It's all relative and proportional to the platform.

In my opinion Socket 7 was the most significant... personally I have an Asus T2P4 v2.1 that went from Pentium 75/90 all the way up to at least K62-400 (maybe further--I only went to 400 BITD). And certainly other socket 7 systems can go to faster K6-2 and K6-3 systems.

These PCI systems really spanned the DOS to NT era.

Runner up might be the 486 era where you had lowly sx25 up to 133MHz and even some Pentium-like chips (such as Intel PODP83 and cyrix offerings).

It's just how do you compare?

This is a bit out of left field, but technically the Motorola 68000 is in a socket that could be considered vintage. Admittedly, one could consider it to have been "retired" long before the Pistorm was conceived .

https://www.hackster.io/news/hands-on-with-th … ga-449ef0634f3e

Well, Amiga had insane upgrade path and you can loosely consider upgrade slot as "socket".

Ok going by the CPU series and first chip to last and just clock speed it is the 486 series. First was Intel SX-16 last was the AMD-133 for a ratio of 8.3. No other series got above about a 6 (P2-233 to p3-1400 is a 6, etc)
Of course that is only by clock speed not true processor power... just an observation...

keenerb wrote on 2022-05-09, 17:39:

I guess I'm really talking about any socket prior to multi-core CPU support. […]
Show full quote

I guess I'm really talking about any socket prior to multi-core CPU support.

Socket 7 was P75 - K6 233?
Super socket 7 was up to K6-3 550?
Slot A, Slot 1?
Something else?

I probably wouldn't consider a slotket in the Slot 1 speed category.

well technically socket 7 can host a k6-3+ 500mhz at 83x6 or 66x6 for 400mhz I think there was even the oddball socket 7 that could unofficially hit 95mhz, though good luck with that on a non ss7 chipset

There is also this 386 upgrade kits and the Transcomputer 486HPi - Am5x86-160 anyone?

Super socket 7 is not a different socket afaik.

kolderman wrote on 2022-05-11, 05:23:

Super socket 7 is not a different socket afaik.

yeah, honestly socket 5-7 have really similar and overlapping small changes

Most notable:
-The addition of dual plane voltage. This started in socket 7 but not all of them got it. by ss7 all of them had it.

-The addition of the various BF pins. there are 4 BF pins BF0 BF1, BF2, and Intel BF2 each with their own pin. the Intel BF2 was kinda forgot to history. Some motherboards got it but the only cpu to ever use it was the tillamook. which didn't work in most motherboards. instead, History marched forward with the AMD/Cyrix BF2 pin and Cyrix even went so far as to use the intel BF2 pin for another function.

-And then of course there is the fsb difference the SS7 was designed to run up to 100mhz stock

-Between socket 5 and 7 the added a key pin

-at some point the processors started using a inner row of pins also I am trying to recall which cpus had it though, I know the last cyrix MII 2.2v chips had them but not the 2.9v

-One nice thing though, while forwards compatibility was rife with issues(hence the various tweakers, see my signature), backwards compatibility was very good. you can run about any socket 5 or 7 cpu you want in a ss7

btw, just to get the word out there I'll include this: cyrix media gx and amd geode use a socket that physically is socket 7 but is absolutely not socket 7 electrically and those chips should never be installed in a regular motherboard or vice versa.

Con 2 botones wrote on 2022-05-10, 12:55:

Just considering the socket alone is a bit tricky to me.
Take socket A for instance, a board from the the KT133 chipset era would not let you go above Palomino.

my Asus A7V worked fine with Thoroughbred 2200+ @ 2GHz (20.0x100fsb). There were even people running Bartons on those.

rasz_pl wrote on 2022-05-11, 12:53:

Con 2 botones wrote on 2022-05-10, 12:55:

Just considering the socket alone is a bit tricky to me.
Take socket A for instance, a board from the the KT133 chipset era would not let you go above Palomino.

my Asus A7V worked fine with Thoroughbred 2200+ @ 2GHz (20.0x100fsb). There were even people running Bartons on those.

Same. When I first got one I only had a Barton to test it with. Ran without complaints (though because of the 100MHz FSB not at rated speed mind you!)

I had a non-IBM compatible 286-4mhz Unisys machine
286-25 came out near the end for a 6.25 ratio

If we count the strange Evergreen overdrives you could take the systems to a 486slc2-66

In my Amiga 1200 i can go from a 14Mhz 68020 to a 300Mhz of the Powerpc 603. Can probably the most wide increase in performance.

Core 2 Duo.
Ivy Bridge and Haswell
Pentium M
Athlon XP
Pentium and Pentium MMX.
Partially 486 especially with external cache.
286 because 386 is not much in performance but useful for 32 bits software and helped some with cache.
PIII had good bump but not as dramatic.

Cheers,

The Serpent Rider wrote on 2022-05-10, 05:21:

Socket 3 was released for new DX4 CPUs, so it has very short upgrade path. It's not correct to include pin-compatible CPUs which […]
Show full quote

Socket 3 was released for new DX4 CPUs, so it has very short upgrade path. It's not correct to include pin-compatible CPUs which are older than socket itself.
Now Slot 1 is a strong contender, but only if third-party mods for BIOS and adapters are considered. Same goes for LGA775, because it had pre-Core 2 and post-Core 2 revision, latter shuffled some pins.

flupke11 wrote:

Socket 370: 333 (iCeleron) to 1,4 Ghz (iPIII-s)

Actually 300 Mhz

Strictly speaking, Mendocino Celerons and Tualatin Celerons speak different bus protocols (AGTL+ vs AGTL).
So when a 370 pin socket from 300 MHz to 1500 MHz counts, a 168 pin 486 socket from 16 to 160 MHz counts too.

Well, socket 5-7-ss7 was really one socket if you consider socket 370 did sinilar changes through its life.

Some configurations along the development required interposers that is true.

There was an extra pin added to some s7 dual plane cpus, aswell some cyrix chips added an entire row of pins to reach pins on the socket that had always been there but I think were defined later on.

In any case, I group them all together 5-7-ss7 because they can all be made to work and fit backwards and forwards. Sometimes with aid of interposer.

It spanned
-75-570mhz, though 7-700mhz is possible.
-Multipliers ranging 1x to 6x
-Voltages ranging 1.6v to 3.6v
-TDP of 4w? To 29w
-Generations 5x86-quasi6x86
-FSB 50-100 though unofficially 7-150mhz is possible.
-L2 cache: none to 2mb on motherboard, 256k on chip(board would become L3)
-Many companies: intel, AMD, IBM, ST, Cyrix, IDT, Rise, and intel even made a fpga version of the intel atom to work with it as a experiment.

If I had to pick one that defined a generation it would be 775 and the launch of Core processors ...That CPU Uarch saved Intel's arse quite literally, the topic is about sockets but I find it impossible to separate 775 and core from each other.

It would be slot 1 wouldnt it?
I mean it went from a Pentium II 233 to a 1GHz Pentium III.

Socket 775 saw a larger change in CPU but its hardly vintage yet.

ElectroSoldier wrote on 2024-02-04, 23:03:

It would be slot 1 wouldnt it?
I mean it went from a Pentium II 233 to a 1GHz Pentium III.

Socket 775 saw a larger change in CPU but its hardly vintage yet.

How do you account for adding cores? LGA775 went from single-core P4 HTs to quad-core C2Q 9xxxs, while Slot 1 only ever had the same core count (1)...

I don't know my AMD sockets very well; after 939, how many cores did sockets like AM2/AM2+ reach?

My unresearched guess is that if you were looking at single-core performance only, you're right, slot 1 would come out ahead of LGA775.

There were some big boosts in 386 and even 286 sockets.... though that might be better called "footprints" since some of the CPU swap out options are dismounting a QFP and swapping for another.

Anyhoo, AT launched the 286 socket with 8mhz max and 6mhz the cheap option. What could later be put in that, were not just the legendary Harris 25Mhz 286, but also 386 and "486" which were enhanced 386 upgrades. The IPC of those over the intel 286 and 386 CPUs was 1.5x due to enhancements and CPU cache. Therefore on a 286 platform you might go from 8mhz to 66Mhz "486" which was at least 99 286 megahertz in performance, so 12.375 times. On 386 platforms you could go from 16Mhz to 100Mhz SLC3 so effective 150mhz boost for 9.375 times performance.

However often you were trying to put QFP CPU in sockets, so they had to have interposers, though also lower voltage support was needed in some too. It's a bit like the situation of putting a powerleap into an old single plane socket 5 or 7 though.

You can argue you need very particular boards to do an sx16 to SLC3 upgrade, but you also need particular boards to run both a P75 and a K6-2 @ 400 or 500, yes you can rework and mod a 430FX board, but if that is allowed, you can mod 286 or 386 boards to take their best of the best upgrades too.

Disruptor wrote on 2024-02-04, 18:45:

Strictly speaking, Mendocino Celerons and Tualatin Celerons speak different bus protocols (AGTL+ vs AGTL).

same protocol, lower voltage