Anyone else missing the MHz wars....

[Stickman]

Back in the day I remember how (to compete with each other) Intel and AMD were constantly bringing out CPU's with higher and higher stock clock speeds. Whilst it was (and is) well known that pure clock speed alone does not equate to best performance (as architecture also plays a huge role especially in these days of multi-core systems), does anyone else feel it's been far too long since we've had any real noticeable increases in raw CPU stock clockspeeds, and miss the good old days of seemingly constant speed increases?

Just check out this range of speeds from 90nm prescott processors around that time - 2.8 to 3.8 GHz - it could just as easilly be the speeds of the as yet unreleased haswell refresh or broadwell processors ten years later!

Even with the fact that sure, more cores means more heat, with all our technological advancements, I'm suprised we haven't seen higher stock speeds yet? With the way computing tech usually speeds forward, isn't an entire decade a hell of a long time for stock speeds to have stagnated for?

 

[eddiew]

To be fair, if you put a 1 core Haswell up against a Prescott at the same frequency, the Haswell would thump it into the ground

Thing is that power is directly proportional to both frequency and voltage; there's a limit to how low you can take the voltage as determined by the process size and quality, and there's a limit to how much power a cooler can sensibly dissipate (about 150W seems the upper edge for air), which means there's a limit on the frequency you can run a chip at. To avoid heat death, manufacturers instead focussed on instructions per clock, meaning frequency and voltage stay the same. In fact I'm pretty sure Intel's Atom processors are nearly Pentium 4 speeds, at half the frequency and tiny power draws.

That said, I do miss the numbers races, but the architectures are now so different that it's meaningless. It's like comparing cars on engine RPM - means nothing without knowledge of gear boxes, vehicle weight, torque figures, BHP, etc.

Though to be fair to them, AMD have their 5 GHz monster xD

 

[Krisali]

It's to do with entropy and the limitations of silicone. When they change materials processors will get faster

 

[tbh]

I still remember reading Anandtech's review of the first Pentium 4 to break the 3ghz barrier, all the way back in 2002 (I've linked it below, for nostalgia reasons); if you had told me back then that 12 years later we'd still be releasing processors at around the same clock speed I'd have thought you were crazy!

http://www.anandtech.com/show/1031

As Krisali said above though we've hit the wall as far as silicon is concerned; when foundries switch to using silicene/graphene then more big gains will be made, don't expect that to happen anytime soon though.

 

[Five.Stars]

Agreed, back then I just assumed that the frequency of chips would continue to increase and that by 2014 we would all be rocking 6ghz+ chips!

 

[Steviepunk]

I remember the old days as well, back when it was MHz and GHz, and measured in 2 digits! (I'm sure someone will remember single digits though...)

They were good days and the numbers race was entertaining, but in reality I don't miss it. Like most things, when a 1 dimensional element becomes the driving factor, real innovation get's a bit lost. They focused so long in the MHz (or GHz) that other worthwhile tech was slow in arriving - for example, multicore and 64bit.

 

[eddiew]

I remember the old days as well, back when it was MHz and GHz, and measured in 2 digits! (I'm sure someone will remember single digits though...)

I remember the 7MHz 68000 in my Amiga 500+. Later, my Amiga 1200 had a 68020 that ran at a mighty 14MHz, later upgraded to a 25MHz 68030. Such power!

The best bit was the "maths co processor" - floating point was done on a completely separate chip xD

 

[tbh]

It's to do with entropy and the limitations of silicone. When they change materials processors will get faster

Incidentally since we're talking about materials, silicone is what they make breast implants out of; what you mean is silicon, or else you've been buying some very iffy processors

 

[Ajaxx]

Incidentally since we're talking about materials, silicone is what they make breast implants out of; what you mean is silicon, or else you've been buying some very iffy processors

i7 34DD won't be coming out then???

 

[mmj_uk]

Not really because Mhz was rarely relative to performance anyway, there were always architectural differences which made Mhz useless as a performance barometer.

Nowadays Intel keep clockspeed the same and increase core efficiency rather than the overall clockspeed.

 

[dvs8]

I think more telling is the fact that there have been about 20 threads on here recently (including one I started!) the conclusions of which were that in fact, no, there is no noticeable performance upgrade to be had if you own a CPU that's even 4 years old (example: my 4 year old 3.8GHz i5 750 won't yield enough benefit for the cost of a new haswell chip (+mobo))

I can't remember a time when huge gains in performance weren't to be had by buying 4-years-new CPU technology (remember Pentium II (1997) -> P4 (2000), K6-2 (1998) -> AthlonXP (2001))

 

[tbh]

I can't remember a time when huge gains in performance weren't to be had by buying 4-years-new CPU technology (remember Pentium II (1997) -> P4 (2000), K6-2 (1998) -> AthlonXP (2001))

This is certainly true, the steady stream of performance gains has slowed to a trickle; if you bought an i7 920 back in 2008 then there's still not really any compelling reason to upgrade, a situation which would have been unthinkable a decade earlier.

 

[HangTime]

Yes, it is frustrating. I knew something was afoot when I ended up having a chip of the same clock speed 3 upgrades in a row (P4-1.8A, A64 Venice 3000+, C2D E4300)!

Obviously modern CPUs are way faster clock-for-clock but one can't help but wonder what things would have been like with fewer limitations.

One benefit I suppose is that power consumption has probably been reined in a bit compared to if we were all packing 10ghz+ cpus.

 

[almighty15]

Some of AMD and Intel chips come with 4Ghz+ stock speeds.

I remember not that long back on the Core Duo/Quads that 4Ghz was considered a massive overclock and is what everyone targeted.

Now 5Ghz is the thing that everyone targets.

 

[spoffle]

I think more telling is the fact that there have been about 20 threads on here recently (including one I started!) the conclusions of which were that in fact, no, there is no noticeable performance upgrade to be had if you own a CPU that's even 4 years old (example: my 4 year old 3.8GHz i5 750 won't yield enough benefit for the cost of a new haswell chip (+mobo))

I can't remember a time when huge gains in performance weren't to be had by buying 4-years-new CPU technology (remember Pentium II (1997) -> P4 (2000), K6-2 (1998) -> AthlonXP (2001))

There certainly have been a lot of performance gains, the issue is that the things people generally are using their PCs for aren't really getting that much more demanding, which means that hardware has surpassed software requirements for a lot of people.

Which is why old hardware is holding up very well, as there isn't a perceivable difference between some hardware because it's taxing neither CPU.

 

[Krisali]

Incidentally since we're talking about materials, silicone is what they make breast implants out of; what you mean is silicon, or else you've been buying some very iffy processors

Ooooops

I must have breasts on the brain

 

[Krisali]

hardware has surpassed software requirements for a lot of people.

Which is why old hardware is holding up very well, as there isn't a perceivable difference between some hardware because it's taxing neither CPU.

Yes I agree and this was my argument in one of the other threads that popped up recently. It's a curious situation, good for the pocket but boring because the only real reason to upgrade is for forum kudos/benching. I think Star Citizen running on Cryengine as it does and being massively multiplayer will tax my current system to the point where I must out of necessity splurge on a Haswell-E and X99

 

[Damo667]

I was updating my profile on an online racing league forum the other day and I signed up in 2010 sporting a 3.0ghz pentium 4 and an nvidea 8800. The next pc we had in about 2012 was a 2.7ghz amd athlon ii x3 so a slower clock speed but treble the cores meant it was so much better performing. My current system in my sig is my first custom built machine and blows any previous pcs we've had out of the water. I'm used to waiting for folders to open haha this is instant.

 

[zyzzyzus]

I remember the 7MHz 68000 in my Amiga 500+. Later, my Amiga 1200 had a 68020 that ran at a mighty 14MHz, later upgraded to a 25MHz 68030. Such power!

The best bit was the "maths co processor" - floating point was done on a completely separate chip xD

We had a C64 which apparently had a 0.985 MHz processor.

As to the thread topic, I don't really see the need for increasing MHz at the moment. Intel could start using better TIM and a better stock cooler so the stock clock could be higher but then they'd struggle to sell the marked up unlocked processors.

Would there not still be a big problem dealing with heat at really high clock speeds if something like graphene was used instead?
Though with something like liquid nitrogen you could probably get something crazy.

 

[Hades]

I remember when brr 1ghz was massive news.

I also remember my VIC20 running at 1.1mhz.