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.
fpg, chipguy: re: You think it's impossible for AMD to figure out how to keep a second FP unit fed?
K8 already has three separate FP exec. units, i.e. FPMUL, FPALU, and FPSTORE. They make use of the same single FP pipeline though up to the moment the instructions are executed.
Kind regards,
Rink
Grimes, just to be complete here's a link ($$) to your story: http://online.wsj.com/article/0,,SB109173698642384153,00.html
Kind regards,
Rink
Joe, I think what is more likely is that Sun will commit to purchasing certain minimum quantity of special purpose chips, and for that, Sun will get some exclusivity for a period of time.
That's likely, I agree.
These things could include:
1. More L2
2. More cores
3. More HT links (new socket)
4. Wider HT links (new socket as well)
Why not:
5. Chip to chip interface via capacitors on the side of the chips instead of routing via the pcb (SUN invention, first mentioned about a year ago). It's certainly a rather nice fit for Opteron, whether it'll happen remains to be seen. Note that this will allow them to be faster, and wider, though not more ht links.
Almost would not mention this one as it's so unlikely but as I wrote the above I migth as well mention this one too:
6. The Rock - Opteron hybrid (i.e. Opteron with Sun cells added to it). Rather remote possibility but one that comes back to my mind every couple of months when I'm trying to think out of the box.
Kind regards,
Rink
Kpf, tx. Do you know when that's supposed to be due?
Kind regards,
Rink
Joe neither 64b Linux nor Win64 runs on Nocona yet (maybe there will be new beta releases of both soon but I don't count on it). So for now it makes perfect sense that they only test it in 32b mode (sigh).
Kind regards,
Rink
Paul, re: Test 64 bit Xeon 3.6GHz.
They test dual processor systems with single processor benchmarks?!
Kind regards,
Rink
Keith, fwiw this was the link to the $75 transcript:
http://www.fulldisclosure.com/conferencecall.asp?date=20040714&sortby=startdate&client=cb#
It turns out the WSJ transcripts that I posted the link of in previous quarters were in fact done by fulldisclosure / cnbc under a contract that allowed WSJ to republish them...
On another subject: That was good news btw on that quad Sun server!
Kind regards,
Rink
Keith, I've checked a couple of times and the WSJ hasn't published any cc transcript that I can find.
I've seen somewhere though a site that has cc transcripts for sale (around $70), but as I've listened to the cc myself I found it too expensive. Via google search on something like 'cc transcript amd'.
Kind regards,
Rink
Tx DDB, I suddenly remember some stuff from college again...
Guess CJ is right too though, even though I also think the cleanroom enlargement did lead to a bit higher possible wafer starts (despite the additional metal layers). Hard to say when you're on the outside.
Kind regards,
Rink
CJ, re: What they didn't do is add all of the equipment to run SOI completely and in parallel with the bulk Si.
How do you know this (you sound quite sure of yourself)? Is it all based on the presumption that AMD can't afford two lines, or because you know the facts?
It's just that knowing it's either one of these would give me more peace of mind on this question.
Another question (again trying to drain your knowledge...) I always guessed that one could add all metal layers using basically one machine (via multiple passes inside the machine), as I simply couldn't imagine 9 to 12 machines in serie (one for each layer) for all metal layers as that strikes me as possibly inefficient. So, should I take it I'm wrong here too?
Kind regards,
Rink
CJ, re: They just don't have room to have different equipment for the different processes.
I think you might be wrong: AMD enlarged their fab30 cleanroom summer 02 or 03 (which one it is slips me right now; think it was summer 2002 in order to prepare for 130nm SOI), and additional cleanroom space isn't worth anything without additional equipment. They will at least have partially different lines, but likely just have more than one.
I don't know AMD went with new scanners either, as I think I read they reused the ones they used for 250nm (for a short while) and 180nm also for 130nm.
It's possible though I misinterpreted what I think I read a long while ago, as this is not my area of expertise (still think I did read it however...)
Kind regards,
Rink
Keith, cpu rev was down 7%, asp's were down 'slightly', don't recall hearing units being down but per their other comments it is likely they were down a bit too.
Kind regards,
Rink
UpNDown, that's cool.
Double posting apparently is sometimes not intentional (your case), and sometimes it's needed to complete another discussion (my case), and sometimes it's just convenient (important news that everyone would appreciate to hear as soon as possible). As long as it's done showing proper respect to the poster that dug up the information I think that's ok. I fully respect your opinion and appreciate your posts too.
With regard to the remarkedly more important Sherlock matter I think, with all due respect for him ofcourse, that he's only mostly right: The dog did in fact do nothing which means he didn't get sufficient stimula before black out period. So far so good, but it doesn't necessarily mean nothing happened, just that - especially any negative - net effect before backout wasn't material enough. Q2 is historically rather back end loaded. So up to the very last moment this Q is hard to predict (both positively and negatively). My gut feel is something did happen (transition to 90nm effected capacity to a tid bit more than the planned extent???, did those movements in on-shelf retail products during the Q that I think I saw signify something reflecting this???). Continuing to guess in the dark I'd say they might have had some either painful or happy barking stimula, but that at the very end it turned out that at least the net effect was simply not material enough to make barking a must. For the moment I slightly think about the former more than the latter. Both cpu demand and supply (capicity) might have been a bit awkward (not regular). AMD manages storage tightly and I think we've witnessed some shortages related to slightly wrong projection of demand and/or available supply. Flash is less in question and can well have met top of guidance. All in all it certainly won't be a disaster and if they grant us the insight we might hear some rather interesting stories. Also the future will look at the very least reasonably ok, because both product and production lines are basically approaching optima forma. Lastly I think that hence the current stock price reflects more uncertainty then there really is. FWIW (probably almost nothing as that's how much I know)...
Kind regards,
Rink
UpNDown, Mmoy, please understand my perspective.
I tried to answer a point from another poster over at SI, namely why AMD was down so much in the morning with this post about Lehman here: http://www.siliconinvestor.com/stocktalk/msg.gsp?msgid=20282638
From that post some of us tried to get a somewhat better picture about AMD's situation, and a couple of posts later I found Buggi's info rather useful so I copied it crediting him for the information.
About copying your intellectual property, I realize it might comfort you I never copied any information either of you posted (just teasing...). Also there's enough info from SI copied over here without asking. For me it's a non issue, and here's why:
It is common practice to copy posts from posters at other boards without asking about it because the internet is a public forum. If someone tries to be polite he credits the originator by mentioning him/her by name. Even better link to the original post so people can read and add to the original thread. I always do that. Mostly I also don't copy the whole post, but only a teaser or the core of it. Also, to be complete, for as far as it concerns SI and IHUB I usually continue the discussion at the tread I see the information first, and if the topic is started over here I frequently discuss it further here. I always start at SI though.
If Buggi would ask me not to copy any part of his posts I will honor it, as I don't like being considered inpolite for a reason.
Similarly I won't copy your information Mmoy because you made it clear you don't want it - that's fine by me. Lastly, if you don't like other people than me copying your posts too you might find less discomfort in a private chat room.
Kind regards,
Rink
EDIT: This is too weird UpNDown. I think you might agree that it's at least peculiar that you posted the time of the next earnings call and CC 30 mins after DARBES did that on SI (without reference). Don't like checking any other posts so I'll stop here (post is too long anyway).
Tx for the update Buggi. Copied your post to SI too (disadvantage of communication crossing msg board borders): http://www.siliconinvestor.com/stocktalk/msg.gsp?msgid=20282893
Kind regards,
Rink
re: Dothan SPEC results. One reason is L2 cache.
True, but it doesn't really matter: The quite huge 2MB L2 cache is part of relatively small (87mm2) sized die.
Separately the comparison chipaul makes against same frequency (2GHz) comparable memory bandwidth (single channel) Athlon 64 isn't completely relevant either: It's comparing a top ended (for now) 90nm cpu, with a mid range 130nm cpu. Newcastle now, and Winchester later scale rather significantly higher in frequency than Dothan. Compare top end Dothan with top end Athlon 64 single channel and you'll get a better picture of the performance difference. Winchester (the 90nm A64) will (my guess) be 512KB L2 chip, and be comparable in die size. When it arrives we can equally well compare top end Winchester with top end Dothan and get an idea how much better Winchester performs than Dothan. To be complete: It's currently not possible to compare Dothan with an equally low powered mobile A64 - for that we have to wait till 90nm mobile arrives.
Kind regards,
Rink
CJ, Sgolds, the article says first dual cores are same socket. That's in 2005. It separately says there will be a new socket in 2006. So what part is incompatible with previous Ruiz' remarks that customers will be able to plug in a dual core in their current single core systems?!
Kind regards,
Rink
OT, Paul,
Congratulations!!
Kind regards,
Rink
Chipguy, re: What the heck interest is EM64T to someone in the market for a Superdome class machine? Anyway HP reported greatly increased IPF sales in its most recent fiscal quarter. The EM64T announcement was made less than a month into that quarter yet IPF sales growth was the same or higher than the previous quarter.
Remember your own expectations for Itanium systems sold turned out to be a little off for Q1, especially for Itanium (more than a factor 2). A bit like this rather knowledgely person that was mistaken too: http://www.realworldtech.com/forums/index.cfm?action=detail&PostNum=2368&Thread=1&roomID...
Further the number of Itanium systems sold might actually have decreased. http://www.siliconinvestor.com/stocktalk/msg_multireplies.gsp?msgid=20173933
BTW, I know there are indications that in Q1 HP sold more (in dollar terms) Itanium products than half of all last year. But I doubt might not be that much of an increase sequencially.
Kind regards,
Rink
Re: AMD64 chipset's.
In addition to what CJ said herewith a link to the AMD-Broadcom announcement (via Doug/SI): http://www.siliconinvestor.com/stocktalk/msg.gsp?msgid=20182244
Herewith another one from Rupert earlier on this board: http://www.investorshub.com/boards/read_msg.asp?message_id=3214684
Kind regards,
Rink
Chipguy, you *might* have seen this observation by Demone over at RWT, about possibly booming IPF sales: http://www.realworldtech.com/forums/index.cfm?action=detail&PostNum=2342&Thread=1&entryI....
I just copied the contents of it below.
Kind regards,
Rink
Here's the link
http://www.hp.com/hpinfo/investor/financials/quarters/2004/q2.html?mtxs=home-corp&mtxb=B2&mt....
"Enterprise Storage and Servers reported revenue of $4.0 billion, up 8% year-over-year, led by 15% growth in Industry Standard Servers, which posted record revenue and unit growth of 32%. HP Superdome server orders and shipments achieved another quarterly record. Revenue in Intel® Itanium®-based HP Integrity servers almost doubled sequentially."
In comparison HP's IPF sales in 4Q03 were up 60% over
3Q03 sequentially.
So much for the "February surprise" at IDF.
Tx Buggi.
I think it's immage and corporate acceptance. Just like business people prefer being seen in a a4 or a passat instead of a octavia because brand is seen as value too.
Kind regards,
Rink
Darbes, it's only the ')' at the end of the link. Copy into the address field of your browser and delete the ')'... The link goes to Pravin's post that says AMD might be in trouble (shrink delay + power explosion).
Kind regards,
Rink
(Edit: You're echoing Rupert now virtually verbatim...)
Tx Buggi. Those slides are more marketing than content. Is their a reason for citing the same quote from Gordon Moore 3 times? (in 4 presentations)?. These presentations were definitely all out to impress everyone with the don't-forget-who-we-are syndrome. It was a near waste of time to read them.
Kind regards,
Rink
Joey, mirrorbit is technically superior to strataflash (both from cost and performance perspective). AMD has aligned customers by being trustworthy with prices and performance. Those customers will not go to Intel for the same price (Intel does absolutely not have better prices than AMD, at least not publically that I know of), and with AMD's roadmap looking better, and even though Intel's financial situation is looking better customers are not likely to run away (just my opinion). There's some talk and anecdotal evidence that AMD is continuing to gain market share, and that is likely to continue. I think Intel will again gain market share as well but not at the expense of AMD. There are more NOR flash providers with less strong roadmaps / financial positions.
Kind regards,
Rink
Keith, tx. I think that's reasonably perceptive. Still all of this does not justify the extent of the downwards move, or maybe it does for the big ones but not for me. AMD has an edge for at least a year. They don't sit still, which is what these research notes presume (the analysts don't compare against where AMD will be; not one of them; at least there's no text on that). Intel might increase flash a bit and be set for profitability in 2005. AMD is that now, has recently voiced they're continuing to gain market share, which will simply lead to more profit (more share, lower cost because of 110nm ramp, and next gen mirror flash; this is not difficult - this is simple). AMD will gain 2 digit server share by Q4. 4GHz Prescott isn't exactly real news for end this year. 4GHz Nocona will very unlikely have a 1066MHz FSB (always follows desktop speeds with a safety margin of 6 months or more).
XD was a surprise though, for me at least. Didn't think they had it in Nocona. They still need to show it ofcourse, but they even look good when they do it end Q3 (tx to MS). Also Itanium roadmap looks rather strong for big tin that is. Eventually it might come down after all but not before end 2005 if at all.
Just some thoughts.
BTW, I posted your previous post on SI because I think it's insightful.
Kind regards,
Rink
Low Power Mobile AMD Athlon™ 64 Processor Model Numbers and Feature Comparison
From Joe / SI: http://www.siliconinvestor.com/stocktalk/msg.gsp?msgid=20128064
512KB L2, 144mm2, 35W, 68.5M trannies.
Kind regards,
Rink
Correction to previous post.
Acer Ferrari 3200 does have a Mobile Athlon 64 2800+, but it can't be Odessa as it has 1MB L2. This one has a battery life of approx. 3 hours.
Kind regards,
Rink
Re: Acer is due with an Odessa machine this month (thin and light performance segment)
Acer Ferrari 3200 much linked to on this board (so I'll just be joining the club now): http://www.acer.com/APP/AKC/INTERNET/aacpubli.nsf/Web+Pages/D28B2A87421364B788256E8B005D0867?OpenDoc....
BTW, my rather much lower budget but high power Acer Aspire 1511Lmi (a DTR) is just absolutely great - got it just yesterday. Paid just under 720 euro's for it as it was indirectly subsidized by the company I work for (+-50% via a employee private pc project; keep in mind prices for all pc equipment in Europe are higher than in the US). Except on cpu performance and some minor things it otherwise can't compete with this Ferrari though (Ferrari 3200 has a better screen, bit better DVD because of -RAM, one step better gpu, and the low power Odessa). This 1511Lmi is mine now though: http://www.acer.co.uk/acereuro/page4.do?dau22.oid=4810&UserCtxParam=0&GroupCtxParam=0&dc...
Kind regards,
Rink
Thanks Pete. WBMW, what you're saying is currently not sufficiently based on facts (Pete's figures are much more detailed than yours, plus you don't seem to be able to say anything about northbridge power consumption).
An indication of whether you're right or not will be battery life of comparable (mem, gpu, etc...) Intel / AMD low power notebooks.
Kind regards,
Rink
OT - Billy, at least that's what everyone calls him (eom)
(edit for Rupert) Here's the latest Osha comments I could find - Typically Osha I'd say:
(Edit: Rupert, just saw your comments. You might be right in your explanation of 'althon'. It's at least understandable. It does however go slightly against the general picture painted below, but that might still be correct as the general pc picture is bound to be tainted by Intel).
Joe Osha, Merrill Lynch Semiconductor Analyst, made these remarks regarding first-quarter inventory data for the semiconductor industry:
After two years of grinding down, inventory levels moved broadly higher in the first quarter of 2004. The most notable increases were in the distribution, electronics manufacturing and wireline-communications-equipment segments, all of which saw first-quarter inventories increase in days of sales and dollar terms (albeit off of low bases). Personal-computer (PC) original equipment manufacturer (OEM) and wireless handset balance sheets look better. Semiconductor device manufacturer inventories continued to steam ahead.
The overall picture is surprisingly benign, especially given the market's negative reaction to first-quarter reports. We think that investors, spooked by a return to bubble-style inventory overbuilds, may be over-reacting to supply-chain activity that is, in reality, normal for a recovering business.
The PC supply chain spent the first quarter in inventory-reduction mode. Demand was weaker than expected as well, so the result was a slight increase in days of sales from 53 days to 56 days. Dollar inventories declined by 3% sequentially following several quarters of growth.
Reported wireline-communication-equipment days of inventory increased from 67 days to 85 days, which is slightly above the pre-bubble average of 83 days. Aggregate sales declined by 9% sequentially, while reported inventories in the supply chain increased by 10% on a dollar basis. We think that some of the inventory growth is attributable to zero-cost inventory being replaced by new products on OEM and EMS balance sheets.
Wireless handset OEM and EMS inventories remained flat during the first quarter on an absolute dollar basis. However, days of inventory increased from 41 days last quarter to 49 days, which is consistent with historic trends during the first calendar quarter.
Looking at our 30-company sample of semiconductor device makers, inventory in days of sales terms ticked up two days, to 73 days, after declining for 12 straight quarters previously. Inventories have been building in dollar terms for two years, but sales have grown more quickly. We note that device makers appear to have borne much more of the inventory risk this cycle, probably as a result of having decreased bargaining power vis-à-vis their customers.
http://askmerrill.ml.com/res_article/1,2271,18883,00.html
Kind regards,
Rink
Rupert, your post has a very low probability of being accurate because:
1. ML's Joseph Osha didn't write it. (Note: ML does not have to my knowledge more than one analyst covering AMD).
2. Athlon is spelled consistently wrong as Althon.
Kind regards,
Rink
Buck, # of layers remains constant during transition. At later stage 2 layers will be added though (for a total of 11; 9 now).
Kind regards,
Rink
CJ, 83mm2 for 512K and 114mm2 for 1MB don't add up.
130nm 1MB Opteron: 193mm2
130nm 512K A64: 150mm2
So on 130nm 512KB cache takes 43mm2 die area.
90nm 1MB Opteron: 114mm2
90nm 512K A64: 83mm2
So EITHER:
- on 90nm 512KB cache takes 31mm2 die area (in case A64 512KB is a half cache Opteron) OR
- on 90nm A64 is a distinctly different die with different circuits with different functionality (different from 90nm Opteron that is).
My point is I can't explain why although the cpu goes shrinks from 150 to 83 (i.e. 45%), 512KB cache only shrinks from 43mm2 to 31mm2 (i.e. 28%).
BTW, Keith (and me too) just repeat what Rivet said in an investors conference call which was that the 90nm A64 (with unspecified amount of cache) came in at either 98 or 102mm2, and that he couldn't tell which from the top of his head because it wasn't his area of expertise. JFYI.
Kind regards,
Rink
Paul, we'll pray for you too (eom)
Paul, tx but ATI does not provide notebook drivers for user download, but instead typically refers to the notebook vendor (I checked). Plus ATI is known to be somewhat slow with drivers, compared to nVidia that is. Plus once they provide one to Acer, Acer still has to tweak it as you said. In total that's too many question marks for me to even estimate the chance that it'll happen. Hence my question.
Kind regards,
Rink
Keith, Mobility Radeon 9600 driver question.
Would you happen to know if it's likely that Acer will provide an updated driver for their Aspire 1500 notebook (A64/Mobility Radeon 9600) after the release of the 64b version of WinXP? I want to buy one but would like to know if I'm limiting myself for the future. I asked support via email but they didn't answer me so far (that was more than a week ago).
Normally I'd probably wait till the 1710 but as this purchase will need to be part of our companies employee pc project I can't wait longer than end of this month so I'll miss that one.
Kind regards,
Rink
Dan, would you happen to know anything about Intel's Palladium and/or LaGrande that are possibly present in Prescott/Nocona and that are only talked about in Intel presentations?
Kind regards,
Rink
Second quote on half wafer starts in Q1: http://www.siliconinvestor.com/stocktalk/msg.gsp?msgid=19953350
Kind regards,
Rink