Probably not Barcelona, as that is the server version and the quantities are not vast (assuming adequate yields and bin-splits). What it will affect is the consumer roll-out (Athena?) where there need to be a lot more wafer starts.

mmoy, on cut reductions

They're talking mainly about reduced capital expenses. This can be done simply by delaying delivery and installation of new machines and/or some work items on the conversion of Feb 30.

Elmer, on deleted emails

Yes, but what situation would he be in if a customer were to send him an email that discussed an off-limit subject, complete with reply history? Like "Heard about the AMD suit, but I expect the same deal we had last quarter!"

CombJelly, on the new DTX standard

Yes, and the new DTX standard is ideal if you just happen to make processors that only take 35 watts. Hmmm., wonder who makes CPUs like that?

alan81, on statements

The Acer president already said this allegation is false.

Did he say so under oath? And with the possibility of follow-up questioning?

fastpathguru, on ATI accelerator

The ClearSpeed COO also knocked FPGAs and GPGPUs saying developers are underestimating the work and transistor budget necessary to support double precision floating point.

"How do you do your rounding, handle infinites and other complex operations," McKinnon asked. "You can't do that overnight. These are pretty hard devices."


Seems to me that AMD might know a thing or two about implementing double precision. I would think the former ATI divisions would have clear access to this knowledge now. A point where nVidia will trail.

muzohub, sorry

I didn't mean to chastise him. Just wondered what his point of view was. As has been suggested on this board (or was it SI?) the first Conroe stock appeared to be a risk start. I just wondered what his ideas were on the matter.

And perhaps he isn't worried about the cache erratum. All chips have them; I'm not downgrading Conroe as a design -- this will be fixed in due course.

mmoy, on Conroe performance

Any chance your Conroe performance is unsatisfactory because you're using an early Conroe, with some workaround for the cache erratum published much earlier?

And, I hate to ask it, but why would you buy such an early Conroe with the seriousness of the cache erratum? Conroe seems to be nice, but I think I would wait until that erratum had been resolved. Otherwise, every time the system locked up, I'd be wondering if that was the cause.

pfosse, great post

Thanks for the comparison; great post!

alan81, on $20K per employee

That's the number that struck me as odd. $20K is only enough to get someone to leave if they already have a better job lined up. Sort of a "moving bonus" :). If you really have to pare away the problems, it's going to cost much more per person.

The cost numbers quoted just don't pass mustard, to me.

wbmw, on AMD/ATI

Since AMD owns ATI now

Really? My, time flys. I could have sworn it was still July, not December. And to think I missed both the AMD and ATI shareholder votes. Must be getting senile in my old age.

Jules, on posting entire articles

Please don't post entire articles from The Inquirer here. A snippet plus a link for those who want to read the whole thing would be better. The Inquirer is a business and produces their material with the aid of advestising revenues; if the advertising revenues are denied, the business can't survive.

There are other text sources where it might be appropriate to post whole text -- ones that do not keep material on-line for an extended period, for instance.

jhalada, on "reverse HT"

I think "reverse HT" has somehow gotten mistranslated as "reverse hyperthreading", when it really is probably "reverse hypertransport". "Reverse hypertransport" would mean that a coprocessor is in control of some aspects of the crossbar actions (through the hypertransport link) rather than the CPU being in control of them. This may mean the use of downloadable microcode from a coprocessor to control the crossbar and SRQs, rather than the built-in system of control from the CPU chip.

chipguy, on cache associativity

Isn't that their (AMD's) instruction cache? I thought the data cache was different, perhaps 16-way.

fastpathguru, on FPGA API support

Support for standard APIs as a first step into FPGA use provides the prospective buyer with a "quick win" and "instant payback" that can be used to justify the purchase. Once the device is purchased, the application can be modified more radically if that is of advantage, but it need not be so. There are always cost/effort trade-offs...

wbmw, on coprocessor take-up

Well, maybe the developer already used a standard API like:
http://www.amd.com/us-en/Processors/SellAMDProducts/0,,30_177_4458_3505%5E869%5E2272%5E8716,00.html

Then, perhaps the coprocessor comes with a drop-in replacement for that library. Then how much work would it be to support it?

alan81, on Yahoo evaluating Woodcrest

They mentioned that Yahoo is already evaluating woodcrest servers.

That doesn't stop Yahoo from evaluating Opteron servers as well. In fact, if they are evaluating Woodcrest servers, surely the benchmark they're comparing against is Opteron.

mas, on AMD's future.

I don't think you'll hear anything from AMD that Osbornes the current production. In particular, when K8L is talked about in June, it will be stressed that it will use the same infrastructure as the then-shipping AM2 devices. So you can buy an AM2-socket system this year and upgrade later. Not that many will, but it provides a great peace of mind to buyers.

avatar, on Intel option pricing.

Yeah, makes sense. Intel has no volatility towards the upside, and only $19 to go on the downside. :)

Ixse, on Turion X2 socket

I thought Turion X2 was going to be on socket S -- a revised 754-pin socket that supports dual-channel memory.

wbmw, on your numbers

I think your numbers look good. Now lop 10% off all the segment ASPs and I think you've got a decent model at an overall ASP of $98.

Dictionary feed is from MIT Dictionary server

wbmw, post from Browser Appliance running under VMware Player

Just for fun, this post is from the Browser Appliance running under VMware Player. Both free, of course. The Browser Applicance offers a secure browser environment.

Might be of interest to you, since you are well-spoken in these forums, to check out the dictionary supplied with the browser appliance. I don't know where they're getting their feed, but it's a joy to behold.

Fabulous | Fab"u*lous |
(fa^b"-u*lu^s), a. L. fabulosus;
cf. F. fabuleux. See Fable.

1. Feigned, as a story or fable; related in fable; devised;
invented; not real; fictitious; as, a fabulous
description; a fabulous hero.
1913 Webster

The fabulous birth of Minerva. --Chesterfield.
1913 Webster

2. Beyond belief; exceedingly great; as, a fabulous price.
--Macaulay.
1913 Webster

3. Exceptionally good; extremely pleasing. informal
PJC

Fabulous age, that period in the history of a nation of
which the only accounts are myths and unverified legends;
as, the fabulous age of Greece and Rome. --
Fab"u*lous*ly, adv. -- Fab"u*lous*ness, n.
1913 Webster

wbmw, on virtualization

I consider that Xen and VMware are the leaders here: which approach do you think will generate more mind-share?

On the one hand, you have Xen, freely available open-source, with the ability to run unmodified Windows XP if these articles are to be believed. All I can say for this is "wow"!
http://news.com.com/2061-10791_3-6061102.html
http://www.linuxinsider.com/story/UtCarQ0LA3aFjt/Novells-Xen-Hypervisor-Promises-Systems-Management-...

On the other hand, VMware, which offers a powerful product, commercially based, and reasonably priced ($189 for workstation, I believe). Who would want the hassle of dual-/multi-boot if this could offer multiple VMs simultaneously?

It really seems to be an ideal richness of choice. Have you any thoughts on the matter?

Here's a good Security Focus article on VM futures:
http://www.securityfocus.com/columnists/397

wbmw, on Mac VT article

Truthfully speaking, I can't tell you how well AMD-V will do, in comparison with the Anandtech article.

In fact, I really don't care how well it does in the Mac world at all, I'm not a Mac-o-phile (but that's just me, others have their own preferences, which I respect).

I'll be very interested in seeing results in Xen or VMware on both AM2 (and Turion X2) and Intel VT platforms. And yes, video seems to be a real sore point -- it will be extremely interesting how well things have progressed.

From what I've read in snippets, Xen and VMware each have been working extremely closely with both Intel and AMD. So this should be a technology rollout to watch. I expect great progress on all fronts.

Xen virtualization support for AMD-V

Gotta post here, I'm sure nobody can read all the posts on SI :).

http://news.com.com/2061-10791_3-6061102.html

Xen now supports AMD Virtualization
April 13, 2006 3:36 PM PDT

Programmers released version 3.0.2 of the Xen virtualization software Thursday, adding support for a hardware assist called AMD-V coming in future Advanced Micro Devices processors. The feature makes it possible to run unmodified operating systems such as Microsoft Windows, even though Xen relies heavily on Linux.

...

mmoy, on "AM2 non-event"

As far as AM2 goes, it looks like a non-event so I may just
go ahead and buy something now.

You must be joking! AM2 is a huge event. If you buy an AM2 system you'll be on the next-generation AMD platform, with next-generation memory and next-generation virtualization. Buy the workstation version of vmware (once the new version is announced, probably coincident with AM2 launch) and you'll have a great system on which to run Windows and Linux simultaneously. Build it with a Linux base system so you only need one Windows licence.

Any CS student, certainly any Open Source developer, any serious science major, should hold out for such a system. As a matter of fact, AMD marketing should line up several CS profs to recommend such systems: 64-bit, dual core, Linux/Windows dual OSes simultaneously. No dual-boot crap any more, this is the future, just in front of you.

mas, on pending battleground

We have ourselves the mother of all cpu price/performance battles about to occur with x86 domination and marketshare the goal with excess capacity removing pricing constraints. What has gone before will be mild and a cakewalk in comparison.

Armies are always built to fight the previous war. In this case, the server battleground may change to one of whose virtualization is best. AMD and Intel may well end up comparable in performance, performance/watt and price. But if AMD is able to pull off a coup in virtualization, the server area is theirs to own. And that's where I'm betting (plus, of course, Intel's lateness to the party).

jhalada, on abondoning the Itanic

It seems that some are jumping the Itanic even before the inevitable collision with the iceberg.

Oh, the iceberg has hit already. Its name was AMD64. The delay is just in having each watertight compartment give way in turn, until there aren't enough left to keep the ship above water. The first compartments breached were Opteron, 64-bit Xeon with EM64T, then Opteron Dual Core, Sun support, HP support. There can't be more than 2 or 3 left before it slides beneath the waves.

mmoy, on Intel buy

Good luck with your Intel. I don't see it getting above $25 any time soon. As for that $37 target, I don't see that happening in my lifetime :)!

Why will Intel not go up much? Because it is now clear that AMD will never be folding its cards and will continue to play in all the markets Intel has milked over the years. Intel will never again be able to charge near-monopoly prices. And AMD is the more efficient competitor in a duopoly situation.

jackthex, on board usage

There are lots of us that have been following the discussion, and you've handled it pretty well. To some of us it looks like you take the devil's advocate position. No matter.

There are posters on each and every board you find who will be annoying and useless. Yes, we have some here. But that doesn't mean that there aren't posters that are worth reading. I hope you'll become one such, with regular or irregular posts.

As for the debate subject matter, I too have not seen a *personal* need for XP64 yet. I don't run anything right now that requires it, and my TV tuner card would be useless if I were to switch over entirely. I don't game, I don't do video editing, I really don't watch much on the TV tuner (that's more for my daughter). So I can see where you're coming from.

I bought my daughter a Turion notebook recently -- it's running XP Home (32-bit) simply because I don't have time to make changes where they are not required, and I don't want to have more support required when I might not be able to do it in the time I have.

But when I buy a computer, I'm not going to buy a 32-bit-only one. You can't plan what you're going to do with that certainty, that you'll only need 32 bits in the time you own the machine. For instance, the TV tuner software I use can compress programs after recording -- I'm sure that it *could* run much quicker in AMD64 mode (extra SSE registers would probably make a big difference). Whether it would or not depends on whether they take advantage of AMD64/x64. So there is a potential advantage that I'm foregoing by not going with 64-bit computing right from the start.

My next buy will be running XP64 right from the start. Probably a Linux distro too. And it will be a dual-core AMD machine. And as far as I'm concerned, anyone who buys less is making a rash decision.

Oh, and from Google I find Visa. May be right, may be wrong...

gollem, on swap file size

I don't know if Vista does something differently, but usually the swap file will map the entire virtual address space -- the sum of virtual address spaces of all running apps (plus parts of the OS that are swapped). So if you expect Vista to use 2GB of virtual memory and apps to use another 2, the swap file size you would use would be 4GB.

In fact, if you want to use large-footprint 64-bit apps, a swap file size of 4GB is quite small. Remember that the footprint of everything running on your system must fit in that 4GB. I think Windows dynamically allocates the swap file, but don't be surprised if you want one to be 10 or 12 GB if you're doing video editing.

I_banker, on amateur mistakes

Yes, almost as amateur as doing a demo against the opponent's existing machine and having the BIOS report "Processor Model Unknown", LOL.

wbmw, on Richard interview

And yet, he didn't take the opportunity to note that Conroe is in the future and only AMD can supply 64-bit compatibility throughout their line now. He didn't pick on Core Duo being 32-bit only.

He also didn't pick on the Pentium 4s, he might have said that Intel may be performance-competitive when Conroe comes out, but is not now.

I think it is important to remember that this interview was for Digitimes, which is an industry organ, so time frames are different (in that the industry must be thinking about what they have to do 6 months down the line).

All in all, an interesting article. I look forward to seeing the next 4 parts.

wbmw, patent sharing/cross licencing

That was my understanding. See, for instance this report, admittedly dated 2004:

http://news.com.com/AMD+compatibility+no+problem+for+Intel+chip/2100-1006_3-5159067.html

Under the terms of the settlement, both companies gained free access to each other's patents in a cross-licensing agreement. AMD agreed to pay Intel royalties for making chips based on the x86 architecture, said Mulloy, who worked for AMD when the settlement was drafted. Royalties, he added, only go one way. AMD does get to collect royalties from Intel for any patents Intel might adopt.

It is my understanding that Intel can no more make a modern microprocessor without using AMD-patented technology, than AMD can. For instance, AMD has the patent on RISC-like x86 microcode. How do you think Intel would do without using that?

But I may be completely wrong. I'd like to hear from anyone who has the definitive answer.

gollem, on when Intel would be a buy for me.

Not until their shareholder's equity + dividends increases at a rate high enough over the treasury bill rate to merit the risk. As the current rate of change of those is below the inflation rate, the real rate of return is negative. So things will have to change quite a bit before Intel becomes an attractive holding to me.

Of course, if the current decline continues, the dividend might start looking attractive. That's when they'd start thinking about cutting it though.

mmoy, on macro ops fusion

What makes you think that AMD may not add macro ops fusion to their up-coming devices? They do have a patent-sharing agreement with Intel, n'est ce pas?

gollem, on Intel stock

at some point Intel stock is bound to go up, but when?

Really? How high? How'd you like to buy some Nortel from me? How are your SGI holdings?

"Bound to go up". Wow, new concept to me.

Tell you what, when it gets back to $25, give me a yell. I won't hold out for $37 :).

[Edit:]

Re the $37:

http://www.investorshub.com/boards/read_msg.asp?message_id=7917828

mmoy, on Intel buy

Lots of folks here trying to figure out how
to go long on Intel given the news.

Not me! If you want to believe this dog-and-pony show, go right ahead though.

wbmw, on Anandtech story

What FX-60? The first screen shot clearly says "Main Processor : AMD Processor Model Unknown". So Intel is unable to set up a computer with BIOS properly updated to recognize the AMD processor they installed? What a bunch of loons.