Intel kills plans for 4GHz Pentium

[SDK]

Intel is dumping plans to release a Pentium 4 processor that runs at 4GHz, saying it will boost performance on next year's chips using other means than clock speed.

The Santa Clara, Calif.-based chip company said it plans to brief PC manufacturers Thursday on the latest changes to its processor road map. The main change is that the 4GHz Pentium 4--scheduled for release early next year and originally due out at the end of 2004--won't come out at all now.

Instead, Intel will boost performance on its chips by increasing the size of the cache, a pool of memory located on the processor for rapid data access. Current mainstream Pentium 4s now have 1MB cache. In the future, these chips will have 2MB of cache, like Intel's Xeon server chips and the "Extreme Edition" Pentium 4s designed for gaming PCs.

Intel will continue to come out with Extreme Edition chips by boosting the bus speed and cache size, said Bill Kirby, director of platform marketing at Intel.

The first mainstream Pentium 4 with 2MB of cache will run at 3.8GHz and come out early next year, an Intel representative said. Larger caches will then cascade down the Pentium 4 product line, the representative added.

The chipmaker also intends to emphasize more sharply technologies such as 64-bit functionality, HyperThreading and a security technology called LaGrande. It will also increase development efforts on dual-core chips with the goal of a 2005 release.

Intel has already "taped out," or completed, the design on its dual-core Pentium-style chips, Kirby said, a major milestone that Intel has not announced until now. Typically, chips start to come out a year or so after taping out.

Behind the shift is Intel President Paul Otellini, who wants the company to move away from focusing on increases in chip speed, measured in megahertz, as the primary way to increase performance. Intel has talked about such a shift for years, but remained fond of the clock-speed approach until recently. Speeches by executives about moving away from megahertz were often closely followed by announcements of faster chips.

Technically, Intel could likely come out with a 4GHz chip, said Dean McCarron, principal analyst at Mercury Research. Hobbyists are already running Pentium 4s at 6GHz, and game machine specialist Alienware is selling an overclocked 4GHz system.

On the other hand, Intel would have to put so many engineering and testing resources into qualifying a 4GHz Pentium to work in all conditions that the cost would far outweigh any financial benefit. Because Intel's dual-core chips are set for release in 2005, only a few thousand 4GHz chips would likely have been sold, McCarron said.

"It's somewhat embarrassing, in that they have promised this product for a really, really long time," McCarron said. "But it was the right decision to make."

Problems with power consumption and heat that accompany megahertz increases are likely another spur for the change. Processors with larger caches or two cores can run at lower speeds than conventional chips--hence they produce less heat and consume less energy, but provide better performance. "Hot spots"--high energy centers on chips that crank out heat--can also be spread out or reduced.

To further reduce power consumption, the company plans to begin in 2007 to produce a desktop chip code-named Merom. The processor is based on the more energy-efficient architectures from Intel's notebook chip line-up.

Chips with larger caches can be more expensive to make, because fewer can be produced out of a single wafer. Two factors, however, will soften the overall impact of a larger chip size. Earlier this year, Intel shifted to the 90-nanometer manufacturing process, which leads to smaller chips. The company also said that, because of an inventory surplus, it will scale back production in the current quarter.

"When you aren't running the fabs at 100 percent, there is no (financial) penalty to doing this," McCarron said.

Source : http://news.zdnet.com/2100-9584_22-5409816.html

haha! I think the processor industry just hit the end of their golden age for 32 bits. AMD is having trouble also so…

[phishphreek]

I would guess intel is going to focus more on the 64bit processors.

They have to keep up with others like AMD.

AMD relased a 64bit processor that is backwards compatible with 32 bit and it even has the northbridge chip built in! Intel's 64 bit processor is not backwards compatible and they still have the northbridge chip on the mobo. AMD dealt a blow to intel. Now they have to play catch up.

[kurt_der_koenig]

always wondered why they stopped at 3.6ghz.

[Danielsd]

The problem with 64bit processors is that there are no 64bit operating systems. LONGHORN please come faster!

[Und3ertak3r]

Problems with power consumption and heat that accompany megahertz increases are likely another spur for the change. Processors with larger caches or two cores can run at lower speeds than conventional chips--hence they produce less heat and consume less energy,

the heating problem and the problems with testing /quality control ..

AMD has been Benchmark clocking their chips for the past couple of years


so has Moores law been beaten?

no 64bit operating systems.

huh none? none whatsoever? realy?

[(V)/\><]

Originally posted here by Und3ertak3r
huh none? none whatsoever? realy? [/B]

smart arse.

there does exist 64bit OS's.

but most mainstream software that would benefit from this would probably need a makeover. this would take time. and money.

and intel? they don't seem to be publically planning for any radical changes. perhaps pulling out of the 4GHz development leaves alot more money for secret, room-temperature superconductor research!

[nihil]

I think that dual cores and much larger caches give a better cost/performance payback than raw clockspeed at this moment.

I remember the old PIII and PIII Xeons..........................the Xeon slaughtered the standard PIII performance wise

AMD and Cyrix were always into "performance ratio"...............looks like Intel are going there as well?

[Tim_axe]

I'm sorry to say, but I don't have hopes for extra Cache and less MHz to be the cure for solving heat problems. Think of it like this: L2 Cache runs at the same speed as the core. Like the 3.8GHz 2MB L2 example the article gave. Currently Prescotts are running with 1MB of L2 cache at the speed of the processor. At 3.2GHz they are pretty heated. Think of adding 2x more Cache size on the die, and having all of that running at 3.8GHz. I don't know how big the L2 cache is compared to the rest of the die, but IIRC 512kb was a pretty costly chunk of silicone (at least for 130nm). And I'd imagine that it running that fast will produce some heat...

But Intel's change of strategy is sure a showing that they have the balls to turn their hugely successful microprocessor business 180 degrees out of the GHz race. Props to them for it




Also, I think Moore's Law was reinterperted at some point in time to have something to do with the features on a processor. I think more Cache and Dual Core are improvements that keep Moore's Law true today.


Besides, it appears that Intel is working on the 45nm process as we speak: http://www.eet.com/semi/news/showArt...cleId=50500191

LEUVEN, Belgium — IMEC has demonstrated the integration of fully-silicided NiSi gates on top of high-k gate stacks, claiming a potential breakthrough that would enable development of a 45-nm manufacturing process.

...

Intel Corp. claims to have developed a high-k gate stack, but has declined to identify the materials. Intels' high-k gate insulator is thought to be hafnium-based, something it would be using in common with most other research groups. It is not known whether Intel is using a dual metal gate or a FUSI scheme.

[chsh]

Originally posted here by phishphreek80
I would guess intel is going to focus more on the 64bit processors.
They have to keep up with others like AMD.
AMD relased a 64bit processor that is backwards compatible with 32 bit and it even has the northbridge chip built in! Intel's 64 bit processor is not backwards compatible and they still have the northbridge chip on the mobo. AMD dealt a blow to intel. Now they have to play catch up.

AMD dealt a blow to Intel by outselling them, nothing more, nothing less. Just FYI, the Athlon-64 does not have a northbridge on the processor, it has the memory controller only, which was most of what made up a north bridge, but not all of it. Certain other things sit on a north bridge, like the AGP bus, as one example.

Originally posted here by Danielsd
The problem with 64bit processors is that there are no 64bit operating systems. LONGHORN please come faster!

Umm, there are plenty of 64 bit operating systems, only two are pertinent at the moment. WinXP 64bit is in beta at the moment, and Linux has supported the AMD64 architecture pretty well since it was introduced. I think it's funny that Microsoft has come full swing with their NT-based OSes and 64-bit processor support. Had it, ditched it, have it again!

Originally posted here by Und3ertak3r
the heating problem and the problems with testing /quality control ..

AMD has been Benchmark clocking their chips for the past couple of years


so has Moores law been beaten?

Moore's law has nothing to do with speed, and everything to do with transistors. It hasn't been broken, nor will it likely ever be, unless we move past transistors and on to some other technology that can't be described as such.
Anyway, This is what Intel has to say about it.

Originally posted here by nihil
I think that dual cores and much larger caches give a better cost/performance payback than raw clockspeed at this moment.

They make desktop operations and multitasking run smoother, but so few applications are written with SMP in mind that it is useless if you are singletasking.

Originally posted here by Tim_axe
I'm sorry to say, but I don't have hopes for extra Cache and less MHz to be the cure for solving heat problems. Think of it like this: L2 Cache runs at the same speed as the core.

Only if it's full-speed L2 Cache. On many processors -- excluding the top of the line current ones -- it runs at half the speed.

Like the 3.8GHz 2MB L2 example the article gave. Currently Prescotts are running with 1MB of L2 cache at the speed of the processor. At 3.2GHz they are pretty heated. Think of adding 2x more Cache size on the die, and having all of that running at 3.8GHz. I don't know how big the L2 cache is compared to the rest of the die, but IIRC 512kb was a pretty costly chunk of silicone (at least for 130nm). And I'd imagine that it running that fast will produce some heat...

But Intel's change of strategy is sure a showing that they have the balls to turn their hugely successful microprocessor business 180 degrees out of the GHz race. Props to them for it

It's kinda hard to not be hugely successful when your chief competition has only had decent products for about 7 of the last 20 years. It's not really balls either, they know that they can only push it so far before they end up with another PIII/1133 launch (within I think a month and a half after "launch", they recalled them).

[Tim_axe]

Only if it's full-speed L2 Cache. On many processors -- excluding the top of the line current ones -- it runs at half the speed.

From what I've seen, all of Intel's recent (Pentium3 on up) processors have a full-speed L2 cache. Even their Celerons. I think half-speed L2 Cache ended with the Pentium II era. Feel free to correct me on this for Intel's Desktop offerings. ( Processor Specs @ Intel: http://processorfinder.intel.com/scripts/default.asp )

AMD did however use slower L2 Cache on their Athlon 750 & 800MHz processors. It ran at 2/5 of the core speed. ( http://tech-report.com/onearticle.x/328 ) AMD's website has been less than helpful at showing me the speeds of the L2 cache on their newer processors, although I know that CPU-Z is able to tell the speeds/ratios of the L2 cache and my AthlonXP-M 2400+ has a full-speed L2 cache. I am sure their other new processors use full-speed L2 cache too.



Well, the announcement of canceling 4GHz was kind of unexpected because they've been pushing further into the high GHz speeds despite the problems with heat they saw moving to 90nm. People did notice Intel was having a hard time pushing even further, but I thought they'd continue pushing because they're, well, the Intel giant. There is speculation into other reasons for scratching 4GHz though. Namely, Dual Core's debut around the time 4GHz would have come out. Dual Core would probably take money away from the launch of a 4GHz part, so they're putting priorities on things and seeing that dual core CPUs have a better future than a 4GHz part. I happen to run a Dual CPU AMD system, and I absolutely love it