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.
petz
"tens of thousands" to "hundreds of thousands" -- I'm assuming 50K to 250K.[/]
Exactly the numbers I have as well.
K.
elmer
The most simple I could think of for is that increased number of connections cannot not be done without an additional layer due to architectural layout-restrictions. (That was the humble thought behind "needs another layer")
But additional layer(s) could be necessary for all sort of reasons, most of them are foreseen and considered designwise. But then unexpected issues occur as well, e.g. what AMD encountered in 130nm node, induction phenomenon requiring not only one, but two additional layers from Thoroughbred A to B stepping to design it out.
These sort of things can be very nasty as you can hardly guess how long it will take to understand what is going on. K.
elmer
Gift. Make it "Prescott has one more metal layer".
For whatever reasons, which can be below or beyond
yield considerations as well. K.
sgolds - 6 vs 9 layers of interconnect
It is 7 versus 8 metal layers
Prescott needs a layer more than Northwood
(see pic 4 here: http://www.theinquirer.net/?article=12373)
And K8 is nine layers including poli, makes 8 metal layers. K.
wbmw
Ah ok. Thanks. K.
'The Duke of URL' - nice one
I can feel the light streaming on me already. K.
'The Duke of URL'
....price reductions to make way for new releases.
Did i miss something? I mean, which release the recent price reduction made way for? TIA for enlighting me. K.
smooth2o - Comments?
I really wonder why iNTEL chose to match exactly AMDs prices for their A-64 3200+ and 3000+ models listwise.
What might be the intention to make AMD look as price-leader here?
K.
elmer
I can only tell you what they should be able to produce if yields are at or near industry standards
That is what you keep posting again and again and again.
However I doubt you will ever do so..
K.
borusa
Chapeau! Playing Intel against its CFO who sold the majority of hit position last week is courageous. Good luck! K.
elmer
if so, they would not have been particulary bothered about as I never saw any number in your posts.
Anyway, it ist nearly impossible after the consolidation of FASL to extract process-information out or the quarterly numbers. Retail-pricing now is the only data to look at, which they can make very misleading if they have the OEM-supplies under control to prevent from grey-market parts swapping into retail-channels. So, definitely less visibility going forward thisrepectively.
K.
upndown
I dont think any company could survive our dynamic economical system for long with a static understanding of laws and contracts. If this would make any sense at all, all enterprises would be lead by lawyers. K.
upndown
Sure, no problem. How much do you think HP and SGI would ask for Intel to be able to slip out of such a restriction? $10 billion? $40 billion? $100 billion?
If contracts exist which prevent Intel from introducing another 64-thing for a while to protect IPF, it would most probably be limited to serverspace. (Otherwise quid pro quo-terms would not allow HPQ to offer Athlon-64 as well, dont you think?)
K.
wmbw
Tooling is completed as of today. For 2003.
Btw, the first step from 95K to 120K CRS was for C4 previously done in Austin and therefore does not increase capacity. As for the rest, dont expect AMD to make waferstart-data public anymore, they realized it would be too easy to figure out frontend process results for everybody. Thats why they do not deliver unit shipments per quarter as well. K.
UpnDown
Whatever clauses in whatever contracts might exist - contracts can always be renegociated if market conditions require it.
K.
upndown
but there's no chance it would ever be answered.
Well, in this interview http://www.eweek.com/article2/0,4149,1362947,00.asp
eweek did not dare to ask the question, nevertheless Craig made this statement:
There is not a desktop OS that really takes advantage of that 64-bit stuff today. If the marketplace expresses a demand for that then we'll look at it carefully and decide what our next move is. We haven't been asleep at the switch on this. Our strategy is to meet what the market needs.
Now, you can read whatever you want from this. But it is definitely a significant course-change of communicating Intels position on 64-bit computing in desktop-space. [Which course would that be? Rowing back is what you mean? Ed.].
K.
elmer
Sure.
After extensive construction activities in 2001, Fab 30’s SMIF clean room now measures more than 11.000 m2, or almost 120,000 square feet. An additional clean room expansion project, started in 2002, is designed to increase the total production and development area to more than 14,000m2, or 150.000 square feet, in 2003.
http://www.amd.com/us-en/Corporate/AboutAMD/0,,51_52_502_509,00.html
wmbw
Sure. You just forgot to mention it is easy to combine two clearspeed processors for 64-bit-computing.
Welcome to Legoland.
wmbw
How would you explain AMD getting more capacity?
30% additional FAB30 cleanroom-space have been tooled recently.
wmbw
Ah ok. For friendly chat it is irrelevant what is feasible or not.
http://www.investorshub.com/boards/read_msg.asp?message_id=1603536
K.
wmbw
Did you get any reply from Clearspeed? K.
keith
Well, from previous consensus for 04 (-19 cents) to current (+19 cents) its at least a step in the right direction.
I mean, you cannot expect these guys to call for a Dollar earnings from a previous loss in just one step, can you?
You know, two months left for this year, two more steps - and
Joe Sixpack will sure be impressed how hard these guys work day and night on their forecasts, updating it every month..
K.
smooth2o
Lunches
It is not recommended to eat big meals in the evening, right?
K.
elmer
Well. I am thinking of many many camels lining up to go through the eye of the needle.
Maybe a lit bit too early to put money on it, though. K.
sgolds
This raises the question what the heck is China doing with all those PCs!?!
Quite a lot of them (actually i guess it is safe to say the majority of 2500+ and 2,4C models) end up in areas outside china. A lot of Chinese exporters use this way to hide earnings and at the same time convert it to SwissFrancs, USDollars and Euros. This money remains outside China, no way to follow by means of statistics. On the AMD side, you have to add some remarked stuff leaving China as well headed for South East Asia and East Europe.
As you take any statistic with a grain of salt, you better take any chinese statistic with a ton of salt.
K.
AMD's Athlon FX51 on allocation
http://www.theinquirer.net/?article=12210
Hmm. Its always good to leak some future shortages to sell FX-51 this month for 600 USD instead of next month for 400 USD.
Seems like AMD is not forced to price-protect its Top-Models anymore - finally these guys get a real grip on distribution channels.
Good.
In this context: This http://www.theinquirer.net/?article=12200 can only solve the problem partially- In USA and Western Europe it might work, but not where the bulk of remarked CPUs are shipped to.
Expect a package without possibilities for Multiplier-Mods by means of bridges as soon as AMD will have a mix which allows to walk away from serving remarkers. (AMD can still allow Pin-Mods to keep the overclocking-community happy.)
This could happen pretty soon: For the first time since 2001, AMD could afford not to flood the bottom-value-channels end of last quarter. (which is, imo, the main reason ASP were up). And which bodes extremely well for this quarters ASP.
Seems uptrading of AMDs product mix is already in full swing not only on the Top, but as well on the bottom side.
Excellent.
K.
gb
Do they share a common die?
Is the pope a catholic?
Sounds crazy
Completely agree.
K.
wbmw
I even talked with the presenter afterward, and he thought it would be a good pair with Itanium. So why do you think otherwise?
A 64-bit System paired with 32-bit-Coprocessors is beyond my imagination. But well, I know there are things existing beyond my imagination. Could you reconfirm this tidbit pls? (And maybe you can ask how this architecture would be feasible?)
TIA
K.
wbmw
They could also be paired with Itanium systems.
Aha. I am keen to learn how that could be done.
Suggest you go through http://www.clearspeed.com/downloads/ClearSpeed_MPF_2003.pdf
K.
wmbw
Let's face it - it's a niche product right now
Hmm. If you consider HPC as a niche...
At any rate, he's not looking to compete with Intel, but rather to complement Intel hardware.
Hmm. Yes and no. I could easily imagine Xeon-Systems partnered with Clearspeed Silicon to build 32-bit HPC-Clusters and Workstations for scientific appliances.
The dilemma is, such systems would be cheap and performant enough to be a hell of a threat to Itanium. For this reason, if we will see such systems next year or so, I dont expect them to come from Intel....
Anyway, the approach to combine "commodity"-Silicon with Coprocessors for HPC is the way tomorrow's supercomputers will be built imo. Not sure the Clearspeed-product announced is what it takes though - 64-bit and Hypertransport-bus is what it would take e.g.) Hope somebody will come up with such a thing. K.
K.
sgolds, iceberg ;)
dont know how easy or difficult it is to make such a product. Maybe someone can provide an educated guess?
K.
http://www.clearspeed.com/news.php?pr=17
Imagine a package of an Opteron-Core and eight of these goodies around it - Opteron HP?
You could call it iceberg as well.
K.
paul, pre-beta
Means pre-beta, nothing else. From the context I assume it is not the build distributed now, but an earlier version.
As for the benchmarks, Dresdenboy posted earlier today in another forum:
Compare:
http://www.tweakpc.de/hardware/tests/cpu/athlon64_fx/athlon64_fx_bench1.php
with
http://www.tweakpc.de/hardware/tests/cpu/athlon64_fx_64/athlon64_fx_64_bench1.php
http://www.sisoftware.co.uk/index.html?dir=news&location=2004_release&langx=en&a= dated 10/02
Shareware-Version-Download
http://www.simtel.net/product.php?id=70476
SiSoft Sandra going the Babco way with their 2004-suite....
Not really surprising at all - a good target to choose, with all the credibility built up over years.
Well, guys, enjoy the money. I cant blame you for that - I probably could not have resisted as well.
K.
wmbw - certainly. K.
wbmw
For example, how do you quantify successful marketing? Is there a specific ROI?
Sure there is. But it is way more than that: The crucial point about it is: Intels business model essentially depends on it.
If Intel cut all marketing efforts to be cost competitive with AMD, what would be the end result?
Nothing less than a broken business model.
That is why Constantine (and others) are right on the spot with their remarks that looking at COGs alone (although correct from an accounting perspective) can be very misleading.
How about anywhere in between?
Maybe possible to a certain, very limited extent. But dangerous. Very, very dangerous: You got to pay people for singing your song.
Business-models like Intels are fragile at exactly this point; that is well known in Santa Clara.
For this reason, expect job- and CapEx-cuts to improve Intels cost-position, not marketing cuts.
K.
wmbz - Bingo <eom>
blauboad
Forbes
http://www.forbes.com/finance/mktguideapps/compinfo/CompanyTearsheet.jhtml?tkr=AMD
Yahoo
http://biz.yahoo.com/p/a/amd.html
Stock Selector
http://www.stockselector.com/profile.asp?symbol=AMD
Wall Street Courier
http://www.wallstreetcourier.com/technician/market-indicators/nyse-short-interest.htm
K.
wbmw
Lot's of silly ideas out there.
Totally agree. The net is so ful of it these days. K.
wbmw (edit)
This feature was in the errata list until recently. And we have the reports of handwarm coolers for long. So I dont think it is that.
Could be the characteristics of SoI-manufactured chips are so that it consumes way less power in idle state that we are used to from bulk-silicon chips - pretty independent from supply voltage.
If so, to get your the warm or even hot it would require maybe ten minutes of full load operation. Which probably nobody did before touching the cooler. Could be as trivial as that, I dont know.
K.
Edit: Your brillant analysis about sorting bins by current fits in here. And Elmers explanation as well: SoI is known for reducing leakage significantly, and by means of that static current.
Edit2: One step ahead: SoI should make great notebook chips then although max power consumption is high. OTOH, the benefit of PowerNow! (or Cool'n'Quiet) is much less than it was for K7.
Anyway, i understand PowerNow! sells, so its ok to keep it.