LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   General (https://www.linuxquestions.org/questions/general-10/)
-   -   Is moore's law coming to end ? (https://www.linuxquestions.org/questions/general-10/is-moores-law-coming-to-end-943766/)

nec207 05-07-2012 07:23 PM

Is moore's law coming to end ?
 
So we hear that every 18 months CPU processors double in speed !!! This is a big myth !!! They get smaller every 18 months but this has nothing to do with speed !!

If you had CPU in 2006 and CPU in 2012 they will not be 2 or 3 times faster !! But may be 30% faster.

This is big shock for the public that thinks a CPU in 2012 would be 2 or 3 times faster than a CPU in 2006.

Why is this the case ? Are we coming up to brick wall with CPU speed?

jschiwal 05-07-2012 07:34 PM

Moore's law says that the number of transistors doubles every 18 months.

Clock speeds have stabilized after the Pentium 4 due to its power consumption. Today the emphasis is on multiple cores and integrating gpus.

Increasing the CPU clock speed doesn't pay if the CPU has to wait on busses to catch up.

Kustom42 05-07-2012 07:35 PM

Moores law stated the number of transistors that could be placed into an integrated circuit would double every 2 years. You are thinking of the 18 months that was spawned by an Intel CEO who quoted moore's law and stated double cpu performance. Two separate things...

273 05-07-2012 07:40 PM

I think you need to educate yourself about what Moore's Law is.
Aside from that, CPUs are not "just 30% quicker" now than they were in 2006. The complexity of CPUs increases every generation and does lead to impressive real world increases in power.

nec207 05-07-2012 07:45 PM

Quote:

Originally Posted by jschiwal (Post 4672833)
Moore's law says that the number of transistors doubles every 18 months.

Clock speeds have stabilized after the Pentium 4 due to its power consumption. Today the emphasis is on multiple cores and integrating gpus.

Increasing the CPU clock speed doesn't pay if the CPU has to wait on busses to catch up.

Clock speed or number of cores is not only think that going to make a CPU faster.


Quote:

I think you need to educate yourself about what Moore's Law is.
Aside from that, CPUs are not "just 30% quicker" now than they were in 2006. The complexity of CPUs increases every generation and does lead to impressive real world increases in power.
What do you mean?

273 05-07-2012 07:58 PM

CPUs come with more instructions and can work on bigger numbers than they could and there are more cores per CPU. Per core each CPU may well process instructions faster than the previous generation despite the "speed" being the same. Registers grow larger and more numerous.
I'll admit I'm not the most tuned in on CPU architecture but using a few PCs and a quick google tells anyone that there has been more than a 30% increase in real-world speed since 2006.

cascade9 05-08-2012 02:45 AM

I wish you would read and remember what people have already told you nec207-

Quote:

Actually, Moore's Law doesn't say anything about speed (and he said two years). It is about the count of inexpensive integrated transistors. The timespan was reduced to 18 months later by an Intel executive and he changed the doubled transistor count with a doubled processing speed.
But twice the amount of transistors doesn't double the performance, have a look at the example here.
http://www.linuxquestions.org/questi...4/#post4602327

Quote:

Originally Posted by nec207 (Post 4672826)
If you had CPU in 2006 and CPU in 2012 they will not be 2 or 3 times faster !! But may be 30% faster.

Got any evidence? Because I completely disagree with you.

Circa 2005/2006 CPUs are P4 HT and Athlon 64.

If you compare even 'clock limited' single core only performance, circa 2012 CPUs are up to 3x faster than the 2005/2006 CPUs.

Quote:

Our CPU core performance charts compares various different processor architectures at a normalized clock speed of 3.0 GHz and only using a single processing core per CPU. This allows a direct comparison between different processor models that are otherwise hard to compare, as todays products run different architectures, clock speeds or core counts.
http://www.tomshardware.com/charts/x...marks,128.html

http://www.tomshardware.com/charts/x...rake,2767.html
http://www.tomshardware.com/charts/x...-CS5,2775.html

The CPUs are slightly underclocked, increasing the clock would just make the newer CPUs look better (due to greater memory bandwidth, etc.)

If you enabled multicore on the newer CPUs, you would see much more than 3x times faster performance.

suicidaleggroll 05-08-2012 12:16 PM

Quote:

Originally Posted by nec207 (Post 4672826)
If you had CPU in 2006 and CPU in 2012 they will not be 2 or 3 times faster !! But may be 30% faster.

Have any numbers to back that up?

My company has a large multi-core atmospheric model that is heavily dependent on both processor and bus speed, and runs at an incredibly consistent speed for days/weeks/months per run.

I have a system that was built in 2006 using some of the best Xeon processors of the time. I know, because I built it. It runs this model at 24.6 seconds per timestep on 1 core, or 9.6 seconds per timestep on all four cores (twin dual core procs)

I have another system that was built in 2010 using some of the best Xeon processors of the time. I know, because I built it. It runs this same model at 10.6 seconds per timestep on 1 core, or 2.7 seconds per timestep on six of the eight cores (twin quad core).

That's a factor of ~4 improvement in processing speed, in 4 years. The two machines cost nearly the same amount of money when they were built, the only differences are the advancements made in processor architecture, bus speed, and RAM speed over those four years. In fact, the newer machine only has a 2.67 GHz clock speed while the older machine has a 3.2 GHz clock speed.

nec207 05-08-2012 01:49 PM

Quote:

Originally Posted by cascade9 (Post 4673059)
I wish you would read and remember what people have already told you nec207-


Got any evidence? Because I completely disagree with you.

Circa 2005/2006 CPUs are P4 HT and Athlon 64.

If you compare even 'clock limited' single core only performance, circa 2012 CPUs are up to 3x faster than the 2005/2006 CPUs.
.

What is Circa CPU?

Quote:

That's a factor of 9 improvement in processing speed, in just 4 years. The two machines cost nearly the same amount of money when they were built, the only differences are the advancements made in processor architecture, bus speed, and RAM speed over those four years. In fact, the newer machine only has a 2.67 GHz clock speed while the older machine has a 3.2 GHz clock speed.
What ? 9 times faster ?


It hard to believe that CPU double in speed every 2 or 3 years.

suicidaleggroll 05-08-2012 01:55 PM

Quote:

Originally Posted by nec207 (Post 4673594)
What ? 9 times faster ?

Sorry, used the wrong number as a reference, I've fixed my earlier post. It's closer to 4x faster...9.6 seconds on all 4 cores of the old machine versus 2.7 seconds on 6 of the 8 cores on the new machine. Even running single-threaded on one core the new machine is over twice as fast as the old, AND it has double the number of cores, AND that's on a slower clock speed (2.67 vs 3.2 GHz).

Kustom42 05-08-2012 02:09 PM

Quote:

Originally Posted by nec207 (Post 4673594)
It hard to believe that CPU double in speed every 2 or 3 years.

For the final time... Moore's law stated nothing about CPU speed, simply the number of transistors in integrated circuits would double. And CPU speed is very subjective to the other hardware you have it surrounded with. I remember someone got a P4 Prescott core over 6GHZ about 4 or 5 years back. The Intel exec stated the double cpu speed as a marketing gimmick and a play off of Moore's law please research this before you re-post with the same statement.

cascade9 05-09-2012 07:35 AM

Quote:

Originally Posted by nec207 (Post 4673594)
What is Circa CPU?

Circa-

Quote:

Circa (from Latin, meaning "around"), usually abbreviated c. or ca. (also circ. or cca.), means "approximately" in the English language, usually referring to a date.
http://en.wikipedia.org/wiki/Circa

Quote:

Originally Posted by Kustom42 (Post 4673603)
I remember someone got a P4 Prescott core over 6GHZ about 4 or 5 years back.

There was at least sceenies- no benchmarks, to unstable at the time- of a P4 560 (Prescott) @ 6.0GHz+ about 3-4 months after the chips were released, late 2004. Of coruse that was using liquid nitrogen. I have no idea if anybody has got any P4 over 5GHz on 'non-exotic' cooling (and I'm counting even water cooling as 'non-exotic'). If someone did it would be outclassed and slower than more modern CPUs, even if they have much lower clock speeds.

syg00 05-09-2012 08:08 AM

moore is less ... :D

Sorry, .... no really, I am .... :p


All times are GMT -5. The time now is 12:49 PM.