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.
Thanks, i will try that next time. I am not sure if this was the case in the past, but may work fine. I will be setting up another computer with combo of nVidia 6150+430.
Joe
chipguy,
The new instructions seem most helpful for operations
likely to be performed by specialized libraries rather
than compiler generated from application source code.
going on memory here, I recall that bunch of instructions are there to get stuff in and out of XMM registers faster, or more conveniently, which would be of interest of standard compilers (if the developers) if there was a universal install base of SSE4.
But you make a good point about use of these instructions in specialized libraries, and the differentiation which code path would be taken can happen at the time the library is loaded (or inside of the function calls for slightly more overhead).
I think the library approach is what AMD may be targeting with the Torrenza coprocessors.
The advantage of SSE4 (which AMD will copy, probably within a year of Intel's released product) is that some narrow interest apps may start to appear to take advantage of them. I think some interest may be in the string handling instructions. I wish Intel. But I don't see any neat cut off on the horizon as there is with 64 bit windows, which cuts off MMX, 3DNow and FP stack instructions in favor of SSE1,2.
Joe
combjelly,
Well, if your Linux installation has nVidia drivers build it, it makes all the difference in the world. AFAIK, Windows XP, or SP2 do not have up to date nVidia drivers.
But I will test if for myself in the next couple of days...
Joe
combjelly,
It all depends on how the SATA chip manufacturer chose to implement their SATA controllers. Smart ones, like nVidia and others made them 'look' like their PATA controllers as much as possible. So no special driver is needed to get basic functionality.
That news to me. Are you talking about installation of the OS?
Joe
mmoy,
My Core 2 Desktop has a SATA disk. Not sure how the CDROM is hooked up but the BIOS sees it with no problems.
Some 99% of CD/DVD ROM drives are IDE, so BIOS sees them just fine. Well, BIOS sees SATA just fine too, but Windows does not have a built in generic SATA driver, while it does have IDE, which is where the problem lies, not in the BIOS.
Perhaps an
install could be done over a LAN.
As far as over the LAN installation, while you can hack the installation files to add support for a particular SATA adapter, it is a PITA (I have done it). It would be nice to get a pre-made XP (32 or 64 bit) installations with bunch of new drivers released in last few years for new hardware types.
Joe
Tenchu,
The way it looks, SSE2 will become the lowest common denominator for all 64 bit executables, we will not see much movement (outside of very narrow nices) for half a decade.
SSE2 certainly beats the alternatives prior to SSE2, but using further extensions any time soon (through separate executables or logic inside of the program) seems like a very poor business proposition for a software developer.
Joe
Keith,
Great to see the first 1P tower from major OEM. Somehow, it is telling me that no products are available. Could it be that they are available in other geographies but not yet in the US?
Joe
Edit: never mind. I came to the screen from a different direction and I got 3 models, all in stock.
Edit 2:
New (at least new to me) 2U dual socket and 4U quad socket are listed here:
http://www-03.ibm.com/systems/x/rack/index.html
mmoy,
Do you have a link for the costco setup? Somebody is asking me for a recommandation for something under $1,000, not really for mobility, but compactness (vs. desktop).
So recommended the friend to look for something with 17" screen. I would add 64 bit CPU, which would mean Merom or Turion. The Merom at costco would certainly be interesting if it had 17" screen...
Joe
sgolds,
I am assuming that you needed a driver for SATA disk / controller. Just FYI, this problem is not unique to 64 bit Windows install. 32 bit windows has the same issue with SATA.
I think MSFT should create a generic SATA drivers, in order to be able to install the OS without having to have access to a floppy...
Joe
sgolds,
The other consideration is how much of that 11.7% IPF market is due to HP's forcing their customer base to it. When a major RISC company switches architectures then there is a certain forcing of the issue for many users. To really gauge IPF growth against competitive RISC one has to look at the non-HP portion - after all, it is only the non-HP segment where IPF has direct competition.
Well, SGI went bankrupt selling Itanium in direct competition with other technologies, IBM got out of that market, supporting their Power and AMD64 (from AMD and Intel), Dell also got out of Itanium market in favor of AMD64 (from AMD and Intel). So Itanium is still at mercy of HP to channel business, credibility - and growth to it. Excluding HP, the sum of second tier Itanium suppliers does not add up to minimum credibility threshold.
Joe
Tenchu,
"Oh, nobody really wants 'fake quad-core.' They'll just wait for 'native quad-core.'" It all depends on what your definition of "nobody" is. Intel's definition is at least one million ...
"fake quad core" is better than dual core (certainly for servers), and if AMD had infinite resources, they could probably allocate some of them for an MCM "fake quad core". But cost benefit analysis of limited resources must have come against such a short term project, which would come at expense of other, long term projects.
Joe
mas,
The "rocket" lots are processed much faster. AMD may already have the first wafers packaged and is starting the testing. If everything goes right, I think AMD may demonstrate running chips during November analyst meeting.
Joe
What's the target of Fab 38 conversion?
Joe
Tenchu,
Welcome to four and a half years ago, Joe. Can I take you on a tour of Yamhill?
The contingency plan for the scenario that Itanium turns out to be a failure certainly came in handy when Itanium turned out to be exactly that.
Intel is certainly better off retaining 75% of the server market, throwing Itanium more or less overboard, than sticking with Itanium, and being a minor player in the server market today...
Joe
wbmw,
Anyone expecting it to set the standard for volume server architectures is going to be sorely disappointed. But it's obvious that it has carved out a pretty successful niche at the high end, that's growing at the expense of Sun and IBM, and has a sizable contribution to revenue in the server market. So proclaiming it to be a winner or loser depends on what ends you perceive it as needing to achieve.
Good point. Intel wanted Itanium to be its only 64 bit chip, meaning it would have to be a mainstream server chip today, replacing Xeon everywhere except the lowest end. By that standard, Itanium is a complete loser.
Only based on the new, vastly scaled down goal (for it to be a high end niche product) it is not a loser.
Joe
chipguy,
Given that IDC's mijudgements about Itanium were legendary, I am not sure I suddenly trust their methodology.
As far as suddenly trusting their methodology now, because they talked to some IT folks, are you suggesting that their past predictions came from one guy locked in a cubicle, rather than surveying the industry and its participants?
Hiring IDC for opinion on Itanium for the purpose of getting the right answer is equivalent of hiring Bob Shrum to run your political capmaign for purpose of winning it.
2. As far as your record goes, I have lost count - have
you killed off IPF 200 times or 300 times to date? :-P
Zero is certainly closer to actual Itanium shipments than IDC estimates.
Joe
JJ,
Good point about OEMs who now feel less intimidated by Intel under the umbrella of the law suit.
Joe
Chipguy,
Wow Joe, don't you ever get tired of being dead wrong
about IPF time after time, year after year?
I certainly have a better batting average than your source, IDC on the subject of Itanium.
Joe
Tenchu,
Wishful thinking. Not only do you believe that Intel is guilty, you also believe the execs know it, because to you, Intel's alleged guilt is as clear as day.
Yes, I already said yes.
I am not sure if you ever posted this, but do you believe that Intel is innocent of all violations alleged in AMD complaint?
Joe
chipguy,
I guess that depends on whether you say it in front of a
group of Fortune 100 CIOs or a bunch of web site fan
boys who get their information from tabloid tech news
sites.
I think you would get a more positive reaction from web site ban boyz, because it would be nearly 100% don't care.
With CIOs, IMO, you would have more people with opinion, more negative than positive.
Joe
Tenchu,
Oh please, you seriously can't believe that Intel is intimidated by the lawsuit, can you?
Yes, I can. Mainly because Intel executives know Intel is guilty as charged.
Joe
wbmw,
So if AMD loses market share in the next few quarters, do we get to say that the trial is a failure?
No, it would be a natural thing for Intel to gain share at the time it had superior product. It was unnatural for AMD to not gain share or lose share while it had a superior product, which is the condition the law suit intended to, and had improved somewhat.
Joe
wbmw,
Right, and just like others like him, he gets to define what "failure" means. It doesn't matter what IPF is, or what it becomes, because he already has it decided in his head what he's going to call it.
It is a free country, you can keep your own score. Just don't be surprised to see a smirk on people's faces when you say that Itanium is a winner.
Joe
Tenchu,
If all of this merely costs Intel eight figures after the trial is over (and you know the conclusion of the trial is over five years away), then that will be an EMBARRASSING loss for AMD. Oops ...
Any award is a bonus. Chage of behaviour of Intel is the real prize. The fact that the trial is in 2009 only stretches the time Intel has to be at the best behaviour. If I were an AMD shareholder, I would be happy about 2009 date rather than, say, 2007.
Joe
chipguy,
Good point about about the DEC/Compaq/HP legacy apps migrating to IPF. I was forgetting about those.
Joe
wbmw,
I stand by my point that there may be one or 2 dozen real apps with active teams of developers supporting them, keeping tham up to date.
The rest is a hodge podge of one-offs, in-house utilities, recompiling some standard source, failed ventures etc.
As far as numbers such as 10,000, it is only a question of how bold face Itanium promoters want to get. Using their standard, I could search my hard drive and find a few "Itanium applications" like those listed on Intel site. <g>
Joe
wbmw,
you're grasping at straws.
There's bound to be a few useless apps out of a list of 10,000, just like any other platform out there.
Grasping at straws is claiming that some old C source code is Itanium app, not me pointing it out.
Joe
wbmw,
That's a useful link.
Yeah, that's a useful link to find out that the so called Itanium apps are crap. Let me give you an example of two of the so called Itanium applications listed:
Phrap
by University of Washington
http://www.itaniumsolutionsalliance.org/kshowcase/view/view_item?item_key=638c2a38abea729fe53ab545b4....
Phred
by University of Washington
http://www.itaniumsolutionsalliance.org/kshowcase/view/view_item?item_key=9e00c9cde1346ad3851078bae3....
Ok, looking for what it is you get, I found this:
http://www.phrap.org/phredphrapconsed.html
We distribute phred as 'C' source code: in order to run it you need a 'C' compiler.
So some C source code is suddenly one of 10,000 Itanium applications.
Here is something from the documentation:
# Copyright (C) 1994-1999 by Phil Green.
# All rights reserved.
#
# This software is part of a beta-test version of the swat/cross_match/phrap
# package. It should not be redistributed or
# used for any commercial purpose, including commercially funded
# sequencing, without written permission from the author and the
# University of Washington.
So this C source code may be from 1999, before the first Itanium was even released...
Joe
wbmw,
The first seems like filling some gaps, reducing the instruction counts in the compiled code, and also making SSE2/4 more accessible, and reducing the overhead of vectorizing certain routines.
What I really like though are the string functions, that can have a drematic effectt on performance of databases and text processing in general.
Joe
chipguy,
Economic/technical factors favour merchant processors in
the long run. From 1995 until a few ago years x86 was the
only realistic merchant processor offering but it was only
suited for the low cost, high volume servers. But now IPF
is a realistic merchant processor offering for mid to high
end servers.
I agree on this, as far as hardware is concerned, and IPF is in the middle vs. the OEM processors, that I also see being long term losers.
OTOH, on the software side, the Sun and IBM have some advantage of installed base, and some automatic sales as the installed base adds to / upgrades their hardware.
It is interesting to speculate how competition between
x86 and IPF will evolve as the two alternative merchant
MPU families face off against each other more and more
as the non-merchant processor families gradually lose
share to both. IPF grows ever more price, price/perfor-
mance, and performance/ Watt competitive
I am not sure it is a foregone conslusion that that IPF will grow this way vs. x86.
while x86 gains
more scalability and RAS features as time goes on.
I think the scalability of x86 - breaking roughly 4 socket cap that has existed since x86 has invaded server market - will be interesting to watch. It will be the third wave of assault on the high end in a very short time. First wave was breaking of the 32 bit addressing limit, second ongoing wave is the multiprocessor wave, the next one, in 1 to 2 years will extend caps on the number of sockets (thanks to DC 2.0 and CSI).
gains
more scalability and RAS features as time goes on. Once
software would have been the key advantage for x86. It
still is of course on the client side but IPF now has over
10,000 apps, a figure still growing rapidly. IMO it isn't
clear what software advantage, if any, x86 will have in
servers in a few more years.
I don't, of cours, buy the 10,000 apps. There may be 1 or 2 dozen full featured, fully supported IPF apps out there, which is enough to sustain IPF in the narrow server niche. But I view the x86 vs. IPF in a similar way as merchant CPU vs. custom OEM CPUs. IPF software development lacks scale of sales, making it more expensive to develop (per unit sold), more expensive for software customers to acquire etc.
IMO, ranking the way of advances in server CPU development, IPF may gain relative to custom OEM processors, but I see it losing to x86
1. 32 bit addressing limit of x86
IPF vs. custom OEM processors: neutral (no effect)
IPF vs. x86: negative, ongoing
2. multicore
IPF vs. custom OEM processors: likely positive
IPF vs. x86: at best neutral
3. breaking the socket scaling caps of x86
IPF vs. custom OEM processors: no effect
IPF vs. x86: negative
You could claim that IPF may go from severely software starved to less severly software starved, but IMO, the ratio will not change much.
Joe
chipguy,
Are you merely saying that the decline of non-x86 is not a straight line or that non-x86 is on the rise?
Joe
sgolds,
11.7%? That is a start.
But it is 11.7% of a segment that is shrinking relative to x86 (or should I say x64).
Joe
Mike,
My sources are all in the top 25 but not the top
6, so... there are other places to get them. I ususally
have no problem getting product from one of my sources.
Thanks for the reply. I knew there had to be sources that have the X2 in stock. But low distributor stock levels show that AMD is selling all they can make. The question is how many AMD can make (how the transition to DC and Rev F is affecting the unit volume).
Joe
PS: I missed out on last 2.5 points of the latest runup, and I am out of CPU stocks for the moment...
wbmw,
Wouldn't this be the case if online vendors were full of stock and stopped ordering new parts?
I am not sure I understand your comment.
Mike's data shows distributors' stock and orders. Some / many / most of the online vendors get their processors from the same distributors...
BTW, the distributors show healthy orders for X2 parts, just nothing in stock. The first thing I would guess is that AMD has trouble delivering enough of these X2 parts. Anyway, end of September through end of November should naturally be the leanest months - inventory-wise.
But online vendors seem to have all mainstreem X2 processors in stock, 5000 is rare (and marked up), 5200 is non-existent (probably outside of the boutique vendors).
Joe
Mike,
I was out of the loop for last week or two, but I noticed your update. I noticed that with the exception of S939 4600, neither of the 2 distributors have any X2s in stock. Yet, I can find them available for order all over the net. Do you have any explanation for that?
Joe
wbmw,
Sure thing, Joe. Whatever it takes to fit your fantasy....
I take it your benchmark will now become presently shipping processors. Wait, not right now. In about 2 months.
Joe
wbmw,
Now you're saying it's ok for them to be behind, as long as they finally adopt AMD processors? That they have "learned" how to sell crap, so this "knowledge" will help them to sell AMD? LOL.
Dell was missing high end Athlon 64, X2 FX and Opteron processors, and cheap Semprons. HP did not, so HP made some modest gains as more competitive products offset Dell's more competitive distributions system.
If Dell decides to adopt AMD processors, it does not mean Dell is losing on anything. Entire portfolion of Intel processors will be available to Dell, just as it is available to HP. There will be areas where Intel has a better solution, other areas where AMD does, when it comes to price, performance, power consumption, system architecture.
And most importantly, Dell will have the customer the opportunity to chose.
Hypothetically, Intel can sell every one of those CPUs to an eager HP, who would be more than willing to sell something more competitive and bury Dell.
It has nothing to do with Conroe, if that's what you are thinking. Dell will receive its share (if not more) of Conroes. Intel will sell all the Conroes. Dell would just purchase fewer Netburst, possibly Dothan and Yonah CPUs.
Do you really think HP will remain loyal with AMD, now that they are flirting with the competition? Who is going to ignore Intel's products, when they are the superior ones?
HP has offered AMD products alongside Intel products (which have always been in majority). HP just gave customers a choice. I know that word causes major pain at Intel headquarters, since Intel spend half a decade trying to illegally deny customers choice.
As far as HP, I see no reason for HP to stop offering customers choice. The novel idea here would be for Dell to offer choice as well.
Face it, Joe. Intel's going to sell Core 2 to whomever would like to buy one. And the rest of the product lines (P4, PD) will be sold at prices low enough that someone will buy them, too. So where does that leave AMD? Overcapacity.
Didn't you argue just yesterday that AMD would not be able to compete because there was not enough capacity to switch to DC?
Joe
wbmw,
Do you know something the rest of the world does not? If not, then I would argue that Intel has superiority in the bag, at least until AMD delivers K8L next year.
Based on your posts, Intel has had superiority since last fall IDF when Intel released some benchmarks of future WMC core. Therefore, AMD will have a superiority on the day when AMD releases benchmarks of a future processor that are better than Intel's processors at the time of the release of the benchmark...
Joe
wbmw,
end-users buy them over Core2, good for Dell. Intel can take their business to HP and others. If Core 2 ends up being preferred over Athlon, Dell will end up with a lot of effort for little return.
Dell has gained a lot of experience selling inferior product, while being Intel only OEM.
Hypothetically, if Dell replaced 1 to 2 million $150 Intel CPUs + $40 Intel chipsets with 1 to 2 million $100 AMD CPUs, Intel would lose $190M to $380M of revenue. That's a pretty strong negotiating position to be in, IMO.
Joe
wbmw,
I think the decision to sell Xscale has primary to do with deteriorating profitability of Intel (which is related to AMD). It is easier to sweep problem divisions under the rug, when profits are high and rising, much harder during lean times.
If Intel could afford it (without sparking a rebellion of the investors), I believe that Intel would have preferred to have more choices in the growing market where XScale is participating, rather than less.
I think Intel may now do as AMD has said it would do, which is to work on getting x86 to address some of the markets within reach.
Joe