When will we shatter the current 3GHZ speed ceiling?

N

nist7

New Member
It's amazing how much technology has progressed over this past decade....and now it's even more amazing how it has seemingly hit a ceiling with respect to CPU clock rate speeds. An Intel Pentium4 first hit the market with an advertise 3GHZ clock rate and after almost 7 years...it seems that has gone no where with the highest speed of 3.33GHZ for the E8600 Wolfdale (at least according to newegg...although there may be faster business/commercial CPUs that I'm unaware of). Now obviously, there's more to it than simply the GHZ clock speed in terms of performance and newer CPUs are faster performers when benchmarked with Pi calculations and whatnot. And obviously enthusiasts have pushed clock rates into the 4GHZ territory, but imagine if the clock rates came stock at 4GHZ...!

I'm just curious as to why, with regard to the basic clock rate of CPUs, they seem to be hovering at around 3GHZ. I've heard that this is because they couldn't physically manufacture for higher clock rates at our current production technology (hence, they just put more cores into each CPU: dual core, quad core, etc.)

So is this a limit to our progress in computational power? Or are we just going to add more cores to each CPU? Or maybe we need to dig up some alien technology they've been hiding under Area 51? :D
 
it seems that has gone no where with the highest speed of 3.33GHZ for the E8600 Wolfdale
Click to expand...

P4's went as high as 3.8 stock. The higher Ghz you get, the more heat is generated. Higher speed is not the problem - the heat, however, is. With each higher clock obtained, the amount of heat generated increases exponentially - cheap HSF's would not be able to dissipate it fast enough.

I'm just curious as to why, with regard to the basic clock rate of CPUs, they seem to be hovering at around 3GHZ. I've heard that this is because they couldn't physically manufacture for higher clock rates at our current production technology (hence, they just put more cores into each CPU: dual core, quad core, etc.)
Click to expand...

That is incorrect. The technology exists. You can still purchase (new) CPU's 2.2 and below. 3Ghz is the 'default' for a few reasons; marketing and bang-for-the-buck-return being the top two. Again, heat is a factor. It's actually more efficient to utilize multi-core CPU's than higher clock speed (to a certain degree, there are exceptions).
 
Many overclockers constantly run their CPUs at 4GHz+, and even 7GHz has been reached with certain chips (NetBurst-based). People have also hit 5GHz on Core2 CPUs and the upcoming Phenom II have hit 6GHz (on dry ice I believe) in a demonstration of the CPUs capabilities. Also, Pentium 4/D CPUs shipped at stock speeds above 3GHz (3.8GHz was the highest they ever did), as do several Athlon CPUs (X2 6400+ Brisbane core runs at 3.2GHz stock AFAIK). The problem isn't that the speed can't be achieved, it's, as said before, heat generation and power consumption. It wouldn't be very economical to sell a 4GHz CPU when you'd have to include a massive cooler that costs almost as much as the CPU itself. And power efficiency is something that matters nowadays - CPU manufacturers are focusing on efficiency rather than clockspeed, so that they can make faster processors without having to increase power consumption or heat generation to a significant extent.

Of course, there is a limit to the clockspeed as well... to get higher clocks you need higher volts and smaller manufacturing node, and higher volts kill chips while smaller manufacture makes chips even more voltage-sensitive...

Also, no, clockspeed is not a limit to CPU power. Even in single-core applications Core i7 beats Core 2 which beats Phenom which beats Athlon 64 which beats Pentium 4/D, even when all of those CPUs are running at the exact same clockspeed and even when only one core is utilized. Efficiency... CPU manufacturers are making CPUs to get more done each clock cycle, so performance scales up without having to force the clocks higher. And then, we keep adding more and more cores... the clockspeed barrier is by no means a serious issue as far as performance scaling is concerned.
 
imsati said:
P4's went as high as 3.8 stock. The higher Ghz you get, the more heat is generated. Higher speed is not the problem - the heat, however, is. With each higher clock obtained, the amount of heat generated increases exponentially - cheap HSF's would not be able to dissipate it fast enough.
Click to expand...

Wow 3.8GHZ? That is quite impressive. Doesn't some OC enthusiast in Japan hold the record for the absolute highest clock rate using a single core P4? Something like 8GHZ....

imsati said:
That is incorrect. The technology exists. You can still purchase (new) CPU's 2.2 and below. 3Ghz is the 'default' for a few reasons; marketing and bang-for-the-buck-return being the top two. Again, heat is a factor. It's actually more efficient to utilize multi-core CPU's than higher clock speed (to a certain degree, there are exceptions).
Click to expand...

Ah interesting. So it seems that heat is the main issue, of course since modern day CPUs won't even stay cool enough to boot into BIOS w/o its HSF, one can argue that the "technology" is still insufficient as cooling is so integral to the process. But it's good to know that we're not at the raw CPU limit, as long as we find more effective ways to provide practical cooling solutions.

But with that in mind, I wonder if Intel/AMD/etc. have done experimental testing of raw CPU power, and then adjusting the cooling appropriately. ie, pushing a CPU to double digit GHZ and then using whatever coolant neccessary to maintain temp, maybe by using liquid helium or something crazy extreme :eek:

diduknowthat said:
Uh, there are plenty of P4's that broke 3ghz, and even a handful of Core 2 Duos.
Click to expand...

Indeed. Although I was referring to the 3ghz ceiling as in there haven't been any CPUs coming from the factory with stock speeds of 4GHZ. So I guess I should've revise the title to 4GHZ ceiling....eh stupid me :o

hackapelite said:
Many overclockers constantly run their CPUs at 4GHz+, and even 7GHz has been reached with certain chips (NetBurst-based). People have also hit 5GHz on Core2 CPUs and the upcoming Phenom II have hit 6GHz (on dry ice I believe) in a demonstration of the CPUs capabilities. Also, Pentium 4/D CPUs shipped at stock speeds above 3GHz (3.8GHz was the highest they ever did), as do several Athlon CPUs (X2 6400+ Brisbane core runs at 3.2GHz stock AFAIK). The problem isn't that the speed can't be achieved, it's, as said before, heat generation and power consumption. It wouldn't be very economical to sell a 4GHz CPU when you'd have to include a massive cooler that costs almost as much as the CPU itself. And power efficiency is something that matters nowadays - CPU manufacturers are focusing on efficiency rather than clockspeed, so that they can make faster processors without having to increase power consumption or heat generation to a significant extent.

Of course, there is a limit to the clockspeed as well... to get higher clocks you need higher volts and smaller manufacturing node, and higher volts kill chips while smaller manufacture makes chips even more voltage-sensitive...

Also, no, clockspeed is not a limit to CPU power. Even in single-core applications Core i7 beats Core 2 which beats Phenom which beats Athlon 64 which beats Pentium 4/D, even when all of those CPUs are running at the exact same clockspeed and even when only one core is utilized. Efficiency... CPU manufacturers are making CPUs to get more done each clock cycle, so performance scales up without having to force the clocks higher. And then, we keep adding more and more cores... the clockspeed barrier is by no means a serious issue as far as performance scaling is concerned.
Click to expand...

Indeed, there seems to be a lot more than meets the eye. Efficiency in terms of "how much can a CPU do running at this speed" does mask the higher "speed" in the clock rate number. Although, I completely understand your point, it seems that heat (and voltage sensitivity) is the only big enemy when it comes to clock rates. As you said, CPUs are more efficient these days and do more per cycle than they did before. So I was just wondering how much MORE they can do if they could just up the cycles per second. As I've said above, I'd be interested to know if Intel or AMD have eliminated heat from the equation and just tried to test for raw CPU power.
 
The highest I have heard of was somewhere around 15.1 ghz. A university overclocked an Intel (can't remember the model) quad core, and they literally had to keep it cooled with liquid nitrogen. Sorry I don't have a link, I saw it on Discovery somewhere.
 
of course since modern day CPUs won't even stay cool enough to boot into BIOS w/o its HSF
Click to expand...

Many can and will. Tit-for-tat, modern CPU's are much cooler than earlier versions. Better internal architecture and more efficient use of voltage are to thank. Lots of Big Brand systems only have a HS, no fan. Depending on the CPU, you can boot into Windows without even a HS. I would *not* recommend it though.
 
The reason intel and AMD haven't manufactured higher ghz processors is that there is no need too. Ghz doesn't matter. It comes down to the CPU itself that determines the speed. If you took a higher end cpu today, turned off all but one core and clocked it the same as a P4, it would still dominate it. GHz means nothing. Intel just gave the conception when they manufactured the P4 because the only way they could produce faster CPUs is to up the frequency, that is all.
 
Motoxrdude said:
The reason intel and AMD haven't manufactured higher ghz processors is that there is no need too. Ghz doesn't matter. It comes down to the CPU itself that determines the speed. If you took a higher end cpu today, turned off all but one core and clocked it the same as a P4, it would still dominate it. GHz means nothing. Intel just gave the conception when they manufactured the P4 because the only way they could produce faster CPUs is to up the frequency, that is all.
Click to expand...

hmm...


You say "GHZ means nothing." and then in the very next sentence, u say "the only way they could produce faster CPUs is to up the frequency."

All else being equal, a CPU with a higher frequency will be faster than a CPU with a lower frequency no?

Obviously, today's CPUs can do moer per cycle than yesterday's CPU; so the effective frequency of modern day CPUs are higher and the i7 with a speed of 3GHZ is faster than a 3GHZ P4.

26GHZ graphene transistor: http://www.dvhardware.net/article32106.html --> The beginning of the end of the silicon-based CPUs? Or does GHZ mean nothing? :o
 
Well their is a ceiling and till they find a way around it they will continue to add processors vs ghz.

I was reading some where a review which stated its far easier to make tack on more cores then it would be to up the ghz after 4ghz.

Sure processors can do that now but not without a insane cooling method that couldn't be used 24/7

I7 and Deneb have impressive speeds but they haven't found a cost effective way to get us a 5-6ghz quad core with temps/wats that wont make us cry.

So I imagine the future till they find a solution will be running 8 core + cpus.
 
You must log in or register to reply here.
Back
Top