Intel Drops Tejas, Xeon To Focus On Dual-Core Chips

PunkerTFC writes "Reuters has an article about Intel dropping the fourth-generation P4 chip (codenamed "Tejas") and the Xeon server processor. Intel says they want to concentrate on their new 'dual-core' technology for desktop and notebook systems. This is essentially putting two processors on one chip, allowing for a doubling of performance with less energy use. The introduction of this technology was not expected for another year and a half. Rival chip maker AMD says they have the capability to produce dual-core chips and will introduce the technology when they "feel there is a market need.""

Whoa, deja vu

[gooberguy]

This [slashdot.org] has been discussed before.

Parallel?

[Uber Banker]

Intel says they want to concentrate on their new 'dual-core' technology for desktop and notebook systems. This is essentially putting two processors on one chip, allowing for a doubling of performance with less energy use.

Is this a parallel implementation then? In that case performance is only doubled for processes that can be performed in parallel.

I think this is more related to moving to the PM from the P4 architecture as the M series is more scaleable - taing P4 any further requires a lot more power and generates a lot more heat.

Dual core opterons

[Mdalek]

This seems to be the new trend,
AMD will have dual core opterons next year: [arstechnica.com]

Re:Whoa, deja vu

[ciroknight]

Yes, it was discussed before, but it is confirmned now. Instead of industry insiders saying it, it's news professionals. Who do you trust more?

Re:dual-core

[SiMac]

Hyperthreading and dual core are very different things. A dual core processor is basically two processors put onto one die. There are twice the number of execution engines, just like two separate cores, but on the same chip. This means it's easier and cheaper to make and install than two separate processors, and it has approximately equal performance.

Hyperthreading takes one physical processor and makes it appear to be two logical processors. There's still only one core and one execution engine. It appears to be two processors, but a 3.2GHz Pentium with HT will have nowhere near the performance of 2 3.2GHz Pentiums without HT.

Dual core explained

[tpengster]

Transistors are getting smaller and the chipmakers can fit more and more onto a chip.. However, it is much cheaper (less design time) to simply run two cores with some "glue" hardware than to design a new core that is 8-way superscalar instead of 4 (for example).

One way to look at dual-core is to view it as a dual-processor (MP) system with a very low communications cost, since both cores are on the same die. The disadvantage is similar; since the two units are not perfectly synchronized, such a system runs best with multithreaded code. A single-core CPU with the same number of transistors will run faster, while the dual-core is not quite "double the speed" of one of its cores.

Re:Real impact

[Anonymous Coward]

I wonder how this will play out with XP Home which only supports one CPU?

XP home and win2k3 do correctly recognize xeon's with hyperthreading as only one processor for licensing.

Win2k thinks each logical processor in a HT xeon is a real processor. So if you want a quad-xeon box to run win2k, you have to get the win2k advanced or enterprise version. Regular win2k only supports 4 processors.

Re:And right after AMD....

[Gherald]

> Interesting statistic, but you didn't cite the source. Where did this figure come from?

Four days ago from a completely unreliable source :) [slashdot.org]

Sun, IBM, other major vendors also going dual-core

[brg]

The UltraSPARC IV processor is also essentially two UltraSPARC III processors on a chip, integrated using chip multithreading (CMT) technology. Here is an article [internet.com] and some marketing blurbs [sun.com] about the UltraSPARC IV.

The current IBM POWER4 and upcoming POWER5 chips are both dual-core chips. Here is a nice presentation [hotchips.org](PDF format) about the POWER5; you can see in the die photos where there are two cores. There have also been rumors of a dual-core PowerPC [theregister.co.uk] based on it, but nothing concrete yet.

Broadcom (which bought SiByte) markets a dual-core, 1GHz 64-bit MIPS chip called the BCM1250 [broadcom.com] which has a lot of integrated networking goodies.

Finally, it bears pointing out that on the other side of Intel's severed corpus callosum [disenchanted.com], they're also working on a dual-core chip [theinquirer.net].

Re:Real impact

[IceFox]

Incorrect! It Hyperthreading is turned on in the bios, XP wont install on a quad box because it is a "license violation" stating that only four cpu's are supported.

-Benjamin Meyer

Not so interesting

[eRacer1]

I mean, Intel is a year a head of schedule.

The author incorrectly states that Intel's dual core CPU is "more than a year ahead of schedule". Six months ago during the Intel fall analyst meeting Intel claimed (slide #40) [investorshub.com] dual core for the home computers would arrive in 2005.

This is a rather interesting bit of information from the article: "This strategy was not expected for at least a year-and-a-half, said Dean McCarron, the head of Mercury Research."

Well, how is this news? Intel is claiming that they will go dual core by the end of 2005. A year and a half from now is...2005, just like the Intel presentation from six months ago said.

Re:Remember

[mikis]

I mostly agree, only AMD already announced their dual-core CPU strategy even before Intel. In words of Mr. Ruiz [eweek.com]:

"One of the most powerful things next year is going to be our dual-core product. To me, that's going to really shock the hell out of everyone, because it's going to be hardware-compatible, infrastructure-compatible, pin-compatible. I mean, people that have a 2-P system can slap in a dual-core product and end up with a 4-P system for the price of a 2-P. That's been the biggest drawback, everyone tells me. What keeps them from going from a 2-P to a 4-P system? It's price"

Re:Interesting

[mikis]

Um, AMD announced this [xbitlabs.com] in September last year.

"With coherent HyperTransport, it is inevitable that we will have multiple cores on a single chip. This is a tremendous opportunity because with our architecture the scaling is far superior to anything else that's out there., The Register quoted Mr. Sanders."

Also, see this: AMD CEO: "Dual-Core Opteron Will Shock the Hell Out of Everyone" [xbitlabs.com]. Ruiz confirms dual core Opteron in 2005.

They say that Intel Tulsa (dual core Xeon) will arrive in about a year [xbitlabs.com] and Jonah (dual core Pentium M) is planned for 2005/2006 [xbitlabs.com].

So, nothing new here for AMD.

Re:Interesting.

[AlecC]

Sooner or later you are going to bottleneck on the memory interface. Dual cpus are going to give more capability than hyperthreading, at more cost. If they are strangled by the memory interface, there is no advantage to it. But if it gets more throughout - and Intel have probably simulated it to death - it could be the way to go.

because it would be a waste of cpu time

[rebelcool]

contrary to slashdot ignorance and FUD, the OS doesn't spend most of its time running the CPU.

Most of what the OS does is IO, which idles the chip while waiting for the IO to complete. Tthis is why all operating systems switch to the next task while waiting on IO. If your CPU is running at less than 100% usage its because every program is waiting for IO for most of the time.

Re:Exactly.. Market Need.

[dfghjk]

Can you substantiate the claim that "Apple and other Mac developers have spent more time working on dual CPU optimised apps"? Apple didn't offer a true multitasking, multithreading platform until OS X. PC's have had them since OS/2 1.0, over 15 years ago. True multiprocessor support came shortly thereafter with NT 3.1. Macs are relative newcomers at this and the Apple/Mac developer base is relatively small compared to Windows. The NT kernel, basis for current Windows platforms, has always worked well with multi-CPU systems and was designed for it from the start. BSD, in contrast, has added multiprocessor support rather recently.

It's hard to believe that a platform with 20 times the marketshare and over a decade head start can't compete with Apple when it comes to dual CPU support, so I think you're imagining things. Dual CPU Wintel machines are not uncommon if you look in the right places. They just aren't typical of desktops.

Re:Parallel?

[RzUpAnmsCwrds]

"Intel can make Pentiums cheaper than AMD can make K8s"

That's a big assumption. AMD has always tried to keep a per-die cost lead on Intel. They do this by keeping their die-size down so that they achieve higher yields. That's why most of AMD's future low-end Athlon 64s have 1/2 as much L2 cache as the Pentium-M or P4 Prescott.

"If Intel drive the cost of PCs down a couple hundred bucks, AMD will be marooned in the high end workstation market"

Huh? If the costs of PCs decreases by a couple of hundred bucks, they will be free.

This makes no sense. AMD has *always* had a price lead in the low end. My Athlon XP 2600+ was $84 and it beats any Intel chip under $150. Moreover, AMD has consistantly pushed the price of Atlon 64 down since introduction. You can now get an Athlon 64 2800+ for around $175.

"The P-M is even cheaper to make than the P4."

Where did you get that fact? Northwood Pentium 4s are certainly cheaper to manufacture than the Pentium-M (die size, mostly). Intel does not disclose their per-die costs and thus it is impossible to determine which is cheaper to manufacture.

"Intel can also demand a much higher profit margin."

Not necessarily. Intel has been forced to drop their margin substantially since AMD introduced the Athlon several years ago. That's why the P4 2.6GHz is only $156.

"be forced to sell at a loss (which they've done before)"

It is not clear if AMD ever sold CPUs at a loss. They have made substantial losses as a corporation in the last few years, but that is likely because of fixed costs like the development of K8 and Fab 30.

Re:Dual processors are nice.

[AlecC]

Fractionally worse than dual proc, but not much. Only a single cache, so two processors will do better at different jobs. On occasions the two dual CPU may go faster by sharing cache but this will be rare, unless you are heavily into simulation or. similar.

Dual CPU chips is better - but much more expensive. Like anything else in this business, if you have got it out of the packaging, it is obsolete. When this chip comes to market, everybody will have what you paid good money for at $200 less.

Re:CPU's becoming more like GPU's.

[ciroknight]

Depends on how far you want to go back. If you go all the way back to the day of the original Pentium Processor, then yes. But if you go forward to look at the Pentium Pro to the Pentium 2's and the first Xeon's, then you'll see this "passing back" that I'm talking about. Most of their Xeon improvements are actually put into the Pentium !!!, and the Pentium !!! Xeon tested a lot of things out that they used later when they moved to Flip Chip packaging. AMD had the same vision of Intel, make cheap chips, but they started a little late and mostly played catch up. They were never too keen on innovation (I give them 3d Now! and it's extensions, but this only came after MMX, and never got as popular like SSE and MMX did (yes, I do remember when they declared MMX as dead technology). AMD's in the same position as Intel was in around 1996 *which for me, was ages ago, almost 10 years..*. They created their form of the Pentium Pro (P6), the 64-bit x86-64 archetecture. Right now though, x86-64 is too expensive to take home, but is great for midrange servers. See what I mean?

Re:Parallel?

[NutscrapeSucks]

Huh? If the costs of PCs decreases by a couple of hundred bucks, they will be free.

Why not $50 computers? There's a lot of things in this world where the assembly cost is basically nothing, and you're paying for marketing, packaging and support. I think Bill Gates already predicted this. Not for everyone of course, but the people who need speed will pay for it.

That's a big assumption.

More like a small assumption. Compare the die sizes: K8 - 193mm^2, Northwood 127mm^2, Pentium-M - 83mm^2. Intel is reputed to have the best process technology. Also, there's been some reporting on The Inq to the effect of Intel wanting to drive down costs. There's a reason Prescott is dead end.

My Athlon XP 2600+ was $84...

Yeah, and they sold it to you while losing money because they're such nice guys. Look, the OLD AMD tried the value angle, and failed. The NEW AMD is trying to reposition themselves as a high-end provider, and it's working very well. The question is if the market is going to go in the ultra-value direction or not, and if it does, the P-M is probably the best positioned.

Lots of misconceptions here....

[mr i want to go home]

Dual Core is NOT Hyperthreading. What happened to all the tech savvy people this weekend?

Hyperthreading (or simultaneous multithreading - SMT) creates the ability to run 2 virtual threads on a single chip. This can be as simple as running an 'integer' and a 'floating point' thread, as the SIMD/FPU are really still separate units anyways. In a more complicated form, it means multiple threads are split across different pipes/units/stages - making more use of the processor if you like.

Dual Cores are 2 actualy processor 'cores' on the one chip. That's 2 LSU's, 2 FPU', 2 IFU's, etc plus as much L2 cache you can shove on to feed 'em.

It is like having 2 real proc's....it's just more efficient to make dual core chips than 2 separate ones. Ofcourse you still need an OS that is good at handling multiple processors - which windows isn't really (compared to OSX for example). Speaking of Apple - IBM have been making dual core chips for some time now. The POWER4 was dual core, and the POWER5 is dual core and multithreading! Lots of Apple rumor sites are saying that the next Mac chip (G6?) will be IBM's 980 - supposedly based on the POWER5, ie dual cores etc etc.

Here's a link [utexas.edu] to an IBM presentation with a bit of info on SMT & dual cores (pdf, sorry).

Re:Dual processors are nice.

[karnal]

Somehow I think you are talking out of your butthole.

One of the hardest things to do in current multi-processor setups is keep memory and cache in coherence. Why is this important, you ask?

Well, just like in a database, you do not want to have 2 seperate accesses to a certain location for an "update". If processor 1 and processor 2 go for the jugular on a certain memory location, it's all over....

Now, with the 2 cores sharing a cache, the board logic will not have to deal with this problem. Hence board prices go down. And, if it's true, AMD should be able to produce these close to what the high end chips are today (pricewise).

I'd buy that!