Pentium Socket 5 : This was UNEXPECTED!!

I take a look at an original Pentium Socket 5 motherboard with Intel FX Chipset and compare it with a 486 at the same frequency and a Socket 7 board.

I also take a look at the output of a Genoa SuperEGA card using an RGB to HDMI converter.

[ilembitov]

Opti Viper was by many accounts slower than FX even though it came out just a little bit earlier than 430VX/HX. It found some use in laptops and low-end IBM Aptivas, but otherwise it was the last nail into Opti's chipset business. I believe the biggest problems with Opti were memory bandwidth and PCI implementation.

To add insult to injury, your Intel Zappa board is one of the early 430FX boards that used 486-style async L2-cache — most 430FX boards that came after that actually had pipeline burst cache. Without PB cache 430FX effectively has performance of an earlier 430NX (Neptune) chipset. In Quake that would result in a difference of 5 FPS (about 15%) for the same amount of cache. A 430FX motherboard with 512KB PB cache would be about 20% faster in Quake compared to your Zappa.

I don't have a Viper motherboard, but I have Socket 5 boards using three Opti Pentium chipsets that came before it and competed against 430NX and 430FX: 571/572, Cobra and Python, all of them VLB-based, PCI is only available via a bridge chip. Cobra and Python could only match Zappa in most of my tests under these conditions:


- a lot of cache. My Python board supports up to 2MB of L2 cache which I think is not supported by any 430NX or 430FX board. My Cobra board supports up to 512KB — still double the amount Zappa has. In 1994–1995 I think 2MB of L2 cache would cost about $200 while Zappa itself would be around $250 :)


- VLB videocard and IDE controller — because all of my Socket 5 Opti boards have PCI implementation in a form of a 82C822 bridge chip. That chip is essentially a VLB device and its job is to try and pass PCI requests whenever it can, making PCI a second class citizen in the system and adding latency to PCI cards. Viper is the first and only Opti chipset to have a native PCI implementation, but I'm not sure how good it is.

For Quake at 320x200 I had 24.1 FPS with Zappa, 23.5 FPS with a Cobra board, 24.4 FPS with a Python board. PC Player at 320x200 gave me 25.7 FPS for Zappa, 25.5 for Cobra and 27 for Python. Pretty much the same dynamic was seen in Speedsys memory and cache tests: Cobra being a bit slower than 430FX, Python being a bit faster thanks to 4x the cache.

Note that I used Ark Logic 1000VL on VLB boards and 2000PV on Zappa for my tests — those are among the fastest DOS videocards, but unfortunately I don't have a PCI 1000PV card. 2000 uses a 64 bit chip, 1000 is 32 bit, which is probably why at 640x480 Zappa was faster in both Quake and PCP.

Bottom line: at the very least Opti Pentium chipsets had crappy memory architecture which you had to compensate with a lot of high-quality cache. Opti was a "budget" choice that didn't really save you money. In 1994 a Pentium 90 PC would cost at least $3000, so saving $50 on a motherboard and then buying $200 worth of cache and messing with BIOS settings certainly didn't make any sense at all. I suspect the same is true for your Viper board, so it would probably need a 512KB cache stick to match Zappa a bit closer. But it would still never match a full-featured 430FX board.

Sorry for the long rant, hope it could be somewhat useful :)

[Ice_Karma]

For the little it's worth, I'm Canadian, and I'm even fluent in French, but in spite of at least three decades of effort to retrain myself, I'll still say "caysh" if I'm not careful.

[Fredjoe]

The "PBA" means "Printed Board Assembly", a typical label found on Intel-manufactured boards of the era. It basically denotes a particular product type and is often found accompanied by another similar-type label with "AA" at the start, the AA standing for "Altered Assembly", which indicates that it's a particular configuration of the main PBA. So, for example, if you had a PBA which had 3 different variants, say no cache, 256k cache, and 512k cache, each one would have the same PBA number, but different AA numbers. The 3- numbers after the dash is the specific revision of that PBA or AA.

The yellow label in the upper left corner is the serial number for the specific board, and the letter at the start of that number denotes where the board was manufactured. If that number starts with a "K" or "U", it means it was manufactured in Ireland, although most of the boards I saw were K serial numbers. I don't remember the other codes very well these days 😆.

So, you've got a genuine Intel-manufactured board there.

[techdistractions]

We had a school lab back around 1996/7 that had a mix of pentium and 486 pcs. We’d get into physical altercations when it came to lunchtime quake sessions.. lol

[DarrenHughes-Hybrid]

People had a tendency to skip every other generations of PC's/CPU's because of fear and they were already invested in a PC. In other words, People did not invest in 8088's because they were new, but started to see their value, so they invested in the 80286 gen. Then the 80386's got ignored because people had just bought the 286's. Then when the 80486 rolled out, people were ready to invest in a new machine again. The same thing happened when the 1st Pentiums rolled out. It's true they were hot, but everyone already had a "fast" PC, so they waited for the 2nd gen Pentiums. Same thing happened with Windows. If you look back, it was every other version of Windows that was successful.

[JohnDoe-ipoq]

Back in the day, I found a discarded slim Pentium 1 machine, I upped the ram, installed a voodoo 2, and was able to carry it around to play multiplayer quake 3. Many games had CPU requirements, but running the game in glide always bypassed the checks.

[jfidel]

Messing around with Pentiums is always fun! Thanks for continuing to make interesting videos

[adg]

3:40 - Socket 5 CPUs had 2 multiplier selector pins, for socket 7 it was 3 of them - x2 was repurposed for x6 by AMD.

[drzeissler]

I startet at 8088 later V20 EuroPC, 386-DX33, Pentium 75. The P75 was a lot faster than a DX4/100 some of my friends had back in the day.

[molivil]

Thanks for another amazing video. Love the thorough research!

[thatlinuxguy]

I guess back then with the memory prices coast modules where cash on a stick.

[gamingthunder]

your experience with early PCI boards reflect my own. i bought a asus PCI board and expected huge gains but instead was extremely disappointed when my vesa local bus system smoked the PCI board and i returned it next day.

[JVHShack]

The Intel Pentium actually went to 233MHz on socket 7 boards, but I happen to have a 266MHz socket 7 Intel which is an OEM CPU for Compaq computers installed in a "mid-range" socket 7 motherboard.

[EternalxFrost]

What I would be really curious is to see a Socket 6 board in action. They are freaking rare, as it was a kind a transition platform between the socket 5, 7 and 8.

[olivierpericat]

It look like you have a very very early revision of the Intel Advanced/ZP (Zappa) motherboard with Triton chipset. The first revisions could only go to 100 MHz (90 Mhz officially supported). Later they pushed the support to 120, then 133 Mhz. The latest available BIOS is 1.00.11.BS0. You have the first BIOS (1.00.01.BS0). This motherboard was also distributed in the OEM channel such as Gateway 2000 (very very widely widespread).

[IkesVintageTech]

That looks like the Intel Zappa board, Gateway along with many other companies used it.

[TheRetroRaven]

When my Dad upgraded my gaming PC back in the day, from a 486DX4-100 to a Pentium 100, one of my friends also upgraded his machine soon after, but he got a Cyrix 5x86 clocked at 120.
We were both surprised how much faster the Pentium really was. Quake wasn't really playable at 100MHz, even in 320x200 ...
But even in games like Doom and Duke Nukem 3D, my P100 easily outpaced his Cyrix.
I was also surpised that my P100 overclocked to 133 and was still stabile as a rock. I didn't dare push it further.
The cooler actually had an issue later on, and the motherboard would play für elise :D

[Chriva]

This is my time to shine :D
That board is a intel Advanced/ZP (aka zappa). Got one myself and it's horrible. Can't get it to work reliably with usb addin cards no matter which os (tried w95 r2, w98, linux etc).
Another annoyance is its hdd drive support. Forget anything larger than 7-8 gigs :)

[ml.]

We are Pentium of Borg. Division is futile. You will be approximated.

[Thales_WH]

The 430FX Triton I (1995) is a typical chipset for the first generation of Socket 7 boards. It is possible to use it for Socket 5 as well, but if we want to research technologies from the time when the first 3V Pentium P54C (75-100MHz) were introduced to the market (1994), then it would be better to use the chipset from that time, i.e. 430NX (1994).