Intel to break the 4GHz barrier in 2008

[~>Dg<~]

thankyou for some decent arguements for and against what i put but 99.9 percent do not know what there cpu manufactures are doing in there labs. what they can do and cant do its all guess work by plain joes not calling anyone on here but a lot of peeps claim to know what theyre gunna do next do you work in these labs, i guess not neither do i . in my opion dual cores are being forced on us .theyve being trying to force them on us for 4 or five years now gaming is not by in large dualcore and it sickens me to the back teeth when peeps say theyll soon be multi threaded . right how long before theyre all multithreaded its been roughly 4years since they come out maybe another 5 years till they are if they ever do. probably some other cpu idea will arise maybe faster single core for instance . pls dont put crappy reply im just seing what you think as me personally dont think dualcores are that good your just being told they are bit like agp being told it was dead long while ago when they could have used it for a lot longer ,how many said long time before it was dead (not yet) that there would b no more cards being made and year or so on and there still making higher cards if the consumer wants faster single core thats what theyll build for us .might take a while and i hope in a year or so i can say told you so.reason for my arguement is cause a highly rated gaming source asked a certain company if they could do it and they responded they could and it was thing they were looking into.my english owns all

 

[Slam62]

and breath ......

I am not aware that a single core cpu will be available to purchase this time next year therefor one would hope that most of not all cpu intensive apps will be written for multi core.

We obviously need multi core, however we are currently still in the transition from single to multi.

Surely at the moment what we are seeing is an increase in processing speed alongside multi core which has got to be good for everyone in the end.

 

[p4radox]

When the transition to PCI Express was made, we all knew that the AGP bandwidth wasn't saturated yet (and still isn't). But with dual core there are advantages today. There are multithreaded applications that take advantage of multiple cores. They get far more of an advantage from multiple cores than they would from a faster single core. Games will be the same.

 

[PinkFloyd]

dgmug, the main people are crapping your posts is because there so damn hard to read as the English is poor, there's barely any punctuation and the spelling is, well, not there.
Guarantee you will not see powerful single core cpus in the future. there is no point.

Why would you make a single core cpu, when you can make a dual core cpu that is 2 single cores stuck together, and therefore is just as fast at single threaded things as the single core. It just defies logic....

 

[FrankJH]

mind you with more and more cores becoming available is there much point in HT being "re-employed"?

After all HT was in many ways a cheap 2nd core.... but by the time HT is meant to come back it looks like we will have 4-8 cores readily available anyway, I just cant see the point myself.

Unless the architechture completely changes I cant see single cores coming back for quite some time - with firewalls, av scans, let alone actual activities you want your PC to do - its much better to have numerous cores rather than just one

 

[NathanE]

FrankJH:

Of course there's still a purpose for HT It is meant to extract every last possible drop of performance from each core.

Single threaded CPUs will never be back. SMP should have been adopted on the desktop years ago but the market forces to do it just weren't happening back then.

 

[jaykay]

really depends on the length of pipelines if they are really long then ht will be back

 

[NathanE]

Nope, there are many different types of HT. The type used on Netburst did depend on a long pipeline. The type of HT coming in Penryn will be totally different

 

[FrankJH]

Nope, there are many different types of HT. The type used on Netburst did depend on a long pipeline. The type of HT coming in Penryn will be totally different

You may be right you may not - all I was thinking was that some top priority "core" function could be pushed onto HT and potentially give you worse performance ... and with multi core's being so prevelant (by the time this 2nd Gen HT comes out at any rate) I just cant see it being anything other than a waste of space.

 

[Cyber-Mav]

The enter key. Its the biggest key on the keyboard. You should press it occasionally.

i got a microsoft multimedia keyboard, and its the spacebar that has the highest surface area.

 

[Joe42]

i got a microsoft multimedia keyboard, and its the spacebar that has the highest surface area.

Heh... i knew someone would point that out.

I was making a point btw.

 

[Cyber-Mav]

well 4ghz by 2008 is a much better target than their "10ghz overnight" statement.

 

[ByteJuggler]

but 99.9 percent do not know what there cpu manufactures are doing in there labs.

From the horse's mouth then, if you won't believe it from anywhere else: http://softwarecommunity.intel.com/ISN/ISNVideoRepository/Video.aspx?fn=1031

 

[Joe42]

From the horse's mouth then, if you won't believe it from anywhere else: http://softwarecommunity.intel.com/ISN/ISNVideoRepository/Video.aspx?fn=1031

Theres some very interesting videos there, nice find.

 

[muon]

From the horse's mouth then, if you won't believe it from anywhere else: http://softwarecommunity.intel.com/ISN/ISNVideoRepository/Video.aspx?fn=1031

His mathematics is really bad. Power usage goes down 150% lol.

He has to use compound formulae for it to work.

 

[Jonny L]

4ghz is good but I heard from a very reliable friend that IBM has allready produced a 20ghz processor, however the cooling is worth more than the processor. They recon retail processors will be hitting 10ghz by 2010.

 

[Joe42]

His mathematics is really bad. Power usage goes down 150% lol.

He has to use compound formulae for it to work.

Yeah it wasn't very elegant but he made his point. I understood it. What he was saying is its best to make a core where the clock speed is such that power usage is fairly low but performance is fairly high, somewhere just before the point of diminishing returns. Then put lots of those cores together and you get a very efficient high performance cpu, you just need the software to take advantage of it. The end of the free lunch for software developers i think he said, which is very true. Their software won't get faster as processors get faster anymore.

4ghz is good but I heard from a very reliable friend that IBM has allready produced a 20ghz processor, however the cooling is worth more than the processor. They reckon retail processors will be hitting 10ghz by 2010.

Watch the video. They can't make the transistors go any faster. They probably could make a 10ghz processor but it would need an insane cooling system and it would have abysmal performance compared to a multi core cpu.
I suppose its possible that IBM are able to produce some better transistors, but i think its unlikely and even if they could do it its cheaper to do multicore.

 

[FrankJH]

IBM have found a way (according to anandtech) of slowing light down enough for it to be usable in CPU's instead of electricity, as they believe internal communication speed is the biggest barrier currently.

sounds interesting but maybe not usable in the timeframes posted above

 

[muon]

IBM have found a way (according to anandtech) of slowing light down enough for it to be usable in CPU's instead of electricity, as they believe internal communication speed is the biggest barrier currently.

sounds interesting but maybe not usable in the timeframes posted above

Doesnt electricity travel at the speed of light?

ediy: nvm, i see what they mean now.

 

[Joe42]

Doesnt electricity travel at the speed of light?

ediy: nvm, i see what they mean now.

No it doesn't, only light travels at the speed of light.

The problem is you have to make a transistor which works with light and thats kinda difficult.