Poor AMD

[jeffrey]

thank you amd

new releases

 

[NOS440]

Rectalfier man you have a short memory. Just because Intel had some problems last year they are not the only company to have them. Before the Athlon AMD was on there last leg period. They release CPU after CPU that were almost complete failures and performed Horrible and Also were known to be flaky as hell. In my opinion the early Athlon was flaky also. If the whole track record for the last 5 years were taken into account AMD would loose hands down for good products.

 

[MadRat]

Funny how that K6-amajigger that everyone claims was "AMD's downfall" is what became one heck of a cash cow. I'd go so far as to say it inspired alot of people to avoid new systems which really put a peg in Intel's shoe as far as their earnings progress went.

 

[Rectalfier]

I'm not talking about the past five years, I'm talking about NOW.

 

[NOS440]

Well I hate to tell ya but intel hasn't had any problems for almost 9 months and they are still the Number ONE cpu manufacturer in the world and out sell AMD by almost 2 to 1. So lets talk about NOW !!!!!!!

 

[MikeyP]

If Intel hasn't been having any problems, why are they laying off about 5000 employees NOS? (I am not trying to put down Intel here, it is truly unfortunate for those involved)

I'm starting to think that AMD is doing a major rework for the Palomino. When their roadmap was first released, Palomino was stated to have more stages in the pipeline, a vastly superior BPU, lower power consumption, and a 256 bit bus to the L2 cache. Although these rumors were later denied, I'm starting to think we may have a pleasant suprise on our hands

Another point, on the Thunderbird core, AMD absolutley cannot afford to get in another Mhz war. Sure, they may be able to pull out the 1.4 or 1.5 on this core, but the P4 architecture is much more scalable. Intel will raise the bar along with them. Therefore, they aren't "milking" the market as much as possible. By releasing chips slowly, they maximize their profits. It's simple economics.

 

[NOS440]

MikeyP please !!!!!! First off the entire tech industry is tightening there belts and alot of other companies. Also we were talking about technical problems what the hell does lay offs have to do with it ?????

 

[Wingznut]

Intel is not laying off 5000 people. They are cutting 5000 jobs (~ 5.9% of the workforce) via attrition, over a nine month span. BIG difference.

 

[Rectalfier]

Man, I don't even know why I am arguing with you. All I did was quote Tom, from Tom's hardware. I didn't mention anything about technical problems. All I implied by that quote is that AMD is pacing the release of their CPU's, not delaying them because of technical problems.

 

[MikeyP]

<< Well I hate to tell ya but intel hasn't had any problems for almost 9 months >>


Nowhere does that say &quot;technical problems&quot; NOS. Intel is &quot;struggling&quot; (they are still generating huge revenues!) at the moment, as is most of the tech industry. One of the most notable exceptions happens to be AMD Intel may be outselling them in the desktop market by a 2 to 1 margin, but AMD is rapidly gaining. In Europe, most processor sales are actually AMD. Also AMD is starting to spread it's wings into the low-end server market and the mobile market.

Also, I thought they were cutting the jubs, but I didn't want to overstate the problem, seeing that I wasn't sure I used the term laying off.

 

[Wingznut]

&quot;but I didn't want to overstate the problem&quot;

Cutting jobs through attrition is not nearly as bad as layoffs. It just means that as people leave the company (for whatever reason), Intel won't be rehiring for the position. 5.9% over nine months is typical.

 

[MikeyP]

Well, I guess you would know better than me I was under the impression that they were actually firing people. I got that idea from an earlier pm post, where he said something to the effect that his job was not in danger. Thanks for clarifying!

 

[pm]

<< In Europe, most processor sales are actually AMD. >>



Actually I'm pretty sure that this should read &quot;most consumer processor sales&quot; - not overall processor sales.

I'm not especially concerned about losing my job, but I am honestly touched (seriously) that people seem concerned about this prospect. It gives me a nice feeling to think that some of you guys seem to care what happens to me.

 

[Soccerman]

hey it wouldn't be Anandtech forums without such fuzzy feelings now would it?

 

[seewhy]

As an AMD investor, I am happy with its progress in CPU market. Actually I am more worried about their flash memory business as flash memory price is getting hammered. AMD do not release faster CPU because of business reasons, actually the 1.3/1.333 cpu was available to many people months ago, this is unlike Intel few months ago when they paper lunch to compete with AMD. I would like to see Palomino and laptop cpu sooner, but in this kind of economy, AMD isn't gonna help itself by lunching products few people may buy right now.

Actually, I see Intel and AMD kind of working together in slowing the pace of introducing new processors. I personally think it is a wise business move in this kind of economic environment, yeah I know, we the power users aren't gonna be happy, but companies will do what's best financially for themselves.

Based on Taiwanese mobo manufacturers, AMD based mobo are selling like hot cakes compared to P4 based mobos, so it is definitly doing well against Intel right now. International market wise, AMD is gaining market shares in consumer markets and slowly getting into corporates with help of NEC and maybe IBM when the 760MP is out, that is good since the economic problem hasn't hit international as bad as the US.

Looking ahead, I think AMD will do well and keep up with Intel. Well, now I don't want Intel to do bad either, my bro is working for Intel as well. But looking at mature industries, there is always a few big players, for example car/drug...etc. There is definitly room for AMD to grow to size close to Intel and that's gonna be the day I retire...

 

[Bremen]

Great discussion guys, loved reading the whole thing. Now an economics lesson for those of you clueless berks out there.

Suppose Intel could have released a 1 GHz P3 instead of a 386 all those years ago. What do you think would have happened? Well we'd all have bought P3's of course!!!! Intel could have charged a song and made out like a bandit. However then comes the problem... when would you buy your next computer? Sure they'd sell alot of those P3's however it would be years before anyone felt the need to buy another computer. Hence, no more profits for quite sometime.

Now of course that example is heinously(sp?) exaggerated, however the same still holds true. It is far better to sell a lesser processor now, and a better one later, than to sell the better one now, and not sell any later.

 

[pm]

No one would ever hold back a faster CPU without a good reason. The speed improvements that come out of a fab (that are not actual process conversions to new technology) are generally approximately free. A fab engineer like Wingznut will one day find that if he sets the plasma etch on mask step #2 to .1s longer, CPU's go ~2% faster. Or the boron implant is extended and that gets a 2% speedup. The point is that people are improving the recipe - not that new equipment is needed.

When a processor is released on a given process it shows up at a certain speed. Then over time, that speed generally increases and a company sells these faster parts. How do you think this happens? Do you honestly think that Intel had 1GHz Pentium III's just sitting around when Coppermine was first released on 0.18um? No, it was a gradual improvement in the process (and often to a lesser extent, the design) that allowed faster and faster CPU's.

But a company will always take advantage of process improvements to improve speeds - why wouldn't you do this? What point is there taking 1GHz CPUs in the beginning and marking them as 733MHz CPU's? So that you can later sell 1GHz? But by that time you will be getting even faster stuff out of the fab. This is where theories like Bremen's fall apart. If you could sell 1GHz Pentium III's back in the 386 days, then we would be way faster now. People are always inventing, improving and innovating.

People who come up with theories about holding back designs don't really understand how semiconductor manufacturing really works. You are getting the improvements in speed gradually (through tweaking and experimentation) and essentially for free (you just have to change a few steps slightly to get faster units). It's not a matter of maximizing profits - it doesn't cost any more to produce a 1GHz Pentium III than a 733MHz Pentium III - they are just chips on the same wafer. And you always want to increase speed - even if all of your 'usual' competition is behind you in performance, there are always people ahead. And if you get far enough ahead of everyone else, you can charge an even higher premium.

Frequency improvements over time on the same process are essentially &quot;free&quot; - a company will always take advantage to them to either improve profits or to increase market share. No one will ever just say &quot;what we have is good enough&quot; - because it simply doesn't make sense to do so.

 

[Bremen]

I have to disagree there pm. Now I agree that proc improvements are a steady slow thing. However minor improvements like you describe can only take a core so far. Eventually the core will get a radical redesign or a totally new core will be released, ex Katmai -> Coppermine or P3 -> P4. Now say AMD could release the Palomino tomorrow, and did. Then you're left with the question &quot;How long will we have to live with this core?&quot; Similarly if AMD released a 1.4 tomorrow &quot;How long until we're able to release the next speed grade or update to the next core?&quot; In my example (which I agree was a bit silly) you can see what the effects of going too long without improvements are, sales go dead.

Anyone around here have marketing experience to educate us in our follies?

 

[pm]

Maybe I've lost the train of the discussion, but about mid-way down people started musing that AMD felt that their current 1.2GHz CPUs are fast enough and that's why there haven't been in any intros. They are capable of 1.5GHz, but they aren't releasing them because what they have is good enough.

And I'm saying that if it's the same core, then there's no point in holding back a design. If you are getting 1.5GHz out of the fab, then you will release 1.5GHz CPU's and you will do this as soon as possible - because on the same process, with the same core, a 1.5GHz CPU will cost the exact same amount to make as a 1.1GHz CPU. So what possible benefit is there to not releasing the 1.5GHz as soon as possible? So you can wait and hook consumers with a better product later on? But you have that product now, and the longer you wait the closer the rest of the world is to catching up to you.


But if we are talking new cores, it's the same thing. When a new design/core is ready, you ship it. You don't wait around for your competition to come up with something that's as good or better, you start selling it as soon as you possibly can. There's no possible reason for waiting around and stockpiling them, or saving the design for a later day. The design loses value by the day.

There's a saying in CPU design that missed schedule is lost performance. Meaning that if the design remains the same, and you can beat schedule then you have a higher performing part than if you merely met schedule - because if you miss schedule, the rest of the industry simply moves right past you. If you release a 100SpecInt CPU when everyone else is shipping 50SpecInt CPUs, this design is worth vastly more than if you release the same design when everyone else is shipping 100SpecInt CPU's as well. So, there's a huge incentive to get a design out the door as soon as it is ready. Too early and you risk missing a bug or a manufacturing problem. Too late and you lose money. There's no incentive to hold on to it and hoard it - it loses value by the day

 

[Bremen]

Ok then, why hasn't Intel released faster P4's ?? Given the number of people doing FSB overclocking it seems reasonable that they could easily release a 1.6 or 1.7 GHz part today. So why don't they? Unless you can verify that there actually are problems with releasing those parts today :0)

 

[Vrangel]

<< , I see Intel and AMD kind of working together in slowing the pace of introducing new processors >>


My resume is in the mail.

 

[pm]

Why aren't there 1.7GHz Pentium 4's available? Because the design isn't ready. If it was ready, it would be shipping. Just because someone can overclock a part to 1.7GHz doesn't mean that that same part could be marked as a 1.7GHz CPU and sold in the retail channel.
First off, there could be some bit of code that this overclocker hasn't stumbled across yet that will crash his/her system. I remember when Unreal was first released a whole bunch of &quot;stable&quot; OC'd systems crashed all over the place. They'd phone Epic, Epic would say slow down the clock, and everything would be fine. There have been others. Recently a whole bunch of people complained that 3DMark2001 crashed their stable systems. In any case, customers who paid money for a processor don't want it crashing on random bits of code, and generally get upset and demand recalls when such things occur.

Second, OC'ers are very careful on heat. They use well ventilated cases, they put on big heatsinks, they monitor their temps. This isn't practical in the real world where systems need to be used in very hot climates.

Third, OC'ers often push up the voltage. They aren't concerned that the little bump that they just did shaved a sizeable percentage off the life of the part, they want the speed now and they think it will last long enough for them anyway. Corporate customers want their machines to work for the advertised amount of time. So there is a limit to how high a manufactuter can set the voltage and keep reliability high.

And the last one that I can think of right now is that chips slow down with time. It may run at 1.7GHz today, and probably tomorrow, but in two or three years, it may not, and in 5 years probably not. A company needs to mark a product at a rated frequency and a customer can rely that it will work at that speed even over a long period of time.

There are probably more reasons.

I can assure you that when 1.7GHz Pentium 4's are ready, they will ship. This is the way the industry works.

And the comment that AMD and Intel are working together to slow down progress - this is almost laughable. If they did, then some startup would show up out of nowhere with a new wonder-product, have TSMC and IBM fab it for them, and they'd make a fortune. I can assure you that Transmeta did not originally intend Crusoe to be a mobile/laptop part, they wanted something that would compete in the desktop space, but they weren't fast enough and are now in the mobile arena. There are other Transmetas out there. Even so, if AMD and Intel somehow came together to slow down progress, Transmeta and Via would keep moving forward and would improve their products and their products would eat market share.

 

[seewhy]

Well, what Wall St. don't like to see is cut throat competition between Intel and AMD. Because that really reduce the margin for both company. Let's say AMD introduce 1.5 Gig Athlon today, which has to be limited in quantity because of the yield. How much can they charge for 1.33?? Not as much as they are charging now I would say.

That is the whole point of managing your product mix, and based on the availability of 1.33 since month ago, I will have to say AMD has the capability to introducing faster chip, they are just holding back to increase the margin.

 

[Bremen]

Well I guess we won't agree anytime soon so I'll let it rest :0)



<< chips slow down with time >>



I've never heard of this before (or I forgot). How much of a concern is this, and how fast does it happen? I've been under the impression silicon is mostly forever (unless you start scratching it ;0)

Oh, and I find it hard to believe that in the 5 months? since the P4 was introduced that Intel has been unable to produce a faster version. The thing was designed to easily scale the MHz ladder...

 

[Remnant2]

I just wanted to say that those were a couple very good explanations PM, I know I learned something.

I think another bit of evidence for those who need it is in the P4 introduction. The P3 core ends at 1ghz, and yet the P4 is introduced at 1.4 and 1.5, leaving a large frequency gap. According to the &quot;hold-back&quot; theory, Intel would have released a 1.1, 1.2, 1.3ghz P4 first, milking those as they went, until finally reaching the 1.5 they did launch at. Of course, there is the problem of performance -- a 1.1ghz P4 would definitely have NOT have made a good first impression -- but I still think this example is valid.