Microsoft early bird doesn't necessarily catch worm

Microsoft's early patch warning system

I happen to do some work for one of the 3,500 Microsoft customers that gets early security warnings. Here is an example of what you get:

Cut here

To: [insert customer here]

Microsoft Confidential Information - subject to terms of Non-Disclosure Agreement

Our PSS Security team attempts to provide advance notice to our Premier customers on the number of security updates being released and the products affected. This is intended to help our customers plan for the deployment of these security updates more effectively. The goal is to provide our Premier customers with information on soon-to-be released security updates.

On Friday 30 JULY 2004 the Microsoft Security Response Center is planning to release:

One Microsoft Security Bulletin affecting Microsoft Windows. The greatest maximum severity rating for this security update is Critical. This security update will require a restart.

Although we do not anticipate any changes, the number of bulletins, products affected, restart information and severities are subject to change until released.

At this time no additional information on these bulletins such as details regarding severity or details regarding the vulnerability will be made available until 30 July 2004.

Cut here

It's practically useless. The only benefit really is you can tell your IT staff in advance they will be staying late to apply some patches for the latest security problem. Maybe on lower severity problems they have advance technical details but I've never seen that.

If you publish any of this, don't use my name.

Xerox as a verb

Here in Poland, people since the begining of photocopying have used the term Xerox as a verb. It has fixed itself in the Polish vocabulary transformed and "polished" to fit even more, the word Xerox has become "ksero", since we pronounce "x" as "ks".

I, as an English speaking Pole, have had a number of situations where shop attendants did not understand me when I asked for something to be photocopied (translated into Polish, of course).

My two cents...

Peter Skonieczny

Dear Letterman,

Xerox is correct to protect their trademark Xerox from falling into generic usage. A better generic alternative is to use xerograph or xerography. Xerox overlooked this in their list of suitable generic alternatives.

Xero- means dryness and is used as a prefix to many word forms. Graphics made with a dry process would be xerographic. Google lists many examples of xerographics which have nothing to do with Xerox.

Xero- is a good prefix and we should not let it fall into disuse. A xero-example would be…

The Inquirer brings xero-humor to the xero-details of my xero-existence.

Thanks,

John Owens
Evanston, IL

Intel Inside holds up Dell Opteron adoption

I just noticed this past article at the bottom of the page while reading a more recent article of yours. I didn't feel compeled to write in at the time, but nobody else has brought up what was on mind when I had originally read it. The article basically infers that it's in Dell's best interest to remain Intel only as Intel pays great sums of money for Dell advertising via the Intel Inside deal.

It seems you, or whomever deduced this, are trying to convince us that Intel would spend less on advertising if all of the tier one vendors were pushing their Opteron boxes to the same extent as the Intel boxes. It may just be me, as nobody has contested this point of view yet, but does it not seem more likely that Intel would have to just accept the fact that Dell and IBM sell AMD systems too, and keep forking out the dough to advertise their own?

At this point, Intel not only has to spend big bucks on advertising just to move their wares, but they're now in deeper competition with AMD. Would Intel lie down and play dead with AMD nipping at their heels, or would they start pumping even more money into their advertising budget to differentiate themselves from the competition? I'd assume the later, though I seem to be alone in that assumption.

Regards,
Jesse

AMD Opteron outsold Intel Itanium by 10X

Mike

You will, with the same juvenile glee I hope, note that (imminent) moment when the number of Intel Xeon (Nocona EM64T core) processors sold passes the number of Opterons FOREVER, NEVER TO BE SURPASSED AGAIN, right? Because that's the meaningful comparison.

Beating Intel up because Opteron outsells Itanium makes no more sense than beating up any of the big RISC processors Opteron also outsells. Ask any market analyst that knows what they're talking about and they'll say comparing x86-based extended-64-bit processors (Xeon and Opteron) makes sense, but why compare Opteron to Itanium? Why not SPARC or Power or MIPs or ... - BECAUSE THEY, LIKE ITANIUM, ARE NOT VOLUME SERVER PROCESSORS.

I can't believe how dense you guys are about this - your David vs Goliath enthousiam for pushing AMD and slamming Intel really warps your thinking...

Kevin Mitchell

Emachine dream

I bought an Emachines M6805 for my son at the beginning of this year when it first came out. He needed a new machine and this one caught my eye because it felt a lot faster than the Toshiba mid-level system I was using (2.4GHz P4M, 512M, nVidia graphics), it had a better screen and graphics than the Toshiba (or indeed most laptops) and it was surprisingly cheap.

My son uses computers (too much, actually). He likes to play and modify graphics intensive games. The M6805 seems to have an unlimited amount of power for this purpose; it starts to break into a bit of a sweat with Doom 3 but its probably not alone in that. (Lets face it -- the mere fact that you can actually play D3 on a cheap laptop is pretty remarkable -- the Toshiba is barely good enough for GTA3.) I think it outperforms anything out there except systems that have been specifically assembled for gaming (the only laptop in that category that springs to mind is the Alienware and that was well over twice the price).

The performance of the system is significantly reduced when running on the battery. This is to be expected -- I think of it as a desktop replacement, not a true portable. (If you really need to run it on the move then you can buy an inverter to power it from a car for $35 at the local hardware store.)

This isn't our first Emachine, we've bought some of their desktop systems. None died, we just ended up giving them away as they were replaced. This system has proved reliable so far except for it getting filled up with crud (dog hairs and the like) which caused it to overheat. My son learned a valuable lesson about room cleaning from this -- he needs to clean his room not just because parents are nagging him but because all that crud will end up somewhere, in this case in a nice mat around the processor cooling fan. Getting it cleaned caused a bit of a scare because we have the system on a service contract from the big-box retailer that we bought it from, and since the BBR no longer carries these systems ("Thanks, Gateway!") they were threatening to replace it with an "equivalent". I had an interesting conversation with their service people about this -- their idea of equivalant performance would start at a Celeron...........I don't know what they're smoking, such a system wouldn't be worth the trouble taking home from the store (I don't have any table legs that need propping up).

The M6805 is an incredible machine. Everyone who has played with it has been impressed with it, and I know at least two of his friends bought these units because of his system. I am a bit worried by what feels like the Intel Marketing Machine -- it almost feels as if Intel's prepared to pay outfits like the BBR to swap out and crush these systems to prevent them from penetrating the market, because once you've used one you just don't have the respect for a P4 system like my Toshiba -- its just plain klunky.

Martin Usher

Intel Inside holds up Dell Opteron adoption

I told you a year ago that IBM accepted an $18 million check to kill its work on mainstream Opteron servers.

Couple of things to note for you:

1) Whenever Intel actually starts to deliver i86-64 parts they will run slower on the same program running in 64 bit mode than in 32 bit mode. The reason for this is "add" unit in microprocessor is actually cut in two and in 32 bit mode only bottom half runs, when you go to 64 bits the carry propogation from the lower 32 bits to the upper 32 bits takes an extra cycle increasing the typical instruction execution time by 3 cycles or 10%. What is worse is that the pipeline gets yet another three stages deeper, sigh, and ties up even more scoreboarding space (register rename, reservation stations and branch target and prediction slot). Overall it may be as much as 20% slower on some programs.

2) Remember that Intel has not implemented all of iAMD64. They do not do 64 bit I/O in any current or announced chips. This means that all I/O transfers must be staged through the first 4GB of memory, if you need to read or write data to an area above that you have to copy it to or from lower memory and then do the I/O form there. This makes for massively poor performance and the loss of critical memoriy to nonsense buffering really screwing the type of programs you need 64 bits for most, graphics and SQL Server.

3) Early reports put the dual core Precott (Nocona actually) power dissipation at 150 watts, even at a reduced clock speed. This is hardly a drop in replacement - even with the server class die on heat spreader technology they are running out of coupling area to get the heat off of the die. Contrast this to the Opteron where, once it is shrunk to 90 nm the dual core dissipates about the same as a single core 130 nm.

4) The E-0 stepping is carrying the family code from the original Tejas chip and possibly used that as a base. This would indicate that the Prescott is much more of a mess internally than previously thought. It has become apparent over the last several months that the i86-64 implementation in Prescott was done in such a way as to work program wise but be impractical as a "production" architecture (emasculating it so that it would not threaten the Itanium, give the rubes the ability to run simplistic 64 bit work but don't design in any of the features that would make it a superior server). Stepping D0 and now E0 have been correcting this as the market has forced changes (such as the NX bit). Note however, that the original compromises cited above still exist and must be fixed. It should also be noted that as I told you last year there is now no difference between the Prescott and the Nocona in a DP configuration, they have the same family codes, steppings and errata, they are the same chip.

5) The 2MB L2 cache chip is carrying the family code and stepping ID of the Xeon MP chip (which is also the Extreme Edition chip). This would imply that some insane person has taken all of the EMT64 stuff and re-integrated it back into the 130 nm chip, doubled the L2 cache and then shrunk that (as opposed to doing the sane thing and moving the cache coherency logic to a fresh copy of the E0 code base). This cost a fortune and it is no wonder they are a year behind in releasing the chip. Even though it is in the price book don't look for it on your dealer's shelves, it is going to be very late and very expensive. And you wonder why AMD is making mincemeat of Intel engineering.

Name, email address supplied