P4 will scale above 10GHz

[Raystonn]

"nowhere does it mention that that technology would be implemented in the core of the p4"

Very true. I was just pointing out that Intel has the technology to scale up to 10GHz using standard silicon practices. This will likely be incorporated into shrunk down versions of processors using the P4 core. The article indicating the P4 core will go beyond 10GHz is here:

http://quickenexcite.cnet.com/news/0-1003-200-5742941.html?tag=pt.intuit.newsfeed..ne5742941?tag=

"what is your perspective on the belief that others in this thread seem to think that there motherboard may actually be usable all the way up to ten gig"

It's possible, but having the latest motherboard always increases performance of your system. Keeping the same motherboard for 4 years wouldn't be the best idea from a performance frame of mind. The Abit BH6 motherboard worked on Celeron, P2, and P3 systems from around 200MHz to about 1GHz. That's a multiplier of 5 on clockspeed using the same basic core. (You can shrink the die, add a couple new things, and still use the same core. These were the changes from P2 to P3.) 2GHz to 10GHz is also a multiplier of 5. Was the BH6 the best performer on 1GHz processors? Probably not, but it worked.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[TheAntipop]

i was attacking your argument, you are trying to sound technical; part of youre argument. and if you think that im attacking you personally, then i suggest you step away from the computer and go outside, then you might find out what a real personal attack is. what im doing isnt even close. oh and you dont have to quote my whole post every time in order to get your points across, thats why the reply button sits in the top right hand of each one.
btw, you didnt address any of my logical questions. just attacking your argument, not you

<A HREF="http://www.512productions.com/lobstermagnet/" target="_new">Hyakugojyuuichi!!</A>

 

[Raystonn]

"you are trying to sound technical"

I don't have to try. I have technical information and merely supply it.

"if you think that im attacking you personally"

Look up the term 'ad hominem'.

"you didnt address any of my logical questions"

On the contrary, I addressed them all.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[ejsmith2]

Raystonn, do you use a company computer to read through this group, or is it at a private residence?

 

[Guest]

Yeah.. If you dont mind, i'd like to know too.

 

[Raystonn]

Private of course. I don't know many people who sit in an office on a Sunday night, except perhaps lawyers.

Why?

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[girish]

there are many physical limitations to consider.
first, motherboard may not go much beyond 133~166, at the most 200 MHz.

we have already reached very close to similar limitations on silicon. how many MHz can silicon handle?

well, as fabrication processes migrate from 0.13 to 0.7 or even 0.3, it would be more and more difficult and expensive to produce chips since producing masks will require higher and higher accuracy and fabs to be cleaner, silicon to be purer... contrarary to what experts say i guess. or it might take a while to get these production processes to be cheaper -thats increased gestation period.

so now, will it be worth to wait? will it be worth to actually produce a 10 GHz (physical) chip on silicon when there are other exciting and promising technologies like GaAs or Carbon nanotubes or optical or even bio showing up?

i think 10 GHz will really mean 10 GHz performance, as I said in my earlier post. reduction in tramsistor size, die size (upto its practical limit) will allow to pack more circuits on the same size chip - allow us to make innovative organisational enhncements to the chip so that it will *perform* as if it is running at much higher physical speeds. for example, inclusion of cache, and the L2 cache on the chip itself allowed to increase the performance, the barrel shifter introduced in the 386 allowed multi-bit shift operations to be done in single cycle, increasing the number of pipes in the superscaler pentium which were actually two independent execution units almost doubled the performance at the same MHz as that of the 486. now there is VLIW (Very Long Instruction Word) that could actually scale the performance of the processor to *effective* levels of GHz speed yet working at much lesser physical (and practical) speed.

such could be the impact of shrinking transistors on cpu performance, apart from brute force MHz increments.

girish

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>

 

[Raystonn]

"motherboard may not go much beyond 133~166, at the most 200 MHz."

After this you lost me. 133~166 what? 200MHz for what? To which post were you replying? We're already well beyond 200MHz for processor, FSB, etc.

What exactly does this have to do with the clockspeed the P4 core will be able to attain?

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[girish]

hey thats real Mhz I am talking about, which is still 133 (what you call 266 MHz) and P4 still works with only 100 MHz (thats quad pumped - dual channel 400 MHz) and which might go to 533 (133*4) in near future. BUT the board will still work at 133 MHz. thats what i was emphasising on - the real physical frequency.

FSBs greater than 150~166 could be achieved, but 200 MHz is difficult though not impossible. there are physical limitations which you might know of. there are similar limitations on silicon also, although on much smaller scale that allows us to run it higher than 166~200 Mhz. we are already near that *scale* and it would be difficult to drive silicon any faster.

and if we cannot drive silicon any faster, how will any processor run that fast???

still i insist that whatever 10 GHz we are talking about, cannot be reached on current technologies and fabrication processes ON silicon. what 10 GHz we will be seeing in future would be effective speed of 10 GHz, when it would actually be running at 3.33~5 GHz. thats it.

hope you guys agree with me

girish

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>

 

[TheAntipop]

well clock speed is all relative really. i mean an athlon at 1.33 ghz runs faster than a 1.7ghz p4, its the same type of equivalent speeds you are refering to.

and raystonn, while none of us are denying that a p4 can get to 10ghz, we are merely questioning the pratical relevance of its applications versus other alternatives. and while i do agree that a 10ghz chip would be cool, i think were getting close to the period when the latest and greatest isnt the fastest clock, but the best designed.

<A HREF="http://www.512productions.com/lobstermagnet/" target="_new">Hyakugojyuuichi!!</A>

 

[Raystonn]

"the latest and greatest isnt the fastest clock, but the best designed"

I agree there. That's why clockspeed isn't relevant. Those who complain about a P4 1.7 being beat by an Athlon 1.33 in 'certain' benchmarks really aren't saying anything substantial. Performance is clockspeed x IPC.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[Raystonn]

"the real physical frequency"

That's not of any real consequence. Performance = clockspeed x operations per clock (IPC). In fact, the industry is probably going to move to clockless systems in the future.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[girish]

true, clock speed is relative. what matters is performance - and nobody will deny (even fugger and meltdown i guess) that Athlon 1.33 GHz does outperform P4 1.4 Ghz.

10 GHz on your desktop is sure cool, but I choose to deviate from your point about practical relevance of this speed. such superpower will have its applications, and todays's supercomputers may become tomorows desktops. then this GHz will power tommorows supercomputers.

there is still a lot of awe and craze of MHz, but really it is too much relative. SOME P4 SSE2 optimised photoshop FILTERS run 70% faster, so could e3DNow! optimised ones.

of course, I do agree that clock is reaching its practical limits, so better designs will win.

now about P4 getting to 10 GHz - i guess silicon wont go much beyond 4~5 GHz in near future, so 10 GHz P4 seems too far fetched. it would be nice if i am proved wrong, but i doubt how long it will take. P4 core (the netburst architecture, whatever it means) is said to be designed for 10 GHz that large number of stages in the instruction pipes allow for faster clocks at multi GHz. but is this potential Ghz really achieviable? giving it a serious thought answers this as no. there are other organiastional improvements that will allow the *relative* performance of the processor to reach 10+ GHz. i dont have to repeat my last posts i guess.

and it seems almost everybody is leaving AMD out of this discussion. AMD already has 1.5 & 1.7 GHz Palominos slated for release next quarter, which will actually compete 2+ GHz P4s.

girish

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>

 

[Guest]

"Intel Corporation researchers have achieved a significant breakthrough by building the world's smallest and fastest CMOS transistor. This breakthrough will allow Intel within the next five to 10 years to build microprocessors containing more than 400 million transistors, running at 10 gigahertz (10 billion cycles per second) and operating at less than one volt."

What is the point of this quote ? I'd be suprised if VIA would not be able to launch a 10 Ghz Cyrix VII in 5 to ten years.

You should be smart enough to understand such a press release is nothing but PR to keep stock holders happy.. Not saying its not true, just that it doesnt matter.. We've come a long way since 8 Mhz 286 cpu's, and no one is expecting this evolution to stop real soon. You dont need a crystal ball to predict faster cpu's over the next 5 to 10 years. When More's law will stop is anybody's guess.

Raystonn, though you may be polite, informed and not entirely stupid; to me, you're nothing but an intel troll.. Either you have some serious intel stocks, or your mother works there.. ;-)

Why are you so feverishly (hmmm.. is that correct English ?) promoting intel on this forum ? You really sound like a smart Intel PR guy.. Tell me, if AMD would issue such a (pointless) press release, would you start a new topic saying "AMD has the technology to produce 20 Ghz cpu's in 10-15 years ? " (I'm guessing the article is based upon a press release, Im too lazy to read the article)

 

[girish]

the real frequency does matter a lot.
its the same frequency you base your 200, 266 MHz and processor internal speed on.

yes, we might move to clockless systems some time later - and that means either the clock speed will not have much relevence because it will already be too hight for most applications, or there would be some other technology that wont work on clocks.

btw performance is not alone IPC, its how every cycle is managed. IPC somewhat took over from IPS (Instructions per second) and became the topic of choice. while i think the latter is more relevent, IPC just gives a hint about the internal architechture of the processor and not its performance. IPC indicates how many instructions per clock are executed per clock cycle, and it really means one, two or three! and that is heavily dependent on the code itself!!! and managing this code sequencing well is the key to performance, not just IPC.

todays processors have multi-stage multiple execution pipelines, and there is one instruction in each stage of each pipeline at some stage of execution. for example, P4 has a twenty stage hyperpipeline that handles as many as 20 instructions one in each stage at any given clock cycle, it does not mean it is executing all of them at once.
then there are cache misses, branch miss issues, data dependency issues that are harmful for the optimal performance of the processor units.

hence the architecture of he processor is more important than MHz alone which it still very important though.

girish

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>

 

[TheAntipop]

ok, were all converging on the same thing. most of us are arguing the same point from different angles at this point, so its pretty much useless to go on. maybe you guys feel differently but im just tryin to help us all out by making sure we all dont sound dumb by arguing the same side of an argument.

<A HREF="http://www.512productions.com/lobstermagnet/" target="_new">Hyakugojyuuichi!!</A>

 

[Raystonn]

"What is the point of this quote?"

To refute someone else's claim that going to 10GHz is not possible with current technology. That is all. Read the thread in context please, and keep the flaming out of it.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[Raystonn]

As I've said before, Performance = (clock speed) x (average instructions per clock). That computes to 'average instructions per second' which is the real figure indicating how fast a CPU can run an application. Also note the word 'average'. IPS depends greatly on what instructions are used in the application as well as other optimizations allowing for parallel execution. Having a low average IPC figure is not a "Bad Thing" (tm) if you can make up for it by allowing clock speeds to soar. Having a low clock speed is also not a "Bad Thing" if you can make up for it by having a high average IPC figure. (Just look at the G4.)

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[TheAntipop]

quit double-posting, just edit your previous so you dont look like youre trying to up your post count, i hate when people do that.

<A HREF="http://www.512productions.com/lobstermagnet/" target="_new">Hyakugojyuuichi!!</A>

 

[Raystonn]

I haven't made a single double-post in this forum. I have no idea what you are talking about.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[TheAntipop]

the two posts above mine, timed 5 minutes apart? i would call that a double-post.

<A HREF="http://www.512productions.com/lobstermagnet/" target="_new">Hyakugojyuuichi!!</A>

 

[peteb]

Even when responding to two different people/threads? - I think Raystonn is perfectly justified to make separate responses to separate people/posts.

If you are looking in exploded view, they seem overkill, but if you use the threaded view - it makes sense...

Raystonn - I don't mean to fight your battles for you either. It just really bugs _me_ when people get jittery about upping post counts.

If you want to get funny about post count - go and have a go at those in the other/other forum posting all sorts of meaningless (but often fun) stuff. Who cares about post count? Some of the highest counts on here are from people who's computing opinion i hold in very low esteem. (That is the exception, not the rule - before everyone jumps on me!)

-* This Space For Rent *-
email for application details<P ID="edit"><FONT SIZE=-1><EM>Edited by peteb on 05/01/01 02:24 PM.</EM></FONT></P>

 

[girish]

and what are those angles?

the basic argument of this thread is whether P4 will go on to 10 GHz or not, based on some statement by an intel marketing official and a news about some new fabrication technology. There are people saying it will, and there are people saying it wont.

in fact, i guess the whole discussion has come to a single point that 10 GHz is achieviable, maybe not with P4, maybe not with real clock, maybe not with silicon, maybe not in 8~10 years!

there is a divide among opinions that might continue till we actually see 10G machines.

girish

<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>

 

[Raystonn]

Thank you. I was about to say the exact same thing until I read your post. I wouldn't want to repeat what you said. He might accuse me of double posting again.

-Raystonn

= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[peteb]

Raystonn,

You are welcome - and I don't care if I get accused of double posting!! ;o)

-* This Space For Rent *-
email for application details