Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
hw,
Guess you won't be buying stock in Apple anytime soon.
You guessed right. The market seems to be (so far) giving the deal (mild) thumbs down.
Joe
Keith,
In addition to the ones you mentioned, ATI must be taking share as well, mostly from VIA, I guess.
Joe
Can somebody remind me what Bensley is? Is it one of the chipset, platforms or yet another CPU codename?
Joe
Keith,
That´s all nice and all, but I just do not see IBM and AMD combining their respective CPUs, which is what Opteron + Cell would be.
Time to market on this would be something like 1 to 2 years, IMO, which, unless this started long time ago, is kind of irrelevant. Cell-like CPU from AMD is IMO a much more likely possibility, as the next generation AMD CPU.
Joe
T64,
Doubtful...my guess is that Intel must have promised Apple generous amounts of Yonahs for the Dual Laptop Market given that Laptop sales are now currently exceeding Desktops 53% to 47%...my WAG is that Apple comes out with x86 laptop before all else because Apple is sucking wind in the portable market...
Are you saying that now, in 2005, apple will take a further detour on the way to x86-64 market via x86-32? Nothing from Apple will surprise me any more, so it may just as well be the case.
Joe
T64,
Are you thinking this through or are you just shooting this out there...usually they design this stuff for 2-3 years and are very specialized forms of processing with low cost designs; this is not something that you would swap out in six months
The length of the design cycle, and the fact that the steps are dicrete, with not much continuity makes it harder to keep the design wins. Where are those who made the CPUs for the last set of console, and why don't they own this market?
if IBM truly loses this console market it will truly be incompetence and all of IBM management should just quit their jobs...;>)
Nope, intead, they will just stand around, and wonder what hit them, just like today whe they lost Apple business.
Joe
Rink,
I did that on SI too 10 times or so during the last 18 months, although I called it Opteron with Cell elements, either derived from Niagara/The Rock or Sony's Cell. It looks though AMD will go quad core before implementing a cell-alike processor if the Inq is to be believed today.
I think Opteron with Cell like features is a lot more likely than Cell with Opteron core, and, BTW, I have some doubts that either Niagara or Rock will see the light of day standalone, much less part of Opteron design.
Chipguy, Keith, before saying this is nonsense, Ruiz mentioned a couple of months back that they were not sure at that point whether to go quad core first, or implement different cores on one die first (cell-alike). It's not exactly out of the question that AMD will still go that route. When is even a bigger question.
Cell like approach seems to be the direction AMD is moving. As far as what comes first, it seems to me that quad core has to do with manufacturability and some limited tweaks some components (SRQ, XBar).
The other one, cell like approach, seems to have a need for new cores or core components to be ready for market. Also, there may be a time to revisit the communication infrastructure between the cores. That seems to be much more open ended, longer term project. It will come at some point, but quad core was apparently a much lower hanging fruit.
hw,
That's the point. Are you trying to say Apple cannot grow?
From the pitiful place where it is today? Even from there, any kind of growth is way down the road. We will first see a shrinkage of that low market share to a fraction of what it is today. The question from that point will be that of direction: down to zero, or up, on a slow road to recovery that may take another year or more just to get back to where Apple is today.
Joe
hw,
That is just pure unadulterated crapola! Of course they care and have a large stake in making it a great success.
Where do you think 8 out of 10 former Apple customers would go if Apple were to go out of business tomorrow?
Apple means nothing to Intel. Making things difficult on 1 of the 2 remaining competitors means a lot.
Joe
T64,
I don't think so...IBM will have the game console market all to itself...if they can make it for $25 and sell it at $75 to sony, Microsoft, Nintendo, at 10 million per quater that's 500 million net profit...
IBM used to have all kinds of markets all to itself, and still managed to piss all of them away. Why should Cell and the console market be any different?
Joe
Keith,
maybe, but that doesn´t look like a sane business model for the next year.
Apple should have had, at minimum, Dothan based notebook for sale right now.
Joe
Keith,
what does this move say about Cell wrt. PCs?
Pravin suggested Cell with Opteron core (on SI thread)...
And, when will IBM learn who their partners, and who their competitors are?
When their pink slips have Chinese characters on it?
Joe
Keith,
Now honestly: What sane person that knows of this switch will still buy the current Macs?
Now, you are not talking completely sane crowd to begin with. It may well become a nostalgia item. Last Mac before Apple sold out to the Evil Empire. Coke Classic vs. new Coke.
Joe
decaw,
Apart from stupidity, and I don't think he is stupid, I can only conclude that Intel is putting up $$$ to support AAPL during the transition.
In which case isn't this a secret takeover? I mean this really smacks of anti-competitive activity.
Which ever it turns out, it is a brilliant move by Otellini. Whatever it costs Intel to eliminate IBM from being a competition in merchant chip market, IBM is the one ultimately paying the bill, since IBM hands over to Intel just about all of the profits from its x86 activities.
So IBM is paying Intel to eliminate IBM Microelectronics from competition. How much more briliant can this get? To Intel, it doesn't really matter if Apple is out of business next year. IBM surely is out of the merchant CPU business (well, outside of maybe some embeded systems and cell).
Joe
mas,
HP to outsocket Sun
looks like the multi-core multi-socket heavens have opened up and are raining hard !
I wonder where that leaves Niagara. What's the ETA and how many cores (much simpler cores, I understand) it is supposed to have?
Joe
chipguy,
If I was an IPF niceguy I would be playing up Apple/IPF
rumours not shooting them down.
That was one of Dvorak's off the wall idea.
My take is that it already risky for Apple to change the instruction set (yet again), and flush a lot of code written for Mac down the toilet. At least the transition to x86 would be last transition, which would add some comfort to companies crazy enough to still develop for Macintosh.
Going to Itanium would probably lose Apple good number of the few remaining software developers.
Joe
chipguy,
IPF top frequency in 130 nm - 1.6 GHz, in 90 nm 2.6 GHz.
You are an eternal optimist, when it comes to Itanium. If you were a little less condosending, you could be Itanium's very own niceguy.
Joe
wbmz,
I am going to side with Tench on this one.
I don't think I was entirely disagreeing with Tench. He just clarified a some things for me.
I use cool n quiet so I can see what the cpu is doing in real time and I rarely see the cpu run faster than 1Ghz.
What utility are you using for showing the speed?
Joe
Tenchusatsu,
In the Advanced Settings, you can change the transfer mode from "DMA if available" to "PIO only."
Thanks for the reminder. I remember doing it some time ago, but I completely forgot about it. These days, Windows seems to default disk drives to "DMA if available".
It has been a few years when I was helping a friend with a project, when we realized that his IDE disk transfers alone consumed upwards of 50% of CPU utilization.
Joe
mmoy,
which version of Office does not install under Windowx x64?
Tenchu,
DMA does bandwidth does outweigh the bandwidth of I/O transaction sent from the CPU, but I am not sure how universal DMA is in low end stuff. For example, I haven't checked this for a while, so it may have changed in latest chipsets, but last time I looked into it, IDE drive reading and writing used to result in high CPU utilization. So I would guess that the data is going through the CPU.
Joe
Tenchu,
I don't understand what you mean by "separation of memory and I/O busses." Are you saying Intel sticks the I/O onto the same bus as memory? LOL, that hasn't happened since the 80's.
I think he means that CPU communication to memory and I/O devices goes over the same bus (FSB), which is true. In case of single CPU K8 system, those are separate. But when it gets to > 1 way, memory accesses and data to an from I/O share HT.
Joe
Keith,
power consumption of individual CPUs of the same core can differ quite drastically from one to another these days.
I remember people cheering and panicing when some reviews / previews were released of Venice being either slightly above or slightly below Winchestern power consumption.
I have a feeling that AMD will use the best parts for dual core Opteron for HE parts (when available), blades (68W), regular Opteron (90W), X2 in that order (110W?). I may be off in last 2 numbers.
Later, when X2 outnumber Opteron DC by some 5:1 or more, some of the better parts will make their way to X2 as well.
Joe
Keith,
That doesn't seem to make a lot of sense, except if the the variable voltage can be set to lower values more aggressively on single core than on dual core parts, and most conservative on dual core.
Joe
chipdesigner.
Nonsense Keith, they consume less than 2x the power. 1 memory controller, 1 crossbar, 2 venice or san diego cores.
I am pretty sure Keith was kidding.
Joe
Tenchu,
Where's the alternative to the low-end PC for the developing world?
What's wrong with low end PCs, and why should there be an alternative for one in developing world and not in the developped world?
Joe
Keith,
147 and 199mm^2 (2x512KB and 2x1MB)
So that´s extremely close to Newcastle and Clawhammer die sizes @ 130nm.
This has been disclosed when the first "previews" appeared. Since the general comments about 90nm (and yields) have been very positive, bin splits are good as well, AMD is very well positioned on the manufacturing side. The Pentium D distraction may turn out to be a blesing for AMD - or a trap that AMD lured Intel into (as someone posted on these boards).
Joe
Tenchu,
Intel doesn't have much of a choice with Smithfield. Time-to-market requires taking whatever you've got and making the best of it. No reason to wait around when you can just reduce single-threaded performance by 20% in order to get a multithreaded performance hike of up to 60% (counting both the penalty due to clock speed and the gain of an additional core).
Since AMD did not originally plan on introducing dual core for desktop until later, Intel was under no pressure to introduce it now. If you have a pretty bad processor on your hands (Pentium D), why not wait with its introduction until it's ready and/or until it is necessary?
I think this introduction of Pentium D is strictly a defensive PR move on part of the Intel, trying to compensate for AMD's introduction of dual core in servers. The hot air (figuratively and literally) introduced by Pentium D may help help slow down defections on the server side.
Why introduce Pentium D on desktop instead of server? Because Intel has taken a some beating in its reputation of invincibility and inevitability in the server market. Intel can't risk making things worse by unleashing the Frankenstein processor on the server market, where the buyers have at least some clue.
Hence, desktop Pentium D: for the the clueless buyers and for influencing clueless business journalists and analysts.
Joe
Dan,
And if presler isn't competitive enough, or if it's late, Intel could become number 2 in the CPU business.
Intel remarkably insulated itself from the possible negative effects of having an inferior product. Intel has gone more than a year with Prescott for majority of its unit sales, without much of a loss of the market share to AMD. Another year wouldn't make Intel #2, but AMD would probably start to make greater gains.
Joe
wbmw,
When AMD makes enough X2s for Intel to worry, they will have a more competitive answer with Presler.
Let's just wait and see who makes what. I don't see the dual core parts as percentage of total processor shipments any different between Intel and AMD in 2005.
What do you think, BTW, will be Intel's % of dual cores sold in Q3 and Q4, and that of AMD?
Joe
wbmw,
Intel first surprised AMD by working on a dual core variant in secret, and they further surprised them by the project being a simple and clever joined die approach, rather than spending the time to create a crossbar between cores.
I am not sure much of a secret, since Intel announced dual core as plan B, when plan A (4 or 5 GHz Netburst) did not materialize.
As far as clever, it is as clever as making lemonade out of 2 lemons, but it doesn't lose it's lemon-like nature.
No, it's definitely Dual Core. Don't try to obfuscate things by making up personalized definitions. Whether or not you include a crossbar, or even if the cores are physically touching, it doesn't change the fact that it offers the performance of a dual CPU system in a single package.
It's a dual Xeon in a package. It just fits (barely) into the broadest dual core difinition, but you will unlikely see it in a text book as what a dual core processor is.
Take a step back. Smithfield is Intel's response to AMD being ahead in dual core designs, and it's a brilliant answer. Instead of playing AMD's game, Intel came out with a design that works nearly as well as a "true" dual core Prescott, but it came out much earlier and it's cheaper. If you were to look at my point of view, why shouldn't I turn this around and say that if the design says Intel Inside, you will be condemning it, indiscriminately? You are equally polar on most debates.
If you look at all back at all of the Banias / Dothan debates, I was not on the other side, because I recognize it as an elegant CPU. I haven't said too many bad things about Montecito, which may turn out to be a good processor. Both have a problems (in case of Mentocito severe) in terms of support of strategic instruction sets.
But any way I look at Smithfield, I see a montstrocity, a Frankenstein like design. (BTW, does anybody remember the Inquirer article of some kind of demo of a stacked die connected with wires? Looks like the Inq was right on the money that time as well. The wires have since been neatly layed out inside the MCM).
For you to call this clever and brilliant, while it is consuming 2x to 4x the power of higher performing AMD processors, and at the same time praise the virtues of low power consumption of Dothan (and likely Yonah) is at best contradictory.
Joe
chipguy,
From a PC-centric x86 myopic POV that is arguably true.
IPF competes with RISC based servers which are pricier
than x86 gear and don't run x86 software. But non-x86
servers represent 51% of server hardware sales in 2004
by Gartner's reckoning.
Do you ever wonder why the x86 went from 0 to 49% in 2004 (or > 50% last 1 or 2 quarters)?
IPF is Intel's rapidly expanding
beachhead into the half of the market it doesn't own yet.
It may be a good idea to step back, to see where that beachhead is. It is on an iceberg and this iceberg is about to encounter the Gulf Stream.
Joe
Tenchu,
I can think of two:
- Inadequate response to HyperTransport.
- The Prescott design, even when heat was called out as an issue way back in 2001.
Ok, that's fair. Except, Smithfield = 2 Prescotts, and heat is very much an issue with it...
Joe
wbmw,
Just look at how Intel's Pentium D 820 2.8GHz "joke" gives AMD's 2.6GHz FX-55 a "whooping" in these benchmarks:
If there is a joke on that chart, it is the 840 (both regular and extra crunchy), struggling (and mostly failing) to keep up with the slowest X2 processor AMD is going to introduce.
Joe
Tenchu,
Smart decision, if you asked me.
Have you ever commented on any current decision of Intel management that it is dumb? (just to keep your comment in a proper perspective)
Joe
wbmw,
From a strategic point of view, Intel planned from the beginning to make Smithfield easy to make and cheap.
I think Yonah will be easier to make. Will it be cheap as well? Or could it be that Smithfield is cheap because it is crappy, Yonah will not be cheap because it will not be crappy?
That's why they got it out to market faster than AMD, in spite of starting the design late.
The fact that it is not a dual core, but a dual processor may have something to do with it as well.
It's also what has enabled Intel to start with mainstream price points, while AMD has priced themselves into a small high end niche.
I could have a field day with these contradictions, but let's just summarize it that it can be crap, but if the crap says Intel Inside, you will be right behind it, indiscriminately.
Joe
combjelly,
Maybe. But 84% looks really impressive, doesn't it?
Maybe it is 84% success rate in its class.*
* Means class of processors that are overpriced, and incompatible with existing software.
drjohn,
Actually my predication for availabilty of 64 bit celeron,s will be almost bang on!
Except, your prediction was for August 2004, and as of now, even August 2005 is not yet clear.
Joe
wbmw,
Precisely. AMD couldn't ramp 130nm K8 without heavy volumes of half cache and K7 Sempron parts. Similarly, ramping X2 will require that the majority of parts be 90nm or half cache.
Half cache Newcastle is roughly the same size as half cache X2 4200 and 4600. Full cache Clawhammer / Sledgehammer is roughly the same size as Dual core Opteron, X2 4400 and X2 4800.
Re: I said I would not be surprised if AMD beat the 1M target for a the year.
Which is ridiculous,
250K Opteron + 750K X2 does not seem too far fetched.
unless they drastically reduce prices.
Depends on much of a factor dual core is in the marketplace. First of all, AMD should introduce at least one, 1 speed grade lower processor, such as 2 GHz 512K with model number of 3800 X2, price it below $300, and just deliver surplus parts at this speed.
I am sure there will be a price cut between now and beginning of Q4, which will bring the prices down.
But I don't thik AMD should chase these design wins on the X2 side. They should do so on the Opteron side instead. On the desktop / notebook side, AMD should instead go for more volume, and design wins, to lay the ground work for higher unit production capacity (as many X2s as market desires) in H1 2006.
First, I don't think you have a realistic break-down of full cache 130nm parts.
Breakdown does not matter, because even the half size cache 130nm have the same size as half size cache X2s.
But FYI, K8s were 100% 1MB until late Q1, early Q2 2004, when the rate was already around 1M units per quarter.
Second, 8-9M units would cause AMD to lose market share, given that the market continues to grow.
Normally, the profit would be a priority, not any particular unit shipments. But as I said, in Q4, AMD should be preparing the market for higher unit sales to follow in H1 2006.
In order for AMD to keep up, they have to manufacture even fewer large die parts.
Please explain what you are talking about, because we don't seem to be comunicating. Baring yields much worse than 130nm yields, AMD can replace every 130nm part with dual core 90nm part (full or half size), every K7 Barton with 1MB 90nm part, and every low end Thorton with Venice, and maintain die size parity.
In Q1, more than 1/2 of the product was K8 based, not much of it was 90nm.
If you add this up, AMD could be selling well over 1M dual core units per quarter, and not disrupt anything in any way, actually still have room to grow units.
Then the question is if AMD was at capacity in Q1, answer to which is no, IMO.
Joe
wbmw,
Intel is forcing AMD's hand. The larger, more expensive X2 processors will be a draw on AMD's manufacturing.
How so? X2 is the same size as the 130mm equivalent.
They can do well if they price them high and sell limited volume, but if they try to ship millions like you think they can, it will quickly drain their capabilities. Just you watch.
I am predicting? I said I would not be surprised if AMD beat the 1M target for a the year.
As far as draining AMD capacity, AMD shipped 1M+ 130nm K8s per quarter since Q1 or Q2 2004, without draining anything - quite the opposite, the units went up to 8-9M unit range per quarter.
Joe