You've not a shred of integrity

Hi Mike,

I notice with alarm the way you minimize the fall out from AMDs recent calculation error bug. The bug is much more likely to actually show up than the Intel FDIV bug a few years back, yet you blandly accept AMDs assurances that this is no big deal, that they are working on a BIOS fix - fat lot of good that'll do for all the punters who already have the chip and who don't know how to do a BIOS upgrade. In fact, from a technical point of view, this bug plays out very similarly to the way the Intel FDIV bug did. But not from a press reaction point of view.

Had this been Intel, you would have been fanning the flames of discontent, talking about how that big bad Intel was trying to get away with the commercial equivelent of murder.

If you had a shred of integrity, you would be demanding that AMD do what Intel ultimately did with the FDIV bug: Offer to take back every single last chp with the bug and replace them with a new part that would work with the old BIOS, and you would be reporting on this every day until AMD responded.

Your bias is showing.

Steve Williams

Steve Williams writes in today's letters that the Inq. isn't playing fair, because they are toning down the bug on Opteron whilst they made a big fuss about the FDIV bug.

I believe that the Inq. wasn't really there when the FDIV bug was revealed, but that is beside the point.

Steve's argument for this being as bad as the FDIV bug is interesting, because it's about the probability of failure.

I did some analysis, and first of all, I'd like to make clear that it's very unlikely that a compiler would ever come up with a piece of code that causes the problem, because compilers are, first of all, rarely able to come up with a reverse (using eflags.d = 1) REP MOVS.

Where the most likely scenario of this would happen is in some memory copying code that copes with overlapping sections. But this code needs to also use another microcoded instruction prior to using the REP MOVS instruction. I'd like to challenge Steve to find some code that ACTUALLY uses a pattern that matches this.

So, we're looking for some code that does: ... [1]
std
... [2]

rep movs
...

where either [1] or [2] has a microcoded instruction that is not serialized (all serialized instructions will automatically make sure that this bug doesn't happen, since it means that the instruction is finished before the next instruction can be performed).

And where the count (ECX) is either unknown (possible bug) or known to be less than 20.

I haven't been able to find one so far, but I have only analyzed some of the more important and bigger DLL's in Windows.

Seeing as there haven't been any apparent reports of failures due to this (admittedly, the FDIV bug went undetected for quite some time), I'd say that the conditions to make this happen is not exactly a common occurrence.

Finally, there is one BIG difference to the FDIV bug, and that is that the fix for this is to upgrade have the BIOS load a microcode fix. This method of fixing processor errata was not available when the FDIV bug occurred (but may have caused Intel to start thinking about how to implement the Microcode patch feature).

Mats Petersson
Software Engineer

Consumers revolt against LCD monitor pricing

I think alot of people are not willing to invest big money in these lcd monitors until;

a) Refresh/rise/fall issues are addressed, I would really like to see a LCD do a "true" 120+ hz refresh rate. And of course brag good rise/fall times.

b) More options are needed for handling non-native resoltions. Is it much to ask that a native lcd at 1280x1024 could do 640x480 with double sized pixels instead of that nasty blur mode every non native mode seems to use. and no, I do not mind those pixels between 961-1024 being ignored for the sake of gaining 1 good looking non-native mode.

How about this option, say you got a smaller non native mode that is not an easy multiple of the native mode. How about allowing me to choose to simply to place the imagine in the centre while simply turning off the outside borders? ... Instead of blur mode..

Are these things much to ask for?

Part B) is a easy do in firmware

Anyway, fix these problems and I will slap $1000-2000 on a good LCD.

Name supplied

Gaming mags taking bribes for early code?

Here is a solution to “Gaming mags taking bribes for early code?”. Don't buy the game based on reviews. Download the game, play it and if u like it that much buy a legal copy of the game. No harm done. Or just play the demo….

Name supplied

The top 500 supercomputers

In regards to your recent articles about the 'top 500 list' of supercomputers.

According to the actual people doing the counting of those flops, at this address, the #1 Baby of the computer world is the 'Earth simulator center' in Japan doin' a whoppin' 35,860 GFlops Max, with a peak of 40,960 Gflops. Now for a little perspective. The earth simulator cost about $400 Million according to this page.

Now, for the flip side, we have the Seti@home project. According to those wonderful college geeks at Berkely, at 9:45 CST, the current total statistics for the past 24 hours are sitting at 6.760178e+18 (78.24 TeraFLOPs/sec) Or roughly 78,240 GFlops Max on average.

The cost of that project? Well, for the information I've been able to find, based on an article in Wired written on 1/7/2000 "The budget for the project is a few hundred-thousand dollars. " Keep in mind as well, at that time, They only had 1.5 million users, with about 500,000 active at any time. Currently they have over 5 million, with about the same # of active users. (also important to remember that the work units have changed, as well as the definition of an 'active user'.)

I'll tell ya, for my tax dollers, I'd rather the US GOVT. put some more marketing dollers into those folks at Berkeley, and make a few more BOINC projects, then trying to build the fasted supercomputer. On a price/performance ratio; they even outstrip Apple and AMD by doing almost twice the amount of work at 1/400th of the cost of their closest competitor.

Hope you found this informative!

Martin M

Sanitising DVDs
Hi Mike,

Love your site. Keep up the great work.

I believe Congress is simply providing consumers with additional choices by allowing the market to decide what it wants to see.

Today, parents may allow or prevent their children from watching films with blood, sex and violence (in their own home, at least). Under the Family Movie Act of 2004, parents still have exactly those same two choices in addition to a third choice because now the market will include DVD players (or modified DVDs) that provide "selectable" levels of blood, sex and violence--meaning either yes or no in the most basic sense.

I see two positive outcomes from this act:

1) Hollywood sells more DVDs to consumers because people who would not purchase specific films on due to content they found objectionable now have the ability to view those same films in a sanitized form. Do they miss content that may be critical to the plot of the film? Possibly.

2) Parents have more family entertainment choices.

The market will determine the appropriate level of sanitization through the choices consumers make.

If Hollywood was focused on the marketplace, they would already be including flags in DVDs to identify various content according to some rating system for sex and violence. They would be licensing the use of these flags to DVD manufacturers and already have their hands on the money and the market ClearPlay and other ventures are attempting to capture. An even more market focused approach would be to film alternate lines of dialogue and/or scenes to provide a seamless film and complete plot at two or more rating levels.

The argument that the Family Movie Act limits artistic freedom is either disingenuous or a lie depending on how charitable one feels. Studio executives and MPAA ratings boards already tell writers and directors what will be allowed to be in a film or at least the location of the boundaries of acceptability depending on the rating they want for the film.

No choices are being removed from the marketplace. No artistic freedom is being curtailed. Hollywood can still make any movie of any rating they desire. When they sell a DVD they are still selling the rights to view the film in whole or in part. They cannot, however, dictate how the consumers watch the film in the privacy of their own home.

Fundamentally, this is no different than the fast forward button on the remote.

Hollywood is just mad because they refused to go down this path first based on either hubris or the belief they didn't operate in a marketplace that wanted choices.

Cheers!
Chuck K

Family Movie Act II

The Family Movie Act would make it lawful for a person who is watching a motion picture on a DVD in the privacy of his or her own home to use software that filters out certain types of content that the person would prefer not to see or hear.

This is no way implies that the filter will be required, or even recommended. It merely states that using such a filter (which already exist) is not against copyright law (as movie companies are claiming at the moment). If one cannot choose to filter, than that is as bad as forcing one to filter.

Thanks,
Jeremy Kane

More on Monitors

Hi Mike,

I wonder if this is related to the fact that nobody seems capable of producing a 17" TFT monitor with a native resolution greater than 1280 x 1024 pixels? The whole point of buying a bigger screen is so you can see more information on it, yet I can get a higher resolution than that on my laptop's own 15" TFT screen. Right now, my laptop is not only the fastest computer I own (I have two desktops too), but it's also got the highest-resolution display, even though my main desktop PC has an 18.1" TFT NEC Multisync!

I can get a 17" CRT for much less than the price of a 17" TFT, yet a modern CRT will run comfortably at resolutions up to 1600 x 1200 or so.

A 17" TFT monitor displays only a few more pixels than a 15" TFT monitor. Worse still: a 19" TFT monitor has *exactly the same native resolution* as a 17" TFT monitor! The 19" TFT just costs more, takes up more space and uses more electricity. It may be useful to people (like myself) who have less than perfect eyesight, but for the vast majority of people, bigger TFTs just aren't worth the money.

Regards,
Sean Timarco Baggaley

Zombies are Guilty!

I see from your piece:- ISPs call for Zombies attack clone that there is some inherent assumption that the original owners of zombies are somehow blameless and innocent. Here's an analogy: anyone can buy a car and drive it round their garden but to take it on the road is now so fraught with danger that we insist on training and an driver's license; there is a need for social responsibility.

Likewise, I suggest that anyone can buy a PC but when you connect it to the internet, a lack of expertise can cause considerable inconvenience to others. Fortunately no new legislation is required; simply an acceptance that for someone (such as an ISP) to disconnect a zombie until it is sorted is a perfectly reasonable thing to do.

The owners of these machines are not so much hard-done-by innocents as thoughtless and reckless drivers who believe, because nobody has ever explained otherwise, that they have a divine right to drive inexpertly in a public place and disrupt the lives of others.

Having said that, I appreciate that in many cases, it is the fault of the dealer who sold them the car without explaining the importance of using the key to stop the engine and lock the doors when leaving the vehicle. Many of the older models were delivered in a "hot-wired" condition to save trouble with unlocking.

Mark (George) Rainer

Microsoft and FAT patents

Nick:

This is in regards to your recent article "Microsoft's FAT patents aimed at Linux" for the Inquirer. It seems you misinterpreted Eben Moglen's commentary to eWeek, and have made the situation sound much worse than it actually is.

First of all, the Microsoft patents in question do not cover the entire FAT filesystem, but only a technique for using long file names instead of those limited to 8 characters. There is a common misconception that without rights to these patents, all FAT formatted media would be unusable.

Microsoft does not own any fundamental rights to the baseline FAT format itself, only specific extensions to that format. Microsoft's recent attempts to get digital camera media manufacturers and others to pay up only capitalizes on this misconception (since these parties rarely if ever actually use any of the patented features.)

Secondly, your comments about the enforcement of the Microsoft patents being a "goodnight Vienna" scenario for Linux and open source are wildly exaggerated. As Moglen mentioned in the interview, the GPL forbids the distribution of software encumbered by enforced patents. However, the Linux kernel consists of millions of lines of code, of which the FAT long file name support is only a tiny subsystem.

Therefore, any enforcement action by Microsoft would simply result in the removal of the long file name related code from Linux until the situation was resolved. We in the Linux development community have encountered similar issues before, and at no point have they affected Linux as a whole.

The worst case scenario from removing the allegedly infringing code would mean that Linux users would see long filenames as short 8-character versions, but still fully accessible under Linux. This is also unlikely to affect access to most removable media from digital cameras and the like (which use only short filenames not covered by the patents, as mentioned above.)

Of course, all this assumes that the patents are not immediately invalidated by the USPTO after their upcoming re-examination, and that Microsoft decides to actually enforce them. In this case, as mentioned above, the Linux code would be rewritten before their lawyers could even figure out who (if anyone) to sue.

Thanks for taking this information into account; I hope it clears up any confusion on the scope and impact of the Microsoft FAT patents.

Matt Yourst

Toner Costs

Dear Mike

I read the article on toner cost and I like you to know that more than 50 % of the toner used in a printer are dropped in a container ( called used toner container) so when you end up your toner cartridge, it remains 50 % of your toner in it.

15 years ago, in prehistoric time of computing, I bought a Tandy (radio shack) laser printer, with all those containers open, so regulary, I drop the used toner in the new toner container, and my printer worked fine for a very very long time,

So the price of toner is really very very expensive.

best regards
Guy Leclercq