backtop

Print E-mail del.icio.us 45 comment(s) - last by masher2.. on Mar 31 at 8:25 AM

Sam Naffziger
AMD grabs 9 PA-RISC engineers

AMD has managed to wrangle away nine of Intel's top processor engineers. Samuel Naffziger and his eight colleagues were apart of the design team that helped to develop the Itanium server processor.

Naffziger was an Intel Fellow and the Director of Itanium Circuits and Technology. He and his colleagues have been with Intel ever since Hewlett Packard abandoned Itanium development at the end of 2004.

As is the case with the typical PR machines, Intel doesn't see the move as a big blow to its Itanium development even though Naffziger has been instrumental in the launches of McKinley and Montecito.

The real question is what the team at AMD’s Fort Collins design center focus on. Considering that most of the engineers in the area have worked on PA-RISC or Itanium, it seems reasonable to expect that they will concentrate on the server side, rather than on desktop or mobile MPUs. Either way, this should open up some interesting possibilities for AMD, and some new opportunities for Colorado residents.

With K8L due to be released in 2007, this latest development is an indication that the competition between AMD and Intel just became more intense.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
A different view
By masher2 (blog) on 3/30/2006 9:39:08 AM , Rating: 2
> "Considering that most of the engineers in the area have worked on PA-RISC or Itanium, it seems reasonable to expect that they will concentrate on the server side..."

I don't think this follows. AMD doesn't have a server-oriented instruction set, and I don't know that they're planning one. I imagine this team may be tasked with developing the K10...for both server and desktop use.




RE: A different view
By Targon on 3/30/2006 9:53:40 AM , Rating: 2
The K10 has been in the works for quite some time. The K9 was dropped quite a while ago from what I remember reading a year and a half ago.

If anything, I suspect that AMD is working on a whole new CPU design, not just some design tweaks to the Athlon design. Remember, the Pentium-M is based on the Pentium 3, which was based on the Pentium 2, which was based on the failed Pentium Pro. The Athlon 64/Opteron are based on the original Athlon design.

AMD knows that any design will eventually run out of room for development in a competitive environment, it's why the Pentium 3 was dropped for the Pentium 4, because at the time, Intel couldn't get it to run faster with AMD pushing the Athlon faster and faster. Instead of waiting for that day, AMD has been working on a whole new CPU design(that will still be x86 compatable), but it takes time. Getting more talented engineers into the project will speed up the development of that all-new design.


RE: A different view
By masher2 (blog) on 3/30/2006 9:59:25 AM , Rating: 2
> "The K10 has been in the works for quite some time...."

It was reported last year that the K10 project was "delayed or dead". Either way, experiencing some serious issues.

> "The K9 was dropped quite a while ago from what I remember..."

I think AMD is considering their dual and quad-core K8s as the K9. There was a separate K9 project (Greyhound) but it was cancelled a good while back.


RE: A different view
By Knish on 3/30/2006 12:52:16 PM , Rating: 2
quote:
It was reported last year that the K10 project was "delayed or dead". Either way, experiencing some serious issues.

By the inquirer...


RE: A different view
By JackPack on 3/30/2006 1:33:03 PM , Rating: 2
Gotta love it when the uninformed make blanket statements about the Inq. As if there was only one writer working there....


RE: A different view
By masher2 (blog) on 3/30/2006 3:03:05 PM , Rating: 2
Hey, I'm the one who quoted the source...and I have to agree with the prior poster. The Inquirer is certainly the bottom of the journalistic barrel. Still, they have at least a nugget of truth in the story. K10 may not be "dead", but its certainly been delayed.

And to put this all back on topic, whats a more probable project for these ex-Itanium engineers. The K10...or some wholly new server-only architecture heretofor unmentioned by AMD?


RE: A different view
By Knish on 3/30/2006 3:29:31 PM , Rating: 2
If i remember correctly, SN was really involved with Intel's Foxton stuff for Itanium -- power saving things. Whether or not he will be doing K10 stuff might be difficult to pin out, but I would not be surprised if he is involved with some of the K8L stuff, particularly with power reduction.


RE: A different view
By CorrND on 3/30/2006 1:07:07 PM , Rating: 2
I don't really know anything about the K10 but the blurb I just read on wikipedia. Interestingly, it says:

"...the K10 could prove to be a surprising chip, looking much more like a classic RISC processor than has traditionally been the case with x86 systems, although the AMD64 instruction set is still a CISC instruction set."

If you believe wikipedia, the grabbing of Intel's top RISC engineers could be very interesting for K10.

http://en.wikipedia.org/wiki/AMD_K 10


RE: A different view
By Oderus on 3/30/2006 10:07:36 AM , Rating: 2
The Pentium Pro did NOT fail. It was their best processor, IMHO and clock for clock, it owned. You take ANY P2 and under clock it to the same as a PPRO clock and it gets bitch slapped. It worked so well they've been using it a platform for the P2, P3 and P4 as you said. I still have my Proliant 2500 server running with 2 PPRO 200MHZ CPU's running Linux. It kicks ass.


RE: A different view
By masher2 (blog) on 3/30/2006 10:13:47 AM , Rating: 2
> "The Pentium Pro did NOT fail. It was their best processor.."

You have to understand what "failed" means in this context-- the Pentium Pro did very poorly in the market. It had poor 16-bit performance in a time when most code was still 16 bit. It was a very low-yield chip and thus, despite being sold at a significantly higher price than the Pentium, was a very low profit chip for Intel.



RE: A different view
By defter on 3/30/2006 3:29:26 PM , Rating: 2
quote:
You have to understand what "failed" means in this context-- the Pentium Pro did very poorly in the market.


You have to understand what "market" means. Pentium Pro was launched in November 1995 at speeds 150-200MHz. At that time fastest Intel Pentium chip was running at 133MHz and fastest AMD's chip was a 486 clone running at 120MHz.

As you can see, Intel's Pentium didn't had any serious competition at the desktop market when Pentium Pro was launched, thus Pentium Pro was targeted for server/workstation market where 16bit speed was irrelevant. In that market it offered huge improvements compared to earlier x86 CPUs including SMP support and cache, for example, intially Pentium Pro featured 256KB/512KB of full speed L2 cache and 1MB model was added later. Pentium Pro succeeded in capturing marketshare from RISC CPUs in low-end server/workstation market.


RE: A different view
By Samus on 3/31/2006 4:06:52 AM , Rating: 2
You guys do understand the Pentium Pro was core-for-core identical to the Pentium II, with the exception of cache implentation and packaging, that being a socket opposed to a slot. They moved the core to a slot to move the cache off the die because it was too costly and at the time, the process was .25u, which made the die power hungry and thus hot as hell. They would have never made 350-400MHz that they did with the Pentium II on the Pentium Pro socket packaging.

The Pentium III Katmai (or Katamai) cores were simply tweaked Pentium II's now with SSE, on-die cache and a new .18u process. Later they were so successful at keeping the die cool they decided to abandon the slot packaging as it was no longer neccessary, and ironically more expensive to produce (ironic because they originally moved to slot packaging because it was cheaper)

The Pentium Pro failed in the market because it was too expensive. Pentium II was the answer.


RE: A different view
By Bull Dog on 3/30/2006 10:34:05 AM , Rating: 2
K9 was never dropped. AMD decided to skip that combination because it sounds like a dog. (Canine)

Well at least thats what I read a while back. And if K10 is having problems thats to be expected, it's still in early stages of development.


RE: A different view
By devolutionist on 3/30/2006 1:00:27 PM , Rating: 4
I agree - think what it would take to lure people like this away from Intel - the chance to clean up a design that's already into it's development lifecycle?

No way.

They're being brought in to do something totally new from the ground up. Guys like this aren't hired to be the janitors of someone else's work.


64 bit support
By OrSin on 3/30/2006 9:44:36 AM , Rating: 2
AMD has pretty good 64 bit support not but I would not but it pass them to be developing a server chip that is 64 bit only. In the high end server market AMD is really starting to shine and in 2-3 years almost all the application for that market will be native 64 bit. Have an native 64 bit chip would be nice thing for them. Also cripple the already dieing Itanauim platform is nice touch.




RE: 64 bit support
By masher2 (blog) on 3/30/2006 10:05:21 AM , Rating: 2
> "Also cripple the already dieing Itanauim platform is nice touch"

Dying? Sales of Itanium servers tripled in 2004, and nearly doubled over again in 2005. In the high end server market, they're now equal in revene to SPARC based systems...and still growing.


RE: 64 bit support
By Jharne on 3/30/2006 10:49:25 AM , Rating: 2
Maybe but there still undersold against original estimates by a huge amount, and are unlikely ever to break even.


RE: 64 bit support
By masher2 (blog) on 3/30/2006 11:00:41 AM , Rating: 2
Predicting the future is dicey business, but if Itanium keeps up the momentum they've gained in the past two years, they'll not only break even, but turn into a profit leader for Intel.

As die sizes shrink, Itanium is going to wind up scaling better than X86. At 45nm and below, the architecture is really going to shine.


RE: 64 bit support
By Griswold on 3/30/2006 10:58:51 AM , Rating: 1
Just ignore the scale and buzz words like "triple" and "double" make bad things magically look good.

If you put that in perspective to the real cash cows in the server market, you will realize, that itantium is nowhere near a healthy product. You dont need to drop 10 billion dollars on a 12 year old design in the next 4 years if its not on the brink of total failure - you do that when you believe that it might once become something that makes more money than it costs. Or when you just cant admit defeat.


RE: 64 bit support
By masher2 (blog) on 3/30/2006 11:18:47 AM , Rating: 2
Respectfully, you don't have a clue what you're talking about. In the first half of 2005, revenue from Itanium-based servers exceeded that from Opteron-based servers. Is Opteron a "dying platform".

The facts are simple, though they may not be clear to people who think a $200,000 server is "expensive". Itanium and Opteron COMBINED are a small share of the high-end market. But both are growing significantly faster than their competition. Neither are, in any way, shape, or form, "dying".


RE: 64 bit support
By trooper11 on 3/30/2006 12:41:04 PM , Rating: 2
Itanium may be making progress, but it hasnt gained any acceptance in the market, in fact it has lost much of its support from everyone except HP. The problem wasnt the architecture, it was the implimentation and the constant delays when releasing improvments. New versions of Itanium get scaled back and end up not having the big impact it would have had.

I dont think Itanium is dying, I just dont think itll be a competitor in the end. Itll languish in the high end RISC market, propped up by Intel for as long as they can afford it. Then it will be replaced with a real competitor. Intel wanted Itanium to lead to a revolution in the industry where all system would use its instruction set. Well that hasnt happen and most likely wont happen. Thats why I dont see it ever turning into a cash cow for Intel.


RE: 64 bit support
By masher2 (blog) on 3/30/2006 2:14:56 PM , Rating: 2
You state the situation perfectly-- as of early 2004. Itanium has seen some surprising growth the past couple years however.

> "Itll languish in the high end RISC market"

That's a plenty big market to 'languish' in, combining the total sales of Power, SPARC, Itanium, and PA-RISC. Roughly $20 billion/yr and growing.


> "it has lost much of its support from everyone except HP".

And SGI, NEC, Unisys and Hitachi. Oh, and Sun has been talkinga bout porting Solaris to Itanium.

That's today. What about in 4 years when all server software is 64 bit, and we hit the 32nm node? Anyone who understands Itanium realizes it benefits much more from processes shrinks than does X86.



RE: 64 bit support
By Viditor on 3/30/2006 6:19:13 PM , Rating: 2
quote:
Itanium has seen some surprising growth the past couple years however

A few things to remember here...
1. Intel and HP have expanded the "free trial" program. This allows companies to take possesion of an Itanium-based system on a trial basis for an extended period of time.
2. Intel accounting practises list these trial systems as revenue, even though no money has changed hands.


RE: 64 bit support
By masher2 (blog) on 3/30/2006 7:38:12 PM , Rating: 2
Mm, I don't know anything about a free trial...and I purchased 3 Superdomes just last year. Got a link?

Also, the statement that they're being counted as revenue without "money changing hands" seems very odd. If Intel wasn't receiving real revenue, it'd mean they were giving free chips to HP, who was then using them to build trial systems. Stranger things have happened I suppose, but I'd like to see some info on this before I bite.



RE: 64 bit support
By Viditor on 3/31/2006 1:22:23 AM , Rating: 2
I'm surprised you haven't heard of the trial program! I'll see if I can find a link tonight, but I heard about it from several IT guys here.
As to the revenue, Intel has always counted any CPU shipped as a sale. This is how "stuffing the channel" occurs and why it effects the revenue statements. For example, when Intel ships inventory to OEMs, they account for it as a sale (accounts receivable) even though they get paid after sales are made by the OEM (though excess inventory is returned and listed as accounts paid).
This is a fairly standard practice...AMD used to do the same up until about 3 years ago when they completely cleared the channel and modified their accounting practices to match (they did this more to increase credibility with analysts than because what they were doing was wrong in any way...).


RE: 64 bit support
By fic2 on 3/30/2006 3:37:35 PM , Rating: 2
According to this:http://www.channelregister.co.uk/2 005/11/23/server...
3rd quarter '05 itanium server sales were $619M (1.75x '04s
358M). They don't give the amount of $$ from Opteron server
sales, but do say that the number of Opteron server systems
shipped was 174,000 (2.28x '04s 76,000). If it is true that
itanium server revenue > Opteron server revenue then the
average price for an Opteron system is less than $4k. Since
the Opteron has only been out 3 (?) years vs. the itanium
which has been out 10+ years I think it is doing quite well.


You seem to like quoting server revenue, but I would guess
that the amount spent on the actual chip is significantly
less than the amount spent on HD, memory, etc.

What is going to happen once AMD virtualization and HA and
other big iron things get into the opteron chips? What is
going to happen when significant itanium engineers jump
ship to AMD?


RE: 64 bit support
By masher2 (blog) on 3/30/2006 3:46:45 PM , Rating: 2
> "What is going to happen once AMD virtualization and HA and other big iron things get into the opteron chips?"

Don't try to frame this as an "AMD vs. Intel" argument. Opteron's competition is Xeon, and Itanium's competition is IBM's POWER...which, despite Itaniums rapid gains, is still outselling it by over 3 to 1.

> "You seem to like quoting server revenue, but I would guess that the amount spent on the actual chip is significantly less than the amount spent on HD, memory, etc"

Given that Itanium chips are far pricier than Xeons or Opterons, such a comparison actually hurts Itanium. The price of the CPUs in such servers are, comparitively, a larger share of the total.





RE: 64 bit support
By rocketboy78 on 3/30/2006 1:04:16 PM , Rating: 2
Many of the PA-RISC vendors have set end of support dates on RISC systems. This is forcing IT departments to get their feet wet with Itanium. I suspect this is the most significatn reason Itanium sales are picking up.


contract
By brownba on 3/30/2006 10:33:09 AM , Rating: 2

isn't it common to put an anti-competitive clause in your contract where you can't go to work for the competition for at least a year or so?




RE: contract
By Acanthus on 3/30/2006 10:48:58 AM , Rating: 2
yes, but they really dont hold water in court.


RE: contract
By Phynaz on 3/30/2006 10:56:49 AM , Rating: 2
Actually they hold buckets of water in court.

Most likely somebody at the level of Fellow didn't have a contract. These are the top of the elite, the superstars, and if you want them you have to give them pretty free reign.


RE: contract
By masher2 (blog) on 3/30/2006 11:09:54 AM , Rating: 2
> "Actually they hold buckets of water in court. "

Actually, they do not. Its anticompetitive to prohibit people from earning a livelihood in their field, which means a clause preventing a chip designer from working for Intel's competition would bar them from all work whatsoever.

No-compete clauses are only upheld by a court when they're judged reasonable, which usually means for a very limited geographic area and/or duration. A clause barring a salesman from soliciting your customers for a year is reasonable...a clause barring a store manager from working for any competing business within the same city might be reasonable.

But a clause preventing a CPU designer from working for any competitor anywhere in the globe-- when all such firms are in competition with Intel-- is most certainly not, and would be laughed out of court.


RE: contract
By Phynaz on 3/30/2006 11:36:20 AM , Rating: 2
A clause preventing a CPU designer from working on another CPU for another company would certainly be held up in court. This is covered under contract law, which has no definition of the term resonable. A contract is a contact, is a contact.

Look at the recent MSN/Google case. The guy Google hired is prevented from working on search, because that's what he did at MSN.

Now if said CPU designer went else to design other chips, then that would be allowable (depanding, again upon the wording of the contract).


RE: contract
By masher2 (blog) on 3/30/2006 11:52:36 AM , Rating: 2
> "This is covered under contract law, which has no definition of the term resonable."

US case law does, however. No-compete clauses have been struck down as unreasonable countless times. As I said earlier, they are only upheld in certain, limited circumstances. Don't even attempt to argue this point; the facts are incontrovertible.

> "Look at the recent MSN/Google case. The guy Google hired is prevented from working on search"

You're confused. The suit regarding Kai-Fu Lee was quite different. He wasn't a developer, he was upper management (Corporate VP) with access to business plans and strategy.

Prohibiting a manager from managing related projects isn't anti-competitive. He can easily find work in his field in noncompetitive areas. Furthermore, a VP is defined as a "highly compensated key employee" and thus the presumption of reasonable changes slightly. Space prevents me from explaining this in detail, but the rules are slightly different for upper management.

CPU design, though, is extremely specialized. Such an employee could not find equivalent work designing chips which were not CPUs; the entire bulk of his professional work depends upon that specialized knowledge. A noncompete forces him into an alternate field entirely, at presumably a much lower rate of pay.

Finally, if you don't believe me-- believe Intel. If they felt a noncompete would stand up, they would have included it in the employment contract. I've seen $10/hour secretaries forced to sign noncompetes...the fact that this person was "only" an Intel Fellow has no bearing whatsoever on the lack of the clause.


RE: contract
By johnsaw on 3/30/2006 6:08:49 PM , Rating: 2
Well, Intel's Itanium is protected with patent law. No Intel's employee can go to the competition and make a clone without paying royalties to Intel.


RE: contract
By masher2 (blog) on 3/30/2006 10:03:26 PM , Rating: 3
Exactly, which is one of the reasons that upper management positions are much more likely to have noncompetes upheld by the courts. Technical information can be protected by patent or (in some cases) copyright, but business strategies and customer lists are a different matter. Secondly, information like that "ages" far faster than patentable data, which means a brief noncompete period can be both reasonable and efficacious.

For technical trade secrets though, a protection period of even three years is hardly enough to accomplish anything, besides being wholly unreasonable.


RE: contract
By KristopherKubicki (blog) on 3/30/2006 12:55:22 PM , Rating: 2
I actually agree with masher!!! Let us rejoice!

So long as you do not copy the technology from one company to another, there is no way one company can prohibit you from working for another company AFTER you no longer work there. I went through a similar experience a few years ago and I was fairly surprised about this myself.

On the other hand, there are some regulations where given reasonable cause, the original company has the right to do some audits to make sure you didn't just steal their work and give it to the other company.


RE: contract
By kaborka on 3/30/2006 2:25:45 PM , Rating: 2
I think it varies by state. Non-comp clauses for individuals are unenforcable in CA.


?
By z3R0C00L on 3/31/2006 7:58:05 AM , Rating: 2
Think about it... it's no BIG deal.. these guys probably weren't all that great. Instrumental? Why were they the guys who printed the Itanium boxes?

Intel has the capital to compete on salary and benefits. Had they wanted to keep these guys they could have easilly beat what AMD was offering.

Man... soo much AMD fanboyism even in the news... I run AMD computers cuz they're better for games.. but guys common.. this is stupid.




RE: ?
By masher2 (blog) on 3/31/2006 8:25:12 AM , Rating: 2
> "Intel has the capital to compete on salary and benefits"

You don't typically attract and retain people like this on "salary and benefits". They're drawn more by the challenges and opportunities of the position.

If AMD went to this team and said, "forget the Intel bureaucracy...come to us and we'll let you design a chip the way YOU want", they may have well taken a pay cut for the opportunity.


I like...
By Spartan Niner on 3/30/2006 9:37:02 AM , Rating: 2
More competition to help heat things up between AMD and Intel is always welcome. Now we can sit back and watch the price war begin ;)




Spelling
By WizzBall on 3/30/2006 12:59:27 PM , Rating: 2
I think it's a bit ridiculous that Brandon managed to spell Naffziger first as Naffziger, then Naffzier and then Nazifigger. Just what standards are we talking about here anyway?




By n2 on 3/30/2006 9:02:48 PM , Rating: 2
Dresden has more fabs than just FAB30/36. Infineon has a DRAM fab there
and there is another one from a different company and other
infrastructure facilities like a mask shop. There are also universities
there specializing in fab tech. If Intel wanted to open a fab
in Germany the Dresden area would be an obvious choice.


note sarcasm
By margon on 3/30/2006 10:32:22 PM , Rating: 2
maybe amd can enjoy the same financial that these guys designs have brought intel




"I want people to see my movies in the best formats possible. For [Paramount] to deny people who have Blu-ray sucks!" -- Movie Director Michael Bay

DailyTech Poll
Which web browser do you use on your primary personal machine? 






44 Comments









botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki