Thing is that So5 and So7 are pretty much the same standard, so it's an academic discussion.

As far as the pins go, the difference is that So7 has a second multiplier pin (allowing 2.5x and 3x) and it allows for (but does not mandate) split voltage. So an So5 board will take any So7 CPU that uses single voltage and max 2x multiplier.

On the CPU side, the P54C is officially So5, the P54CS So7, but the P133 only uses 2x multiplier and single voltage so will happily work, particularly as it draws much less power than the earlier P54C CPUs. In fact, an So5 board will happily run P54CS up to 200MHz, so long as you find a way to set the second multiplier bit.

Socket 5 "authentically" will work with everything up to Pentium 200mhz. Socket 7 was designed for Pentium MMX.

6.2. Socket 5
Two upgrade sockets have been defined for the Pentium processor-based systems as part of the processor architecture. Socket 5 has been defined for Pentium processor 75, 90, 100, and 120 MHz-based systems and is defined in the Pentium®Processor Family Developer’s Manual, Volume 1. Socket 5 does not support upgradability for 133 MHz or higher processors.

6.3.Socket 7
Socket 7 has been defined as the upgrade socket for the Pentium processor 133, 150, 166 and 200 MHz in addition to the Pentium processor 75, 90, 100, and120 MHz. The flexibility of the Socket 7 definition makes it backward compatible with Socket 5 and should be used for all new Pentium processor-based system designs. The Socket 7 support requires minor changes from Socket 5 designs – an additional key pin, 3.3V clocks, additional supply current, etc. Contact Intel for further information regarding the Socket 7 specifications.

http://bitsavers.trailing-edge.com/components … ume_1_Jul95.pdf

Socket 7 has been defined as the upgrade socket for the Pentium processor 0110\133) in addition to the Pentium processor (610\75, 735\90,815\100, 1000\120). The flexibility of the Socket 7 definition makes it backward compatible with Socket 5 and should be used for all new Pentium processor-based system designs. The Socket 7 support requires key changes from Socket 5 designs; split voltage planes, voltage regulator module header, 3.3 volt clocks, BIOS updates, additional decoupling, etc. This information is not provided in this databook. Contact Intel for further information regarding the Socket 7 specifications.

I think Socket 7 was released (June 1995) after Intel Pentium 120 (March 1995) so anything after that is technically a Socket 7 CPU.

dionb wrote:

Thing is that So5 and So7 are pretty much the same standard, so it's an academic discussion.

As far as the pins go, the difference is that So7 has a second multiplier pin (allowing 2.5x and 3x) and it allows for (but does not mandate) split voltage. So an So5 board will take any So7 CPU that uses single voltage and max 2x multiplier.

On the CPU side, the P54C is officially So5, the P54CS So7, but the P133 only uses 2x multiplier and single voltage so will happily work, particularly as it draws much less power than the earlier P54C CPUs. In fact, an So5 board will happily run P54CS up to 200MHz, so long as you find a way to set the second multiplier bit.

great, but i think i will go with the official p120 way..😀 becuase... for the p133 looked the 430fx authentic, but with official socket7 board and pipeline cache and i would pair the higher clocked versions with 430hx.

there were just months between the solutions:
1995q2 - 430fx, socket5, async cache (p120)
1995q4 - 430fx, socket7, pipeline cache (p133-p166)
1996q2 - 430hx (p166+)

appiah4 wrote:

I think Socket 7 was released (June 1995) after Intel Pentium 120 (March 1995) so anything after that is technically a Socket 7 CPU.

it is possible - the p133 released in june 95. if i would like to choose a cpu for the socket 5 zappa (maybe the first triton board), i think the p120 would be more authentic.
the board released around may 95, maybe earlier - i will read the early 95 pcmags after the clues.

i will do more benchmarks, but it doesnt look too fast 😀
https://docs.google.com/spreadsheets/d/1eXksI … dit?usp=sharing

1990\3DBENCH
1990\3DBENCH2
1992\WOLF3D
1993\DOOM
1995\PERLMARK
1996\CHRIS3D VGA
1996\CHRIS3D 640x480
1996\DUKE3D VGA
1996\DUKE3D 640x480
1996\JEST VGA
1996\PCPBENCH VGA
1996\PCPBENCH 640x480
1996\QUAKE VGA
1996\QUAKE 640x480
1997\AI_BENCH
1997\MDK
1997\NTS
1997\SKBENCH 640x480

100% - p120, 430fx, async cache, millenium (Jun 1995)

period correct socket 7 comparisons:
130% - p133, 430fx, pb cache, s3 trio64v+ (Dec 1995)
154% - p166, 430hx, pb cache, et6000 (Sep 1996)
204% - p233mmx, 430tx, pb cache, riva128 (Sep 1997)

Don't underestimate the impact of 60MHz FSB (and 30MHz PCI) on those scores. Unless you are doing a pure CPU-limited benchmark, a P120 hardly scores any better than a P100 (which does run at 66MHz FSB).

The gap between the P120 and P133 is huge, but it's multiple factors:
- 13MHz clock speed
- 10% faster FSB (and L2 cache, RAM & PCI)
- PLB vs asynch L2

You could control for those differences by underclocking the P133 to 120MHz or benching the P133 on the So5 i430FX, so you're just seeing the asych vs PLC difference. I'd guess that PLB is probably about half the difference in scores.

i did it previously (in the list as hidden column) 😀

* p120-133 is a pure ca 10%
* the async-pb means ca 10-15% (or it can be timing differences too - the intel ami bios is very dummy)
* the matrox is surprisly slower than the s3 trio 64v+ in this board (i measured same speed in the mmx) - maybe it depends on the not too good vesa bios (i tested with univbe there) ca 5%
* and the 430fx is a fast chipset compared to the previous socket5 solutions - the early 94 pentiums would be even slower

a plan to compare the boards, cpus, videocards with same conditions.

i430FX is vastly faster than the previous Intel chipsets like the i430NX. What forced Intel to clean up its act was the SiS 501 chipset, which had almost the same feature set (just no PLB), but was released a year earlier and completely outclassed i430LX and i430NX. It remained the fastest single So4 chipset (it was dual voltage 5V/3.3V), and on So5 with asynch was only a bit slower than the i430FX. Unfortunately for SiS, once Intel really allocated engineering resources, they stayed ahead.

is there any difference between the sync and pipelined burst cache? the old pcmag says, the pb cache known as sync cache, but the wikipedia says they are different, successor the of the sync (and the older async) cache.

i found some pentium roundups in the 95 pcmags:
may 1995 - no info about the cache (async?), p120
sep 1995 - async and (mostly) sync caches, p120-p133
dec 1995 - pb cache was typical, p133

as i know, the first tritons came with async caches (advanced/zp or the asus), but the asus has a coast, which was much more typical in the vx/hx age.

Async came first, then you got async with COAST to allow for upgrade to PLB, followed by PLB soldered on the motherboard. However certainly during the COAST era you saw all three at the same time, depending on brand, age of stock and high/low-end products.

Not sure what sync as opposed to PLB would be, if it exists, I've probably seen it without recognizing it...

Here's another difference between Socket5 and Socket7, though this has more to do with the boards they are on rather than the sockets themselves. With socket 7, you are guaranteed the board will work at 66MHz, but with Socket5 perhaps not, particularly if we are talking about the Neptune chipset. Even after the move to 3.3V, 66MHz was still a pain in the ass to get working reliably. The P100 was around, but somewhat rare as you needed a special "66MHz" qualified motherboard to make it work. P100 didn't really sell much until after the P133 and 430FX came out.

I think with the 430FX, some of those boards had Socket5s while the later ones used Socket7s. Anyone know if there were 430FX boards that couldn't handle 66MHz?

I collected two 430fx boards, both uses cr battery (that was important):
Early Socket 5 with async cache
Late Socket 7 with pb cache

Both supports well the 66mhz fsb.

get that bad boy a COAST module!

Anonymous Coward wrote:

Anyone know if there were 430FX boards that couldn't handle 66MHz?

i've been wondering about these elusive "FX-60" marked chips that are mentioned in the 430FX datasheet... they seem to be extremely rare, if they were even put on actual motherboards at all.

Video card recommendation focused to the dos performance? Until the socket7/p133 release (jun/jul 1995).

I have a millennium, i found the first test from jun 1995. I am quite impressed, the dos performance is not the best, but the video output quality and the vbe2 is very good.

Stealth 64 (trio) with 2.xx bios (vbe2?)?
The ark2000pv is from aug 1995.
Trio64v+ late 95.

i created a new topik for the build:
Socket 5 @ 1995.05 - Pentium 120, S3 Vision968, Sound Blaster AWE32, Gravis Ultrasound