Intel's brilliant marketing..
Intel has really been brilliant at marketing their processors, for better or worse the've gotten the majority of the population to believe that the ghz. number is the only way of determining the speed of the processor.
I believe for Apple to really increase it's market share it will have to develop future Powermacs around Intel processors.
By the time the first IBM processor Powermac's might appear Intel will probably be at 4Ghz. or higher.
Powermac's topping out at 2.5gig will still look underpowered to the vast majority of people, even if in fact it is not true.
Would it not be in Apples best interest to employ Intel processors for future Powermac's?
I believe for Apple to really increase it's market share it will have to develop future Powermacs around Intel processors.
By the time the first IBM processor Powermac's might appear Intel will probably be at 4Ghz. or higher.
Powermac's topping out at 2.5gig will still look underpowered to the vast majority of people, even if in fact it is not true.
Would it not be in Apples best interest to employ Intel processors for future Powermac's?
Comments
So if Apple went to 2.5 GHz within six months, that wouldn't be too bad at all. Intel might get to 3.2 or 3.4 GHz by then, which is still a higher rating but not THAT much higher. And it's above AMD's current 2.1 GHz chips.
Intel has aggressively 'backed-off' its runaway train of jacking up the frequency. Even they realize its fruitless. It was only the intro of the P4 and the ease of it scaling up so early in its lifespan that contributed to runaway clock speeds. The p4 was designed to clock high and gave up a whole lot in efficiency, hence AMD holding the performance crown for so long with much lower frequencies. The next GHz will not come so easy for intel, its not an easy linear scale of frequency increase.
Thats a large part of why PC sales are in the dumps... people and businesses are not seeing value in another gigahurtz yet. Granted, savvy new PC buyers buy as fast as they can afford in the market they choose (pc, mac, etc). But how many people are dissapointed with how fast they can browse the net, check their email and run office 2000 on a 1-2GHz computer?
Intel may have a 5GHz processor next year but they wont sell it as a computer but as a radiant-space-heater as it wil consume 80 - 120 Watts of power (= big heat).
Consumers are getting more and more frustrated with intel because to keep ramping these clock speeds up which requires them to change sockets like murbot changes macs! Next year, Prescott (P4-II) will require a 775-pin slot. the THIRD slot-type for a P4 already. Essentially the P4 is a throw away platform processor-wise.
[ 03-01-2003: Message edited by: I-bent-my-wookie ]</p>
[ 03-01-2003: Message edited by: MacsRGood4U ]</p>
Intel is and will remain the undisputed champ of 32bit desktop computing. The real opportunity lies in IBM's or AMD's ability to supply faster 64bit machines. That's something IBM (or Apple) and/or AMD can market effectively in the dumb numbers game. 64 bit? that's twice as good as 32 bit, right?
Itanic has been bad. There's no way it's going to get into a consumer desktop near you any time soon, even if they cut the prices drastically. And given the power characteristics it doesn't looklike it's going to go into anything but a bulky tower, and certainly not a laptop or slim desktop. 970's OTOH, are on the cusp of possibility for a laptop. Yes, you laugh, but the 1.2-1.4Ghz PPC 970 numbers are nothing that X86 notebook makers haven't had to deal for the last 2 years. And the point of 970's is that they're cheap.
64bit is where Intel will either keep or lose their crown, and a lot of it depends on M$ -- right now M$ doesn't indicate that it will (ever) support X86-64, that could kill AMD. But IA64 windows doesn't look like it'll ever see wide-spread use, given the way Intel has targetted them. X86-64 actually fits M$'s obsessively backwards compatible (hardware) designs. Who knows. The Wintelon 64 bit world is up in the air. The chip which M$ supports will win, but the product Intel has isn't suited to a proliferation of 64 bit chips, and the company that AMD has isn't likely to get M$ support.
Only Apple-IBM have the right combination of a suitable CPU and a tightly controlled platform to make a total, relatively painless switch to 64 bits over the next 2-3 years. If Apple doesn't get stupid on price (THAT"S A BIG BIG BIG IF!) ... If they don't get stupid on price, this could be the last chance they ever see to get marketshare back up in the 5-10% range. Exciting times ahead.
I don't think that's marketing. That's Human Nature. Good marketing "appeals" to Human Nature and that is what Intel has done. It's no different than the myriads of Maccies who think 64bit is going to be faster than 32bit by default.
[quote] By the time the first IBM processor Powermac's might appear Intel will probably be at 4Ghz. or higher.
<hr></blockquote>
Only if you expect the first PPC 970 computers to ship H2 2004
<a href="http://www.anandtech.com/showdoc.html?i=1752"; target="_blank">http://www.anandtech.com/showdoc.html?i=1752</a>;
Prescott will probably debut at 3.2Ghz and Intel doesn't look close to hitting 4Ghz until late 2004. As Wookie said Intel is backing off of the Mhz Myth itself.
[quote]Powermac's topping out at 2.5gig will still look underpowered to the vast majority of people, even if in fact it is not true.
Would it not be in Apples best interest to employ Intel processors for future Powermac's?
<hr></blockquote>
Consumers don't care about numbers. Apple Marketing's job is to Hype the 970 like the an Apple Computer has never been hyped before. You don't wan't people focusing on numbers. Does a typical consumer ask for an 80GB PVR or do they say "I want to buy a Tivo"
Apple would do well to come up with a nifty name for these ala "Powermac xxxxx) something catchy.
They have the stores and apps to show these things off. It's time for Apple to make a serious thrust into moving boxes like they haven't done in years.
Intel is NOT the answer
This is Apple's last real chance to get back to 5-10% of the market, possibly as high as 10-13% if you count a longer service life and 9 to OSX switchers. That's it. The 64 bit wintelon path is in serious disarray, if Apple doesn't regain ground in the next 2 years they never will.
In case I haven't stressed it enough on these boards, it is absolutely imperative that Apple pretend to be someone other than Apple when they price these things. PRICE PRICE PRICE!!! It MUST come down early to give an immediate boost to demand and they must spend whatever it takes to keep feeding that demand. Intel will not stray long with a hole in their lineup. Nobody though pentiums would catch the PPC, and look how that turned out. DO NOT underestimate Intels ability to circle the wagons, load the shot guns and come out blasting. If Apple doesn't make its impact immediately, they won't make it. Even AMD failed to keep the momentuum of the first athlon launch.
Like I said, interesting times ahead.
Same with the desktops. I want a new G4. Apple's doing just fine. I'd like to see the next incarnation NOT be called G5 though...
<strong>More generally, people want a "PowerBook".
</strong><hr></blockquote>
me too! mmmmm powerbooks.
The next GHz will not come so easy for intel, its not an easy linear scale of frequency increase.
<hr></blockquote>
Note that Intel has increased the pipeline depth of Prescott by several additional stages which presumably include several additional drive stages to allow for operation up to 5GHz. Intel will continue to ramp their product in frequency as high as necessary to repel competiting products from AMD.
[quote]
Thats a large part of why PC sales are in the dumps... people and businesses are not seeing value in another gigahurtz yet.
<hr></blockquote>
No, PC sales are in the dumps because we are in the midst of a huge economic downturn which has consumer confidence at record low levels and businesses severely slashing their IT budgets as a means of cost cutting. Most companies will upgrade their corporate PCs in approximately 3 year cycles. They are looking at x number of new PCs at a price point of y, not the minimum amount of GHz they can get away with.
[quote]
Intel may have a 5GHz processor next year but they wont sell it as a computer but as a radiant-space-heater as it wil consume 80 - 120 Watts of power (= big heat).
<hr></blockquote>
No, because it will be built on Intel's 90nm process which incorporates some very interesting technology (Strained Si and low k) and operate at a vastly lower Vcc.
[quote]Next year, Prescott (P4-II) will require a 775-pin slot. the THIRD slot-type for a P4 already. Essentially the P4 is a throw away platform processor-wise.
<hr></blockquote>
Tejas, due in 2005 and likely to be dubbed Pentium 5, will require the new socket. Not Prescott.
And I'm no fan of Intel for the record
I don't care what runs my software, it could be a huge mother of a die SP machine or a quad of lithe efficient CPU's, the only metric that matters is performance per dollar in a given system. We'll have to see what 970's allow for, versus P4-5-6...
<strong>Just a note...
Intel has aggressively 'backed-off' its runaway train of jacking up the frequency.</strong><hr></blockquote>
Er...nobody told you Intel's lengthening the P4's execution pipeline? Intel plans on getting the P4 to 5 GHz by 2004 and ~10 GHz by 2006, barring any funny stuff.
EDIT: Damned Eskimo!
[ 03-01-2003: Message edited by: Eugene ]</p>
There's an opportunity for Apple/IBM here, but they absolutely MUST compete on price. Don't I always mention that?
Eskimo: I'm sure this is old news to you, but I'll say it anyway. If I remember correctly, IBM's anticipated Vcc for it's own 90nm process is .75W. Since the properties of Silicon as a semiconductor dictate that the Threshold Voltage is 0.7V normally, I don't imagine that the fab size is going to have such a huge impact on power dissipation. Unless someone can doctor the Si to reliably achieve Vt = 0.5V, which I have seen in the lab before in individual samples, further efforts to decrease feature size will be more for shrinking the dies than anything else. A GaAs process might help.
Lastly, since power dissipation increases with higher clocks, maybe the best solution IS to produce lower clocked SIMD architectures. . . Power dissipation per square inch on processors is approaching that of the sun, and it's already beyond Fission reactors in some cases. cooling will only become more of an issue.
[ 03-02-2003: Message edited by: Splinemodel ]</p>
What!? <img src="graemlins/surprised.gif" border="0" alt="[surprised]" />
<strong>There's an opportunity for Apple/IBM here, but they absolutely MUST compete on price. Don't I always mention that?</strong><hr></blockquote>
What? Nah... I think that's the FIRST time you've ever mentioned having to be competitive on price.
Since it is Intel's plan to release IA64 as a successor to x86, I don't think we'll see Intel continue to use clock speed as a be-all-end-all. <hr></blockquote>
Not until circa 2010 though, until then they are still backing x86.
[quote]
Eskimo: I'm sure this is old news to you, but I'll say it anyway. If I remember correctly, IBM's anticipated Vcc for it's own 90nm process is .75W.
<hr></blockquote>
I think you meant to write V since it wouldn't make much sence for a Vcc to be expressed in watts. .75V is way to low for a Vcc even at 90nm. At 180nm we were near 1.8V and at 130nm we moved nearer to 1.5V. I believe Intel has stated a Vcc of 1.2V for their 90nm process.
I think you might be getting threshold voltage confused with Vcc. Vt is the potential difference between the gate and substrate at which an inversion channel can form to allow current flow between source drain. Technically there are multiple regions of inversion each with their own Vt but the rigors of that can be left to those interested in advanced solid state device physics. Vcc rather is the supply rail voltage to which logic can swing between. It's useful as it represents the maximum potential difference that can be realized for drain to source and gate to source voltages to characterize on-state characteristics.
[quote]
Since the properties of Silicon as a semiconductor dictate that the Threshold Voltage is 0.7V normally,
<hr></blockquote>
0.7 is the normal built in potential of a pn junction under non degenerate doping conditions which is applicable for older gate technologies. But with the incorporation of germanium and nitrogen into the polycrystalline gates or the replacement of polysilicon with a metal gate you are able to engineer the workfunction of the FET to allow for thresholds approaching midgap voltages (0.56V for silicon, slightly lower for SiGe).
[quote]
I don't imagine that the fab size is going to have such a huge impact on power dissipation.
<hr></blockquote>
If we assume that a move to a smaller technology node allows us to reduce our Vcc we realize a reduction in power consumption as a square law relationship to the voltage. Considering that increased frequency only has a linear relationship to power consumption we can see that even modest voltage reductions will allow for significant power reductions.
Of course this is often a balancing act as the voltage is not always scaled in direct proportion to the gate length to allow for increased current drive. Of course the higher your current drive in the channel the faster you can overcome the capacitive coupling of the transistor to the substrate and junctions in order to switch it. The designers will utilize this increased drive to clock their frequency up higher provided they ar not limited by RC overhead in the interconnects. Thus they eat up a lot of their power savings with increased frequency. It's a design choice.
But the thing about the sun is true.