Rink,

It's a bit weird though to hear you say you think that I think Bush is to blame for everything from the birth of civilization till present.

I was (maybe not so) obviously being sarcastic.

According to the list I recall seeing in one article MS executives paid MORE (not exclusively but just significantly more) for Bush' campaign than Kerry's campaign

I don't know about the breakdown between executives and rank and file employees, the figure I recall was for all employees. Assuming what you posted is true, it would hardly be an earth shattering news to find out that executives are more Republican that non-executive employees, regardless of the pending case. But I am not denying that some of the executives (with a lot of stock or stock options) would vote in their self interest.

What I don't see is that future administration policy of a candidate being guided by political contributions from employees from one company favoring it, even less likely so favoring a company who's employees favored the opponent.

Here is what I found:
Microsoft, based in Redmond, Washington: $1.9 million, with 61 percent to Democrats.
Cisco Systems, based in San Jose, California: $489,000, with 53 percent to Democrats. About two-thirds of the Cisco donations come from individual employees, and a third from Cisco's PAC.
Intel, based in Santa Clara, California: $337,000: 50 percent to Democrats. Just over half of the Cisco contributions come from employees.
IBM, based in Armonk, New York: $331,000, 65 percent to Democrats. IBM does not have a PAC--all donations come from individual employees.
EDS, based in Plano, Texas: $299,000: 74 percent to Republicans. About two-thirds of those contributions come from the EDS PAC.
Siebel Systems, based in San Mateo, California: $276,000, 63 percent to Republicans. Nearly all of the money came from Siebel's PAC.
EMC, based in Hopkinton, Massachusetts: $228,000, 88 percent to Republicans. About three-quarters of the money comes from individual employees.
Dell, based in Round Rock, Texas: $217,000, 74 percent to Republicans. About three-quarters of the money comes from individual employees.
EBay, based in San Jose, California: $194,000, 54 percent to Democrats. More than half of the money comes from EBay's PAC.
HP, based in Palo Alto, California: $193,000, 57 percent to Democrats. More than half comes from individual employees
http://pcworld.about.com/news/Aug202004id117487.htm

Anyways, The principle, i.e. (more) campaign money for a candidate that can get you favorable policy, is simple, and therefore it also works. There are many many examples of US corporations trying to influence politics via campaign funding.

I would agree with this as far as legislation, tax law, regulation etc. I just thought your post extending it to favorable rulig by FTC was a bit of a stretch.

Against this background I've always fully understood why AMD tried to get EU and Japan to rule on Intel's monopoly abuse before they would try the same in the US. There are more reasons why I think that is logical but this is one of them (they would not stand a chance in the US).

I think Intel knows how to play the system better in the US, as far as when to push more, when to pull back, when to settle. MSFT ended up in a more confrontational situation because MSFT did not want to give an inch. At least that's my interpretation.

Joe

chipguy,

Dozens? You mean over a hundred dozen last year?

I was obviously thinking quarterly figures, and if Dell sold 1200 per year, it averages to 300 per quarter.

What do you think the ASP of Dell's IPF servers were last year? And what about the ASP of Sun Opteron servers in 2004?

When you are talking about unit sales 2 to 3 orders of greater, it would take a hell of an ASP to make Itanium attractive. Besides, I doubt there is much of a premium to be gained by Dell selling substandard, generic Intel box vs. superior HP designs, and superior hardware and software support by HP.

How do you think Dell and HP Itanium ASPs compare?

Joe

Rink,

The US first as businesses are allowed to financially support the campaigns of those who can bail them out (e.g. Bush and MS) which makes the US system effectively corrupt.

While it is true that everything that has ever gone wrong in history (since the early Mesopotamian civilization) is Bush's fault, there are limits to corporate giving to political campaigns in the US. It is funded mainly by individual contributions. Some companies like their employees to bundle their contributions to make an impact, it is illegal for companies to force or compensate employees for contributions.

I don't know about Intel, but I seem to recall that MSFT employees contributed more to Kerry than Bush.

Joe

Paul,

re: demand limited. Easy - ignorance

People who decide what to put in their PCs (DIY crowd) are not ignorant. OEMs are not ignorant. The difference is that Intel can't bribe individual purchasers, but it can brine OEMs, and due to sheer market size and virtual monopoly in certain segments, Intel can use other tools to discipline OEMs who don't fallow Intel orders.

Joe

Keith,

A lot of Intel "fans" just tow the company line, and deny, deny, deny, even something that is undeniable. The defense used to be that if there was no conviction, it did not happen. So now we have a conviction...

BTW, for laughs, check out posts about excuses why Dell does not offer AMD processors. The pricing / discount deal is set up in a way that AMD would probably have to give (I am guessing here) close to a 1M CPUs for free in order to offset lost discounts / advertising support.

A good case in point is Opteron and Itanium. It is "economical" (due to illegal Intel pricing practices) for Dell to sell dozens of Itanium based servers vs. selling Opteron based servers by 10s of thousands.

Joe

Keith,

How else can a superior, less expensive processor be "demand limited"?

Joe

Beavis, ha ha ha, he said, ha ha ha, Itanium and profit, ha ha ha, in the same sentence ha ha ha.

Paul,

I wonder how much was lost in the translation, going from English to Japanese to Chinese and back to English...

Joe

mmoy,

Thanks, I thought I missed the release.

I think Beta 2 will be a public release, and will get a significant increase in participation.

Joe

wbmw,

Based on your comment on 64-bit development software, it's likely that businesses will skip WinXP-64 entirely and wait for Longhorn.

That may be true for corporate client machines. Server will go big way for 64 bit and consumer market will follow, and then consumer client.

The problem that Yonah will face is in the high end of the consumer market, IMO, because it is one thing for an inexpensive product to be a dead end product, it is a problem for the most expensive product to be a dead end product.

I think 64 the there will be so much press about 64 bitness over the next year that even clueless consumers will be aware of the term, and it will be one of the checkmark items.

Intel may still be ok for Christmas 2005, but if Merom is not out for Christmas 2006, Intel will have some problems in its most profitable market segment.

Joe

mmoy,

Is Beta 2 available?

It does say February CTP here:
http://lab.msdn.microsoft.com/vs2005/default.aspx

But when I click on "download latest builds", it only offers Beta 1:
http://lab.msdn.microsoft.com/vs2005/get/default.aspx

Joe

Itanium cores to get "fatter". Hmm... I thought that a CPU with a lot of small cores was the way for Itanium to dominate x86. I guess not.
http://news.zdnet.com/2100-9584_22-5593047-2.html?tag=st.next

Also, Itanium again misses target:
http://news.zdnet.com/2100-9584_22-5593047.html?tag=st.prev

Joe

chipguy,

re: IBM DROPS ITANIUM SUPPORT

A clear sign that IPF is starting to impact IBM's RISC
sales.

I knew you could do it. Turn dropping Itanium chipset into a positive.

Joe

jokerman,

22 Watt for MC

I think 22 Watt figure (if accurate) is the sum of memory controller, 2 DRAM interfaces and FSB interface. All these are built in (and accounted for) in the power dissipation of K8 processor.

Joe

mmoy,

I read somewhere that the PCI-X controller in Apples is AMD's 8131 chip. I think others are IBM's / Apple's.

Joe

wbmw,

Looks like Toledo will get a 15% power benefit over Smithfield - that's actually quite a bit less than what Winchester gets over Prescott. Yeah, I'm sure Intel's in big trouble!

I wouldn't say Intel is in trouble, but I would not say that your comparison is valid either.

2.4 GHz part has a rating of 4000, while Smithfield is at 3.2 GHz, so to compare the power consumption, you would have to compare 2x4000 Prescotts, or Smithfield running at 4 GHz. And that would still assume that dual core hammer does not derive additional performance benefits from the way interconnection of dual cores.

Joe

chipguy,

If you compare the portion of BCS revenues represented
by IPF server sales vs what Gartner and/or IDC say HP's
IPF server sales are and it seems clear that BCS can not
encompass all IPF server sales by HP's ESS division.

It would not completely surprise me that these endless reorganizations and re-segmentations will result in not very clear definition or what is in each division.

I think an interesting segmentation for HPQ would be:
1. ink
2. everything else

Joe

From the looks of that 1 xbit comparison, power usage may actually be lower clock/clock with the 6xx parts (vs. 5xx). That's a little surprising, so I'm not sure it is completely correct.

Either I am confused or the xbit is wrong. Isn't the 3.73 CPU with 2MB L2? If so, it should be on the same line as 6xx line. I think the labels may be reversed...

Joe

wbmw,

AMD demoed in August, and parts are expected some time in the second quarter. That's roughly 9 months, give or take. Yonah was demonstrated last year, and the public schedule says Q1 2006. Now which do you think is more conservative?

I don't see dual core Yonah and dual core Hammer as direct competitors (even though there will be some overlap). Dual core Hammer is primarily a server chip, and Yonah is primarily a mobile chip.

I am mor interested in processor with code name I believe is Dempsey, which will be Intel dual core server chip. Have you heard any updates on that one?

Joe

Keith,

re: HP quarterly results

Enterprise Storage and Servers (ESS) reported revenue of $4.0 billion, up 9% over the prior-year period. On a year-over-year basis, industry-standard server revenue increased 19%, business-critical systems (BCS) revenue declined 2%

Thanks for the reference. I take it Itanium, which is gaining some revenue, can't even at the best positioned Itanium provider slow down the exodus from expensive propriatery solutions to competitive x86.

I think the pace is going to increase in H2 2005, when Windows is in 64 bit, SQL server is in 64 bit and dual core will be available (at least from AMD). 64 bitness was a big selling point of RISC, and it is about to melt away...

Joe

wbmw,

Says who? The data I've seen seems to suggest that x86 is only gaining market share by growing with the market, rather than from RISC customers transitioning to it.

I posted some links to market research showing that x86 exceeds RISC based on revenue.

As far as growth, going forward, x86 is likely to just grow with the market unit-wise. Any win from RISC to x86 is not going to grow x86 unit market significantly, since it is already larger unit-wise and RISC is small unit wise. It is not even going to make a huge deal revenue-wise, since the x86 replacing RISC is generally cheaper. But I think we will continue to see revenue share shift to x86 mainly due to RISC revenue shrinking.

There still seems to be a rather large and toughly fortified market of users who have and always will buy "big iron".

In a territorial war, there are safe areas, and there are contested areas. So what needs the most attention is border areas, where either solution will do, and x86 is likely to win most of these contest thanks to price, price / performance, software availability, and inevitably software price / performance.

There is, IMO, another variable, not really completely technologically related, more of a human factor. Most of the older, retiring IT decision makers grew up in the world of mainframes and minis. It was fashionable among these guys to "dis" PCs, and technologies that grew out of the PC. These guys alwasy had a big-iron bias. The younger guys have a pro-PC bias. As the young replace the old, the bias will change, and with it, fewer people will think of reasons to waste money on RISC, where better performing and cheaper alternative exists.

Joe

wbmw,

Frankly, I think the platform approach is very earth shattering, but having said that, it will of course be a challenge for Intel to change their corporate culture enough to make it all work as seamlessly as theoretically possible.

There is a trend to move what is relevant to the microprocessor (floating point, memory controller even graphics - with various level of success), and to migrate the various components that are on the motherboard into few. What used to be called chipset can these days be more or less a single chip, sometimes two.

So maybe this platform strategy has been going on way before Intel decided to call it platform and give it a distinct name. Nvidia and ATI are pursuing the same goals for both Intel and AMD systems.

Joe

wbmw,

If Intel is going to market IPF as a RISC replacement, they aren't going to have much success with vendors that are heavily dependent on x86 infrastructure. Rather, it makes more sense for Intel to put chipset and platform development on Xeon, and let the OEMs create their own solutions for IPF (like they already do for RISC). The economics work out better this way, since RISC/EPIC systems carry enough margins to fund internal R&D towards specialized chipsets.

The economics of this has not worked so well so far. The R&D investments result in higher prices, which in turn result in low volume. On the other hand, "generic" solution based on Intel chipset has been well behind (HP, SGI) in performance, again, slowing down movement of Itanium to larger volume (if software was not a big enough obstacle).

I don't know what the strategy will be going forward, since the IPF and Xeon infrastructure is going to merge...

Joe

wbmw,

Unfortunately, even with a near-$100 ASP as in the past quarter (aided by much higher ASP Athlon 64 and Opteron products), they still ended up with a net loss of $30M.

I think there was a good increase in ASPs in Q3, but (I will have to check the CC transcript), I think the ASPs were either flat or down. And, the processor part of business is showing ok profitability while investing more in future designs, technologies and capacity.

The prospects of further increasing ASPs are likely to become more difficult

Hopefully, one of these days, AMD will break into the commercial segment, the way they broke into server market. It is not the quality of the product that is keeping them from there.

Joe

wbmw,

Your software argument applies for some segments of the server market and not for others. Intel put the stake in the ground recently by announcing that Xeon will have EM64T and thus compete with Opteron in all the segments where x86 is currently being used. In segments where RISC is being used, the software story is obviously different to begin with.

Intel may target IPF to RISC market, but RISC market is not shiny city on hill living peacefully. When IPF got there, it found it to be a crumbling, under assault of invaders who got there way before IPF got there - x86. x86 has been replacing RISC in many segments where it exists, and has been doing it for some time.

Going after RISC does not make IPF immune to competition with x86. Once Intel sells the Xeon processor, the OEMs or customers take it where they want to make it. If Intel makes it difficult to take Xeon to some markets, AMD will be there offering Opteron.

Joe

wbmw,

I don't think it's fair to consider the advances Intel introduced prior to their platform shift. Most of the features you mention came years ago, when Intel was still centric on the CPU. I think features going forward are going to take greater levels of investment in the platform, and AMD will be harder pressed to do the same with 3rd parties.

Am I the only one who does not consider the "platform thing" a earth shattering event? It seems to me that the main point, or definition of the "platform" is for Intel to evict everyone else's silicon. Other than that, it is just a set of componets put together.

Dell chose to go exclusively with the volume manufacturer, while Sun is placing all their bets on a temporary interconnect advantage from AMD. 3 years down the line, do you expect AMD to once again reinvent themselves and Intel to be similarly disadvantaged? Can Sun hope for the same miracle to happen twice?

If Sun does become a relevant, or even top tier vendor in x86 server market, they will probably pick Intel processor in areas where it is more competitive in order to provide competitive solution to their customers.

Very much unlike Dell, which offers only Intel, regardless of whether Intel solution is competitive or uncompetitive.

Regarding Sun, I think they probably invested more in software for AMD64 than in hardware, and software is compatible. Sun can add Intel product anytime there is a compelling reason.

Joe

wbmw,

Just like the other OEMs, Sun will eventually figure out that duplicating an R&D effort is a money losing prospect when Intel is willing to do it instead. Over time, with so many adjacent groups and a lot larger R&D budget, Intel will design better solutions than any of the OEMs are willing to spend to compete.

I guess HP's Itanium chipset would not be a good example of what you are trying to say.

Joe

wbmw,

Exactly. This strategy, by the way, is not the one preferred by investors.

Unless you are an investor in a company with ASPs of $50, even a 1/2 of $800 ASP in the server market is 8x of your current ASP.

Joe

wbmw,

Dropping IPF might enable Intel to drive Xeon margins low enough to the point where some RISC users might buckle under the pressure of cheap, commodity parts, but it will be a decade long battle, at least, and the side-effect is an acceleration towards commodity pricing that will hurt Intel's margins going forward.

I don't think it has anything to do with with Itanium, and it is not even completely Intel thing. I think it is software (DOS, Windows) that created volume market, and it dropped margins on everyone, except (near) monopoly vendors such as Microsoft and Intel. The main thing is dropping the margins on the hardware OEMs. It has done a number on just about all of the RISC vendors. Intel managed to maintain comfortable margins on the processors, but there may be some pressure from Opteron, which is going to further accelerate the speed of the freight train. RISC vendors (including Itanium) can't do anything to stop it.

Joe

wbmw,

I would assume that in order for you to make the above statement, you would have to assume that there are zero platform differences between a system based on AMD or Intel, 3 years down the road. I couldn't disagree more.

How much differences do these things make for the end user? Consider PCs. There were some new featurs (MMX, DDR, SSE, RDRAM, integrated video, SSE-2, DDR-2, AMD64, integrated networking, disk controllers, wireless) and companies such as Intel and AMD always caught up with the other one in very short time, or some features started on high end models first, eventually migrated to the rest of the line. How much all these differences really made to the users, as far as being on bleeding edge and getting something 6 months to 1 year earlier than everyone else?

It is very likely that there will be a similar game of catch up in the server market as it was on desktop.

In the end, developing it all in house will be a huge advantage for Intel

Except when the management commits to something like RDRAM, and deny customers choice of PC-133 or DDR.

Sun would be idiotic in their hubris not to offer both Intel and AMD, to at least let their customers make the choice.

Are you saying that Dell's policy is idiotic?

Joe

Keith,

If Newisys hadn´t done all the work for their current line-up, they wouldn´t have anything today.

Which actually validates AMD's decision to fund (partially) Newisys, to do just what they did.

As far as Sun, I believe the workstation is their own design, and last time I looked, it is a very nice design.

But, now is the time to follow up with their own server designs. More than a year into Sun's commitment, they have nothing of their in the server maket shipping yet

Joe

Keith,

I have been complaining about Sun execution as well, and I commented that considering how critical Opteron systems are to Sun (compared to their importance to HP), HP still kicks Sun's a$$ in execution, and HP lineup is now way ahead of Sun.

On a different subject, one thing I was wondering (that was not mentioned anywhere and specs where not posted) is if there is any improvement in # of registered DIMMs and their speed under the E stepping, if there was any improvement.

Joe

wbmw,

The company shipped 13,002 servers with Intel's Xeon processor in the first nine months of 2004, of which 58 percent used Linux. For Opteron servers, Sun shipped 10,566, of which 73 percent used Linux.

I don't know why you keep coming back to this, since it was explained number of times. Here is my post with numbers provided by burn2learn:

http://www.siliconinvestor.com/readmsg.aspx?msgid=20989450

Joe

wbmw,

Features like what? Memory controllers are dependent on the DRAM timings, and I would have thought that these would be mostly optimized by now.

I said CPU features, not features of the memory controller (I understand combjelly did), but memory controller also may have some conservative and aggressive settings.

There have always been utilities around to set certain aggressive features on (even if not officially supported or disabled because of bugs).

I am not claiming this is the case, just offering a suggestion for what combjelly probably meant. We will have to wait for some reviews of the E stepping to see the differences from earlier steppings.

Joe

wbmw,

Look how long it took for AMD to get cross over from Athlon MP to Opteron, in spite of the latter outperforming by wide margins.

Since Opteron and Athlon MP are are completely different products that just happen to come from the same manufacturere, I don't know if it is even comparable to a part that fits in the same socket.

Secondly, I don't know if it makes any sense to even talk about crossover in units. I think this thread started with what products are likely to be competing with each other, and Keith mentioned that Potomac's likely competitor will be dual core Opteron. What does crossover have to do with anything?

More important variable, IMO is price. Price determines which products compete. In that sense, I think Keith is right in bunching Potomac and dual core Opteron together, since dual core Opteron will be closer to Potomac than single core Opteron.

Single core Opterons are much more likely to compete with Xeon DP (Irvingdale and Nocona).

Since you mentioned crossover, I have no idea about AMD plans, but there may never be one. If dual core is priced much higher than single core, single core will continue to sell in higher numbers.

Now if AMD prices dual core only very slightly above single core, that would indicate to me that AMD intends to transition all server chips to dual core, and in that case, we could talk about crossover.

Joe

wbmw,

How would the BIOS have any affect on memory controller logic optimizations? Do you expect a set of registers that need new timing values?

My guess is that BIOS disables certain features of CPU that have errata associated with them. When errata are fixed (in new steppings), the features can be enabled.

Joe

wbmw,

Just how much would you expect it to improve when you add bandwidth to an interface that isn't the bottleneck in a system?

I agree that in 2 way system, bandwidth is very unlikely to be the bottleneck, but there should be a measurable increase in performance from reduction of latency of accesses to the other processor needed to maintain memory coherency.

A heavily loaded 4-way system would probably start to see some gains from the bandwidth.

One thing that is a little puzzling to me is why AMD stopped at 1 GHz HT, if the spec defines the cap at 1.4 GHz. It seems that another transition will need to take place, that could have been taken care of right away. I guess the new sockets coming out in 2006 will have that support.

Joe

wbmw,

AMD cannot produce enough dual core chips this year to transition from Opteron, but I'd expect Intel to transition to Potomac from Gallatin very quickly. Therefore, I would think that claims of "direct competitor" may differ in reality from what is perceived by the enthusiast press.

If you said that AMD cannot produce enough dual core to replace the entire desktop line, I would agree, but what do you think would be the sticking point in producing somewhere between 100 and 300K processors of the die size that is only a little bigger than 130nm Opteron, of which AMD produced millions.

Joe

Keith,

Their tower-server line remains INTEL-only for now

One-way designs from tier-one OEMs don´t exist at all. I´d like to see the DL-365, as that would be the second most popular model.

Good points about the lack of tower based models that some companies without racks perefer.

Also, I think 1 way boxes will become more popular with arrival of dual core. It would be nice to see some el-cheepo models with either Opteron 1xx or Athlon 939.

Joe

NaS,

I didn't have a single problem getting it to install Windows. I still have one item in the device manager without a proper driver installed (Pseudo SCSI something or other that looks to be part of the RAID controller). I'm not doing RAID so I'm ignoring it for now.

I am not familiar with nForce4, but nForce3 (which is similar) has a feature to disable the RAID feature in BIOS, and then, you don't need to install the driver for it. If you leave it on, just locate the RAID driver. If it is enabled in BIOS and Windows sees it, it is better for Windows to be aware of all the resources it uses.

Joe