The prospect of testing a SL6BY (P3-S 1.4GHz Tualatin 1400/512/133/1.45) in the PL adapter is extremely interesting - so much so that I've just bought one! It was only £8, including free shipping. Imagine what it must have cost back in the day! It'll arrive during the week, and I'll be able to test it this Friday. On the PL adapter, there are some Bus jumpers, and I can set them to 133 FSB. On the mobo, I'll set its jumpers to 133 FSB, and that should be it. Thanks a lot for the extra info people!

8 quid 😜

I'd love to know what these chips cost back in the day...

EDIT: Wow that's interesting...

Looking at old articles it seems that the Tualatin was released with the Pentium 4 allready on the market 😕

It seems that Tualatin was allways meant to ship with 512k cache, but these chips where a threat to the Pentium 4s so Intel cut the cache in half and only left the full cache for the server models.

UPDATE:

Found this pricing slide:

Tualatin can be found in the mainstream 1 category (small green triangle) at $800 to $1000. The 1.4 being a server model and the top model it must have easily sold for over well over $1000.

Source: http://www.xbitlabs.com/articles/cpu/display/ … l-tualatin.html

Well there's no doubt that they crippled the P3 desktop platform. 815 can only support 512MB RAM while 440BX can do 1GB.

Some of the notebook chipsets like 830 can do more RAM though.

I suppose the Tualatin didn't really have a future because it couldn't scale in clock rate. Its FSB is also restrictive. Since P4 was really what they planned to use for the next few years, they pushed it hard.

A CPU takes several years to design. So I imagine that Core 2 was probably in an early phase around 2002 or so. Pentium M and Core 1 are Tualatin successors and so is Core 2. Pentium M was likely in the works when P4 was launched.

If the goal was for maxing out CPU-intensive DOS titles, then i think it'd be no worries at all to use a PCI Banshee/Voodoo3 on a P2B with a 133MHz FSB Tualatin-S?

Are there really such demanding DOS games? I though most games from that era used 3D video cards or ran under Windows?

When I had a Pentium 2, most games where allready made for Windows and DirectX. I believe it was DirectX6? I remember some Star Wars Game where you could bring down the Walker and Wing Commander Prophecy (although this might have been glide?).

Mau1wurf1977 wrote:

Are there really such demanding DOS games? I though most games from that era used 3D video cards or ran under Windows?

When I had a Pentium 2, most games where allready made for Windows and DirectX. I believe it was DirectX6? I remember some Star Wars Game where you could bring down the Walker and Wing Commander Prophecy (although this might have been glide?).

Well, only if you want to try something like playing SS1 in XGA/SXGA

Otherwise, i believe the "need" is mostly imaginary 😁

Introduction
I want to see if it's possible to slow down a fast P3 system, using the BIOS set up options, in order to successfully emulate the speed of a 386 DX mobo. The test results below show that my P3 system has some potential to do this, but the speed adjustment range doesn't appear to be flexible enough.

Testing hardware
I'm going to be using the same mobo as before, the Asus P2B, with a 1200 MHz 100 FSB socket 370 Celeron CPU, in the Powerleap adapter.

Test results
It's worth noting that Norton 8.1 sysinfo.exe states that a benchmark score of 35.9 equates to the speed of a 386 DX 33 machine. In my tests, the fastest I could manage was 25.7. That's not bad, I suppose.

Speedsys result
This Speedsys screen shot is taken from the last test featured in the table above, test number 5.

Excellent!

Which benchmark to trust? I have always used Norton SI and a 386DX 33 scores 35 points in that benchmark.

Maybe the other benchmark is more suitable for faster cpus?

Good comment regarding flexibility. I will soon get my Super Socket 7 parts and then go nuts on benchmarking. I believe it will allow me to cover a much wider range of systems and more important be able to emulate a fast 386 / slow 486 machine as well...

Mau1wurf1977 wrote:

Which benchmark to trust?

Indeed. I'm beginning to wonder how important the results are from Norton Utilities 7/8.0/8.1 Sysinfo.exe. I decided to simply run some old 386 era games. I chose Wing Commander, Ultima 6, and Test Drive 3. I could get them all running fine, by just adjusting the "performance settings" (such as memory timings) inside the BIOS set up area. And that was without any FSB over clocking.

It's worth noting that games from the era when the 486 became popular (approx 1992+) don't tend to suffer from speed problems. Therefore, trying to emulate a 486 is less important than a 386. 486 games, providing they can tolerate the Pentium's architecture, can be run on an old Pentium.

Yup that sounds about right...

What I like about the idea of going with the Super Socket 7 platform is that you are slowing down a 200 or 300 MHz chip rather than a 1 GHz. I wish that stuff would have arrived before the weekend, I could have gotten a lot of testing done.

I got 2 Voodoo 2 cards underway, so I think they will end up in a Slot 1 machine and the Super Socket 7 machine will be for my DOS games and ISA Sound Blaster fun. Having said that I will get a real 486 board at some stage as well.

For my benchmarking session I am planning on using Norton SI 8.0, that russian benchmark and also 3D Bench. Not sure how to present it all, some sort of table I guess...

For benchmarking ideas, there is the fantastic mdgx website. Its "speed information" page can be found below. Lots of links, lots of ideas, loads of fun! 😀

http://www.mdgx.com/speed.htm

Wow information overload 😀

Coming back to the benchmark scores. In the manual, Wing Commander 1 actually had system recommendations of a 386 16 MHz machine. I guess it just shows how fast a 386DX 40 MHz actually is (Norton SI score of 43.

That speed seems impossible to hit with a Pentium 3 system.

Hmm we need to look for games that run ideal on a 40 MHz 386 I guess. Wing Commader seems more suitable for an average 386 machine. It runs way to fast on a 40 MHz 386...

Any ideas? What about Privateer? It came out in 1993.

I'm beginning to come to the following conclusion - when you down clock a fast P3, it is prudent not to pay very close attention to what ancient benchmarking scores are telling you. For instance, if you down clock a ~1+ GHz P3, and then run Norton's sysinfo, and it tells you the system is running at "x" speed, don't make this information the only information you base your decisions on.

IMHO, a more successful and practical approach is to run various old speed sensitive games, and then decide for yourself "does it look and feel right?". Games publishers have to decide on what to write on the back of their game boxes, for the system hardware recommended requirements section, and this information is never 100% accurate. There's plenty of room for "plus" and "minus", in the speed range department.

From my testing today with a down clocked P3, the speed range appears to be adjustable between some kind of a mid era 386, up to some kind of rather slow 486. The only negative issue is that there is only a small number of speed adjustment increments you can make. What's really interesting is that I remember adjusting the speed of various 486 mobos using their BIOS settings, in an attempt to down clock them so they ran at 386 speeds, and thinking the same thing: that the number of speed adjustment increments was small.

I guess a solution is to either get a 386, or use DOSBox, or be content with a down clocked 486 or Pentium, with the limited number of down clocking adjustment options available to you. Also, I think it's worth reminding yourself of this question: "How many ancient speed sensitive 386 era games have I actually got, and want to play, but not on DOSBox, and must be extremely close to the best possible speed?" 😉

A slow 486 (lets say a 20 MHz 486SX) scores over 40 points in the Norton Sysinfo test. (A 25MHz 486 scores 54 points). And on such a machine Wing Commander is unplayably fast.

This speed is simply not attainable on a P3 with cache disabled.

So yes I do trust these scores. You could also run 3dbench, as this is closer to running a real game!

A shame games from those days don't come with builtin benchmarks. I only know of Battle Isle which came with some kind of benchmark.

EDIT: Also don't forget that the AGP cards we use are racehorses under DOS. I doubt that even a ISA ET4000 would come close to what a AGP TNT can do under DOS.

OK, please look at the speedsys screen shot about 8 posts above. It's from an over clocked P3 system. The speed score is 11.11. Now read this thread -

Testing 3 Intel 486 CPUs

Do you see the DX/33 and SX/25 scores? They are 12.43 and 9.12, respectively. The P3 11.11 score sits very roughly in between these two 486 CPU benchmarking scores. The over clocked P3 system does feel fast. Wing Commander plays too fast on it! The P3 system is running at approximately 1.6 GHz with an FSB of 133, and that's quite a powerful system. With its BIOS cache and memory timing options tweaked, I think it's powerful enough to emulate a very slow 486. (But it's worth mentioning that the speed sensitive games tend to be 386 based, and so therefore we are less concerned about emulating the faster speed of a 486.)

If you tweak the mobo's settings in another way, by running the FSB at its normal setting of 100 FSB, and also adjusting the BIOS memory timings, you can make Wing Commander look and feel too slow. And if you mess about with the BIOS timings by setting them to "performance" settings, but keep the FSB at 100, you can make Wing Commander look and feel just about right.

I've not tried this, but the FSB can be adjusted in various increments between 66 and 133. This can give you some more "speed adjustment" options.

Thanks heaps for that link!

There is so much good info in past threads, quite impressive.

Ok here is my conclusion. Speed System results are unreliable for slow machines!

Why?

Because Norton SI 8.0 AND Superscape 3Dbench make sense but Speed System result don't match.

A 486SX-25 MHz scores:
Norton SI 8.0: 54 in (This is from PC Guide).
3Dbench: 20.4 (Your test)
Speed System: 9.12 (Your test)

In my PIII thread (YouTube Video: Wing Commander on a P III with L1 Cache disabled) I have a PIII 1 GHz with L1 cache enabled and I get:

Norton SI 8.0: 19 (My test)
3DBench: 11 (My test)
SpeedSystem: 8.5 (My test)

So as you can see between the slowed down PIII system and your 486SX-25 the scores in Norton SI 8.0 and 3D Bench increase drastically. Wheras the Speed System score is almost the same.
What do you think? From my point of view I will go with Norton SI 8.9 and 3D Bench for my upcoming benchmarks.

If you do a run of Wing Commander on your 486SX-25 Machine, you will see should see a confirmation. It should run ~ twice as fast...

Your 1600 MHz PIII score of 25 in Norton SI 8.0 is the same result as a 386DX-25 would give you. Add a fast AGP video card and you have a machine that is very fast for Wing Commander. Remember a 386 16 MHz is recommended for this game (according to the manual). To me it seems that PIII with Cache disabled can reach 386 speed at max. setting and can go as slow as 286 and XT machines. So a PIII is good for emulating XT, AT, 286 and 386.

With Super Socket 7 I am hoping to cover 286, 386, 486 and higher with 1 system and 1 CPU. So goal is a very wide range of targets.

I can't judge the speed of the game on your machine. But for my machine I made a video on YouTube. That system scores 19 in Norton SI 8.0 and Wing Commander still feels a bit slow.

Also my conclusion that a PIII with L1 cache disabled is not the ideal choice doesn't include overclocking. 1.6 GHz is quite extreme and definitly not something I would write a guide for others to build. I really believe that going with a Super Socket 7 board is much more flexible,

Finally don't forget the video card. If a 386 and a slow down PIII score the same score, Wing Commander should still be a bit faster on the PIII because the AGP video card is just a lot faster and this has a great impact under DOS right? I remember seing benchmarks of VLB ET4000 and they where off the charts compared to ISA.

What 3DBench score do you get on your PIII machine? That would be very interesting...

Hang tight, proof will come soon. Incl. Videos and lot's of Data. Just be sit tight, dock onto a space station and enjoy the show 😀

I think that if we can get any faster machine to act like a 386, we've achieved our objective. 😀 If I can get my P3 to act like a 386, and you can get your SS7 machine to act like a 386, then we've both succeeded. Luckily we don't need the speed of a 486, because generally speaking, 486 era games can be run on any old Pentium. They can tolerate the Pentium architecture, and they're not adversely speed sensitive.

BTW, I have a box of 486 and SS7 mobos. I was just concerned about the many folks who don't want lots of different retro systems. It seems that most people want one box that they can call their retro system. If you can "hod rod" a P3, and have it handle all DOS, Win 3, and Win 9x, then that's a useful thing to do. The "all DOS" part can be achieved with a blend of hardware tweaking, and DOSBox.

Regarding the 3DBench test, I only get 14.7. That's with an era appropriate S3 based PCI video card. But I'm happy with that. 😀 Regarding the P3 over clocking, if I fry the Celeron CPU, ebay has plenty more at very cheap prices.

14.7? Very very interesting...

It appears that Norton SI is a good benchmark for testing CPU performance. And 3D Bench is a good benchmark for gaming performance. It appears that in games, your machine is almost 20% faster than my 1GHz PIII.

Yes the objective ob emulating a 386 has been achieved but I believe we can do even better. Results will come soon, we have a holiday today, so no packages, but hopefully tomorrow I can start building and testing.

More performance never hurts. E.g. Doom, Strike Commander and Comanche all prefer 486 performance. So I am expanding my objectives and the new goal is to build machine that can emulate 286, 386 and 486.

EDIT: Also I might get a cheacp 15" LCD soon, so I can finally work without having to move my widescreen back and forth all the time. Can't wait...

swaaye wrote:

Well there's no doubt that they crippled the P3 desktop platform. 815 can only support 512MB RAM while 440BX can do 1GB. […]
Show full quote

Well there's no doubt that they crippled the P3 desktop platform. 815 can only support 512MB RAM while 440BX can do 1GB.

Some of the notebook chipsets like 830 can do more RAM though.

I suppose the Tualatin didn't really have a future because it couldn't scale in clock rate. Its FSB is also restrictive. Since P4 was really what they planned to use for the next few years, they pushed it hard.

A CPU takes several years to design. So I imagine that Core 2 was probably in an early phase around 2002 or so. Pentium M and Core 1 are Tualatin successors and so is Core 2. Pentium M was likely in the works when P4 was launched.

Couldn't scale? The Tualatin is the basis for the Pentium M and Core architectures! There was plenty of room for the PIII to grow yet but Intel wanted to get the P4 out. Now THERE was a chip that had trouble scaling later in it's life. The early P4's were even slower than the PIII when the clocks were equal. They went back to the PIII for the Pentium M and Core architectures when the P4 hit the wall.

sliderider wrote:

Couldn't scale? The Tualatin is the basis for the Pentium M and Core architectures! There was plenty of room for the PIII to grow yet but Intel wanted to get the P4 out. Now THERE was a chip that had trouble scaling later in it's life. The early P4's were even slower than the PIII when the clocks were equal. They went back to the PIII for the Pentium M and Core architectures when the P4 hit the wall.

Pentium M and Core 1 aren't all that exciting compared to Athlon 64 until you put them into a notebook and notice their low power usage. They can't keep up in clock speed because they weren't designed for it. They made compromises for power. They also weren't around when the Athlon started to beat up the P3. So Intel went with the P4. Northwood isn't a bad chip by any measure.

Tualatin can't beat a 1.7 GHz Willamette in much, let alone an Athlon XP, so I think they made the right decision to not put the future of the company on Tualatin. Pentium M didn't show up until 2003.

I don't think Intel realized that semiconductor physics would doom Pentium 4 at the end of its run. But it wasn't a useless experience because Core 2 does have elements that are clearly related to things learned from P4. Core 2 is really the best of both worlds.