• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Will we see 6Ghz cpu's?

Soldato
Joined
18 Oct 2002
Posts
5,381
I know quite a few run higher clocks than 4ghz on insane cooling but do you think we will see 6 cores at 5-6ghz over the next 2 years? We seem to have levelled out on the 'mhz', whats next?
 
Soldato
Joined
12 Nov 2008
Posts
3,378
Location
T.O.
Well multicores are probably going to become more proficient.

Possibly break past the 6Ghz threshold, but I doubt much further than that tbh.
 
Don
Joined
7 Aug 2003
Posts
44,308
Location
Aberdeenshire
I doubt we'll see 6GHz with silicon to be honest, the highest commercially available clock rate CPUs where P4s which were 5-6 years ago now. AMD/Intel are more interested in increasing CPU power by increasing the number of cores these days.
 
Soldato
OP
Joined
18 Oct 2002
Posts
5,381
At what point will the number of cores out strip what they are running though? I mean we still see most new games dont use 4 cores, they can bring out 12 core cpu's but it will take another year or so before 1 game even utililses it, if they even do.
 
Soldato
Joined
20 Oct 2002
Posts
11,293
Location
Derby
I think the way forward is to get inteligent multicore. So you use an app that only uses 1 core the cpu can spread this across all cores and utilise the speed of each core without it being dependant on the software. who needs 6ghz if you can get 6x3ghz all working together seamlessly?
 
Permabanned
Joined
22 Feb 2010
Posts
1,629
I think the way forward is to get inteligent multicore. So you use an app that only uses 1 core the cpu can spread this across all cores and utilise the speed of each core without it being dependant on the software. who needs 6ghz if you can get 6x3ghz all working together seamlessly?

This would be a Super break through :O
 
Soldato
Joined
5 Jan 2003
Posts
5,001
Location
West Midlands
I think the way forward is to get inteligent multicore. So you use an app that only uses 1 core the cpu can spread this across all cores and utilise the speed of each core without it being dependant on the software. who needs 6ghz if you can get 6x3ghz all working together seamlessly?

Multithreaded apps will already grab as many cores as the app has threads, but a single threaded app... they struggle to even make use of all the execution units within a single core of a Core 2 or i7 processor, let alone using execution units within multiple cores. Thats the problem with single threaded applications, they are very hard to execute out of sequence, so you have to execute every instruction in order, so by the time you reach 1 instruction on every clock cycle the gains drop rapidly. The whole reason why Intel added Hyperthreading to P4, and i7 processors is because single threaded applications are so bad at making use of even a single core, that additional program/threads can make use of underutilized execution units, rather than purely timeshareing the cores to 1 application at a time.

6x3Ghz already works seamlessly if you have an application with 6 evenly loaded threads :p, but a badly written single threaded application using more than 1.2 cores at any one time (so about 20% cpu load divided equally between 6 cores).. I just dont see anything better coming anytime soon. This whole AMD reverse hyperthreading... people said Phenom would have it... clearly it doesnt... and I will be extremely surprised if Bulldozer, or any other CPU gets it anytime soon.

There is just too much sequential coding, and too many unpredictable loops in the average single threaded application, so you simply cant get a high degree of parallellism going.

Multicore is interesting, but I wonder why intel(or AMD) dont just make a superscaler processor with a much higher IPC... lets say 16 decoders/execution units, use hyperthreading to share the execution units around for multithreaded apps, and if the singlethreaded app has a fair amount of code which can be executed out of order then it would be allowed to grab as many encoders/execution units as it wanted.

At the end of the day though, unlike graphics processing (big wide GPU's with hundreds of shader/execution units), general purpose CPU's quickly reach a point where execution units are unused and code is waiting for results from the last instructions before continuing.
 
Last edited:
Don
Joined
7 Aug 2003
Posts
44,308
Location
Aberdeenshire
Imagine if the clocks of all the cores were out of phase with each other and each core ran an instruction in turn, probably thinking of things to simply. Obvious issues with branch prediction, IPC and data sequencing.
 
Don
Joined
7 Aug 2003
Posts
44,308
Location
Aberdeenshire
That wasn't the question though, haven't we already demonstrated 1000GHz graphene transistors?
In the next two years?

We won't see it with a commercial CPU and I doubt we will ever see it with silicon either unless someone decides to go down the netburst route again. Would be interesting to see what kind of clockrate a P4 would do on 32nm.
 
Back
Top Bottom