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.
Didn't realise they uses a faster core for XDR. Does that in itself possibly mean higher power consumption?
Maybe not. How do they get a faster core? By selection or design?
(smaller capacitor, shorter charge/discharge times ....)
Maybe when calengineer has stopped being calroofer he can enlighten us
Cheers
Cor
Hey Threejack,
The lead time on that kind of equipment is around 6 months or more, so this stuff they are installing has been ordered quite a while ago. Also it is in tune with earlier statements made by LeFort several months ago.
But a "lucky" confluence of events it is, maybe, for Infineon
Cheers
Cor
Cal and Elixe, do we have a power comparison from data sheets for DDR2 vs. XDR?
One would be interested in standby power as well as power when in use and I guess one has to add the power in the controller and terminators?
I think standby power is mostly controlled by the refresh activity, so as the core is the same would not expect a wildly sifferent result unless a part of the i/f is clocking like mad in one and not the other.
Any ideas?
Cheers
Cor
PS Elpida has a new type of core which has controlled refresh using 1/20th of standby power (nice for cellphones).
Trivia in the morning (afternoon for me) is good.
Shall we compute the price per teraflop for these things?
One thing, that Maui machine would fit in your and mine garage, not sure the Blue Gene would.
Only a two bay cabinet. We used to have Eclipse computers from DG in two cabinets, well the disk and tape drives were pretty big those days:) I remember making good money upgrading 128 KB memory boards (15x15 inch) to 256 KB (because the 128 B were half populated and I could buy the special chips through DG service). Note that is Kilobytes !!
We would swap two of those half populated boards at a customer site where we did the service for one fully populated board, suck all the solder holes dry with a special tool and upgrade and test it. (memory test typically 5 hours)
Worked until DG started delivering these half populated boards with the holes not punched out everywhere:(
I still have a working Nova 4/X 256 KB ni my garage, which has an IO board in it which I designed. Haven't fired it up in a while...
Cheers
Cor
Hey, Elixe,
We jointly edit and I mostly write the stuff to start with plus model etc. Bill is marketing for us.
Considering Bill’s writings while with Hager, and your feelings toward Rambus, that should be one heck of an article. You guys going to play good/cop bad cop?
LOL, well nobody knows my feelings about Rambus. In fact I have no feelings about the Companies only numbers coming out of my spreadsheets and sometimes my ears.
We will also be introducing a new stock index, the SIPI: Semiconductor Intellectual Property Index.
(Threejack this is an Internet scoop, first mention:)
Cheers
Cor
Hi just a small supercomputer, but meanwhile as you are hacking out the specs of the latest future supercomputer the airforce has bought a XD-1 from Cray for Maui:)
It's only a measly 1.4 teraflops, but ... available now:)
http://www.tmcnet.com/usubmit/2005/Feb/1118299.htm
There is a link at the Fool with mostly talk about the rarified atmosphere in the mountain, I believe it is called hulahula, or haleakala (is that kala from Greek: beautiful?) or something like that.
http://boards.fool.com/Message.asp?mid=22073602&sort=whole#22076520
The computer uses 288 opterons and is very inexpensive ($ 23 M). Of course the Opterons run at cellphone speeds, which keeps them nice and cool
Cheers
Cor
Wow, that was a way around to the Jihads
Although I agree with a lot you say let's not get into that.
However if you think SMIC will sign up with Rambus, that's OK with me
Cheers
Cor
Threejack are you suggesting that SMIC (the partner of Infineon in dram) should have been subsidized by the Import Export bank?
I am not saying that they will not do their expansion anyway, but to help from the USA is a bit over the top. China is going to end up owning the USA anyway by way of "paper" bonds and shares as long as they refuse to revalue their currency against the dollar.
A side effect of imploding dram prices by too much capacity would be that Rambus' royalties go down from those who are paying on a revenue basis like Toshiba and Elpida.
Just a thought
Cheers
Cor
OK but with the 4-bit chips you could then have a 1 GByte PC (which is about the minimum people will be using after 2005 imo
So maybe XDR for pCs after all? Thought for the weekend:)
Cheers
Cor
The 256 MB is based on four 512Mb 16-bit chips.
Q. Do the other chips exist or is this it for now?
Cheers
Cor
Elixe, do I read correctly that there can only be 256 MB (now) of external memory?
at:
http://www.realworldtech.com/page.cfm?ArticleID=RWT021005084318
The obvious advantage of the XDR memory system is the bandwidth that it provides to the CELL processor. The less obvious drawback is that it is a memory system designed specifically for a high performance embedded system. Particularly, the maximum of 4 DRAM devices means that the CELL processor is limited to 256 MB of memory, given that the highest capacity XDR DRAM device is currently 512 Mbits. While the capacity of limit of 256 MB or even 512 MB of memory would not limit the expected primary application of the CELL processor, the capacity constraint does present itself as an issue in applications where large memory capacity is needed alongside high DP FP throughput and high bandwidth. Clearly, the CELL processor is currently not suitable for these environments, but the high speed FlexIO may enable the use of an external memory controller where multiple channels of high frequency, Fully Buffered Dimms may be attached to retain the bandwidth capability but solve the capacity constraint issue. Incidentally, Toshiba is a manufacturer of XDR DRAM devices. Presumably it brought the XDR memory controller and memory system design expertise to the table, and could ramp up production of XDR DRAM devices as needed.
Btw indeed an excellent article, thks for the link.
Cheers
Cor
Hi Threejack, on the Intel board at TMF that is the wrong place to look. Look on the AMD board, there is extensive discussion of the Cell (possible) impact.
Cheers
Cor
This sucker won’t be cheap
Hi Elixe,
Don't know whether you refer to he chip or the PS3 :)
The guys on the TMF board who know about such things estimate the cost of the cell chip as shown at approx. $ 100 and for a later 65 nm version at $ 40 approx.
Maybe there is a plan to have versions with less coprocessors and internal memory for the consumer devices. This is rather expensive for a digital TV.
Cheers
Cor
Something I don't quite get about the Cell chip (see Tom's for example) which has twice the area of a current Pentium 4 (and about double the number of transistors too) is this:
If this thingy is so darn powerful, why do they still need NVidea for a graphics co-processor?
Am I missing something? (very likely anyway:)
Cheers
Cor
Tom's Hardware link about Cell
http://www.tomshardware.com/hardnews/20050207_160318.html
snip:
The memory and memory interface for the chip is designed by Rambus and will be built around the firm's FlexIO processor bus interface and XDR memory. The aggregate processor bandwidth of the Cell processor is estimated to top 100 GByte per second. According to Rambus, FlexIO is capable of running up to 6.4 GHz data rates providing bandwidth more than four times faster than best-of-class processor buses available today. XDR data rates will clocked from 3.2 GHz to 8.0 GHz, Rambus said.
It has a picture of the chip (pretty large)
Cheers
Cor
Hey Cal, thought for the weekend:
There is a capacitor in there (per cell)
Cheers
Cor
Thanks Elixe,
I had guessed < 5% for the total, but if it is less than 3% at samsung, it could be around 2% or less for the total.
Are the only "major" users now PS2 and some models Cray supercomputers?
Cheers
Cor
Hey Elixe
Have you seen a recent number for rdram market share (for 2004 e.g.) Samsung used to have those pages with FAQs about dram but I can't find it; do you have a link?
tia
Cheers
Cor
I wonder when Samsung will put up an XDR data sheet?
Isn't mass production a wonderful thing?
(especially if you don't know the specs)
Seriously, I was in touch with Samsung IR and asked them as I was communicating with them anyway what the 10K price for XDR is.
They replied that such prices fluctuate due to market conditions and could not give an answer. Will try again on a different tack sometime later.
Do we know anything about pricing? I.e. has it been published? I thought DramExchange was going to track them, but am not sure about that.
Cheers
Cor
pretty high on the EE scale
Are we talking vitamin E here?
Cheers
Cor
Hey Paul, it may not be important, but Samsung signed that agreement you linked on june 30, 2003 for yellowstone "for use in future memories"; no mention of redwood and none of XDR (which probably did not yet exist as such at the time).
If you look at Toshiba for example, you will see that they first signed up for yellowstone and redwood and later for XDR.
I want to leave it here, not much point to speculate further; knowing how Samsung makes their contracts with partners in other cases, I personally have some reason for worry.
Cheers
Cor
re: "Seems pretty clear cut to me."
Hey Will
But apparently not under a new deal. So what is Samsung paying for it? 1.5% like in rdram deal, or fixed total royalties like in sdram/DDR deal?
Nothing like supposed 5% royalties here as far as I can see...
Cheers
Cor
Yes Mac, but have you seen an announcement of a Samsung deal.
Btw that page is very misty to me: it shows Samsung under XDR memory interfaces. That's not XDR ram in my book.
I sense a problem.
Cheers
Cor
Block trade, yes.
The broker must have been able to find (in advance) sellers. They are probably institutionals reading Erach Desai's newsletter ;)
I worry about the to be expected legal fees. Might bring profits right down or into negative territory. It's not just Intel, something is strange in the relationship with Samsung as well. They go into "mass production' of XDR (well that's a marketing term, because who is using them right now?) without having a contract for XDR.
Cheers
Cor
Mystifies me too, the tepid Wall Street reaction. The volume isn't there yet either. Shoudl be like 20M or 30M volume.
Maybe tomorrow:)
If it's the MMs then monday will be the day.
Cheers
Cor
I expect to see CELL solutions with different memory architectures.
Me too, Elixe. And PS3 does not equate Cell.
PS3 (or whatever it is going to be called, not PS3 apparently according to the interview I just posted) may have XDR and other cell applications DDR2 or DDR.
Cheers
Cor
Kent, the $ 500 M Intel "gave" to Micron has nothing to do with patent laundering to which you reply.
Itel got shares for that on which they got a $ 1.5 B to $ 2 B profit so they were very happy.
Intel has also publicly said at the time they had no problem with MU. MU had a license from Rambus so 1) they have paid the up-front license fee and 2) if they produced any rdram they paid royalies for them.
Micron produces DDR2, Intel cancels the DDR2 with Micron.
Kent, I think you are quite confused here. The ramp up is slower than Intel wanted, that's all, it is still happening.
Intel may be the dark force behind all of this.
That is possible
... but didn't this last one say the cash flow is 8.6 mill. If my numbers are right that means Rambus' cash flow is negative without the 10 mill from Intel.
Correct, so to say something positive about Intel: they helped rambus through a very difficult period by the $ 10 M a quarter. Do you think rambus would have had enough cash to sue Intel as well? I don't think so.
Cheers
Cor
Xbitlabs interview with NVidea exec, about cell gpu etc:
(thks to Sabatino)
http://www.xbitlabs.com/articles/editorial/display/ces2005.html
Cheers
Cor
Thanks Elixe for the extra info; I was rather busy when that played and did not replay it so don't remember the exact words, especially before the interuption.
Who is Scott? I mean who interupted Harold? (Howard?) I get confused now
Cheers
Cor
The analysts are geniuses
Look at this just posted:
http://www.marketwatch.com/news/yhoo/story.asp?guid=%7BB2F3042A%2D0F7F%2D48B6%2DBED0%2DCB2A255062CC%....
Looking ahead to results for the first quarter, Chief Financial Officer Bob Eulau said the company expects sales of $38 million to $42 million, with operating costs and expenses ranging from $31 million to $34 million.
Two analysts surveyed by Thomson First Call currently put the company's first-quarter sales at $40 million, on average.
Hahaha, the average of 38 and 42 is 40, isn't it
Just imagine some young guy coming out of university and doing a stint at a brokerage to get some experience of Wall Street and then reading this. He must be thinking: "am I missing something?"
Cheers
Cor
For example, what if Intel started doing MM stuff using XDR and then sub-contracted it out to Micron.
Hi Kent
I don't think so. What yuo describe is "patent laundering" afaik and that is explicitly excluded in the Intel <> Rambus cross license.
Yes it is a cross license, Rambus needed some Intel IP as well
Of course it is a reason for concern that the largest semiconductor company can have all for free after the next 7 quarters.
Cheers
Cor
Threejack, Rambiskit (Watz) has posted fairly complete transcript of the CC on yazoo, maybe you want to post them here. They are pretty accurate and amazing the speed he has done them at:)
This is part1:
http://finance.messages.yahoo.com/bbs?.mm=FN&action=m&board=4687909&tid=rmbs&sid=468...
This is part2:
http://finance.messages.yahoo.com/bbs?.mm=FN&action=m&board=4687909&tid=rmbs&sid=468...
This is part3:
http://finance.messages.yahoo.com/bbs?.mm=FN&action=m&board=4687909&tid=rmbs&sid=468...
part 4 is a bit later and there is an afterthought, i believe.
Cheers
Cor
Something very positive though, although it was assumed before, we can now safely say it is confirmed that XDR is used with the Sony cell.
About in minute 29:00 of the call there was a correction that almost completed work did not refer to redwood in cell but to XDR i/f.
I now officially proclaim XDR confirmed by Rambus -unwittingly- in the Sony cell and PS3
Although it is not official that redwood is NOT in it, that is likely in view of that CC intervention.
Cheers
Cor
Ha ha Cor - I got the 7% sequential increase in royalty right - that's about it!
Hey Will, congratulations
I got the total revenue and profit closer; also knew where litigation was going and apparently it will get worse:(
TMF consensus from a poll I did was 10.5 ct; I voted 7-8 and my spreadsheet said 7.3 ct. Erach Desai had 8 ct. Better than the (other:) analysts, which does NOT make me happy.
We now know that guidance from Eulau is unreliable. Look at that tax rate, he specifically guided 25-27%. It was 38%!
(imo he made a mistake in the 3rd qtr, but is not saying this, has to correct back and all that, heads should roll instead of options flow, that job is so simple I could do it and I would for a fraction of the options and no fixed salary, LOL)
Tax for 2005 expected at 37%, so all back to normal.
Disquieting and not quite resolved yet is the further decrease of deferred revenue. I see it this way, pse correct me somebody if I am wrong (I was in contact with Desai about this, note his questions on the subject):
About qtr4:
contract revenue is 6.0 M
cost of contr.revenue 5.5 M (back to normal, close to rev.)
decrease of def.revenue 6.1 M
means: new license rev. -0.1 M
Isn't the deferred revenue account a balance sheet account which reflects money received (usually for licensing deals) but not yet earned. As it is earned (amortized) each quarter depending upon work performed, the balance is reduced. Conversely licensing revenues increase on the P&L.The difference up or down reflecting either a growing deferred account because of new deals or a declining one based upon older deals drawing down the balance. Just from memory which worsens daily.
Totally correct, Kenny and this is why I get suspicious when there is very high contract revenue (which is where that does) and at the same time a decrease in deferred.
So 3rd quarter (we'll get to the 4th later where something similar is happening imo) there was 8.3 M contract rev. and at the same time deferred total went down 5.5 M.
Cheers
Cor
Skeptic on the royalty growth:
The qyt-on-qtr royalty growth (ex Intel) was 4.4% the previous qtr. Including Intel the same would be reached now with $ 31.4, just a little above my 31.0
If Rambus sees 7% royalty growth, it puts royalty at 32.6.
That would be 10.3% growth of the non-Intel royalty. the only time recently which was higher than that sequentially was 2nd qtr04 at 12%. The quarters before that not much happened.
Where is your 10.3% going to come from? Maybe serdes?? Don't forget we are talking about the royalty pertaining the the 3rd qtr where there is no PCIe penetration yet worth mentioning.
I don't know on the memory side how 3rd qtr would compare with 2nd; rdram pricing and use?
apply the same 7% growth in COGS, you're at 5.0.
the cogs has to do with contract revenue, not royalty. Historically contract revenue has been about the same as cogs however the last qtr was better; I see that as an anomaly which may return to "normal" at least partially. One reason why contract revenue may eb higher is when a project for which there was a license gets cancelled; the money then gets taken off the balance sheet from "deferred revenue" into contract revenue earned w/o costs associated with it.
Deferred revenue decreased from the 2nd into the 3rd quarter by $ 5.5M from $ 34.5M to $ 28.9M; this has not been explained in the CC nor challenged by the analysts.
Cheers
Cor
Hey Skeptic,
on: Can't get this damn thing to look right - sorry
Use the < pre > and < /pre > tags around your table data, that will do it. *leave out the spaces after < and before >.
Cheers
Cor
re: You Ol' Lowballer, you. Just like an Analyst : )
Oops you're on to me:)
These are my best estimates from all I know and what Eulau told us. But there can always be positive (or negative surprises.
Btw I checked with Erach and he has the same revenue I have and 8 cents/share.
Cheers
Cor
I met Gelsinger and Burns. Good guys.
Cheers
Cor
My estimates, which I posted on TMF for tomorrow's numbers:
This is a bit futile, as we will all know in 24 hours anyway, but I'll have a shot anyway:
Here are my best estimates, first column number for qtr3, second column qtr4:
qtr3 qtr4 comment
contract revenue 8.3 8.0 minute 24.30 mp3
royalties 30.5 31.0 memory flat or down, serial up slightly
total revenue 38.8 39.0
cost of contr.rev. 5.3 5.3 could be a tad higher(India)
gross margin 33.4 33.7
costs:
R&D 8.1 8.2 fixed costs seldom go down when things go well
AG&A *excl.legal* 8.1 8.2 ditto
Legal expense 6.7 8.0 minute 16, guides 6-8 I assume high given what happened
Total costs 22.8 24.4
Operating income 10.6 9.3
Other 1.1 1.3 guidance 1.2 - 1.4
Income bef. tax 11.8 10.6
Income Tax Expense 1.4 2.8 guidance 25-27%, took 26%
Net Income 10.4 7.8
Fully diluted shares 107.6M 107.6M
eps (ct/share) 9.7 7.3</PRE>
Note:</B> where numbers don't seem to add in the first column this is because they come from a spreadsheet which has more decimals; the numbers are rounded correctly.
I have referred to the qtr3 CC, which is still available on my site as:
www.chipstocks.net/rmbs/rmbsCC04qtr3.mp3
when I refer to minute marks it is in that recording.
The guidance is given by Bob Eulau in minute 15:11 and following.
Note that tax was guided bu him to a higher percentage than qtr3; in my estimation that is because they took some more in qtr3 because of the earlier quarters where they had not yet realised they could do it.
Cheers
Cor