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.
Keith,
if it was still a S939 part, would make it a ~5300+ part? DDR-2 causing a big performance hit for AMD?
Now that you mention it, maybe I don't even believe that it is model 5000. The progression should really be that 5000 would be 2.6 GHz, 2x512K. 2.6 GHz, 2x1MB should be 5200.
Is AMD going to discontinue 2x512K parts? it does not make sense, while AMD is at 90nm. Will 2x512K, 2.6 GHz be 4900? I am not sure.
Anyway I look at it, the story does not fit with much of what we know about K8 and the model numbers.
BTW, regarding the DDR-2, I think it in 667 speed, it should gain in vast majority of cases over DDR1-400. DDR2-800, if you can get one with cas 4 will always be at least as good, in some cases (bandwidth limited, DC) much better.
Joe
Keith,
- A64 X2 5000+ clocket at 2667Mhz, reference clock from 200Mhz up to 333Mhz
- HT stays at 1000Mhz (4x250Mhz), DDR2-667 is supported, but the memory controller is supposed to also support DDR2-800 later on with a 400Mhz reference clock
Like the first leak on M2 from vr-zone, this info makes no sense to me, unless there are some unforeseen changes in M2.
All the current chips up to (and including Rev E) derive ther clock crom base clock, which is constant at 200 MHz. From there, it is multiplied by HT multiplier by 4x, 5x to derive the HT clock. It is also multiplied by CPU clock multiplier to get CPU clock speed, and from there, it is devided by memory divisor to get the memory clock speed. There is no reference clock, only derived memory clock.
The only way any of this would make sense if AMD started supplying 2nd clock directly to memory. While it may solve many strange issues (such as uneven clock for memory at some CPU clock speeeds), it also means that some new mechanism must have been created at the intersection of these clocks, memory controller.
As far as their claim of HT at 1000 MHz (4x250MHz), it is simply wrong. HT is 200MHz x 5.
So I don't believe any of the technical info from the source. Only things they could not have screwed up, such as that it is an M2 socket, it is DC, and it will be called X2 5000. I have hard time believing the clock speed. I think it will be 2.6 Ghz, not 2.667 GHz.
Joe
Keith,
Other way to reed this is that Sun is late with its own blades, solaris customers are asking for blades, and Sun has to sell IBM hardware to satisfy them.
Joe
wbmw,
The predictions this year were for a 200M CPU market, and the 3rd quarter is usually smaller than the first and fourth quarters. I think Alan's estimates are spot on, given the market share research data and the projected market size that I mentioned.
As I said, those predictions were too conservative (and as I posted, Q3 is up over Q1 and Q3. Here is the latest data point:
PC Sales Jump in Third Quarter
... Spurred by low-priced computers, global shipments of PCs increased more than 17 percent in the third quarter...
...As for vendors, IDC figures show Dell with global shipments of nearly 9.5 million for the quarter, and with U.S. shipments of 5.6 million. In percentage terms, those figures gave Dell 18 percent market share globally and 33 percent in the U.S., both of which were up slightly over the same period of last year. Year-on-year, Dell had 17.8 percent growth internationally and 12.2 percent growth in the U.S....
[Me: 9.5M / .18 = 52.77M]
... Gartner recorded 9.2 million shipments for Dell globally, for a market share of nearly 17 percent...
[Me: 9.2M / .17 = 54.11M]
Now, add to these sales of servers, workstations. Then add DIY buyers who buy only CPUs, not PCs, and we may be looking at somewhere between 55M and 60M CPUs sold in Q3. This easily supports my claims on AMD unit shipments.
http://www.pcworld.com/resource/printable/article/0,aid,123213,00.asp
Joe
Keith,
Talking of Horus 32-way he said the beast was ready and we should expect a big announcement from a major OEM soon..
We can start guessing who the OEM is. It Teir-1, I guess the order by likelihood is:
1. Sun
2. HP
3. IBM
4. Dell
I would probably be most happy with HP, since we could expect the best time to market (of the 4).
I am not sure if anybody below (Fujitsu-Siemens?) can be called major OEM in the server market.
Joe
Chipguy,
How can GM be above 100%? Back to school for you.
Yup, I caught it in my next post, as in:
I think I did something wrong with my margin calculations
http://www.investorshub.com/boards/read_msg.asp?message_id=8254675
Joe
wbmw,
all that matters is what MDR chose as their basis of reference
I asked you for evidence of that. I provided you with a logical reason why it makes no sense to do that. You have to chose between F-up in chosing a processor and F-up in calculating costs. You can have 1 or 2 F-ups on their part, buy not zero. Tell me where they F-ed up.
The analysis isn't flawed, it's just data without context, unless you choose to purchase their $3000 report. Speaking of which, do you think if their data were totally bunk, that it would successfully sell for years upon years at the prices they charge?
The whole report may provide the series of assumptions supporting it. At some point, (such as now) you are in position to test some of the assumptions and realize which one was wrong.
Was there an assumption about fab utilization, yields? They may have had assumptions about those at the time of publishing. The data came in and they are off.
Ahem, largest volume manufacturer... 300mm wafer plants... 2 year old mature 90nm process without SOI or additional metal layers... cheaper packaging... low cost labor test facilities... the list goes on and on. If you want to live in the fantasy that AMD can beat this because they're JUST THAT GREAT, then be my guest. No one with half a brain will buy your hypothesis, however.
Log of backtracking there... Obviously I did not question intel's capacity to deliver volume, but the cost, that was supposed to unrivaled, but it is not.
Actually, you said (in relation to AMD), "Intel is not that far above $100 either."
I take that back. My estimate was for roughly $140 blended ASP, south of $150, but your estimate of $161 is more reasonable.
AMD's ASPs were $70 before they came out with the K8 core. I really tend to think they are higher at this point. Furthermore, Intel's ASPs were based on limited information, but now that they break down their microprocessor revenues, they are much easier to calculate.
I think AMD is at the high end of the historical range. $95 would not be out of the question. This would still result in > 10M processors.
This may be true, mathematically, but it's not bloody likely, and that was my point. You are comparing Intel microprocessor margins of 70-80% to AMD's margins of 30-40%.
I think I did something wrong with my margin calculations, but as far as costs, it is not at all out of the question that if AMD makes and sells 12.5M processors this Q, is at 100% utilization of the single fab, AMD unit costs would may very well be lower.
In order to debate that to the extent you had claimed (Intel +50%, AMD -50% wrt MDR figures), you would have to assume some very crazy numbers
Intel numbers may be closer, but still low. See my other post. Feel free to use your own assumptions. AMD numbers are way out in the LA-LA land. Let's see if there is a (big) revision downward next cycle these numbers come out.
Joe
wbmw,
Thanks for the breakdown. What do you think is Intel's unit breakdown between Microprocessors in the Enterprise group (Server + Desktop) vs. mobile?
Maybe we can come up with the some rough cost reconciliation with the data from financial statements.
Let;s say we allocate costs in proportion to revenue. The other may be hither on cost than its proportional representation, but most of the negative number is probably composed of some executive pay. Ignoring that for the moment (realizing that not ignoring it would make GMs higher, costs lower - maybe you can throw in your WAG on cost) cost allocation to Enterprise vs. Mobile vs. Other would be:
$2566 vs. 1427 vs. 19
Assuming that Chipset, mobo GM is 50%, their cost would be $701.
That leaves Cost of sales for Enterprise group at $1,864M. Assuming 25% of 45M desktop and server units, or 33.75M units, I get unit cost in this segment of $55 per CPU. This wouild include Xeon and Itaniums, resulting in Netburst desktops at $45 - $50.
With Microprocessor revenues at 4936, cost at 1864, the gross margins, strictly on Microprocessors would be 164%. BTW, this is a number that Intel does not really want to publicise. Understandably.
On mobile side flash complicates things somewhat. Assuming breakeven (likely small loss though), we can take the cost equal revenue, 50% margin on chipsets and I am left with cost of sales for mobile of $535M with 11.25M units. That gives me unit cost for mobile of $47.56. Gross Margins of 335% on CPU mobile
Did Intel give any guidance on mobile vs. rest units? I was assuming 25%, 75%.
Joe
wbmw,
No, I agree with the MDR figures, I just realize that when MDR says "Athlon 64", they mean the 1M cache variety.
And your evidence for that is what exactly? You do realize that 512K Athlons outnumber 1MB Athlons by ratio of 10:1 or more, and it has never been otherwise for 90nm Venice vs. San Diego? San Diego's original intention was only for Athlon FX.
So their analysis would be on 2 products with volume so low bearing on AMD cost in general. So what if the cost of 1MB X2 and San Diego was $200 and $100? AMD profitability is determined by how profitable Venice is. X2 and San Diego derived product would still be profitable even with those costs.
And, I wonder why you are clinging to such a flawed analysis? To give it some defense in eyes of Teletubbies?
The FUD is that Intel is by far the lowest cost producer. Teletubbies live by this, thinking it makes Intel unassailable.
It may very well be that it is not true, or that the differences are too small to matter much in market where gross margins on microprocessors easily exceed 100%.
> 100% gross margin on microprocessors is easily true for Intel. Largely as a function of vastly higher ASPs. Only to a minor extend affected by small differences in cost.
GMs are a very important part of the cost calculation, and they aren't just slightly higher for Intel compared to AMD - they are vastly higher.
That's ludicrous. Think again.
You try to paint a picture of Intel's ASPs being close to AMD's, and their costs as having no advantage... well, the numbers simply don't support either scenario.
I am not at all suggesting that Intel's ASPs are approaching AMD's. (I thought you were when it suited some point you tried to make.)
Conventional wisdom has been for a long time been that Intel ASPs are in $150 range, and AMD's in $70 to $90 range. AMD unit cost could be lower than Intel's, and still have much lower Gross margins than AMD. In this business, GMs are largely driven by ASPs, costs by fab utilization. Secondarily by other factors. Maybe you should listen to Andy Bryant (if you can make it through what he says without falling asleep).
Joe
wbmw,
You got that wrong. The $7.267B number is CPUs only. Intel lists their microprocessor revenue now, so there are no chipsets to subtract.
You may be right. At some point in the past when I looked into this, chipsets were mixed in together with CPUs. Where are the chipsets in the current breakdown?
Joe
wbmw,
What quarter did AMD state that they were not at $100? It's hard to believe this is possible without some very heavy discounting, because their price sheet shows that all their mainstream product lines, and even some Sempron skus, are priced above $100.
There were like million times people at various AMD threads went from price sheets, sales volumes etc. to predict high ASPs, and the next thing, AMD comes in with ASP of like $70, still in Jerry days. So heavy discounting is a way of life.
I don't think this is right at all. Intel lists their CPU revenue separately on their balance sheet. For Q3, enterprise CPU revenue (which include desktop and server) was $4.936B, and mobile CPU revenue was $2.331B. If you take the total of $7.267B in CPU revenue and divide it by 40M CPUs, it comes out with an ASP of $181.68. If you want to imagine a larger market, then divide by 45M CPUs, and you'll still get $161.49, which is a far cry from $100.
My WAG is that the market was ~55M units, with Intel at ~45. Take $161, subtract chipsets, which are some $35 on > 50% of Intel CPUs, and this would result in some $140 ASP, which is below $150 (my claim, not that Intel's ASP is $100).
I guess I missed the Teletubbie celebration. The $48 figure looks right to me, especially given the relative Pentium M analysis I made:
So than you agree with me more than with MDR...
BTW, notice I didn't dispute their claim for X2 DC. While it is high as well, it is not totally removed from reality as their A64 number. DC was supposed to be 2x1M. Originally only Opteron, later X2, yet later 2x512K was released, which may have missed the deadline of their analysis. But there is just no excuse for their A64 number.
Gross margin for Q3 was 59.7%, including the 27% of the company's revenue that is from lower margin components, flash, and networking. Compare that to AMD's GMs, and you'll see that Intel obviously has the lower cost of sales per unit on CPUs.
???
BTW, this is a very common fallacy on this thread, whether it is intentionally nurtured or just widespread misunderstanding.
Gross margin depends on selling prices. Unit cost is completely independent of price the processor is sold at.
You can calculate unit costs by trying to separate cost of sales of the processors. Units are more or less known. You can get a blended cost, from where you can go with some assumptions of volumes and relative costs of various CPUs that Intel sells.
Joe
wbmw,
For you as a non-expert in this field and with far less information to argue with the people who have calculated these costs for more than a decade, it's more than presumptuous - it's foolish.
Let's see what these experts have said:
I believe someone has to point out when the emperor has no clothes, and in this case, the missing clothing is customer demand for AMD's SledgeHammer processor. To be blunt, I doubt there is a significant market for AMD's SledgeHammer, and I think AMD is placing engineering resources on the wrong opportunity. Before I explain why I have come to this conclusion, let me say that I consider AMD's Hammer family an excellent engineering solution. I also see AMD's evolutionary 64-bit migration plan for the x86 instruction set as a good conservative strategy. I believe the ClawHammer processor will gain acceptance as a high-performance processor for desktops, notebooks, workstations, and one- and two-processor servers. AMD probably senses that Intel's Xeon processor is vulnerable because its very deep pipeline and small L1 caches are not efficient on server applications. I have a problem, though, with AMD's chances for SledgeHammer in the low-volume, high-margin, high-barriers-to-entry, and ultraconservative server market for four-way and above processor configurations.
AMD's SledgeHammer is designed for multiprocessing systems that use from 4 to 16 processors. With coherent HyperTransport links between processors, AMD can gluelessly connect arrays of SledgeHammer processors, not unlike the capability of Compaq's forthcoming Alpha EV-7 processor. AMD has stated that SledgeHammer will allow it to penetrate the server market by offering a complete solution, from one processor to many, as Intel does. Although that claim is technically correct, what major OEM would actually design and build a server using SledgeHammer? And, perhaps more important, are any AMD customers asking AMD to build SledgeHammer?
A quick review of the top five server OEMs, which (according to Dataquest) have 70% of the server market, does not provide a promising candidate. Compaq has already committed to Itanium, having killed the future Alpha EV-8. Compaq may shortly merge with HP, and HP co-developed the Itanium architecture, making it very unlikely that HP, or a merged HP-Compaq, would choose SledgeHammer. IBM already has the excellent Power4 and has support for Itanium, giving it no incentive to add another 64-bit architecture. Dell has yet to field any non-Intel solution. Sun is committed to the UltraSPARC architecture but will sell Linux boxes using x86 processors. To avoid competing with Sun's own high-end servers, Sun's Linux boxes will likely be limited to one- and two-processor solutions and be very cost sensitive. Those boxes offer a good opportunity for Athlon MP and ClawHammer. SledgeHammer systems larger than dual-processor systems would be too threatening to the UltraSPARC interests at Sun.
http://www.mdronline.com/mpr_public/editorials/edit16_12.html
So this would not be the first time these "experts" were wrong.
First, there is no reason for AMD's ASPs not to be over $100.
I would be thrilled if that were the case, and I follow the subject intently. Unfortunately, my reading of the tea-leaves is that at best, ASP is in $90 to $95 range. At one point in recent CC (few Q back), AMD was asked about $100, and AMD stated they were not there.
BTW, Intel is not that far above $100 either. If you exclude chipset sales from the revenue from, computational division,
you likely end up on $100 side of $150, rather than $200 side.
The predictions this year were for a 200M CPU market, and the 3rd quarter is usually smaller than the first and fourth quarters.
AMD Q1: $750M
AMD Q3: $969M
Intel Q1: $6895M
Intel Q3: $7267M
Also, I could be wrong, but I think the latest estimates are for 2005 to be > 200M units.
Reread what Alan said. He said that $60 was for the 1M parts
That may have been Alan's assumption. But if you go along with assumption, you are also making an assumption that these so called experts F-ed up again, and evaluated the wrong processor.
He said that $60 was for the 1M parts, while the 512K parts were more like $48.
That is a significant concession from $60 (some Teletubbies were celebrating), making it closer to my under $40, but not close enough.
MDR is saying that Intel's small die equivalent (Pentium M) costs about $33 to manufacture. This takes into account a far larger volume of wafers, so the individual cost per wafer is smaller. These also are manufactured using bulk silicon, less expensive 300mm wafer process, a cheaper package with fewer pins, and probably other factors as well. IMO, this puts MDR's estimate closely on track.
Could be. I don't feel as strongly about their estimates for cost on Intel ASPs, because I have not run a reality check vs. cost of sales. You may want to do the excercise...
Joe
chipguy,
ROFL. Yeah what would the folks who designed the EV7 know
about building high performance processors with massive
glueless scalability.
As if that is what is needed for T&L notebook...
Joe
wbmw,
This is just speculation, but my impression of the Whitefield cancellation is that CSI got pushed out.
My take it that the proposal for CSI handed down from the Itanium people got cancelled and Intel is going to start from scratch on something new, originating from the x86 side of things (folks more in touch with real world). Just my speculation...
Joe
Chris,
Interesting info. BTW, Intel will be able to get the new instruction extensions as part of cross license with AMD.
One curious thing I have notice over time of these instruction set extensions SSE1, 2, 3, AMD64 is that there seems to be a 1 year lag for the laggard. I wonder if that is a coincidence or if it is part of the agreement between Intel and AMD.
Joe
Amy J,
AMD doesn't have anything in its lineup that measures up with Intel on the higher-end
I am not sure if you are up to date with product offerings and their performance. FYI, Intel has a lead in 1 segment - notebooks, AMD has a clear lead in 2 segments - server and desktop.
1. Expand lower-end product lines, when entering recessions.
Intel has a broad product line on the lower-end, in particular, in server and desktop markets. However, it was not by choice. It is the best Intel can do. Intel can still charge more money for inferior products, but I am not sure if I would want to be in that situation.
2. Delay some releases of higher-end product lines, until an economic slow-down ends. ( You don't want to push too many new highend products that could exceed the market's readiness because of an anticipated economic downturn. Intel entered the last downturn with too many plans for highend products but Intel & oems learned the hard way to be slightly more conservative during economic slow-downs. )
Sorry, this may work with potato chips or detergent, but not in technology. Intel is one product failure away from going from a world dominant semiconductor company to just another low margin vendor. That company saving product is due in mid 2006.
Joe
Andy Grave,
You just don't get it. The die size of Intel chips is smaller, allowing Intel to make chips ever cheaper. Since AMD cannot match this, Intel will bury AMD day after tomorrow.
But at the same time, the die size of Intel processors is larger. The large size of Intel chips proves that Intel yields are superior. The superior yields allows Intel to make chips ever cheaper. Because of this, Intel will bury AMD day after tomorrow.
Joe
Alan
The 9M number is obviously wrong, because it would assume AMD ASPs > $100, which is something AMD tends to be some $10 under even at best of times. Also, 50M units is very low for this Q.
There is just no way you can massage this to $60, with reality, and at the same time maintain that MDR has a clue here. 1MB A64 have dwindled, as a % of total to nearly a trickle, and this started sometimes in Q2 2004, still under 130nm. 90nm A64 ratio of 1MB vs. 512K is most likely south of 10%. 1MB A64 has never been much of a factor under 90nm, except in very low volume, high end niches. They just cannot mean 90nm A64 and mean 1MB, if they have a clue. My claim is that they do not have much of a clue.
Besides, Venice core covers about 90% of AMD shipments (WAG). These are 512K cores, and include most A64s all Sempron shipments as well. Would anyone with a clue analyze cost of chip production by examining few niche parts and completely ignore volume parts?
The K8 line was open to FUD, while there were still significant K7 shipments, and boy we there was plenty of it. Now this FUD just cannot possibly reconcile with reality. It's funny to see the Teletubbies cheering wrong numbers. (I don't mean you here, you are generally open to rational analysis).
There are some reasonably valid excuses the authors of this study can volunteer today in order to explain why their analysis turned out to be wrong. They can explain which of the assumptions were wrong, and they can explain how it may have been reasonable at the time to make those assumptions...
Joe
Smooth2o,
One thing you CAN be sure of, is that Dell has seen EVERYTHING under NDA from both Intel and AMD and has evaluated in detail the capabilities of both companies. Intel wins in 2H06.
LOL. Like when Dell, in 2004, saw Intel 2005 roadmap, and Paxville on it, Dell concluded that Intel wins in 2005.
Have another sip of Kool Aid.
Joe
Alan,
Those numbers are obviously wrong. A quick calculation shows that A64 cost below $40:
http://www.investorshub.com/boards/read_msg.asp?message_id=8228423
But I am sure that nonsense like those numbers is keeping the faithful content.
Joe
Thanks for the clarification.
I was just guessing where the problem might be coming from (since others namely wbmw mentioned that a bug was discovered during testing), and beside Foxtron, CMT was the next major, AFAICT. A lot more important too. Dual core and CMT are huge features, and it would be very a major FU if the chip was doomed because much less important feature such as Foxtron.
Joe
combjelly,
re: Foxtron
I wonder if that's the only problem. It seems that the CMT is a complex task to undertake as well, if not more. It would have been a major coup for Intel to pull it off.
I think CMT would have been extra valuable for Itanium, because Itanium lacks OoOE. With CMT, at least you could have 2 threads waiting for a memory access due to cache miss in 2 different threads. Otherwise, ability to do something for Itanium core is restricted.
As far as clock changing technology, IMO, this is one area where AMD is ahead. Even Pentium-M is not so hot when some mixed activity is thrown at it, requiring quick clock speed changes. Pentium-M does well on both extremes (idle and full load), but in between is when Turion closes the gap. And Pentium-M is Intel's best effort in Demand Based Switching. It goes downhill from there (Netburst, Itanium).
Joe
kpf,
Something like that...
Joe
highlandpk,
Pat Gelsinger, enterprise director of Intel, denied, denied, and denied again there was a problem with Montecito at its last Developers Conference.
Pat Gelsinger is no stranger to the situation. Maybe some people just have some natural talents...
Joe
Smooth,
I have to agree with you here.
Joe
Keith,
Chartered to start making AMD processors in mid-2006
Does that mean wafer starts?
Joe
Klaus,
At the conclusion of this long thread, I just want to reiterate my opinion (somewhere at the beginning) that I think you are completely off base on this subject. Feel free to point out, when some facts come out, supporting your opinion.
Joe
Keith,
Re: cost to make chips. I think the MDR guys must be secretly drinking Intel's Kool-Aid.
My WAG is that they are off my roughly 50%. In case of Intel, their estimates are 50% too low (WAG), in case of AMD, 50% too high. So that would make it a very worthy contribution to the Kool-Aid thread.
Just back of an envelope calculation:
Cost of sales attributed to Computational Group is about $495M, the number of chips sold is some 10.5 resulting in blended cost of $47 per chip. That includes 512K Athlon 64 (and all Semprons which use the same die), 1MB Athlons, SC Opterons, DC Opterons and X2.
You can try to play with the ratios of the above chips and the cost derived from the blended average, and I am sure you will arrive at Athlon 64 cost of $40 or below, not $60.
Joe
highlandpk,
Unless I'm mistaken, weren't the single core versions also 55W? I wonder how they will stand up - performance wise - to Intel's "saucy little mobile part masquerading as a server chip"?
I don't know how popular these chips are. There are also 68W chips that AMD provides to blade vendors. I wonder why there 2 categories for somewhat low volume chips...
As far as Intel "saucy little mobile part masquerading as a server chip", without ability to run 64 bit OS and apps, it can not really be considered a server part in 2005.
Joe
Keith,
Athlon 64 X2 pricing coming down significantly ...on Pricewatch. As dual-core volumes ramp up, price cuts would make sense.
From being over the list, to being at the list, to being at the lowered list.
That's not such a bad record. Also, the OEM chips are less common these days.
Joe
Keith,
ProLiant Performance Special Report: ProLiant Servers Sweep Benchmark Records
It seems that, since Xeon can't compete with Opteron, HP has tried very hard not to compare the 2. Instead, HP invented a special Xeon only category.
It's kind of like the Special Olympics for Xeon only.
Joe
wbmw,
What may be as telling as the fact that there is only one review is that Intel made it so - by not sending review samples out. Why would that be?
Joe
wbmw,
They grow by large bounds every year, and businesses typically need to build up their IT infrastructure to support the requirements of the future, too. Dual cores and eventually quad cores aren't going to be a nice-to-have, but an absolutely necessity to compute upcoming workloads. Increasing compute requirements has been one of the only certainties in this industry, so I wouldn't bet against it if I were you.
I hope you are right. I am not betting against. I am betting in favor of increasing computing requirements.
Joe
Phil,
I thought the M2 was supposed to have 1207 pins ?
That may be the next server socket.
Joe
tecate,
Hey been talking with Nostradamus lately? I asked you a question, when is the next recession coming?
I can't predict the future, but I know where to find out current info and what happened in the past. Hint: it is not based on how you personally feel, or how Krugman feels.
I am referring to you comment here:
From what I can tell, we are just really getting out of the last one.
http://www.investorshub.com/boards/read_msg.asp?message_id=8194578
Joe
tecate,
re: Even more so when we get into the next recession
When's that? From what I can tell, we are just really getting out of the last one.
Since NY Times is your news source, I am not surprised that you are confused.
BTW, where in particular do you get your info on what's happening in the economy? Krugman? <g>
Joe
Hmm... Quoting myself here:
The decision for Intel was whether it is better to not show up for the game, until a worthy challenger can be fielded, or showing up and losing badly. The first option offers some dignity to Intel, the second offers some dignity to Dell.
It looks like Dell is calling the shots these days...
And now this is what Charlie says:
http://www.theinquirer.net/?article=27122
It looks like great minds think alike.
Joe
wbmw,
When dual core ramps up and costs go down, the price of a 2-socket, 4 core server will be exactly in line with the price of a 2-socket 2 core server today.
True, but vast majority of these 2 socket servers are very lightly loaded, with 2nd CPU providing just some safety margin that people like to have. 2nd core on a single socket server does it cheaper.
If you think that IT managers would scoff at this and move down to the single socket price range to have equal the number of cores rather than double the cores at equal the price, then we will just have to see.
You are assuming that they need (nearly) 2x performance. If they don't need it, will they not want to save some money by not buying what's not needed?
2 socket -> 4 socket price premium can probably illustrate this better. There is a substantial jump in cost of going from 2 sockets to 4 sockets. If DC can do the job, eventually at price of today's SC, why spend money?
I foresee a shake up in the server market. Even more so when we get into the next recession. In those times, cost savings are high on the list of priorities.
Another variable here is that the additional money can be better spent on more memory, which will be more effectively addressable in 64 bit enabled servers and apps, relieving some of the disk I/O bottleneck, which may result in greater performance than money spent on extra socket server mobo and another processor.
Take a look at Dell submission with single lowly Paxville DC:
http://www.tpc.org/results/individual_results/Dell/Dell_Inc_2005_092605_ES.pdf
That's a lot of work a single lowly box can do. Do you think that typical 2 socket server buyers have 2x need for processing than what this system can deliver?
Joe
Buggi,
So, what could be the number
in Q3? Don't know - seems around 600-800K.
Based on my estimate, AMD sold between 200K and 300K DC processors in Q3. My WAG would be double that in Q4. So it looks like the Intel plan to bury AMD with dual core has not worked in Q3, will not work in Q4, and AMD will have enough capacity in Q1 from the first Fab36 output.
Joe