Meet you back here in half an hour.
What are you going to do?
What I always do - stay out of trouble... Badly.

Tuesday, February 15


Firefox 4 Review

Put the goddam status back, you idiots.

Each new release of Firefox seems to avoid fixing the one thing that has to get fixed (tabs need to run in separate processes, like Chrome) in favour of breaking something that worked fine before (URL bar, now status bar) and requiring a new plugin to get back almost the functionality we had previously.

Posted by: Pixy Misa at 07:07 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 64 words, total size 1 kb.


Argumentum Ad Puppulum

Or as Gir would put it, "Your head smells like a puppy."

We're all familiar with the common logical fallacies like the argument from popularity, argument from authority, ad hominem, and so on.  Even if we don't know their technical names, anyone who's spent more than half an hour online has seen most of them.

There are a few I've named myself; they're just special cases of existing fallacies, but they're common enough to warrant their own entries in the Bumper Book of Stupid Arguments:

Black Knight Fallacy

A special form of the argument ad nauseam: Consists of presenting an unending sequence of false and illogical statements, being rigourously taken apart by the opposing side in the debate every time, and then declaring victory when they get bored and leave.

Invisible Bigfoot Fallacy

This is what you get when you grasp Ockam's Razor by the wrong end.  The name comes from the classic example, explaining that the reason no-one has seen or photographed bigfoot is because it is invisible.

Argument from Dead Philosophers

A form of the argument from authority, but rather than presenting an actual statement from someone, the author just names a bunch of dead philosophers.  "blah blah blah Kant, Hegel, Descartes, Plato blah blah blah".  Extra points may be awarded for misspelling their names.

Posted by: Pixy Misa at 03:31 PM | Comments (4) | Add Comment | Trackbacks (Suck)
Post contains 220 words, total size 2 kb.


Death By Success

Anyone looked at the quality of questions on Stack Overflow lately?


Posted by: Pixy Misa at 02:59 PM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 15 words, total size 1 kb.

Friday, February 11


I Need A New Router

I'm getting 2% packet loss on the local interface.  That's not good.

Posted by: Pixy Misa at 07:21 PM | No Comments | Add Comment | Trackbacks (Suck)
Post contains 17 words, total size 1 kb.

Friday, February 04


Undecodable Chips

The third reason why I'm not typing this on a shiny new Sandy Bridge system (the first two being time and sharks) is that with the new CPU range, Intel have taken market segmentation to a whole new level.

The initial release of the desktop lineup includes 6 mainstream chips, 2 enthusiast chips (with unlocked overclocking), 3 low-power chips (65W) and 3 ultra-low-power chips (35-45W), in three general families: i3, i5, and i7.

The low-end i3 is a two-core, four-thread design.  The high-end i7 is a four-core, eight-thread design.  The mid-range i5 is a four-core, four-thread design...  Except when it's not, as two of the three ultra-low-power i5 chips are actually two-core, four-thread versions.

New features in these chips include hardware encryption - built into all models except the high-end i3 (yes, the low-end i3 has it); VT-x virtualisation, available on the i5 and i7 only; and VT-d I/O virtualisation and TXT trusted execution, available on the non-enthusiast i5 and i7 except for the low-end i5, that is, it's missing on both the low-end and high-end models.

The new chip also has graphics built right onto the CPU die, unlike the previous version which was a two-die package, with CPU and GPU separate.  Graphics comes in two versions: The HD 2000, with six EUs*, and the HD 3000, with twelve.  Oh, and two different base speeds and three different turbo speeds, but that's not important right now.

If you're still with me, here's where it gets good, and by good I mean stupid:

The high-performance HD 3000 graphics are only available on the two high-end enthusiast models.  However, we're talking Intel integrated graphics here, where "high-performance" equates to "barely adequate", and not something any enthusiast would touch with a barge pole.

But...  The big feature of the enthusiast models is that they have unlocked overclocking.  All the chips feature a turbo mode - if you're not running multi-threaded software, so that some of the cores are idle, the chip will automatically shut those cores off and use the increased power & heat envelope to speed up the remaining cores, by up to 400MHz.  With the enthusiast model, you can do whatever the hell you like with the overclocking, with the usual proviso of you break it, you broke it.

Anyway, these overclocking features require Intel's P67 chipset to work.

The integrated graphics require Intel's H67 chipset.

So if you want the faster integrated graphics, you have to pay for the overclocking model, when you can't overclock it at all.  And if you want the overclocking features, you have to pay for the faster integrated graphics, which will be completely non-functional.

And, for some bizarre reason, they turn off the I/O virtualisation as well.

As far as I can see, the only way to get a Sandy Bridge CPU which doesn't have one or more features disabled seemingly at random is to pick up the new (so new it's not out yet) Xeon E3.  Here, while the 1220 and 1225 lack hyper-threading, the 1220L is a suprise two-core special with only 3MB of cache, the 1225 only has 6MB of cache, the 1260L only has the HD 2000 graphics, and the 1220, 1220L, 1230, 1240, 1270 and 1280 have no graphics at all, three out of eleven chips somehow survived with all their bits intact.

Those being the 1235, 1245, and 1275.  There's only 200MHz between the low and high end there, so you might as well go for the cheapest option.

Of course, none of them are actually available; they're expected to be announced on the 20th of this month.  But that's okay, because while they use the same Socket 1155 as the desktop parts, you can't get a Socket 1155 motherboard either.

* European Unions.  Well, probably Execution Units or something like that.

Posted by: Pixy Misa at 02:19 AM | Comments (5) | Add Comment | Trackbacks (Suck)
Post contains 636 words, total size 4 kb.

Thursday, February 03


(Bio)Degradable Chips

That shiny dual-Opteron motherboard still hasn't shown itself in Australia, so in the meantime I was looking into the new Sandy Bridge systems from Intel.  They're not a huge technical advance over last year's models, but they're a good bit cheaper if you want a fast quad-core system, and they support DDR3 RAM - which my current systems don't, and which has gone since I built my current systems from being somewhat overpriced to cheap as dirt, around $100 for 8GB.

I wasn't going to buy a new system until I knew I'd have time to set it up, though - I've done that before and had parts still sitting in their packaging six months later.  So I priced up some configs and waited for the right moment -

And then a shark came along and ate them all.

The exact nature of the problem with the Sandy Bridge chipsets isn't clear - it's some sort of bit rot in the SATA controller* - but it's sufficiently serious that Intel has recalled all of the chipsets - every single desktop motherboard made for the new CPUs is going to be scrapped and replaced.  I was wondering when notebooks were going to show up with the new chips, and there's my answer - it would seem the notebook chipsets have the same bug.

The server version of the new chips was scheduled to launch on February 20th, but I'm guessing that this will either be pushed back or turned into a paper launch.  Though it's possible that the server chipsets are based on a different design and don't have the same flaw; I don't know, and since the product hasn't been officially announced yet, Intel aren't saying anything.

Intel are saying that the recall will cost them $1 billion - but that's just the direct costs in engineering, manufacturing, and repair and replacement of existing boards.  They also have to deal with the fact that sales of their mainstream product family have been stopped dead for at least a month, and that uptake after that will probably be noticeably slower than before.

On the upside, it puts my own occasional technical failings into a very comforting perspective.

* Actually, AnandTech seems to have all the details:
The problem in the chipset was traced back to a transistor in the 3Gbps PLL clocking tree. The aforementioned transistor has a very thin gate oxide, which allows you to turn it on with a very low voltage. Unfortunately in this case Intel biased the transistor with too high of a voltage, resulting in higher than expected leakage current. Depending on the physical characteristics of the transistor the leakage current here can increase over time which can ultimately result in this failure on the 3Gbps ports. The fact that the 3Gbps and 6Gbps circuits have their own independent clocking trees is what ensures that this problem is limited to only ports 2 - 5 off the controller.
They're probably the best PC tech site around; I should have gone there first.

Posted by: Pixy Misa at 10:53 PM | Comments (2) | Add Comment | Trackbacks (Suck)
Post contains 504 words, total size 3 kb.

<< Page 2 of 2 >>
63kb generated in CPU 0.0219, elapsed 0.179 seconds.
54 queries taking 0.1644 seconds, 352 records returned.
Powered by Minx 1.1.6c-pink.
Using http / / 350