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.
mas,
There's no reason Q3 should have been down with increased unit volume unless ridiculous unnecessary price cuts took place which is what did happen.
I am certainly curious to see how Intel and AMD get out of this ASP toilet. They are selling dual core (and even quad core) CPUs with lower blended ASPs than (largely) single core CPUs a year ago...
Joe
wbmw,
No, I said clobber them in sales. That means relative sales, also known as share.
It depends if Intel comes in at the high end of the estimate or the low end. I think the guidance is between 4.5% and 11.5% growth over Q3 vs. AMDs 3% growth.
It would certainly be disappointing if Intel disappoints too. AMD has had meaningful increase in unit sales, accompanied by similar fall in ASPs. It would certainly not be a good sign if Intel ASPs are either stagnant or falling as well...
Joe
wbmw,
I don't find SFF to be synonymous with low cost. In fact, most SFF systems come with a premium. Low power processors also come with a premium, and in the end you are getting more by paying more (in this case, "more" size optimization). Merom would seem to fit this model nicely.
that's true, but one of the things that have kept pricing high is lack of standards. Like those Shuttle XPC boxes. One reason a pathetic vendor (quality wise) such a Shuttle can be a leader is lack of standardization.
Joe
Tenchu,
Probably meet a similar fate. Unless NOR has any chance of competing against NAND in the Brave New World of pervasive computing ...
When you are speaking of computing, or, going in direction of more computing (mobile), it means more NOR. NAND alone can only store data. It needs parallel DRAM to compute anything.
NAND + DRAM combo has not taking over the market by any means, and if anybody is winning the short term tactical battles, it is Spansion with its NOR / ORNAND combo.
As far as Intel, I am not exactly sure what the current strategy is in entering extremely low margin NAND business, lagging in (normally) higher margin NOR, and exiting Arm (that could have a potential of being reasonable margin business one day...
Joe
Tenchu,
Meanwhile, Intel is already making bold moves into NAND. Pretty damn risky, but I think this is the right direction for Intel to take. (More and more these days, I believe that whichever company first adopts a model for pervasive computing will be the company that eventually succeeds.)
What are you envisioning here? There are bunch of cell phones and smart phones out there from a number of different companies. There are component makers, handset makers, software, wireless serviced providers. Way too big for one company to have it all.
As far as memory, Samsung certainly looks the best in data storage, and I placed my bet on Spansion for combo of code and data storage. But it is just one small aspect of the overall picture.
Joe
mmoy,
I would buy but the
X2 5800+ or X2 6000+ seems like a good fit for my application.
Curiously, X2 6000 will be faster in most apps that this Quad FX. X2 6000 is basically identical to FX 74, except X2 6000 will not be slowed down by the multiprocessing overhead.
Look at Anand's review where he compares the system with 1 FX 74 pulled out. It outperforms 2 socket solution by a solid margins in majority of apps. Of all of the apps, only a small minority can take advantage of cores beynd 1st and 2nd.
Still, X2 6000 will still come behind high end Conroe, will probably only match mid range Conroes. So it depends on how AMD prices these.
Since I have 4800 at home and 4600 at work, and most of the lag I experience in daily usage are due to HD (or network / Internet lag) I am not sure how much performance benefit I would experience from extra 600 MHz. Probably not all that much.
I imagine 4 MB of Conroes' cache (on a high end processor) could provide some perceptible gains in select areas.
My current problem regarding performance is the performance of RAID. I set up RAID 5 at home, to increase capacity and to protect myself from dead drives. The problem is that the performance, especially write performance seems to have dropped to about 1/2 of the single drive. I am using built in RAID of nVidia chipset (590).
I came across one review that convirmed it. I don't really get it. Nvidia RAID is actually the best of the motherboard integrated RAID solutions, beating Intel's, but still, it is bad.
I have not really paid a lot of attention to storage in detail, but my perception was that RAID improves performance across the board. At least that's why database vendors includ 100s of drives in their configurations.
What do those RAID solutions do that mobo integrated solutions can't? And can I get it myself in form of an add-on card for within $100?
Joe
PS: I currently have 3 drives. Read performance does not seem to be much of a problem, but writing is.
mmoy,
I have held positions for many years and liquidated all long-term holdings back in May.
I remember when you posted that. You had some dire predictions that apparently turned out to be false. Have you changed your outlook in any way since then?
Joe
wbmw,
It depends on your expectations. If you have high expectations, it is a non-event. If you have low expectations (A0 being initially buggy), a somewhat working silicon is a bit of a relief.
It can certainly counter some FUD, and add some comfort to partners and investors.
But even while there is a working silicon, it is possible that it may not be working that great, and that there are a bunch of bugs still to be worked out, so the schedule may still be tight.
AMD seems to have padded the availability somewhat (while keeping "introduction in Q2). It may be for the possibility of a slight delay.
Anyway, with analyst conference just 2 weeks from now, we may get some more info on where it stands.
Joe
wbmw,
Why not? You posted on the AMD "native" quad core non-event.
Why would that be a non-event? It is a progress report on the most important project that AMD has ongoing...
Joe
mas,
they also selected Power 7 over Rock, read into that what you will .
It does not really surprise me.
Joe
Keith,
So Cray and IBM are in, Sun is out. I asways thought Sun had no chance. Cray's higher award suggests it was the overall winner with IBM coming second.
From IBM press release, it appears IBM is going to be using POWER7:
http://www.marketwire.com/mw/release_html_b1?release_id=187462
Joe
Elmer,
When you can figure out what you're talking about you can try again but keep the abuse to yourself.
Nevermind. All the posts about exchanged over the years about being demand vs. supply limited must have been with the guy doing the late night shift.
Very sad.
Joe
Elmer,
Maybe, maybe not. IBM has had terrible yield problems over the years and AMD's process is closely related.
Since you chose to reply to that post, I was expecting a reply: "I was wrong, you were right". Instead, you come with this bullshit.
It seems that, all this time, I had a higher opinion of you than I should have had.
Joe
sgolds,
Truth is, AMD never had the capacity to expand the market as the fact of the lawsuit would suggest (although I'd suppose they could have done more with Fab 25 if the demand were there).
That's not exactly right. AMD was "demand constrained" from Q4 2000 to Q4 2005. So there was capacity in place to sell more processors.
Joe
chipguy,
I didn't mean IPF was going to push into the sub-$5k segment
but rather major IT users will consolidate some its compute
needs away from fleets of low end servers to fewer larger
servers using virtualization and partitioning. This can reduce
support costs and improve flexibility and security and lead to
overall lower IT costs.
No, the main idea is to retire fleets of old underpowered machines, running oddball legacy apps. All those servers are running x86 software, so Itanium does not qualify.
Good try though. Somewhere along the lines of retaking Baghdad airport.
Joe
chipguy,
With the trend to virtualization-driven server consolidation
and growing concern about RAS and security is even that
segment safe from competition from IPF in the long run?
Are you saying that there are so many Itanium servers out there as an installed base that they are ripe for consolitation?
Joe
Elmer,
AMD can only use pricing to compete here and ASPs will reflect that.
Darn, too bad AMD can't twist arm and make threats like Intel did when behind in performance.
Joe
chipguy,
Only a raving lunatic could think AMD could have dreamed
of offering Apple a tiny fraction of that amount of technical
hand-holding in the areas of system hardware development,
system software porting, and development and tuning tools.
AMD itself largely depends on the kindness of strangers in
all of these this areas.
Which would make it extra sweet if iCon tells Intel to go f. themselves, and starts purchasing microprocessors from AMD, just as Dell did.
Joe
PS: Wasn't the argument that Dell will never use AMD that Intel provided all the engineering and hand holding to Dell?
Keith,
It been logical for Apple to launch their x86 line completely with 64 bit AMD processors, and then switch some lines to Conroe / Merom. Then, there would have been only one code-base for Apple's x86 line, 64 bit one.
But then, I am not CEO of Apple, iCon is.
Joe
fpg,
XT4 is the platform that will accept, in S1207 sockets:
A) Opterons,
B) DRC's Xilinx-based FPGA copro's,
C) Cray's custom Threadstorm multithreaded (128 threads) processors.
D) Various mixes of the above.
You are right. I wasn't sure about c) but this interview posted and highlighted by Robert G:
http://www.siliconinvestor.com/readmsg.aspx?msgid=23010145&srchtxt=cray
confirms it
Original link:
http://www.hpcwire.com/hpc/1086098.html
Very smart approach by Cray, it seems.
Joe
Smallpops,
Have you been able to distinguish the difference between SeaStar and SeaStar2?
I think the only visible change is from 800MHz to 1 GHz HT. But there may possibly changes in interconnect speed between SeaStars (I didn't check). Also, there is some routing built into these communicating chips, so it is possible there were some minor upgrades there as well.
From news.com article, it seems there is going to be the next generation coming soon after this, incorporating HT 3.0.
Joe
SmallPops,
I did a quick comparison between the old XT3 and the new XT4. Seems like Cray starting using the new Rev F socket with DDR2 and decided that was enough to warrant the model number change. What a joke.
There is a little more to it. It incorporates ideas from Octigabay acquisition (XD1 line) into XT3. XD1 had a FPGA coprocessors incorporated in a custom way. The new line incorporates them in a more standard way, in direction of Torrenza, such as either having the coprocessor in an Opteron socket, or possibly on a board that would connect to their SeaStar interconnect (not quite sure about the 2nd part).
Joe
Elmer,
There it seems to tail off pretty rapidly going from 4-way to 8-way or 8 & 16 if you count cores. So how good a predictor is SPECfp_rate when going into large HPC systems? I'd like to understand this better.
The trend seems to be away from large system (as far as # of sockets / cores) addressing one global memory. Those systems are not cost effective compared to large number of 2 socket system.
Joe
mmoy,
A coworker just bought a 4800+ to replace a 4000+ and said that
you can't find them right now. This is socket 939 I think.
I got X2 4800 recently, a month ago approximately, in AM2 format, for a little under $250. Despite the fact that they were supposed to be cancelled...
Joe
wbmw,
Where's the mystery? They FX-62 has been out for months now. Add 100MHz and subtract a meg of cache
I think it is add 200MHz, subtract 1M of cache and also, subtract 20W of power. 6000 should be the same as above, except it has full 2MB of cache.
Joe
mmoy,
In the Northeast, there isn't that much of a building boom.
There certainly is one in New York City...
Joe
Keith,
Gigabyte´s shipments definitely "fell off a cliff". Seems like October was even worse than the GS analyst estimated.
What's your explanation to falling demand for mobos while at the same time, there is a strong demand for CPUs (at least for AMD CPUs)? Do these guys make mainly desktop mobos while someone else (contract builders) make notebook mobos? Just wondering...
Joe
Optiplex 740 just appeared on the US site as well:
http://www.dell.com/content/products/compare.aspx/optix?c=us&cs=555&l=en&s=biz
No 320 yet.
Joe
Keith,
That would be today. But...
If the Dell commercial desktop materializes in the US, it is great news, regardless of the introduction.
Another possibility is a tiered introduction, in smaller geographies before the big ones, such as the US and Western Europe, considering AMD is not in the position in this quarter to make extra commitments.
Joe.
PS: Hard to believe, but I am beginning to warm up to AMD stock again...
Keith,
Doug pointed me to an INQ article where Charlie leaked it:
http://www.theinquirer.net/default.aspx?article=34239
Joe
Keith,
Are these newly announced systems? I don't recall this being announced in the US. What's the UA domain?
Another thing I was wondering is the ACS (Advanced Configuration Services). I have heard some client management technology that a group of companies was going to introduce. Group of companies that included AMD and Dell, but not Intel. Is this it?
Joe
wbmw,
Smithfield's brilliance only goes as far as it enabled Intel to get *something* to market in a very short amount of time.
ACK
mas,
bigger question, IMO, is whether the 4x4 will be able to use cheap unbuffered memory. I think info on this subject has been conflicting.
Joe
wbmw,
You're right. AMD is going to make a killing in the segment of enthusiast gamers that can spend thousands of dollars on 16GB of the lowest latency and high bin overclockable memory, while Kentsfield with roughly 8GB of memory support, really has no chance of competing.
On practical level, high performance memory goes only up to 1 GB (per DIMM). 2 GB DIMM prices go up exponentially (over 1GB) and performance goes down.
So, ability to buy 8x cheap 1 GB DIMM actually is a price equalizer, if not price advantage for users who need 8 GB (or more) memory.
Joe
wbmw,
Equal price, lower performance, higher power. Oh well, I'm sure it will still be wildly successful anyway.
Sounds like Smithfield vs. DC Athlon X2.
By the way, did you catch the last paragraph? When was the last time that Th'Inq staff carried such negative opinion of "brilliant" AMD management strategy?
Smithfield was a brilliant idea, according to you, so I am assuming you would equally enthusiastic about 2 x Athlon FX.
Joe
gollem,
you'd expect they'd have learned by now, it seems to me that that is nothing outrageous to expect from THE big OS company.
Maybe they have learned. They have learned that cost benefit of supporting systems that have been "upgraded" from previous version of the OS is vs. convenience of upgrading is negative.
Besides, not many people upgrade their OS. Most of the sales are to OEMs who preinstall the OS. Retail sales, as a percentage of OEM sales is a tiny fraction.
Joe
wbmw,
AMD is in the same defensive situation that Intel was in earlier this year. They have nothing to carry them forward except for promises of the next generation.
Average shippable part from AMD is still faster than that of Intel, and will continue to be so throught Q4, until Conroe fully ramps. So AMD is not yet where Intel was in recent past.
First 3 quarters of 2007 will likely be a different story, with AMD behind.
Joe
gollem,
there is no upgrade path to vista x64; not from xp 32 and not from xp 64.
Nice, msft.
I hope you are aware of the fact that when there was an "upgrade", it was a bad idea to upgrade, rather than reinstall.
Joe
mmoy,
On another subject, I have seen an article somewhere, that since Microsoft now gives away intro packages of their development tools, that Microsoft makes them (optional) part of standard Windows install. The idea is that, like in the old days, when people could just jump to some version of Basic on their computers, the same way, they could use current development tools. The reasoning is that it could increase interest in programming and tinkering among users, especially young ones, so that perhaps they will have increasing interest in entering Computer Science programms later on.
Joe