VOGONS


First post, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

I have recently tracked down a Socket 4 motherboard and Pentium 60. I decided to put it through the paces and compare it to a much more common system of the era - a 486.

I've always been a bit curious about the original Pentium as released in 1993 - it was a bit of a strange beast at the time. The processors were expensive and ran very hot, requiring a noisy fan; the motherboards were difficult to manufacture since they used a 60MHz/66MHz bus (most 486's used 33MHz at the time), and it had a funny name. Socket 4 didn't have a very long run in the marketplace either; Socket 5 came out about a year or so later which was when it started becoming a more popular system for home users. Socket 5 ran the processors at 3.3V instead of 5V so they were much cooler, and also at higher clockspeeds. The end result was that Socket 4 systems were very expensive and quite rare.

Here are a couple of pictures showing, from left to right, a Socket 7 Pentium 233MMX; a Socket 5 Pentium 90; and a Socket 4 Pentium 60.

IMG_0571t.jpg
click for big

IMG_0576t.jpg
click for big

Here you can see the physically larger size of the socket 4 processor, and its standard pin grid array, compared to the staggered pin grid array of the newer processors.

Note the integrated heatsink and fan on the MMX - I believe 'boxed' processors came with these, as opposed to processors supplied on a tray to OEMs who may want to use their own cooling solution. There were also some Pentiums in the standard flat ceramic package.
Socket 5 and 7 were very similar- Socket 7 had one extra pin and the motherboards were backwards compatible with Socket 5 processors - but socket 7 processors had special voltage requirements and generally wouldn't work in a Socket 5 board. Most people who owned a Pentium had a Socket 5 or 7 system.

On to the motherboard. This is an Acer V12P socket 4 board, with 256KB of L2 cache. It uses the Intel 430LX chipset, which was the first chipset for the original Pentiums. Seems to be about a mid 94 build date judging by the date of the IC's so this is probably a very late socket 4 board. Interestingly it uses a socketed crystal oscillator to set its bus speed, rather than jumpers like many 486 boards do. Currently it has a 60MHz crystal installed. If I had some spare crystals I could try overclocking it although I doubt I'd be able to squeeze much more out of it. I do have an 80MHz crystal in my 386DX40 but I seriously doubt I'd get 80MHz out of this Pentium, especially with no voltage adjustment.

IMG_0566t.jpg
click for big

It's a pretty standard looking board really. Being an Acer there aren't many options to tweak in the BIOS.

One of the interesting things about the early Pentium was the FDIV bug. You can read about it here but basically in certain circumstances, performing division operations on the floating point unit would cause incorrect results. In practice it was unlikely the majority of users would run into this bug but Intel eventually agreed to replace the affected processors on request, which cost them $500 million. Some comparisons can be drawn between the recent Sandy Bridge motherboard recall and the FDIV bug recall nearly 20 years ago 😀

Since processors with the FDIV bug are pretty rare nowadays, I was eager to discover if the CPU I had was an affected model. I downloaded a utility called PENTBUG.EXE which would check for the bug:

IMG_0569.JPG

As you can see this processor is a fixed model and doesn't have the bug. Kind of disappointing (ironically that's probably the opposite reaction most people who ran this utility would have 😜 ) Ignore the blue tinge, my monitor cable is a bit dodgy 😜 Interestingly, the Pentium 90 in the pictures above is S-Spec SX879 which supposedly has the FDIV bug; one day I might fire it up and confirm.

Anyway, I was curious to find how it performed. I fitted it with an S3 Virge DX PCI video card, 2x 16MB 60ns RAM and a Quantum Fireball 1280AT with DOS 6.22 and some benchmarks installed.
I have also benchmarked it against one of my other DOS boxes, a 486DX4 100MHz, on a UM881OP-AIO motherboard (UMC chipset). These systems were available at roughly the same time although the 486 CPU and motherboards were significantly cheaper. I used the same RAM, video card and hard disk on both systems.

Here are the results:

benchmark.PNG

Interestingly, for the average DOS game it appears the cheaper and more common 486DX4 is pretty much on par with the Pentium 60. 3Dbench actually favours the 486 by a significant amount.

PCPbench in SVGA mode gives the Pentium the advantage, but neither are particularly brilliant scores. I tried to play Tie Fighter in SVGA mode on the 486 some time ago and it was nearly unplayable; standard VGA mode was a much more enjoyable experience. I doubt the Pentium 60 would be significantly better in this regard- I would have liked to test this but without some sort of benchmarking tool it would be hard to quantify the difference, plus I haven't had time to do so 😒
The general consensus seems to be that you really need a much faster system, probably over 100MHz for SVGA mode in games to be viable.

PCPBench seems to be a much 'busier' benchmark than 3Dbench, with more polygons and textures etc, so I can see how the higher raw integer performance of the Pentium would be an advantage here, as would the faster FSB (60MHz vs 33MHz for the 486).

Norton Sysinfo's synthetic CPU benchmark shows the Pentium to have close to twice the raw performance of the 486 despite the 486 running at about 50% faster clockspeed.

Doom is the most interesting one to me - it shows that for the average DOS VGA game there is very little separating the two systems.

Later games that used the Quake engine would probably favour the Pentium I'd imagine, as it made heavy use of floating-point instructions- the Pentium has a much stronger floating point unit than the 486. Some say Quake's code was specifically optimized for the Pentium as well. If there is enough interest and I get time I may try to run a Quake benchmark on both systems, however keep in mind that Quake wasn't released until 1996 and I consider it part of a different era in computing. By 1996 Windows 95 was out, and Socket 7 Pentiums were available in clockspeeds over 100MHz.

The interesting thing to keep in mind here is that the 486 motherboard can be upgraded with an AMD 133MHz 5x86 CPU, while Socket 4 was very much a dead end (to upgrade to a Pentium 75 or faster you needed a new, socket 5 motherboard). The AMD 5x86 came out about a year later than the DX4 or Pentium 60 but was significantly faster than the 486DX4 - it was even marketed as being equal to a Pentium 75. Unfortunately I don't have such a CPU to benchmark with but I'm certain it would be a noticeably faster system than either of those benchmarked here.

So, overall, the Socket 4 Pentium isn't much more than a curiosity - you'd be better off with a 486 for VGA DOS games, or a later Pentium for SVGA or Win9x games. Even in 93/94 it seems you would have been better off sticking with a 486.

Reply 3 of 15, by swaaye

User metadata
Rank l33t++
Rank
l33t++

I'm interested in that original P5 hardware too.

The platform really needed to be refined. The early Pentium chipsets are quite unexciting with features not much different than 486 boards. They used an off the shelf IDE controller, often the bugged CMD640, and they were just PIO controllers so they are slow and hog the CPU. They also have plain direct mapped SRAM cache instead of the faster pipeline burst type and probably don't support EDO. So a Am5x86 160 is definitely going to give one of these a run for the money.

I always think of Wing Commander 3 and Need for Speed when these P60/66s come up.

Reply 4 of 15, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Socket 4 actually did have a couple of upgrade options. There was an official Overdrive from intel that ran at 120/133. There was an interesting socket4-->socket5 adapter from dell that was good up to 166MHz, and there were quite a few aftermarket adapters as well. There is a particularly good one that let you go all the way up to a K6-2 400.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 5 of 15, by DonutKing

User metadata
Rank Oldbie
Rank
Oldbie

How's Quake?

Well, unless I stop being so lazy and actually do my own benchmark, you can view some results here http://web.archive.org/web/19961223112037/www … gena/quake.html

Seems the DX4 scored about 7FPS while the P66 scored about 14, both at 320x200... but on that same page, an overclocked 160MHz 5x86 scores 15, a very strange result.
Personally, I waited until I got a Pentium 2 before I tried to play Quake, which ran it silky smooth, I think trying to play it at 15FPS is going to be an exercise in frustration.

Throw some more memory in, a 3dfx card and tell us how Half-Life runs.

Well this means I'll need to install Win95/98.... I do have a Voodoo Rush, and a pair of 12MB Voodoo 2's though so it might be an interesting test. Just a matter of finding the time to do it.

Further to the point about Quake, it might be interesting to install the Voodoo cards and benchmark GLQuake on both systems. I wonder if the 3d accelerator will have much of an effect on the 486 and help even the gap in performance.

They also have plain direct mapped SRAM cache instead of the faster pipeline burst type and probably don't support EDO.

Interestingly, this board has a jumper for 'burst SRAM' or 'Standard SRAM' but the board fails to boot if I set it to burst mode.
And you make a good point, in general the board just 'feels' more like a 486 than a Pentium board.

Socket 4 actually did have a couple of upgrade options. There was an official Overdrive from intel that ran at 120/133. There was an interesting socket4-->socket5 adapter from dell that was good up to 166MHz, and there were quite a few aftermarket adapters as well. There is a particularly good one that let you go all the way up to a K6-2 400.

Now that you mention it I do seem to remember something about socket 4 overdrives... I assume that being an Intel Overdrive, a niche product with no real alternative for people looking to upgrade, their retail price would have been fairly high?
I didn't know about those socket 4 -> 5 adapters though, that's interesting. I wouldn't mind getting my hands on one to see how this socket 4 board performs with a later processor in comparison to an actual socket 7 board.

Reply 6 of 15, by SquallStrife

User metadata
Rank l33t
Rank
l33t

I had a Skt4 Pentium 60 IBM rig at one stage, I used it almost exclusively for mIRC on Windows 3.11. It was too much of a dog for anything else I was interested in at the time. This would have been... like.. 2002? Maybe earlier?

The case was VERY nice to work in, but that's about all it had going for it.

VogonsDrivers.com | Link | News Thread

Reply 7 of 15, by Markk

User metadata
Rank Oldbie
Rank
Oldbie

I've kept some old greek pc magazines, and I found two that have some interesting articles, that are related to that post.

The first one, is from August 1993, when the fastest CPU available was the 486dx2/66. Pentium hadn't been released yet, but they came across an "experimental" model by Nixdorf. The interesting part was that it had the Pentium running at 40MHz. They measured the system, and it was as fast as a good dx2/66, nothing more. But what was really a lot faster, was it's FPU. They estimated that for a Pentium based system, one would have to wait a lot, since even when they would become commercially available, they are going to be so expensive, that the extra power would not be enough to compensate for their so high price. And manufacturers are going to have major problems, because of the nature of the chip(they have to solve some heating issues, and also there weren't available any fast enough subsystems - for example, memory controllers, etc.). And the second problem they would have, is that of the marketing. They wouldn't want to promote systems being 1.5 times faster than a 486, costing 3-4 times more. And they concluded that in any case, the 486 is going to live for long enough.

Now, let's go to the second one, which is from June 1994, when the 486dx4/100 first appeared. They tested 3 systems with that cpu. So they measured that although it is faster than a dx2/66, a lot of functions are in the same speed as the dx2/66. That is because of the memory bus running at 33MHz. But it's FPU was a lot faster, as it was integrated in the chip, and it was running at 100MHz. In general, in dos it was 20% faster than the dx2/66, but in windows 3.11, it was about 35% faster. Then they compare it with a Pentium 66. So they say that the pentium is slightly faster(2-3%), and that is something they were expecting. The pentium works at 66MHz inside and outside, but in fact that doesn't do much difference, as the memory runs slightly faster than 33MHz(In that point, I assume that it was still early for the pentium, and motherboards that used the chip up to its full potential, had yet to appear....) So with the dx4, running internal at 100MHz, it had an advantage over the pentium. They also say, of course the pentium is a lot more powerful than the dx4, and that's why it gets to be slightly faster even when it's running in a lower speed. But they need to see also the 90 and 100 pentiums that were announced, which have an external bus of 60/66MHz, in order to make a more proper comparison with the dx4. And they conclude, that the dx4 certainly will have a position in the market, but they don't know for how long, as Intel announced that is dropping the price of the 60MHz pentium to the level of a dx2 486.

Reply 8 of 15, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

Socket 4 is interesting indeed. I remember a friend of a friend (Lol?) had bought TEH pentium! And it was super...average in everything! Except in how much heat it produced.

It actually wasn't that hot (about as hot as a Pentium 233 MMX) but the cooling solutions weren't as advanced back then.
In those days you had SO many different ways of mounting a cooler, and the clip came later, and at first only for use with those larger passive Pentium heatsinks.

And I think the bugged Socket 4 Pentiums were already quite uncommon compared to the fixed ones, of the 5 Socket 4 CPU's I pulled myself back in the day, only 1 had the bug.

But the upgrade chips for Socket 4...now THOSE are impossible to find!

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 9 of 15, by sliderider

User metadata
Rank l33t++
Rank
l33t++
Anonymous Coward wrote:

Socket 4 actually did have a couple of upgrade options. There was an official Overdrive from intel that ran at 120/133. There was an interesting socket4-->socket5 adapter from dell that was good up to 166MHz, and there were quite a few aftermarket adapters as well. There is a particularly good one that let you go all the way up to a K6-2 400.

Are you sure that last one was for Socket 4 and not Socket 5? I know Powerleap made a Socket 5 adapter that could use a K6-2/K6-III chip. Socket 4 was out for a relatively short time before Socket 5/7 replaced it so I can't see it being too profitable to mass produce a socket interposer for it. Socket 4 came out around 1993 and Socket 7 came out in 1994 with Socket 5 coming out sometime in between so Socket 4 couldn't have been installed in that many motherboards.

Reply 10 of 15, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

Yep. I'm absolutely positive. One is called the Powerleap PL-54C/MMX. You can still find the user manual online.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 11 of 15, by Atom Ant

User metadata
Rank Newbie
Rank
Newbie

Pictures are missing from the first post. I think it is an informative thread from the good old Socket 4 and with pictures would be even more interesting.

My high end of '96 gaming machine;
Intel PR440FX - Pentium Pro 200MHz 512K, Matrox Millenium I 4MB, Creative 3D Blaster Voodoo II 12MB SLI, 128MB EDO RAM, Creative Sound Blaster AWE64 Gold, 4x Creative CD reader, Windows 95...

Reply 12 of 15, by dionb

User metadata
Rank l33t++
Rank
l33t++
swaaye wrote:

I'm interested in that original P5 hardware too.

The platform really needed to be refined. The early Pentium chipsets are quite unexciting with features not much different than 486 boards.

Intel chipsets (i.e. the i430LX) were unexciting as in dog slow and almost no features. But this was before the days of Intel chipset dominance and there was a lot of other stuff out there. The best performer (by far) was the SiS 501, which supported EDO RAM and had a (single channel) decent IDE controller integrated. But UMC, OPTi and VLSI were also in on the game. UMC had the first non-Intel chipset, probably also the worst: it was just a 486 chipset (including cache and memory controller) tacked onto half of the 64b processor bus of the Pentium. So its performance was awful - so bad it made the i430LX look good. OPTi and VLSI also messed around with 32b stuff on the Pentium bus, providing boards with VLB (and awful performance) for people who wanted to upgrade to a Pentium and keep their expensive VLB VGA or I/O cards.

So compared to today, where chipset is almost irrelevant, there was far more diversity and the difference really mattered in terms of performance and compatibility. Quite the opposite of "unexiting". But yes, unrefined and a lot of avenues explored that in retrospect never should have been.

I always think of Wing Commander 3 and Need for Speed when these P60/66s come up.

Heh, WC3 was the first game I bought after getting my P60 back in 1995 😉

Reply 13 of 15, by amadeus777999

User metadata
Rank Oldbie
Rank
Oldbie

The SiS501/2/3 does not seem to be that much faster than the i430LX chipset unless the board is weak in general. Since every notch in performance counts it's of course preferable choosing the optimum, and also quite rare, solution.

For example vogoner Robert's Asus PCI/I-P5MP3 P66 setup is based on the LX and delivered pretty good results, expecially in Doom (time)demo 3 where it accomplished 1440 rtics yielding approx. 51 fps(gfx card is an ET6000 though).
In comparison my, now donated, Scenic P66(LX) only made it to 1515 rtics approx. 49 fps... even though it "houses" an ET4000.
The ECS SI5PI only gave me 1514 rtics with everything fastest which also translates to approx. 49fps. I should test it with an ET6000 but unfortunately do not own one. The system ran with a Matrox MilleniumII, which speedwise is close but no cigar.
The gentlemen's club concerning Doom Pentium66 performance lingers around the 48fps mark.

In synthetic benchmarks the LX is of course slower.
Speedsys - MB/s
Asus(LX) - 124/60/39
ECS(SiS) - 126/74/45

Reply 14 of 15, by dionb

User metadata
Rank l33t++
Rank
l33t++

Interesting results, although as you say it's slightly comparing apples to oranges with different VGA card. I suspect the board vendor also makes a difference - even back then Asus was known for stability and good BIOS, ECS less so.

Thinking of it, my experiences were probably also coloured by that - the SiS board was an MSI MS-5109, a good retail board, wheras the i430LX was a Packard Bell PB520 motherboard, which was an OEM version of the Intel Premiere/PCI Low Profile. Rock-solid reliability, but every setting would have been conservative, with no options to tweak stuff yourself. Of course, this all happened around 2000, so unfortunately none of that hardware is available for any retesting...

Reply 15 of 15, by arncht

User metadata
Rank Oldbie
Rank
Oldbie

basically you should compare the p60 with a dx2-66 😀 the early vlb cards/chipsets were also much slower than the late models. the dx4 came with the p90.
another problem with the p60 - the pci cards are not really available before the p90, and with isa it cant be a too good performer.

so... i would build maybe an 1994q1 pc on p60 base with a vision864, but a couple of months later you could buy a p90 with the 430nx.

My little retro computer world
Overdoze of the demoscene