Intel abandons clock speed chase and drops 4GHz Pentium

[PassiveMan]

QUOTE FROM PCPRO website

"Intel abandons clock speed chase and drops 4GHz Pentium
In another embarrassing admission in a year scattered with them, Intel has announced that it is to cease development of the 4GHz Pentium processor. For now the 3.8GHz part will remain the highest clock speed. Instead, the company says that it will be transferring its engineers to work on the dual core designs demonstrated at the recent Developer's Forum.
The unexpected announcement marks the end of an era for Intel. Ever since the launch of the 4.77MHz 8088 around which the original IBM PC was designed, the company has largely depended on ever increasing clock speeds to boost performance. However, recently Intel and arch-rival AMD have been having greater problems in producing reliable parts that could cope with the amount of heat generated by these clock speeds.

First AMD and now Intel have changed tack away from raw clock speed towards putting more than one processor on a single chip. Early indications show that the dual core strategy boosts performance by up to 55 per cent.

Nevertheless, the announcement will be seen as another blow to Intel's credibility at a time when it has already taken a series of knocks. In July, it announced that the 4GHz part would not hit its shipment dates for the end of the year and put back the launch dated to the spring of 2005. Now, given the increasing technical hurdles, there seems little point in spending huge amounts of money to pursue a strategy which has run its course.

For AMD, this represents a sizable opportunity. For some time, AMD has simply not been able to develop technologies such as 90nm processes fast enough to enable it to match Intel clock speeds in the market and has therefore turned to other strategies such as 64-bit processors and dual core. Now that Intel has abandoned clock speed it finds itself behind in both these areas."

[mathias]

Already heard of it here:

http://www.overclockers.com/articles1123/

and this article is just plain terrible, how can they let this idiot who thinks AMD relies on MHz figures keep his job?

[Straker]

well, he gets the most important thing (indirectly) right: dual core cpu = admission of defeat

poor intel

edit: just read the article, don't think it's too far off the mark, kind of heavy on the rhetoric though, and the first few paragraphs sound google-translated.

i never realized AMD's profit margins and net income were so low; just makes their performance as a company over the last few years even more impressive... Intel's CPUs are making them 30x as much money as AMD and they're generally equivalent at best? wtf?

[Blappo]

http://www.overclockers.com/articles1123/

and this article is just plain terrible

I totally agree, this is a terrible article. Although AMD won't take half the market share anytime soon, they are in a pretty good position. Intel will increase its manufacturing costs by increasing cache, while AMD will be decreasing its costs by going to 90nm. What Intel makes on average on each chip sold will drop by a fair bit if they lose sales in the server market (where markup is very high) and the fact they won't have highend processors to compete with AMD's.

Don't get me wrong, I would hate if AMD captured 75% of the market. The consumers win when there is a competitive market.

[mathias]

http://www.overclockers.com/articles1123/

and this article is just plain terrible

I totally agree, this is a terrible article. Although AMD won't take half the market share anytime soon, they are in a pretty good position.

No, it's the pcpro article I find awfull. I think the AMDroid bashers have a pretty good point. (although I don't always know what they're talking about as I've been paying no attention to hardware developments during the ~700-1800 MHz period when AMD supposedly had the biggest lead)

Don't get me wrong, I would hate if AMD captured 75% of the market. The consumers win when there is a competitive market.

What are you saying? If AMD had 75% percent of the market now, intel would be forced to rush out a desktop dothan, or even a dothan with an integrated memory controller.

[Straker]

er duh, was too tired to notice the first article even, i was referring to the overclockers one too, whoops.

What are you saying? If AMD had 75% percent of the market now, intel would be forced to rush out a desktop dothan, or even a dothan with an integrated memory controller.

or just sell prescotts for $50.

[mathias]

What are you saying? If AMD had 75% percent of the market now, intel would be forced to rush out a desktop dothan, or even a dothan with an integrated memory controller.

or just sell prescotts for $50.

Yay, cheap stove

Okay, not really cheap.

[meglamaniac]

Intel must be getting pretty desperate by now.

Their best performer, the Pentium IV 3.4Ghz Extreme Edition, gets slaughtered by the top end Athlon 64's (never mind the A64FX's) and is very nearly beaten by the Athlon 64 3000+.

If that wasn't bad enough, the price differential is staggering.
A PIV 3.4EE costs £767.15 GBP.
The A64 3000+ costs... erm... £104.99 GBP.

So you'd have to be insane to buy an Intel chip at the moment.
To get very slightly higher performance than an Athlon 64 3000+, you have to spend 7.3 times more than for the AMD part.

Small wonder they've abandonded the Ghz race - their "faster", more serial-orientated processors are being beaten out by AMD's "slower", more parrallel-orientated processors. Intel's design just doesn't cut it.
On top of that, their plans for dual core chips is an embarrassing climbdown from their position a year ago when they were refusing to even consider it - and to add insult to injury, AMD are almost ready to release dual core processors while Intel have yet to design one.

The final stab in the corpse? The Athlon 64 3000+ will easily overclock to 2.4Ghz (the speed of an Athlon64 FX) and still run cooler than a standard P4.

It must suck to be an Intel board member at the moment - yes they may still have greater market share, but that's just because of vendors like Dell who are tied in to make Intel boxes only. Unless they come up with something good fast, Intel are in trouble here.

[Putz]

I agree with everything you've said here, and I agree that it's definitely not a good time to buy an Intel desktop CPU right now, particularly because of the price and heat issues.

However, are you sure about the following?:

[Intel's] "faster", more serial-orientated processors are being beaten out by AMD's "slower", more parrallel-orientated processors.

As I understand it, Intel's CPUs favour code with more mutli-threading and SIMD (SSE, etc.) instructions, thanks to HyperThreading and superior SSEx support. On the flip side, AMDs processors win out in most typical applications, such as office apps and games, because they are very simple and linear in nature, for which the AMD cores are optimized, thanks to short pipelines and the 64's double registers.

[Blappo]

Sorry about the confusion before. I was trying to point out that the best scenario for the consumers would be if Intel and AMD had an equal market share, and had the same reputations with consumers. That way they would be forced to price comparible products at comparible prices.

Next time I'll remember to preview my post before I post nonsense

[Bitter Jitter]

AMD = underdog, yet with a better product, lower price, low power consumption and lower heat.

AMD's problem is there marketing is poor at best, they need to built a strong public image so that joe public go "Let's buy AMD because they are just as good as Intel but not as expensive to buy and run"

Marketing is more important than quality in business.

[silvervarg]

PassiveMan:
Early indications show that the dual core strategy boosts performance by up to 55 per cent.

Where did you find this number?
Is sounds awfully bad, so if it is true something is in desperate need to be fixed. They should be able to reach 80%-90% performance boost if they are well made (at least in theory).

[Tibors]

It's awfully bad because of marketing. That 55% is the increase of performance over the chip before it on the roadmap. They "need" to introduce a new chip every few months. So the first dual cores will run at a lower clock speed. Then they can increase the clock speed over the next few releases, without having to invent something really inventive.

[meglamaniac]

I agree with everything you've said here, and I agree that it's definitely not a good time to buy an Intel desktop CPU right now, particularly because of the price and heat issues.

However, are you sure about the following?:

[Intel's] "faster", more serial-orientated processors are being beaten out by AMD's "slower", more parrallel-orientated processors.

As I understand it, Intel's CPUs favour code with more mutli-threading and SIMD (SSE, etc.) instructions, thanks to HyperThreading and superior SSEx support. On the flip side, AMDs processors win out in most typical applications, such as office apps and games, because they are very simple and linear in nature, for which the AMD cores are optimized, thanks to short pipelines and the 64's double registers.

Sorry my bad for not being clear.
When I say parrallel in relation to AMD i'm refering to the chip architecture, not parrallel computing.
AMD's approach to speed is to use a relatively slow clock speed but process lots of instructions in a single clock cycle.
Intel's approach to speed is to use a high clock speed but process few instructions in a single clock cycle.

Intel's problem now is that they've realised AMD were right all along, because they simply can't keep the heat down on their new processors. They now have to
a) move towards AMD's way of building processors
b) not make themselves look like idiots in the process
which is going to be interesting to see...

[Denorios]

AMD = underdog, yet with a better product, lower price, low power consumption and lower heat.

AMD's problem is there marketing is poor at best, they need to built a strong public image so that joe public go "Let's buy AMD because they are just as good as Intel but not as expensive to buy and run"

Marketing is more important than quality in business.

Amen. It's ridiculous that AMD should have been competing toe-to-toe with Intel for as long as it has, without doing something about its god-awful public image (or lack of one). I bet 90% of average consumers have never even heard of AMD. Admittedly, AMD have never seemed to focus on the consumer market quite as much as Intel, but still, you'd think that now would be the perfect time to make some PR capital out of Intel's difficulties. The "Prescott Survival Kit" stunt barely made the news anywhere outside the IT community, so what was the point? The question has to be asked - is AMD even interested in seriously challenging Intel in the consumer market? Because if they are, they've been going about it a pretty strange way.

If the VHS vs Betamax affair proved anything, it's that where consumer electronic are concerned, good engineering is nothing compared to good marketing. Wake up AMD!

[Putz]

Sorry my bad for not being clear.
When I say parrallel in relation to AMD i'm refering to the chip architecture, not parrallel computing.
AMD's approach to speed is to use a relatively slow clock speed but process lots of instructions in a single clock cycle.
Intel's approach to speed is to use a high clock speed but process few instructions in a single clock cycle.

Well, half-right. I was also referring to chip architecture when I mentioned the SIMD instructions, and even though HyperThreading is technically in the "parallel processing" domain, it reflects the chip philosophy.

Actually, the efficiency in question has less to do with how many instructions they can each perform in a given clock cycle, and more to do with how long it takes for a single instruction to complete. Intel's pipeline is long, forcing even simple instructions to take several clock cycles before "completion" ie. before the result can be re-used for something else. AMD's is shorter. A longer pipeline is better for parallelization, and for cranking up the clock speed. But a shorter pipeline is more efficient for simpler, more linear instruction patterns, and for a better IPC (instructions per clock) ratio. Both have average IPC ratios well below 1.

I do agree with you that Intel took the frequency-race strategy too far. AMD probably would have done the exact same thing (in fact they did -- remember the race to 1Ghz?) if they'd had the resources to throw at constantly shrinking the die and raising the clock speed like Intel. Luckily for them and for us, they didn't, so they've explored other areas of innovation. Intel is now doing the same thing... but they're behind -- the huge momentum they built in their own direction is harder to stop.

I bet Intel is counting themselves lucky that the Pentium M architecture was developed by that semi-rogue project group. Once they get around to really throwing themselves behind that architecture on the desktop, they'll recover from their stumble, and I imagine we won't see a 50/50 market share for some time to come at least. (Don't get me wrong, I love AMD and I'm rooting for them, but I'm also being realistic.)