Who's going to produce the first dual-cored CPU?

[Falloutboy]

I honestly think dothan verse a64 might be darn close mhz for mhz. although I'd give a slight edge to amd because of thier ties with IBM to help them get it right

 

[dudeman007]

We'll have to wait till the second half of 2005 to see the new amds.

 

[jpetermann]

They will both paper-launch and then take 3 months to actually get them to consumers...

 

[Vette73]

Originally posted by: myocardia

Originally posted by: Swanny
I think Intel will, just because they have more R&D and they need it more.

Yeah, I think Intel feels like they're "behind", since the Athlon64 has been such a big seller.

Yea seems like ANOTHER AMD idea Intel is going to have to use. I don;t think intel had and plans on doing it anytime soon, but with the Athlon64 and Opteron doing so ell they need soemthing, and the POS Flamethrower P4 is not cutting it.

 

[alexruiz]

AMD will be fisrt.... Don't forget the hundreds of patents generated by AMD during the K8 development, being a good part fo them related to dual core CPUs.....

 

[Acanthus]

Originally posted by: Marlin1975

Originally posted by: myocardia

Originally posted by: Swanny
I think Intel will, just because they have more R&D and they need it more.

Yeah, I think Intel feels like they're "behind", since the Athlon64 has been such a big seller.

Yea seems like ANOTHER AMD idea Intel is going to have to use. I don;t think intel had and plans on doing it anytime soon, but with the Athlon64 and Opteron doing so ell they need soemthing, and the POS Flamethrower P4 is not cutting it.

That pos flamethrower P4 is probably selling in larger volumes than A64 :brokenheart:

And as for the "another thing intel will copy" comment, IBM already has dual cored cpus.

Edit: Actually isnt the power5 like 8 cores?

 

[Sahakiel]

It seems some people think that SMP support is generic. That's definitely not the case. If the software interface is different, then the kernel must be patched, rebuilt, or replaced. The hardware is exposed to the OS.

HyperThreading required tweaking the kernel a bit on Win2k to prevent an idle thread from hogging the CPU. Dual-core CPUs will likely require software support if they use a different interface. I'm not too familiar with the bus and socket for

Hyperthreading is not quite "simulating" another processor. That's called emulation and is an entirely seperate matter. Hyperthreading is SMT and takes advantage of the unused resources in every single superscalar architecture. The primary reason for developing SMT for the architecture would be the combination of a highly out of order design and near-fanatic focus on raw speed.

As far as I know, Banias does use less power than a comparable mobile AthlonXP. That's just a consequence of differing methodologies. Decreasing voltage on a desktop design yields lower power use, but actually designing for power with even simple gating techniques yields better results.

Financials aside, Intel is widely regarded as the technology leader with AMD following close behind. It's actually quite difficult to think of many revolutionary designs from AMD's CPU division. On the other hand, Intel has a decent list.

Hyperthreading is probably not a priority on the Dothan architecture. The core was designed from the ground up to save power, not maximize performance. Redesigning the basic architecture for maximum performance would likely result in a furnace almost as hot as a Northwood or Prescott, depending on process.

Windows NT support for multiple processors is a bit different than support for multi-core or SMT, as shown by HyperThreading. Enabling MP isn't as simple as flipping a switch on any old piece of software and supporting new chip designs may be difficult or even impossible without more extensive coding.

XP Home does not support multiple CPUs. As far as I know, this was a conscious decision in order to differentiate between target markets. Microsoft is apparently willing to extend support for consumer level multi-core designs, re-defining markets based on socket type and number.

Multi-core CPUs are an old idea and have been on the market for quite some time. Even looking at only Intel and AMD, if I remember correctly, Intel and AMD both drew up plans for dual core long ago and at pretty much the same time.

As for dual-core being shown as Intel following AMD, it should be prudent to note that Intel has rarely followed AMD (to be honest, I can't think of once). Both companies invest a lot in research and keep ears close to the ground. A lot of Intel's key developments were produced elsewhere or came to market as highly evolved derivatives. AMD also has a noted record of not only keeping an eye on worldwide research, but also copying/licensing designs to compete with Intel. Kinda difficult not to do so considering the differential in R&D expenditures.

Generating hundreds of patents in CPU design is also a fairly common event. However, the vast majority of patents are pointless, easily circumvented, or reverse-engineered.

 

[Zebo]

http://www.anandtech.com/cpu/showdoc.html?i=2081

 

[Hans5849]

we need more people like Sahakiel who have a brain and can make a competent argument (i am not one of those people). But i will say i do believe that hyperthreading will be present in a lot of intel CPU's in the future. i can not say how i know because i don't want to get anyone in trouble.

 

[Avalon]

I didn't realize some people were born without brains. Interesting.
I think the newer Intel CPUs are going to utilize HT for one simple reason: marketing.
This is 100% guesswork here, but the general populous associate HT as some great technology that makes your computer a lot faster. If the next core doesn't have it, some people are going to be dissapointed/wondering where that HT ability went. Think about it. You see adds on TV for Netzero that now include Speedband, which they claim makes your internet browsing five times faster. Sounds really good. Then you see a Pentium 4 processor that now includes Hyperthreading. Sounds pretty hyper...hyper awesome. I just figure some people associate HT with the best of CPUs. The new cores can't possibly be the best if they don't have this hyper cool HT ability, could they?
Only a computer literate person would know.

 

[SickBeast]

Avalon, before you go around saying nobody but you has a brain, you should probalby read Duvie's thread on hyperthreading. It's stickied somewhere, or at least it was. He proved there were very large gains from enabling HT on a P4C.

In any event, if the new cores are P4-based, HT *will* be a nice feature. If they are based on the P-M, then HT will not make much difference.

 

[Sohcan]

Originally posted by: SickBeast
I was under the impression that AMD had plans to go the dual-core route long before intel unveiled their plans to abandon the P4 platform and go with the duallies.

Publically announced plans and private roadmaps are two separate issues. The first dual-core Itanium, Montecito, has taped out for introduction next year (I'm on its circuit design team). Intel's COO, Paul Otellini, showed off a wafer of Montecitos a month ago at a analyst conference. To give you an idea how long these CPUs take to develop, Montecito's investigations started in late 2000/early 2001.

Even with compactions of existing designs (such as Northwood, or 90nm Opterons), you can expect around 6 months for a microarchitecture investigation, 12 months to write the RTL (a software model describing the chip, written in a hardware description language), 18-24 months for the circuit design, and 12-18 months for silicon debug. Major core feature freeze starts at least three years before the product release, if not more...adding major changes after that point is most likely going to delay the introduction, and the later the changes occur, the greater the impact of the delay. So estimations of where a competitor will be in four years or so might affect the design decisions in a chip, but publically known or leaked information about competiting chips released around the same time is not going to have any affect on their major design decisions.

I think in general Intel tends to be a bit more tight-lipped about future roadmap details than AMD, so if Intel announces a feature later than AMD, that doesn't necessarily mean that its development started later.

 

[SickBeast]

Originally posted by: Sohcan
The first dual-core Itanium, Montecito, has taped out for introduction next year (I'm on its circuit design team).

Hmm...Itanic eh??? :beer:
As long as you get them for FREE it's all good.

I suppose you couldn't tell us when intel will release a consumer-level dual-cored CPU then? Before or after 2H05 even???

 

[Jeff7181]

Anyone wanna take bets on whether Intel will make use of SOI in their next generation of processors?

 

[Sohcan]

Originally posted by: SickBeast

Originally posted by: Sohcan
The first dual-core Itanium, Montecito, has taped out for introduction next year (I'm on its circuit design team).

Hmm...Itanic eh??? :beer:
As long as you get them for FREE it's all good.

Don't believe everything you read online. I don't want to hijack the thread, we can discuss this off-line via pm if you want.

 

[DClark]

While I don't know who will produce the first dual-cored processor, I've been told Intel's code name for the dual cored Prescott is "Smithfield" or "SMF" for short (anyone else with an Intel connection, feel free to start digging).

Considering the current Prescott's heat and wattage problems, I'm not sure how they're going to cool a dual cored version though.

 

[Jeff7181]

Originally posted by: DClark
While I don't know who will produce the first dual-cored processor, I've been told Intel's code name for the dual cored Prescott is "Smithfield" or "SMF" for short (anyone else with an Intel connection, feel free to start digging).

Considering the current Prescott's heat and wattage problems, I'm not sure how they're going to cool a dual cored version though.

I don't think there will be a dual core prescott... if they make a dual core anything it'll probably be based on the Pentium M core.

 

[DClark]

I know, that's why I was surprised when I got the message from the person I know. I figured the first dual core desktop from Intel would be a Desktop Dothan, but if Intel wants to be first at something then maybe a dual cored prescott it the quickest way to go.

 

[Avalon]

Originally posted by: SickBeast
Avalon, before you go around saying nobody but you has a brain, you should probalby read Duvie's thread on hyperthreading. It's stickied somewhere, or at least it was. He proved there were very large gains from enabling HT on a P4C.

In any event, if the new cores are P4-based, HT *will* be a nice feature. If they are based on the P-M, then HT will not make much difference.

Sickbeast, you completely misunderstood my post, yet now that I look at it I do see how it could be a bit misleading. Let me explain. I'm definitely not implying that I'm the only one with a brain, or that I even have one. I was just poking fun at what I felt was an obnoxious statement made by a previous poster. I've read Duvie's thread on hyperthreading several times. It's impressive. I did not mean for you to get the idea that I think HT is nothing but a marketing gimmick. I can see how comparing it to Netzero, you'd think I made the connection there. No, I know HT has very real and very beneficial purposes. I've got nothing against intel. I'd buy their chips in a heartbeat if my needs were best suited to what they could provide. I made a bad comparison, and I apologize. What I was trying to say in my previous post was that, even though the new p3 based cores with short pipelines will not benefit much from HT, I feel it would, as long as it didn't hurt performance, still be a good idea to have on those new p3 based chips because of marketing reasons. Mainly, name recognition. I know a lot of computer inclined people associate HT with CPU power. At least in my neck in the woods. It would be easier to market the new p3 based cores to computer inclined people as powerful chips still by giving the highest classed chips HT. That's all I was trying to say.

 

[Jeff7181]

Originally posted by: DClark
I know, that's why I was surprised when I got the message from the person I know. I figured the first dual core desktop from Intel would be a Desktop Dothan, but if Intel wants to be first at something then maybe a dual cored prescott it the quickest way to go.

That'd be cool... it'll come with the first STOCK phase change cooler