Which Intel Chip had the Best Performance Leap from its Predecessor?

aphexcoil

Limp Gawd
Joined
Jan 4, 2011
Messages
322
When you go back to the very beginning (to the days of the 4004 chip and going forward), which Intel chip or architecture offered the greatest performance leap compared to its predecessor? What kind of performance leap was it?

Also, which Intel chip do you feel had the best overclocking potential of all time?
 
Celeron 352 and 356 had the overclock records for the longest time (8.2Ghz) that has only been broke recently with 3770k and 4770k(7+Ghz) these not as fast raw but WAY faster usable speed(memory of 2800+)

Core 2 I think was possibly overall the largest all around performance leap for intel far as I know, considering Pentium 3 to 4 then 4 to core 2 was a radical change of direction which also led to Core I.
 
When you go back to the very beginning (to the days of the 4004 chip and going forward), which Intel chip or architecture offered the greatest performance leap compared to its predecessor? What kind of performance leap was it?

Arquitecture? Prescott to Conroe, then Yorkville to Nehalem.

Also, which Intel chip do you feel had the best overclocking potential of all time?
The venerable Celeron 300A. The first to achieve 50%+ clockspeeds. I think some one took it up to 600mhz
 
Last edited:
so far the best overclockers that I recall are the Celeron 352/356 air and extreme, this was equaled by Phenom II 955(MANY died in the process) the new FX first was 8150 then 8320/8350, and now by Ivybridge/haswell. There is clocking potential, then given performance for that clock, potential would have been Pentium 4(and ones based on it) BUT they never could cool it well enough as a whole platform to truly let it fly, next would have been given performance of course that would be the new FX and haswell, both were/are catered to given performance at a high clock(FX) at pretty much any clock(haswell)

Pentium 3(was used as base for Core series based on Pentium M design and further expanded)
So yeh think overall biggest performance wise Intel seems to have been Yonah/Presler moving to Core then from Core to haswell(which has the biggest optimizations and reduced bottlenecks/most advance since core I first came out) so yeh IMHO there was lots of little steps that were jumps, but Pentium M and parts of 3 were used to make Core 2 which also helped for Core i, so Core 2 def is the winner :p

anyways here is nice little line
http://en.wikipedia.org/wiki/Enhanced_Pentium_M_(microarchitecture)#P6_Variant_Enhanced_Pentium_M
 
Arquitecture? Prescott to Conroe, then Yorkville to Nehalem.


The venerable Celeron 300A. The first to achieve 50%+ clockspeeds. I think some one took it up to 600mhz

Agreed on all points. I loved my Celly 300A @ 550 in my Abit 440BX based Slot 1 MB..

Wasn't there a wolfdale celeron that could do a 100% overlock on air?

I seem to remember something about that, but can't find anything on a quick Google search..The early P4s (the first to not use Rambus) could clock to 80-100% if they were a great sample..I believe they were the 1.6-1.7Ghz base clock SKUs...

The I7 920 D0s were also o/c'ing monsters..The highest I recall seeing proof of was clocked @ 4.7Ghz on AIR 24/7 Prime95 stable..that is a 90-92% overclock IIRC (they had a base clock of 2.6Ghz?)
 
Last edited:
Pentium e 21x0 were close to 100% overclock - you had variants with 1,6 or 1,8 Ghz clock that were doing 3 Ghz on stock voltage.
 
When you go back to the very beginning (to the days of the 4004 chip and going forward), which Intel chip or architecture offered the greatest performance leap compared to its predecessor? What kind of performance leap was it?

Also, which Intel chip do you feel had the best overclocking potential of all time?

Biggest performance leap IMHO was going 286 -> 386 & getting protected mode :)
 
Wasn't there a wolfdale celeron that could do a 100% overlock on air?

Yes, the E1200 stock at 1.6Ghz would do 3.2Ghz by doubling the FSB. Wasn't a wolfdale (they were the E3XXX celerons).
 
Arquitecture? Prescott to Conroe, then Yorkville to Nehalem.


The venerable Celeron 300A. The first to achieve 50%+ clockspeeds. I think some one took it up to 600mhz

Actually, the old 80386 and 80486 processors could do 50%+ overclocks pretty easily depending on the motherboard you were using and if the PCI bus could handle the higher speeds.

A 33% overclock could be had by just changing a single jumper on a lot of them, and on stock cooling.

The Celeron 300A was just the first one that was popular to overclock 50%+.
 
I think answering this question with chips that OC is unfair to the question. I think the OP wants to know what chip was the biggest leap in peformance out of the box, purly on technically advancement. OCing is not technical advancement as it can really be up and down based on the chip.
 
My vote goes Prescott to Conroe.

Not to mention my E6300, stock at 1.8ghz, hit 3.6ghz under water.
 
Wasn't there a wolfdale celeron that could do a 100% overlock on air?

also the wolfdale Pentium E5X00. for me one of the best overclockers chip, cold as hell..

BTW i had E5300 when new from 2.6GHZ to 4.0ghz on air (CM N520). with really low temps.. :). that chip its still rocking on a little cousin pc.. xD.
 
I remember having an Abit BP6 with two Celerons. I don't remember precisely what the overclocks where, but the system was ridiculously fast for what it cost. Being able to get a multiprocessor rig back then for next to nothing was absolutely amazing.

Unfortunately, my sister's hamster got loose, crawled into my machine one night and chewed everything in there into pieces...
 
Biggest performance leap IMHO was going 286 -> 386 & getting protected mode :)

I'm thinking it's either what you said or integrating the math co-processor in the 486dx. FP calculations increased by something crazy like 100x. The first true gaming cpu (for IBM/PC of course).
 
Gonna have to go with the innocuously named Pentium Pro, which was in fact the debut of the P6 microarchitecture featuring out-of-order execution and full speed L2 cache. The P6 architecture was the basis for the pentium II and III, Pentium M, Core...Yeah, basically everything worthwhile intel has produced in the past 20 years was based on the P6. So the real question is did they backwards engineer it from aliens, or time travel?
 
Probably the original Pentium. It doubled per clock performance over the 80486, and ramped up somewhat quickly to higher clock speeds.

In modern times maybe Conroe, which cut back on the clock speed, but was also about 2x as efficient as its predecessor.
 
Gonna have to go with the innocuously named Pentium Pro, which was in fact the debut of the P6 microarchitecture featuring out-of-order execution and full speed L2 cache. The P6 architecture was the basis for the pentium II and III, Pentium M, Core...Yeah, basically everything worthwhile intel has produced in the past 20 years was based on the P6. So the real question is did they backwards engineer it from aliens, or time travel?

While this is true, going from Pentium to Pentium Pro wasn't that big of a leap. They both had roughly equivalent 16bit performance, which at the time the majority of programs were 16bit. It did excel at 32bit, but I wouldn't say that is enough to be the biggest generational leap.

Probably the original Pentium. It doubled per clock performance over the 80486, and ramped up somewhat quickly to higher clock speeds.

In modern times maybe Conroe, which cut back on the clock speed, but was also about 2x as efficient as its predecessor.

I was thinking about both of these scenarios as well.

I'd say 486 -> Pentium was a much bigger performance leap because not only was the Pentium much faster clock for clock, it also clocked a lot higher than the 486. Conroe can't quite claim to be the first dual core with it's generation as it was beaten by Yonah and Pentium D. Also clock for clock it wasn't a huge leap over Yonah.
 
Last edited:
Gonna have to go with the innocuously named Pentium Pro, which was in fact the debut of the P6 microarchitecture featuring out-of-order execution and full speed L2 cache. The P6 architecture was the basis for the pentium II and III, Pentium M, Core...Yeah, basically everything worthwhile intel has produced in the past 20 years was based on the P6. So the real question is did they backwards engineer it from aliens, or time travel?


^^^^ This

The P6 was so good that it was even better than its successor. May Willamette never be spoken of again.
 
Core 2 Duo (Conroe), it was night and day difference compared to Prescott and Presler based predecessors in power consumption, overclockability and performance.
 
Core 2 Duo (Conroe), it was night and day difference compared to Prescott and Presler based predecessors in power consumption, overclockability and performance.

But it was only a 2x speedup in per-clock performance, and took a small clock speed hit, so overall performance improvement was around 1.8x. Big, but not world-bending.

The two biggest I've been able to find are the 8086 to 286, and the 286 to 386!

Te 286 was generally held to be 2x faster clock-for-clock than the 8086, and it was clocked about 50% faster, giving an overall improvement of ~3x. The 286 to 386 was closer to the 2x domain, which was still faster than the Core 2 transition.
 
Gonna have to go with the innocuously named Pentium Pro, which was in fact the debut of the P6 microarchitecture featuring out-of-order execution and full speed L2 cache. The P6 architecture was the basis for the pentium II and III, Pentium M, Core...Yeah, basically everything worthwhile intel has produced in the past 20 years was based on the P6. So the real question is did they backwards engineer it from aliens, or time travel?

Only problem with the Pentium Pro back when it debuted was that writing efficient assembler for the old Pentium meant that the programmer had to deal with the instruction scheduling manually by making sure that the A/B pipelines where full. With the Pentium Pro, it was all about maximizing instruction prefetching with the 4-1-1 prefetch cycle. This was so different that code tuned for the Pentium ran like shit on the Pentium Pro in comparison to what the Pentium Pro actually had to offer in terms of performance. It took quite some time before compilers where updated for it and games had their hand tuned assembly snippets updated for the new architecture.

However, the Pentium Pro was not really that special. Almost the exact same can be said about the Pentium when comparing it to the 486.
 
(BTW, a lot of the stuff that Intel introduced in their CPUs had already been present in IBM mainframes for a decade...)
 
But it was only a 2x speedup in per-clock performance, and took a small clock speed hit, so overall performance improvement was around 1.8x. Big, but not world-bending.

The two biggest I've been able to find are the 8086 to 286, and the 286 to 386!

Te 286 was generally held to be 2x faster clock-for-clock than the 8086, and it was clocked about 50% faster, giving an overall improvement of ~3x. The 286 to 386 was closer to the 2x domain, which was still faster than the Core 2 transition.

Unfortunately I'am not old enough to have lived through the 70/80's of Intel based systems. The oldest system I've had experience with was an Intel Pentium 80mhz system with 320MB of Quantum hard drive space, 16MB worth of Edo Ram. Then I jumped to a Pentium MMX slot a system at 266, then a Pentium 3 Coppermine and various Celery's, than the unforgettable Tualatins at 1.4 which was faster than a 1.8 P4 at the time that took rambus. A few Athlon XP rigs came and gone with some older K6 and Thunderbird systems. Then I had rigs consisting of every architecture afterwards from Intel and a few Opterons from the green team.

To me the Core 2 Duo jump was the biggest impact IME, it was also around the time where games started utilizing more than 1 core in processors, same said for applications and than the 8800 took over the graphics card market.
 
Only problem with the Pentium Pro back when it debuted was that writing efficient assembler for the old Pentium meant that the programmer had to deal with the instruction scheduling manually by making sure that the A/B pipelines where full. With the Pentium Pro, it was all about maximizing instruction prefetching with the 4-1-1 prefetch cycle. This was so different that code tuned for the Pentium ran like shit on the Pentium Pro in comparison to what the Pentium Pro actually had to offer in terms of performance. It took quite some time before compilers where updated for it and games had their hand tuned assembly snippets updated for the new architecture.

However, the Pentium Pro was not really that special. Almost the exact same can be said about the Pentium when comparing it to the 486.

Good point. Also don't forget the hit it took processing 16-bit code. That was not alleviated until the Pentium II revision, which also reduced costs with the external cache.

Also, from what I have seen code could be successfully optimized for the Pentium to run almost the same speed as the Pentium Pro (partially thanks to the 32KB cache upgrade of the Pentium MMX):

Graph linked here!

The "P6 Family" of chips had a long and impressive run, but no single chip in that family "wowed" quite as much as previous chips from Intel. Only though core tweaks/redesigns and compiler optimizations did the P6 realize it's true potential.
 
One thing with the old Pentium line of CPUs was that they were easy to understand and thus fine tuning code for them was fairly simple and it was often done. For the Pentium Pro architecture this was much harder, which is why most people just concentrated on the 4-1-1 instruction prefetching and let the out-of-order instruction scheduler do its magic. Actually keeping track of how many units within the CPU was in use at a specific moment was fairly hard. Intel essentially said that they don't expect programmers to do this anymore and that it's the compiler writer's job.
 
Back
Top