PDA

View Full Version : Intel clockspeed ramp slowing......


gaomay
Nov 29, 2002, 08:47 AM
Looks like Intel's rate of increasing clockspeed will be much slower in 2003 than in 2002. Seems that with the 3.06Ghz CPU part running so hot they are reluctant to increase the clockspeed much more until their .09 micron process comes onstream at the end of 2003.
This might give IBM/Moto a chance to reduce the gap somewhat.

http://www.anandtech.com/showdoc.html?i=1752

Sun Baked
Nov 29, 2002, 09:05 AM
The heat is the stick that can bring Intel to a screeching halt - the MHz carrot is getting a little moldy.

Processors already are at a fast enough clock speed (MHz wise) -- lowering heat/power, increasing the chips I/O bandwidth, and increasing benchmark scores at the current clock speeds would be a dandy idea.

But since Intel has backed themselves into their own MHz trap, they're sort of stuck.

gaomay
Nov 29, 2002, 09:32 AM
A classic case of the corporation that cried Mhz.....

Hoist by their own petard........

barkmonster
Nov 29, 2002, 10:09 AM
the MHz carrot is getting a little moldy.

:D

I knew this would happen, inefficient design, fixated on how fast it spins compared with how it performs. The P4 was crap when it came out, by the time it got to 2Ghz and moved up to the northwood core it started to look like a good performer, now it's at 3Ghz with Hyperthreading and it's hit it's peak. It's so funny.

agreenster
Nov 29, 2002, 11:03 AM
Oh yeah. Here comes the fall of Intel. This must mean in 2 years, Apple will be the dominant home computer again. Intel might as well close up shop, since they obviously have no money for r+d to push into the next stage of processing.

Yeah, and the G4 is faster than a P4.

When is everyone going to wake up?
:rolleyes:

Hawthorne
Nov 29, 2002, 12:28 PM
Originally posted by agreenster
Oh yeah. Here comes the fall of Intel. This must mean in 2 years, Apple will be the dominant home computer again. Intel might as well close up shop, since they obviously have no money for r+d to push into the next stage of processing.

Yeah, and the G4 is faster than a P4.

When is everyone going to wake up?
:rolleyes:

Intel's next stage of chips, the Itanium, has significantly lower clock speeds than the P4. Even you'll agree that Intel's backed themselves into a Mhz=Power corner, and even if the Itanium comes out with enough processing power to lift the Titanic, (compared to the P4, which could boil the ocean water away from the wreck :) ), they'll have a fun time explaining to Joe Consumer why this chip is more powerful than the 2.5 ghz P4 (or so) it'll be replacing.

In the mean time, Apple will have a nice little chip from this teeny little company called IBM :). And since they got off the Mhz treadmill long ago, will have a marketing advantage they can exploit right away.

agreenster
Nov 29, 2002, 01:24 PM
I sure as hell HOPE so. But its been years since the Mac has been faster than an equivalent PC.

Hopefully, IBM will be ready to mass-produce in time to pull Apple up out of its speed slump. According to the rumors on this site, IBM wont be ready till late next year or later.......thats a long time to wait..........the G4 is at the end of its life too you know.

NavyIntel007
Nov 29, 2002, 01:29 PM
Wait, so there's a such thing as Intel R & D?

ddtlm
Nov 29, 2002, 03:51 PM
Although I'm sure that gloating is tempting to all the zealots here, at this time no one knows if the PPC-970 will even be available before Intel has chips fabbed on 90nm tech available. Not to mention that Intel can release a few more P4's before 90nm, and not to mention that they are talking about an 800mhz FSB in 2003, virtually gaurenteed to arrive before the PPC-970 does with it's 50% core-speed FSB.

barkmonster:

now it's at 3Ghz with Hyperthreading and it's hit it's peak. It's so funny.
It is getting near it's peak, but it isn't there yet. Big difference.

agreenster:

Oh yeah. Here comes the fall of Intel.
Huh?

Hawthorne:

Intel's next stage of chips, the Itanium, has significantly lower clock speeds than the P4. Even you'll agree that Intel's backed themselves into a Mhz=Power corner
Itaniums and Pentiums have very different target markets and noone is having any trouble with the clock speed issue, despite what all the Intel-haters would like to happen.

NavyIntel007:

Wait, so there's a such thing as Intel R & D?
You stupid troll. Intel's R&D is the best in the industry. Even IBM cannot equal it, although they aren't as far behind as most companies are.

rice_web
Nov 29, 2002, 04:06 PM
First of all, the P4 is nearing the end of its life, but the P5 is just warming up. The P5 will boast such things as increased I/O and lower power consumption, and will be premiering late next year at 4GHz, 667MHz system bus, and 1MB L2 Cache (still rumored--it might be better).

You see, Intel can still play a number's game, even if they don't play on the overall MHz. Intel will have a dominant lead in the system bus and L2 cache, while still maintaining a lead in CPU MHz.

I'd like to see the 970 compete in any category other than system bus and total I/O throughput.

Hawthorne
Nov 29, 2002, 04:12 PM
Originally posted by ddtlm
Although I'm sure that gloating is tempting to all the zealots here, at this time no one knows if the PPC-970 will even be available before Intel has chips fabbed on 90nm tech available. Not to mention that Intel can release a few more P4's before 90nm, and not to mention that they are talking about an 800mhz FSB in 2003, virtually gaurenteed to arrive before the PPC-970 does with it's 50% core-speed FSB.

barkmonster:


It is getting near it's peak, but it isn't there yet. Big difference.

agreenster:


Huh?


NavyIntel007:


You stupid troll. Intel's R&D is the best in the industry. Even IBM cannot equal it, although they aren't as far behind as most companies are.

I think the batteries in your sarcasm detector need replacing. :)

ddtlm
Nov 29, 2002, 05:19 PM
Hawthorne:

Or the batteries in my tolerator. :)

ddtlm
Nov 29, 2002, 05:24 PM
rice_web:

Actually it looks like the FSB will be 800mhz on P4's next year, based on the (rather reliable) rumors on the PC side, so the the P5's will (if the rumors are right) certainly start off with the 800mhz FSB as well. Word is that Intel is prepping the faster bus because the 130nm P4's aren't going to scale enough to deal with the Hammer otherwise.

solvs
Nov 29, 2002, 06:25 PM
Doesn't the 970 have a 900 MHz FSB (~450 DDR)? I'm guessing the P4/5 667 and 800 FSBs will be Quad pumped 167 and 200. Anyone know why they still do that when most of their new chipsets support DDR over RDRam?

Catfish_Man
Nov 29, 2002, 08:30 PM
Originally posted by solvs
Doesn't the 970 have a 900 MHz FSB (~450 DDR)? I'm guessing the P4/5 667 and 800 FSBs will be Quad pumped 167 and 200. Anyone know why they still do that when most of their new chipsets support DDR over RDRam? ...but because of the way the bus protocol works, it has an effective speed of 400MHz DDR (800). So it's ending up looking like the 970 and the P4 (or P5 as it may be called by then) are going to be almost identical in speed. Same bus (800Mhz for both), half the clock frequency (1.8GHz vs. about 3.6GHz) but double the IPC (about 4 vs. about 2), roughly equivalent vector units (the 970 uses the old G4's Altivec layout which is a bit less beefy than the current one). The difference is that the P4 will be on a .09 micron manufacturing process, allowing for more L2 cache (512k vs. 1MB). This should give the P4 a slight lead, until a .09 micron 970 comes out (which would have enough space for dual cores, or on chip memory controllers, or big caches, or any number of other cool things. Plus, it will be cheaper, faster, and run cooler). Overall, I think it's going to be case of the winner being whoever scales the chip fastest in the future. Normally, this would be Intel, but with AMD less in the picture, the P7 core aging, and IBM's huge budget, I'm not so sure. If Apple goes with dual 970s it should be the crap out of the P4 (although it'll cost a bundle).

<edit> to answer the original question, they're planning on using dual channel DDR400 (which is also what the 970 will probably use). That means their ram is 200x2x2, same as their bus. </edit>

ddtlm
Nov 29, 2002, 08:58 PM
solvs:

I'm guessing the P4/5 667 and 800 FSBs will be Quad pumped 167 and 200. Anyone know why they still do that when most of their new chipsets support DDR over RDRam?
I am unaware of any reason why the choice of DDR vs RDRAM makes much difference as far as a quad-pumped FSB being good.

Catfish_Man:

To what you've said I'd like to add that since the PPC-970's FSB operates at 50% of core clock (and then looses 1/9th of what's left to overhead), PPC-970's above 1.8ghz will have more theoretical bandwidth, and those below 1.8ghz will have less theoretical bandwidth than a P4/P5 with a 800mhz FSB. The two FSB's are implemented very differently though, so I don't think anyone here is qualified to say which one is better.

the 970 uses the old G4's Altivec layout which is a bit less beefy than the current one
Yeah I was wondering if that's where people where getting the "two AltiVec units" from. :) Oh well, I can give up a little AltiVec power for some more general-purpose power.

and IBM's huge budget, I'm not so sure.
But is it huge compared to Intel's?

If Apple goes with dual 970s it should be the crap out of the P4 (although it'll cost a bundle).
How about a dual P4, AKA Xeon?

MrMacMan
Nov 29, 2002, 09:32 PM
Um, unless you are like Uber-Hardcore gamer or are running a Super Server you don't get Dual P4's or Xeons.
Microsoft tech can't handle the 2ed processor as well as lets say OS X. :D
And If the 970 Came out not it would kick, but it didn't so we will have to wait...

ddtlm
Nov 29, 2002, 09:59 PM
MrMacman:

Microsoft tech can't handle the 2ed processor as well as lets say OS X.
Being the anti-zealot-zealot that I am, I'm going to think dirty thoughts about you until you can provide support for this statement of yours, which on the surface looks like just more of the usual FUD-spewing and MS-bashing that is so popular on Mac hangouts. In my own experience OSX is not especially good at managing two CPUs, in fact in my experience with 2-CPU-OSX, it stutters at regular intervals in virtually every application where a steady framerate (or the lack thereof) is observable.

rice_web
Nov 29, 2002, 11:20 PM
Dual-CPUs work EXCELLENTLY on a PC. My father runs three dual-CPU workstations at his office, and I can personally say that the performance gain from the extra CPU is all-too-noticeable. Granted, OS X might have a slight upper-hand in dual-processing configurations, but who's going to win: Dual 1.25GHz G4 or Dual P4 Xeon at 2.6GHz? Which is going to cost less?

(BTW, you wouldn't buy a dual-processor rig for gaming on a PC, maybe two or three games will actually take advantage of a second processor)

Oh, and it is 800MHz system bus? Wow, I thought those were only optimistic rumors. In this case, I cannot see how a 970 configuration could beat a P5 configuration in any test, especially with an Altivec unit that is less potent. Granted, the move to 64-bit processing allows some information to travel faster and for memory allotments to increase, but I do not think we'll see enough of a boost to beat the P5 of 2003.

In either case, I think comparing the G4 to the P4 and the 970 to the P5 is irrelevant. If you can do your work on a G4 450, consider the 970 a God-sent item. If you've been working with a P2 450, the P5 at 4GHz is a fantasy that can be fulfilled. I don't see why we compare Apples and Oranges, whichever system you PREFER, you should buy. Only worry about speed if it's actually an issue. For example, in my house are two Win2k machines at 866MHz and a single iMac at 400MHz. I use the Win2K machines for my web design and programming, while my family uses the iMac for its simplicity and software selection (e.g. iMovie, iTunes, iPhoto). Speed is hardly a part of the equation for my family, so they've clung to the iMac; I need speed, so I've hung on to my relatively quick P3 system.

EDIT: Some quick links
- http://www.geek.com/procspec/intel/prescott.htm
Possible P5 with SSE3..... (drools)

- http://www.geek.com/procspec/ibm/power4desktop.htm
The first chip in a long time to make a PowerMac worthwhile.... (drools)

fatalerror101
Nov 30, 2002, 01:09 AM
Hey i was searching the web and found this web site

Anyone think it is reliable?HERE (http://www.arstechnica.com/wankerdesk/3q02/powerpc.html)




PowerBook G4 1GHZ

ddtlm
Nov 30, 2002, 01:11 AM
rice_web:

I wouldn't recommend being quite so pessimistic about the PPC-970 just yet. :) 2 AltiVec functional units or 4, it should still mop the floor with any G4 ever made in any code ever made (pretty much). Going to 4 functional units was nice and all, but pretty much everyone is familiar with the problems the 7-stage 4-unit G4's had against the 4-stage 2-unit G4's until the clock speed difference was > 50%. It was like P4 vs P3, except in PPC land. I'm aware that observation doesn't prove a whole lot, but it provides some reasoning to back up my lack of concern over the "missing" 2 functional units.

In other news, I am going to be very interested to see how the battle of the FSB's plays out. One is pretty simple quad-pumped bi-directional and 64-bits wide, the other is wizbang dual 32-bit unidirectional and packet based. Which will have lower latency? Which will be better in common "streaming" apps? Not a lot of people know the answer to that, just yet.

50% of core clock could be very important after the PPC-970 scales a bit. Of course Intel will not sit down on the job...

ddtlm
Nov 30, 2002, 01:14 AM
fatalerror101:

Yeah, Arstechnica is a well-known webiste with an active Mac sub-community. That article is perfectly reliable... but it's also been discussed before at this site. Note: certain things in that article seem a bit out of date now.

Sun Baked
Nov 30, 2002, 03:11 AM
fatalerror101:

The current arstechnia thread about new processors is here (http://arstechnica.infopop.net/OpenTopic/page?a=tpc&s=50009562&f=8300945231&m=3470943335), of course it's around 20 pages long right now - so it's a lot to look at.

BenderBot1138
Nov 30, 2002, 04:43 AM
The Megahertz Myth (http://www.apple.com/g4/myth/) can't be raising it's ugly head here again???

All the world’s a stage,
And all the men and women merely players;
They have their exits and their entrances,
And one man in his time plays many parts.

:cool:

rice_web
Nov 30, 2002, 09:37 AM
I'm not being pessimistic about the 970; I think it's a great chip and a significant advance over the current batch of G4s. However, the Pentium V will likely still beat the 970.

The true test will come in from Altivec production. If the 800MHz system bus (900 sans the overhead) that the 970 can finally do Altivec justice, we might finally have an Intel killer, but I wouldn't hold my breath (although, with the ability for more memory, Final Cut Pro could become very fun....).

syco
Nov 30, 2002, 12:25 PM
Sorry if this has all been said before, I didn't have time to read the whole thread.

But still. Hold on a darn minute.

This isn't going to be the fall of Intel. They've been slowing down since the 2GHz mark. I still find it funny that it took us 25 years to go from 8 MHz to 1 GHz and then as many weeks to go from 1 GHz to 2 GHz.

If HyperThreading is as useful as they claim it is (which I doubt), then ads will probably plug HyperThreading more than clock speed, especially if 3.06 GHz is as high as they're going to go. It probably won't become much cheaper, just maybe a little cooler, because I'm going to guess that their R&D is probably going to the .09 micron chips.

How's AMD doing? From what I remember, the AMD chips tend to lag behind Intel by about a factor of 50-150 Mhz, so this should allow them to catch up a bit.

Developers will now have a fixed target to shoot for, instead of anticipating what everyone'll have then.

Of course, there is the whole best-case scenario, when Apple does reclaim a significant market share. I can almost guarantee that some people will find Apples attractive over the course of the next year (especially if Steve Jobs delivers on his promise of 2003 being Apple's best year ever; great timing!), but it won't make Apple #1.

Sorry. I love Apple more than anything, and I'd love to see them be successful, but I just can't make my vision cloudy to the truth.

ddtlm
Nov 30, 2002, 04:05 PM
syco:

If HyperThreading is as useful as they claim it is (which I doubt)
The thing about hyperthreading is that there appears to be no downside. In all the reviews I read for the 3.06ghz P4, there were very few things that got worse with hyperthreading. Of course a lot things gain nothing... but the overall effect is positive.

It probably won't become much cheaper, just maybe a little cooler, because I'm going to guess that their R&D is probably going to the .09 micron chips.
The rabid people on these forums would love to believe that the current P4 cannot scale beyond 3ghz, which is not true. Reviewers have overclocked it to 3.6 without doing anything tricky, so Intel has room for a few more speed grades. Also, when Intel goes to 90nm production, that will give them more headroom (just like every die shrink does).

Of course, there is the whole best-case scenario, when Apple does reclaim a significant market share.
I can't see this happening. Apple may get a little bigger, but price and familiarity are huge motivators that will keep most people on their Sams Club cheapo PC's.

MrMacMan
Nov 30, 2002, 08:11 PM
Did someone say there was a P5 ? :confused:
I haven't seen any articles...
All I've heard about was that if IBM 970 were to come out now it would kick the P4.
In 2 years... well we will have to see. Apple has a decent OS and decent for what Teens want, gaming.
So they should hold on to Some market share. :rolleyes:
I just hope we don't go into where Linux went. :(

ddtlm
Nov 30, 2002, 08:20 PM
MrMacman:

Pentium 5 is the likely name of the 90nm chip that replaced the P4 in late 2003. It will be a lot like the P4, but is expected to sport the usual ever higher clockspeed, 1MB of L2 on die and presumably various other improvements that Intel is keeping a lid on. SSE3 has been rumored.

All I've heard about was that if IBM 970 were to come out now it would kick the P4.
Outside of the Apple-worshipping circles the PPC-970 was never really expected to defeat Intel in any particular important area. The PPC-970 is a lot better than a G4, and will be a much better compeditor to the P4/P5 in any case. No longer will zealots have to make fools of themselves when they tirelessly try to defend the performance of Macs.

Apple has a decent OS and decent for what Teens want, gaming.
Teens? Teens are only a segment of the gamer's market.

So they should hold on to Some market share.
Because of teen gamers? I fail to see how they matter very much.

I just hope we don't go into where Linux went.
Linux is doing just fine... I don't know what you think happened to it.

MrMacMan
Nov 30, 2002, 08:49 PM
You sortof blended by comments together...
Many who lead the way into the computing world are Teens and younger.
As I was saying that users of Linux have been ignored and many Linux users have to either:
1 leech of others or 2 Devolop software for their platform. Many like the idea of them helping fellow people out, I on the other hand like that normal devolpers can make stuff for mac's without normal people porting software.
Back to other comments...
I'm try to not fall under the Mac Zelot group but you know... and yeah we do try to defend the mac's as much as possible...
As for the P5 it seems basically the same as the P4 with higher clock speed and some L2 which will not be much better than the P4 other than clock speed.
Apple is gonna need some product that really kicks to get themselves a winner.

rice_web
Nov 30, 2002, 09:04 PM
If the P5 houses SSE3, you'll notice a significant performance boost in 3D Apps (such as Lightwave) and games. The move to a 1MB L2 cache and a faster system bus (up to 800MHz rumored) alone would provide a noticeable performance gain, but with SSE3 also factored in, the P5 will likely gain a 10% increase in performance on some applications at the same clock speed.

syco
Nov 30, 2002, 09:46 PM
Originally posted by ddtlm
syco:


The thing about hyperthreading is that there appears to be no downside. In all the reviews I read for the 3.06ghz P4, there were very few things that got worse with hyperthreading. Of course a lot things gain nothing... but the overall effect is positive.
From what I heard, if an applciation isn't optimized for HyperThreading, it will not use it, and it may even cause the system to slow down. It's similar to AltiVec, although an app non-optimized for AltiVec won't slow the system down.

Still, I don't understand how any processor could act as two. Although, I don't know TOTALLY how a computer works anyway, but it still doesn't make sense...but oh well.




The rabid people on these forums would love to believe that the current P4 cannot scale beyond 3ghz, which is not true. Reviewers have overclocked it to 3.6 without doing anything tricky, so Intel has room for a few more speed grades. Also, when Intel goes to 90nm production, that will give them more headroom (just like every die shrink does).
Well, Intel decided that they aren't going to expand it past 3 GHz because of the heat. Unless you put a 3.6 GHz in a NASA wind tunnel, that thing'll slag just about any computer you put in front of it. :D



I can't see this happening. Apple may get a little bigger, but price and familiarity are huge motivators that will keep most people on their Sams Club cheapo PC's.
I know, thats why its the best case scenario. It won't happen. Best case scenario: People think Intel is stupid, arrogant, etc. for freezing the P4 at 3GHz, jump ship to Apple, Apple ends up with approximately 82% of the computer market.

Of course, this won't happen!! Intel is a chip provider for god knows how many computer makers: Sony, Dell, Alienware, etc. They probably sell thousands of chips a day. They will not collapse.


I'm just giving my (probably incorrect) predictions for this event.

ddtlm
Nov 30, 2002, 10:27 PM
syco:

I am not sure where you are getting the idea that Intel is not realeasing any P4's above 3.06ghz. All that was claimed is that it is getting harder, due to heat problems. Overclocking has demonstated that the P4 has some headroom even with conventional cooling systems.

From what I heard, if an applciation isn't optimized for HyperThreading, it will not use it
Hyperthreading allows multiple apps to run at once even if neither is threaded. The first version of hyperthreading that Intel semi-released in Xeons some time ago did have serious side effects, but reviews of the 3.06ghz P4 have clearly shown that Intel has corrected the original problems. Hyperthreading looks like a win-win enhancement, although it is not necessarily a big deal either. Remember folks, just cause Intel did it first does not mean that it is bad, that it is stupid, or that it is evil. I bet that IBM and AMD will implement hyperthreading some day as well, although it will take time.

This review (linked) is fair and informed:

http://www.aceshardware.com/read.jsp?id=50000319

That website is not especially PC-centric, and their webserver is a Sun machine, so they do not worship PCs or anything (if that concens you).

Sun Baked
Nov 30, 2002, 10:45 PM
Originally posted by ddtlm
I am not sure where you are getting the idea that Intel is not realeasing any P4's above 3.06ghz. All that was claimed is that it is getting harder, due to heat problems. Overclocking has demonstated that the P4 has some headroom even with conventional cooling systems.
There is a lot of headroom left until the cooling and power usage problems become a liability in the home PC market.

It's going to be felt in the commercial market far earlier, as companies find that the fast computers are taking a huge bite out of the electric and a/c bills.

But until people actually start complaining/avoiding classes of machines based on power/heat, Intel will continue to crank out ever faster and more efficient space heaters.

I did like the comparison awhile back that said, Intel has a lot of margin left before the Pentium begins generating more heat per cm^3 (or was it cm^2) than a nuclear power plant.

syco
Nov 30, 2002, 10:54 PM
Maybe I didn't make myself clear. I interpreted the initial article to mean that the P4 wouldn't go past 3.06 GHz without going to the .09 micron process.

I read the first page of that link but not the rest of it (its not _that_ important I know about this technology, as I won't be buying a PC :-) so maybe I'm missing something, but it seems like there is a big potential for backups. Maybe thats what Intel worked on, but it sure sounds like it. For instance, lets say you have 100 integer operations. The integer operation 70 is tied in with a floating point operation. Using SMT, you'd have the FPU sitting around taking up cache memory for 69 integer operations. I mean, cue me in if I'm missing something; I'm not one of those "ignorance is bliss" type people, but the logic doesn't seem to work for me.