re: Note: In this case "non-profit" doesn't refer to AMD.
LOL!

The large majority of Oracle sales are on Intel architecture products, either Intel or AMD. Those platforms are by definition standardized, and any tweaks to improve performance for a specific platform are done by the hardware vendor, not by Oracle. It would not be practical to implement anything that would put HP at a disadvantage without also harming the other platforms.

Oracle's stated intention is to do optimization of the whole stack, hardware and software, a la IBM - something only a vendor who has the vertical integration can do. That strategy is primarily about building and maintaining a very high margin business, but it is not a volume strategy, because for the majority of users, the performance gains don't justify the price premium or the lock-in.

If HP aspires to be 'IBM-like' they need to do something similar, and their new CEO would certainly know a lot more about that than the previous one.

Answering the other question, Oracle has a much more complete application stack than MS, and also a higher-performance database engine - but at a higher price. It's really an apples to oranges comparison - Microsoft aims to be a platform provider and looks to their partner ecosystem for verticals, except in a few cases (like the dynamics packages).

Duke -
I agree that BBD was hard for the SAP culture to swallow. But that's part of what CEOs are supposed to do - if senior management determined that BBD was the key to unlocking the SMB growth SAP needed, but the rank and file couldn't execute, then the CEO should get the management team to deliver, or find executives who can. At a minimum, over a period of years, Apotheker should have adjusted expectations to match reality.

The challenges at HP are in many ways similar - to go to the next level, HP needs to grow beyond the culture that got it where it is today, and get all the dogs pulling in the same direction. Apotheker's track record doesn't give me a lot of confidence that he is the leader to do that.

I'm not sure which Microsoft software you mean - the SAP hooks in Sharepoint 2010?

Duke -
Leo Apotheker seems an odd choice. He was basically canned by SAP in February, and over a period of years at or near the top of SAP he showed no vision on how to get SAP out of the box it was in. As to people skills, SAP during that period was known for paying attention to only the top 100 customers. The only major initiative during that time was Business by Design, which has been pretty much a disaster.

But the whole circus at the top at HP seems to me to be about things unseen - it's just not credible that one of the most successful CEOs on the planet would get canned over some minor expense report issues that he probably had nothing to do with in the first place. So perhaps all will be revealed in the fullness of time.

I had some dealings with Ray Lane when he was at Oracle and was impressed - he was focused, knowledgeable, mentally quick and very accessible. And, as someone pointed out, it's almost as if he and Mark Hurd swapped jobs (with a little latency, as we say in the computer biz).

Another important difference is how a 32 bit OS 'uses' memory - the high order bit is used to separate OS and app regions, so effectively, applications are limited to 2GB of memory.

Some programs (photoshop for example) have both 32 and 64 bit versions - and you need both. Many scanners and other devices don't have 64 bit drivers, so you use the 32 bit version to scan, then switch to 64 bit to work with the images. Kind of funky but better than chucking a perfectly good scanner which doesn't have 64 bit software.

My daughter has an iPhone 4 - and both of these reports are right.
The problem is that the antenna is exposed - so if the user touches it, the reception falls off by 6db or more.

But if the user doesn't touch it, the reception is an easy 3db better than the iPhone 3

She bought a silicone cover for the iPhone 4 - problem solved.

But Apple could have also solved the problem with a little engineering.

More likely she did not support multi-socket designs...

Sun produced some really nice designs with Opteron early on - and they worked closely to get products like the 4600 to provide lots of cores and low latency for large memory configurations. Unisys toyed with the idea of adding an Opteron-based 'cell' to ES7000 based on that design, and even had a prototype custom-designed by Sun, but never went forward with it (they did end up OEMing 4600 directly for a little while, a few years later).

But in today's world, any sensible business manager would drop that product. The former Sun hardware needs to be radically simplified around a common base, engineering and R&D needs to be cut, and the obvious choice for that these days is Nehalem.

EP -
That's a great question - Compared to Hector, Eckhard Pfeiffer is Lou Gerstner. While I think most people would agree that Ben Rosen was a little quick on the trigger to fire three of the 4 founders (Gary Stimac remained, creating the ProLiant server line which was not a bad little product), Pfeiffer was not totally responsible for either the Tandem or DEC purchases - that honor goes to the dynamic duo of former DEC rejects, Earl Mason and John Rose, who I think had only their own self interest in mind. However, I was a big beneficiary of the Pfeiffer era - I first bought in to CPQ not long after Pfeiffer was made CEO, and rode it for a 10X gain over the next 5 years.

Hector, on the other hand, made even Jerry look like a genius.

Duke -
'Prevailing wisdom' today seems to be that Intel and AMD are selling off because Samsung announced it is expanding capital spending, possibly leading to oversupply. Oversupply of what? Samsung does not compete with Intel and AMD - in fact their spend would seem to indicate confidence in exactly the market that Intel and AMD play in.

Is this just nonsense?
Feel free to respond with a reworked movie or song - maybe something from 'get shorty' would be appropriate.

Yea I remember that. While the DDR and some other internals were a clever re-use of some Alpha ideas, apparently the SMP guys were not among the team that came from DEC for K7. The only Athlon MP board I actually touched was from Tyan, and it was pretty flakey. I was trying to put together a budget workstation but on most applications, the 2P system was no faster than a 1P 'regular' Athlon (T-Bird vintage), and the T-Birds also overclocked fairly easily.

Chipguy can probably help me understand some of this - but wouldn't the design standards that produced Barcelona have been in use when this guy was supposedly on the design team? And wouldn't the standards that produced Shanghai and Istanbul have been the ones he is denigrating?

Barcelona was pretty much a disaster. I have played with a few of those chips - hot, slow, temperamental, inconsistent. My experience with Shanghai was the opposite - methods for tweaking performance applied very consistently across most chips in the series, although with different headroom (but that's what binning is all about, I think). The power draw was lower than Barcelona across the board even for the top end chips, and performance for a given set of parameters was consistent and predictable. Not a Nehalem beater, but very clearly a whole different experience than the previous 2 years with Barcelona.

While I don't have any knowledge about Bulldozer aside from what I read here, I doubt that this guy's comments reflect much about Bulldozer's prospects one way or another, and based on his reasoning he might be one of those folks that is useful in a contrarian way - if he is selling, buy, if he is buying, sell.

Tench is right - prior to Opteron, AMD had NO servers with a major vendor - just a few white box offerings. I don't recall offhand what their 'market share' was, but it would have been in the noise level.

The original Athlon, however, did make quite a statement in the consumer space, reinforced by the Socket A products which introduced several advanced concepts (mostly cribbed from Alpha) and held the performance edge from late 1999 until Conroe. Even then, AMD did not penetrate the commercial desktop space to any significant extent.

Maybe they should partner with that WinTel company I keep hearing about

I know Apple has a fetish about product secrecy but this seems like a just plain dumb set of moves. Even if there was some chicanery sending in the goon squad. to a journalist, is sure to generate maximum negative vibes...

Certainly maintaining code lines for products with marginal sales is a challenge - that was the death of Alpha and PowerPC support at MS.

While Visual Studio surely has limitations around 64 bit development, most of those would not be for OS development - especially since Server 2008 R2 is 64 bit only and the rest of the 'enterprise' apps have or will follow suit. I'm sure it would be handy to have some of the emulator and debugger tools run natively in a 64 bit environment, but I don't expect the whole of VS to move until most of the source code is managed code.

The handwriting has been on the wall for IPF at MS for a long time. None of the 'enterprise' apps except MSSQL ran on IPF - so the platform was basically for high end SQL or for 3rd party apps. Many of those installations were on Unisys HW, and with Unisys effectively out of the IPF game, there was probably no lobby to maintain the IPF variant.

re: why conventional drive makers didn't just stuff a lot of RAM for a cache

For any serious enterprise use, the buffer on the disk is disabled for writes, because otherwise the data is vulnerable to corruption. So the primary benefit of on-disk cache is for read-ahead, or consolidating out of sequence reads if the controller does not do that already. As long as the cache has enough space for a couple of tracks, increasing it does not give much benefit.

Controllers like the HP SmartArray series use battery-backed RAM which allows the driver to cut the number of interrupts in half on a write, and does more effective consolidation of requests for large arrays of disks. For example, on a 14 disk array, with as much as 512MB of RAM, sustained I/O rates in the thousands per second are possible using standard magnetic media, and in the event of a power failure or other interruption, pending writes are maintained until the system is restored. You can even remove a controller and array from one system and put it in another, and the write will complete when the system is powered up, even before any OS loads.

The benefits of SSD are primarily when a single operation requires high I/O or bandwidth - such as database transaction logs, or initial loading of large files - in general, where the smaller number of threads or concurrent I/O load would not make an array-based design effective.

re: Not sure why you think its a stunt?

Because the engineers working on it told me it was a stunt. This was a heavily modified Dell notebook, with non-production silicon and the ability to turn power saving features on and off in special firmware to see which ones had real benefit, and to see the impact on performance with different settings. The unmodified production version of that laptop (same battery and features) ran 4-5 hours, the prototype ran 8-10 hours, so effectively twice the life with little visible impact on performance. I don't know exactly what is in it, but when they were done, they said the biggest power draw by far was the display. This is a 'big' notebook, 1920x1200 display, not thin or light by any standard.

IMO Intel is doing leadership work in managing the power-performance envelope - the flip side of a big laptop running 12 hours is turbo boost using thermal headroom. They may have gotten a little carried away putting the APIC timer to sleep on server chips in some C states, but they will sort that out...

I have never been impressed with AMD battery life - I had an Acer Ferrari a few years ago which, when new, lasted about an hour on a full charge doing fairly routine stuff. That's about useless if you're not near a plug. Neo is better than that generation, but AMD has a long way to go to get competitive on mobile power management.

I have a last generation core laptop which was done as a 'stunt' by Intel engineers to demonstrate advanced power management - they gave it to me after their testing. It runs easily 8 hours on a charge doing all the stuff I normally do. CULV presumably brings some of that to the masses.

neo battery life for at least 1 example (a friend's HP neo laptop) is 3-4 hours. That's his real-world result.

Also where HP's HQ for Germany is located - great place to fall in love and get married

I got married in Boeblingen...

Duke -
I was told that AMD imposed a mandatory use of a week's vacation and many people chose to use the Xmas break for this - so at least that part of salary really is a deduction,, vacation being an accrual.

This summer, I read several articles on the Abu Dhabi sovereign wealth fund, which was working to be more transparent. I can't find the articles any more, but the claim at the time was that the SWF had more than a trillion dollars in assets, maybe as much as $1.6 trillion. That is money already taken out of the ground, not possible money in reserves.

If those numbers are anywhere near accurate, they are not likely to be much affected by a few billion here, a few billion there...

I agree - the management change from Ruiz to Meyer was very significant - it was the reason I decided last year to invest in AMD in the first place.

As I understand it, much of the rest of the senior management team is also new - and on the product side, mostly ATI. Does anyone have a feel for the ATI culture? They seem to execute pretty well, and clearly don't have the anti-Intel bias - Intel platforms represent the lion's share of ATI revenue sockets, and I believe the ATI team still maintains a close relationship with their Intel counterparts.

Smooth -
That's kind of what I was suggesting - once a bunch of powerful people get the bit in their teeth, with an agenda, there's no telling where it will go.

Smooth -
I did more than a hundred hours of time prepping execs for the MS - DOJ depositions. The attorneys know very little about the intricacies of the business and are in many cases guided by what they learned in the last deposition. Their assumptions, phrasing, and descriptions of what might be going on were laughably simplistic. Imagine talking about networking problems, to someone from a culture which has no electricity.

Much of the deposition work was defining a 'language' which the DOJ attorneys and the MS attorneys agreed on, before any of the 'facts' were even discussed.

In that case, and others I have been involved with, folks on both sides were eager for 'the truth to come out' before the process started. At the end, what resulted was a third version of events which clarified little, satisfied neither side, but made a lot of money for the legal teams. My guess is that if Dell and Intel executives testify, there will be two completely disconnected starting positions, and three when it is done - what Dell wanted to see, what Intel wanted to see, and whatever mangled version of events the legal teams agreed on.

Joey -
re: AMD would have to have outperformed Intel chips by a significant margin (which they did not) for it to make sense for Dell to switch

Well, the only problem with that is that Dell DID switch - in the sense that they offered AMD products for the first time - and that in 2004-2005, Opteron was clearly superior to Xeon in many areas, and had the 'buzz' value of X64 leadership as well.

Dell was being punished in the server marketplace by competition from Opteron-based servers from both HP and IBM. Dell had worked hard to get credibility in that space after being a perennial also-ran, and the shift put a lot of heat on management.

It did not take Intel long to recover, and Dell ended up getting hit with both ends of the stick - they displeased Intel, AMD could not deliver the volume Dell needed, and Intel came back with a winning product while Dell was coping with all of that. Dell's timing was terrible.

IMHO -
I am not at all convinced of Intel's "guilt" - I agree with the many posters who say that you can prove almost anything with the right selection of email.

But I have enough experience with the legal process to also believe that the same selective process works in that system. I helped prepare a couple of witnesses for the MS DOJ depositions, and it was almost impossible to NOT make some admission that the feds wanted - they just go back and ask the same questions, cut you off if you don't give the answer they want, then go around again. We spent 8 hours on discussion of a single email thread at one point. The government attorneys completely mis-characterized the intent and context of the email, but simply presenting their viewpoint as a question, again and again, created the impression that the MS witness was dodging, when he really was just astonished that something so obviously wrong could not be addressed.

I think the best thing for Intel would be to avoid that Circus if at all possible.

I was responding to the tone of the previous post, not attempting to predict how Dell will testify, if it comes to that. My guess is that they will try their best to just stay out of the fight, but given some of the "evidence" that will be hard to do.

What I was describing is what I see as a fairly pervasive "revisionist history" viewpoint within Dell - it is a lot easier for those at Dell who were near the top in 2002-2006 to find something other than management incompetence to blame for Dell's decline. If (as I would assume) that attitude is reflected in ex post facto email, you no longer have the broad issue of Dell's relationship with Intel, you have the much narrower issue of some exec explaing why he said what, even if it is not germane to the case. I would think that internal politics would trump the need to be best friends with Intel.

I also agree with Duke - a strong Intel and a withering AMD is not the issue that I would want to see raised if I were on Intel's legal or PR team.

Yup - I know Rollins was ousted as much over declining 'perception' of Dell as for declining sales. I also know that as stated, AMD was not able to meet Dell's demand once Dell came on board... but revisionist history is always in fashion with corporate survivors.

Dell pretty much shifted back to Intel as their lead horse after core came out. Their product line also improved a lot in the post-Rollins era - they still have the most compelling portable offerings IMO, especially for commercial machines that do double duty as a home machine.

Dell has certainly backed off on AMD based products. I suspect this just reflects their notion of customer demand - Dell has never been a 'market maker' but they excel at keeping track of the pulse of the customer. Right now, AMD is out-gunned in most of the 'premium' space. You can find AMD offerings, but they don't jump off the page when you visit Dell's website.

Tecate -
re: Intel can EASILY show that Dell lost the hearts and minds of their customers by shoddy made computers

There you go being logical. With the current 'pile on' mentality, Dell will "show" that they were "forced" to exclude AMD, despite "sales and customer demand", and THAT is why they lost their leadership position. I know this has become an established Dell position - a friend who was at Dell in this period claimed that Dell's exclusively Intel product line cost them thought leadership and took them from "Invincible" to "Also Ran". Never underestimate the internal political value of scapegoating, or its effect on a company's prevailing mindset.

re: Apparently, the FTC has looked over the NYS AG suit that has convinced you of Intel's guilt and decided it was not enough to bring charges, at this time.

Link? I think that is wishful thinking and that the FTC will go forward with some kind of action, based on all I see in the press.

I agree - this is simply beyond understanding. A friend who once worked for Ruiz described him as follows:

He has a huge ego, even for a CEO. He likes to slap his d--k on the table and talk about what a big hitter he is. He trash talks other tech leaders.

On top of that he is a terrible manager - he sets his direct reports against each other, defines conflicting charters to drive 'creative discord' and treats his best people as inferiors. He also shows open favoritism to his lackeys and rewards loyalty over accomplishment.

I thought Ruiz was way overrated even before he went to AMD, and I regard his time at AMD as a disaster for the company. Hard to believe anyone could have been worse than Jerry but Ruiz managed it. This is a rare case of a poser finally getting caught out.

My guess would be that Ruiz had motives unconnected to money - and he was already pretty well linked to financial success of AMD and GloFlo.

Chipguy -
re:Unisys is barely in the hardware business at all these days and the services side enjoys far better margins.

Not true at all - UIS HW (systems and technology div) contributes almost all of the profits - they still have margins near 70% in the 'legacy' business. Services is struggling to get 20%.

It all depends on the DVD support. Most Netbooks don't have optical drives (or have an external drive) so the USB route is probably required there. A fast USB 2.0 stick is faster than a 4x DVD reader, so I could see an improvement if that's the race. On a 16x DVD reader, I usually see the install complete in 15-20 minutes - about like what the link claims for USB.

I'm not sure why anyone would want to install Win7 from a flash drive... the DVD install has been remarkably trouble free for me on about 20 laptops and 30 desktops. No problem identifying HW on 90% of the laptops, some going back to 2001. The only issues I had were with a fingerprint reader (had to go to mfg's site to get Win7 driver because of the new security model) and with a non-standard RAID controller on a desktop.

I have done upgrades both from within Windows and booting the DVD - not much difference but the install from within Windows was a little smoother.

Win7 even found drivers for printers and network cards which had no Win7 or even Vista drivers - it identified them by chipset and loaded a driver which worked. That included an Olympus P400 dye printer - that one runs better under Win7 than under XP even though it has not been supported in years.

I would just try running Setup in Windows from the DVD and see what happens - my guess is few if any problems.

<edit> I also question if a USB drive is faster than DVD - unless it is a very old DVD drive. I think the posted guide is for older machines that have problems with DVDs in general... or don't have a DVD. I loaded Win7 on a laptop with no DVD by copying the DVD to the laptop over the network.

EP -
Win7 is at its heart a 64 bit OS. The server version of Win7 (Server 2008 R2) doesn't even come in a 32 bit version. I don't know what the retail packaging will look like but my guess is that both versions are in the box.

BTW I put server 2008 R2 (64 bit) on a 4 year old Irwindale 2P system (a ProLiant DL380 G4), upgraded to 16GB RAM. Performance is at least 2x the 32 bit OS it was running, a lot more in some apps. All of the HW was identified without issue. I think this OS will give a new lease on life to a lot of aging but still viable systems.