The rather unique triple-core microprocessor by Advanced Micro Devices due in early 2008 is nothing else but a way to increase the amount of working processors based on the new micro-architecture and featuring so-called “native” quad-core implementation, claims chief technology officer from Intel Corp.
Recently announced plans to release central processing units (CPUs) with three cores for desktops inspired quite some interest among end-users, as there is certain trend towards multi-core chips, but right now a lot of software relates on single-thread performance and modern quad-core processors may not be really fast there. However, Intel Corp. believes it hardly makes sense to release tri-core microprocessors and the only reason why its smaller rival AMD does it is necessity to increase production yields.
“I wouldn’t make that much of it. This is a yield-improvement technique, plain and simple. IBM and Sony with their Cell processor – they have eight (processors) on that that chip, and they said, ‘well, seven is the actually the number and one is a spare, or one is dead’. I’m never quite sure whether there’s a dead one or not. It’s just like memory chips today… There are thousands of spare memory bits that are there to ensure sufficient yield,” said Intel’s chief technology officer, Justin Rattner, in an interview with Wired web-site.
The new quad-core AMD Opteron and the forthcoming quad-core AMD Phenom processors use monolith implementation, which means that all the cores are incorporated into a single piece of silicon. By contrast, current quad-core processors from Intel use multi-chip-package technique and incorporate two dual-core processors onto a single piece of substrate. Given that it is considerably easier to manufacture two relatively low-power monolithic dual-core dice than to produce one monolithic quad-core product, Intel’s approach does seem to make more sense from economic point of view, due to the fact that potentially AMD’s quad-core chips have considerably lower yields compared to Intel’s. Still, AMD claims that its quad-core CPUs can deliver better performance.
“So, yes, AMD has a four-core product. I’m sure when they looked at their yield losses, they said, ‘Wow, we can offer a three-core version of this if one of those cores are dead or slow or whatever it turns out to be’,” said Mr. Rattner.
But despite of being rather skeptic towards chips with three processing engines, Mr. Rattner indirectly admitted that such microprocessors will be better than dual-core offerings in applications that take advantage of multi-core CPUs. Nevertheless, quad-core will be even more preferrable.
“In terms of software, there’s software for one core and then there’s software for multiple cores. It’s not like, oh, we have a three-core problem but not a four-core problem,” Mr. Rattner concluded.
Comments currently: 35
Discussion started: 10/09/07 08:10:12 AM
Latest comment: 10/23/07 08:02:18 AM
Expand all threads |
Collapse all threads
[1-14]
1.
O RLY?
Who did not know that until now? Right, noone.
[Posted by: 1234 | Date: 10/09/07 08:10:12 AM]
2.
Yea....Intel is a little late on that one. That goes without saying! Maybe Intel thinks the general public is a bunch of dumbasses that know nothing about the industry.
[Posted by: Wingless | Date: 10/09/07 08:36:00 AM]
+ expand thread (1 answer)
- collapse thread
The general public is a bunch of dumbasses...
Why do you think that Intel is still using "Pentium"? Because people are dumb enough to ask for a Pentium when they want to buy a computer, and some are even stubborn enough to only want Pentium or nothing.
[Posted by: mooo | Date: 10/09/07 10:18:03 AM]
3.
With INTEL's current architecture, Tri-Core is imposible to make.
[Posted by: INTEL speaker | Date: 10/09/07 09:07:39 AM]
+ expand thread (4 answers)
- collapse thread
This is incorrect, albeit it doesn't matter anyway since there would be no reason for Intel to want a 3-core processor when 4-core is just as easy for them to produce.
Rattner is also wrong that "there are single-core problems and multi-core problems". Not all applications are either completely non-parallel at one extreme or embarrassingly parallel at the other. There are many examples in which it's possible to generate only a finite number of useful independent threads.
[Posted by: MTX | Date: 10/09/07 10:04:24 AM]
If you feel better qualified than the "chief technology officer from Intel Corp." to determine the nature of single and multi core problems perhaps you should apply for a job?
If you are using applications which take advantage of the number of cores you have, the more the merrier, as they say.
[Posted by: McKnight | Date: 10/10/07 10:43:04 AM]
If you doubt what I said, maybe you should read a book on multithreaded programming? Or even just think a little about it before making rude and incorrect remarks?
As for Rattner: what do you expect him to say? This is a public announcement, after all. I bet he doesn't say the same thing in internal meetings; if he did, he'd soon be corrected, anyway.
[Posted by: MTX | Date: 10/10/07 04:25:55 PM]
Um, excuse me, but "multicore" does not equate "embarrassingly parallel". It means "more than one core". Before referring people to books perhaps you should refer yourself to something called "sense"?
[Posted by: wut | Date: 10/15/07 04:14:15 PM]
4.
Who cares? As long as it's more than 2 cores for the price of less than 4 cores, it makes sense.
[Posted by: shae | Date: 10/09/07 01:49:50 PM]
+ expand thread (2 answers)
- collapse thread
It maybe something strategic AMD has come up with in terms of wrong footing Intel. Imagine if AMD position the price of the Tri-Core the same as Intel Dual Core and the 2 Core Phenom the price of Intel single Core? I think AMD is looking to get an edge over its bigger rival.
[Posted by: Ru4real | Date: 10/10/07 05:21:15 AM]
...or, AMD would be wrong-footing itself should the tri-cores prove to be lower performing than Intel dual cores, yet at the same time priced slighter higher...
[Posted by: wut | Date: 10/15/07 04:15:55 PM]
5.
Complete Celeron lineup of Intel is same thing. And, as shae told, who cares? This information does mean nothing to end users.
[Posted by: BorgDrone | Date: 10/09/07 04:33:19 PM]
+ expand thread (1 answer)
- collapse thread
The answer would be "Appearently, everyone who cares enough to reply on this page"
[Posted by: wut | Date: 10/15/07 04:17:18 PM]
6.
I thought this the minute it hit the web. And I also thought, thats not such a bad idea really, if its useable and wont screw up why not?
Id make this kind of call in a heartbeat its a niche market that never existed before.
[Posted by: verndewd | Date: 10/09/07 05:52:18 PM]
7.
Triple cores is a sign of more scalable architecture of AMD. Tomorrow they will do the same thing with 5 cores. Which means one-core-more at the age of multiple cores.
Current Intel architecture does not allow odd number of cores.
This fact disquiets Intel's CEOs.
In addition Intel has a huge inventory with their fusty architecture.
AMD will succeed!
[Posted by: Sittingbull | Date: 10/10/07 04:03:11 AM]
+ expand thread (3 answers)
- collapse thread
I believe that Intel could make the same 3-core processors with their current architecture... and why would they bother creating like these ones?
AMD's 3-core processor = Intel's 4 core processors at lower price.
[Posted by: PinPin | Date: 10/10/07 05:20:07 AM]
Don't expect a tri-core CPU from Intel until after they get their native quads (Nehalem) out.
With their current Core 2 architecture, they would have to:
* Make a native tri-core chip.
* Make a MCM from a dual-core die and a single core die (or a half-disabled dual).
I don't see the first happening and the second seems a bit (for lack of a better word) ghetto-rigged.
As for Nehalem, Intel will likely pull an AMD on this one and just disable a bad/spare core.
[Posted by: cheeseman | Date: 10/11/07 04:31:17 PM]
With Nehalem, Intel will likely not have to "pull an AMD" at all because there won't be yield issues. Why solve a non-existent issue with a non-needed solution?
[Posted by: wut | Date: 10/15/07 04:32:41 PM]
8.
Justin is right.
AMD's tricore is just a yield improvement technique
An Ex-AMD employee told me
[Posted by: AMD grow up | Date: 10/10/07 10:59:27 AM]
+ expand thread (2 answers)
- collapse thread
Sure. Tricore is basically a crippled, only 75% working, quadcore. Wonder who will buy those.. They might OC bad as well.
[Posted by: rADo | Date: 10/10/07 01:06:08 PM]
If it does OC badly it would not be from being a tri-core design.
[Posted by: wut | Date: 10/15/07 04:35:55 PM]
9.
most of us smart people already knew that.
viva la revolucion VIA!
[Posted by: Joz | Date: 10/10/07 11:59:58 AM]
10.
And sticking two(2) dualcore processor in a package and selling it as Quadcore is what? At least AMD is yeilding actual quadcore processors. Intel yeilds none. Coal pot calling electric kettle "black"!
[Posted by: Taracta | Date: 10/10/07 01:50:09 PM]
+ expand thread (6 answers)
- collapse thread
Both Intel and AMD quad-core processors have four processing cores. That's all that matters in terms of determining whether or not to call it "quad-core"--how the cores are connected may affect performance, but it doesn't matter a jot to terminology.
[Posted by: MTX | Date: 10/10/07 04:29:34 PM]
The whole point of this discussion was the comment by Intel about triple core AMD products. How can they comment on AMD's usage of yield-inprovement techniques when their whole quad-core product line is a yield-improvement product and until the Core series processors so was their dual-core line? My comment was based on this and not the definition of quadcore CPUs (Intel) or quadcore processors (AMD).
[Posted by: Taracta | Date: 10/11/07 12:03:53 PM]
By Intel's method, they can get reasonable yields without sacrificing one of the four cores; while they sacrifice some performance by this, it's certainly not a full 25% in most applications. Of course, this wouldn't matter if AMD's processors were as powerful with three cores as Intel's are with four, but that's not the case.
[Posted by: MTX | Date: 10/12/07 03:51:59 AM]
Intel's method is to use two working dual core processors and package it so that they work in a single socket. Any dual core processors that are not working, one processor defective, will never be used in their quad core CPU. They would probably sell it as a single core Celeron or something. This is where the yield part for Intel comes in. They are able to pick what dual core processors they use for their Quad core CPUs and reject the ones that don't meet their criteria, using these as dual core Core2 processor or even single core Celerons. The only performance that matters for the 3 core AMD processor is whether it will perform better than a dual core AMD processor, yes it will but it will not perform as well as a quad core AMD processor.
[Posted by: Taracta | Date: 10/12/07 01:25:59 PM]
Unless I've been misinformed or Intel has changed their production procedures recently, this is not true. Dies cannot be tested at all until packaged; once CPUs are packaged, they either work perfectly or they have to be down-rated or thrown away. I admit it's possible that they may be able to de-solder one of the two dies in a faulty quad-core part, but this seems unlikely to be economical considering Intel's volumes. The benefit of building smaller dies is that more of them fit on a wafer, and moreover, more of them fit in the _centre_ of the wafer where the highest photolithographic performance is achieved. Consequently, I suspect that Intel takes only dies produced in the centre of the wafer and on its best production lines to become quad-core CPUs, in order to minimise the chances of losing a good die due to a bad one packaged on the same substrate. It's not just about absolute yield--volume counts for a lot in the economics of production as well.
[Posted by: MTX | Date: 10/12/07 04:25:06 PM]
...because AMD is trying to make it seem as though it is something MORE than a yield improvement technique. Which it isn't.
[Posted by: wut | Date: 10/15/07 04:21:52 PM]
11.
By the way, AMD cores are connected to the rest of the processor via a switch. If one core is defective it does not affect the rest of the chip which means that now three(3) core now has the resources that were meant for four(4) cores. Something to think about.
[Posted by: Taracta | Date: 10/10/07 01:55:35 PM]
12.
Yield-Improvement Technique, yeah..of course.. and this is a bad thing.... why ? Because they can't do it ?
[Posted by: Roko98 | Date: 10/10/07 08:48:02 PM]
+ expand thread (1 answer)
- collapse thread
It's not "a bad thing" to have a yield improvement technique. It is "a bad thing" to tout a yield improvement technique as something more than just a yield improvement technique.
[Posted by: wut | Date: 10/15/07 04:24:29 PM]
13.
Intel is just jealous that AMD can create a unique processor that will get popularity.
[Posted by: linuxnerd | Date: 10/10/07 11:47:09 PM]
+ expand thread (1 answer)
- collapse thread
If I were "Intel" (as if one person can be a corporation... but anyways) I would be secretly giggling to myself on the pricing mess that would surely surface later for AMD.
If tri-core could not outperform C2D, then it'd have to be priced less than C2D in the price war, which in turn knocks down Athlon X2's even more............
[Posted by: wut | Date: 10/15/07 04:29:10 PM]
14.
AMD why ? are you making 3- core cpus? in 2008, when the market is moving towards quad-core chips, to be ahead you would need to make your product-range 70% quad-cores format, and have the plans plans for a 6-core-cpu on paper.The time you get round- to rolling out..some of your quad-core, chips" intel" will be on six-core chips. I am at a lost to understand this, 3-cores in not better then 4, thats the bottom line, its that simple.
[Posted by: red lice and beans | Date: 10/23/07 08:02:18 AM]
[1-14]
Make Home Page
Add to Favorites
Print article
Send link to friend
Digg this
Post to del.icio.us
Post to Slashdot
