in write back mode to the similar speed of a 486DX-33.

Reply 32 of 46, by j^aws

User metadata
Rank Oldbie
Rank
Oldbie
philscomputerlab wrote:

Thanks for the answer j^aws.

There are a few tools for the K6, I might check it out. One of them has lots of flags, and I remember some affecting caching.

No probs, if you find the tool, then please keep us posted.

jesolo wrote:

However, this is not what I was trying to achieve. I was hoping that the 5x86-133 would slow down more in write back mode to the similar speed of a 486DX-33.

The L1 WT/WB header works realtime with my K6-III+, so I can connect the Turbo switch to it and basically get the same affect.

Have tried testing your jumper header for L1 WB/WT switching whilst the PC is still on? If this works, you can just rig another jumper switch to the L1 header and use it in conjunction with the Turbo Switch. Or with the right switch, you could rig one switch that works on both headers simultaneously...

Reply 33 of 46, by jesolo

User metadata
Rank l33t
Rank
l33t
j^aws wrote:

The L1 WT/WB header works realtime with my K6-III+, so I can connect the Turbo switch to it and basically get the same affect.

Have tried testing your jumper header for L1 WB/WT switching whilst the PC is still on? If this works, you can just rig another jumper switch to the L1 header and use it in conjunction with the Turbo Switch. Or with the right switch, you could rig one switch that works on both headers simultaneously...

Haven't tried it yet but, I'll see if I can figure out a way. On my motherboard, the write back and write thru settings both require separate jumpers to be shorted (it's not the same jumper that you short and "unshort").

Reply 35 of 46, by badmojo

User metadata
Rank l33t
Rank
l33t
jesolo wrote:

However, this is not what I was trying to achieve. I was hoping that the 5x86-133 would slow down more in write back mode to the similar speed of a 486DX-33.

Why a DX33 if you don't mind me asking? Do you have a specific game in mind?

Life? Don't talk to me about life.

Reply 36 of 46, by jesolo

User metadata
Rank l33t
Rank
l33t
badmojo wrote:

Why a DX33 if you don't mind me asking? Do you have a specific game in mind?

Some older DOS games don't perform well with those high speeds.
Examples are Space Quest III (the dog fight sequence at the end) and Martian Memorandum (music cuts off at certain sections or have hanging notes).

Doesn't necessarily have to be 486-33 speeds. Can be anything between 386-33 and 486-33 speeds as most of the older speed sensitive games I play runs fine at those CPU speeds.

Reply 38 of 46, by RaiderOfLostVoodoo

User metadata
Rank Member
Rank
Member

Hey folks,
first post from a long time lurker here.
About 2 1/2 year ago my buddy and I started building a few retro systems. Whenever we had questions, Google would usually send us straight to Vogons. This forum is so full of useful information. You're doing incredible work here and have already helped us a lot.

We don't really care about period accuracy. Performance and reliability are more important for us. As long as it's 100% compatible, we're fine with combining old and new hardware. For example we use flash (either SSD or CF) on our builds. Also we would never use a period accurate PSU. That's total madness.
If the build has ISA slots, we use Enermax EG365AX or other sizes of the series. They are reliable, safe, quiet and pretty efficent for a PSU that still has -5V. We both snatched up a few of them cheaply and still have spare ones, in case one might fail.
Flexibility is also an issue, because we both are limited on space. He less than I am. So owning 20 builds is out of question.

We both have a Super Socket 7 "Time Machine". We even got the DFI K6XV3+/66 for a very reasonable price (less than 50 bucks). Not because Phil often showcases it on Youtube, but because it's (afaik) the only board with UDMA66 and three ISA slots. More sound cards! We don't need to overclock to 700MHz, since we have multiple builds.
We didn't go for Voodoo3 3000 like Phil, but went for GeForce2 MX400 + Voodoo2 SLI instead for even more flexibility. I currently only have one V2. Can't find a sister for a reasonable price. So I'll probably just sell it and get two Diamond ones instead. My buddy still has his SLI setup from back in the day.
He's also a bit ahead with ISA sound cards, because he earns much more than I do and is often impatient. But they will come with time. For now I have a SB 32 PnP plus a Vibra with genuine OPL3 chip, which is not bad at all.

My buddy is currently working on a Voodoo1 build. I don't have space for that. The five (?) games that benefit from a V1 seem unimpressive.
The next build I'm working on, is the "Time Machine PLUS". A flexible 486 build that lets you travel further back in time than the common SS7 build. Right to the earliest DOS days. I did consider opening my own thread, but this thread has answered many of my questions. So I just resurrect it.

jesolo wrote on 2015-04-07, 18:55:

After my last post, I acquired an AMD 5x86-133 CPU and tested it on my Asus PVI-486SP3 motherboard.
With the motherboard's jumpers set in write back mode for the CPU, the turbo switch does work but, the slow down in CPU speed is negligible (so much that it still out performs a 486 DX4-100).

5u3 wrote on 2015-02-15, 23:51:

I think the main reason why newer 486 boards rely on waitstates instead of actually reducing the clock speed when turbo is off is because of the VLB/PCI buses, which won't work properly at low clocks (VLB: <16 MHz, PCI: <20 MHz).
On old ISA-only boards, a real 8 MHz deturbo was common, after VLB/PCI was introduced, this feature was only found on certain brand-name OEM machines.

At this point you've completely ruined my plans. DX4-120, Trio64, PCI SCSI controller. All pointless now. But better now then when the build is complete. So far I got everything extremely cheap and can easily sell it for more than what I paid.

HighTreason wrote on 2015-02-15, 01:28:

I have a 486 who's deturbo can drag the CPU down to about 1.89MHz...

This is extremely interesting. What board/chipset are you using?
In general I would be thankful for hints to boards which would be suitable for my build. I want to be able to downclock it to IBM PC or 286 level with the fastest processor that can still handle it.
I could just use a 386, but I want a 486 so I can use the sound cards in that system for more modern games without having to swap them to my SS7.

jesolo wrote on 2015-02-15, 18:50:

That is true but, can sometimes become a cumbersome process since it requires a user to either open the case to change jumper settings and/or access the BIOS.

j^aws wrote on 2015-02-15, 19:05:

Yes, definitely cumbersome with changing jumpers. However, I have plans to build an external jumper switch-box for all controllable jumpers (audio/ video/ motherboard) when everything is finalised, hopefully. Otherwise, I personally wouldn't bother to open the case to change jumpers.

I had exactly the same idea. But I want it as a DIP switch on the front. And only for the turbo settings.
I did wonder if there is a plug and play solution. But guess there isn't. So it's DIY time.
Did you finish it? I'm curious and would like to see the result.

Reply 39 of 46, by Stiletto

User metadata
Rank l33t++
Rank
l33t++
RaiderOfLostVoodoo wrote on 2021-11-04, 17:59:

This forum is so full of useful information. You're doing incredible work here and have already helped us a lot.

Speaking as a moderator and one of the founding members - thanks for the kind words!

RaiderOfLostVoodoo wrote on 2021-11-04, 17:59:
HighTreason wrote on 2015-02-15, 01:28:

I have a 486 who's deturbo can drag the CPU down to about 1.89MHz...

This is extremely interesting. What board/chipset are you using?

FYI, this guy got banned, so he won't be back to answer your questions - one of the problems with posting a reply in a discussion topic that has gone without any new replies for six years.

"I see a little silhouette-o of a man, Scaramouche, Scaramouche, will you
do the Fandango!" - Queen

Stiletto