While one or two articles over at Arse-Technica have been almost acceptable (and nearly technically accurate), most of their writers seem to be half-witted teens without the ability to read or comprehend an article -- much less a technical concept that goes beyond MHz or MIPS. I would ignore their stuff, since it seldom offers anything of value or interest, and is as insightful as a car nut telling you your car sucks because he has a blown-454 in his Pinto. All that I can do by responding is draw more attention to their idiocy and half-truths, so I've been ignoring their trolls. But I've been getting email about it -- so I will respond.
Their inability to comprehend was noted by me when they blindly attacked my response to Carmack. Their response completely slaughtered (and slandered) what I said -- and paraphrased me so bad, that I was left to wonder if they had read past the first paragraph (and questioned their understanding of that as well). But of course I had read other half informed drivel, like their first look at OS X Server, which seemed to get many observations wrong, and spin things poorly. But I first took note at stunning stupidity with their fluff piece that rants about how you can bolt a RISC-like hybrid core, on an antiquated CISC ISA, ignore the advantages of a load-store architecture and the disadvantages of being register starved and so on. The CISC vs RISC article on their site was so bad, that they have two more articles (RISC v CISC II, and RISC v CISC III) claiming that despite everyone who argues to the contrary (technical experts in the field or not) that they (Arse-Technica) still know better (their words, not mine). This whole series sounds more like shrill and desperate attempts to obfuscate the truth -- both the truth that their original article was oversimplified and error ridden, and that their intent was to implying that the P6 Core is RISC and the PowerPC may not be. Then in the utmost irony (or hypocrisy), they later attack me for explaining far more on this subject than they did.
The latest flame of theirs was because of an article I wrote on the Pentium III. I get many people asking about the Pentium III and is it the "revolution" that Intel claims in their ads, and what it all means. So I wrote an article that points out the facts that it is just new marketing on a Pentium-Pro core (which all major PC resources seem to concede), and I point out there is little performance to be gained by using it (for most), and that it is mostly hype. Something that most PC Magazines seem to admit (reluctantly and with lots of attempts to diminish that truth).
But Arse-Technica felt the need to respond with overemotional and under-researched hype (which seems par for the course). These guys may mean well, and may almost have a technical clue, but it is hard to see that in their usual opinionated one-dimensional ranting that is devoid of any deeper balance or understanding. They seem to take great offense at my articles and comment on them -- yet the writers (and editors -- if they exist) seem unable to get the very basics of what I say. The latest is from their appropriately named "Wanker Desk". They then attack me for oversimplifying when I write an article on the Pentium III -- without ever doing the first shred of research. If you doubt me, how about this stellar quote:
David Every of Mackido fame decided he would address the relevance of the PIII to the P6 core and the x86 architecture in his trademark style. He starts off with an oversimplified discussion of the Pentium Pro, the first processor from Intel based on the P6 core... There's a large body of excellent work that deals x86 ISA's shortcomings; the above statement, however, ain't in that body. Every doesn't point out any actual flaws, nor does he offer any real criticism. All he does is offer a condescending and downright criminal oversimplification of some solid and innovative work done by the P6 team.
So they were incapable of reading the following articles on the subject by me -- many linked to in the very article they are whining about.
But they were disinterested in information or learning and seem incapable of the most basic research -- they seem more interested in immature teen-ranting. I "brushed over the topic of the PentiumIII or Pentium Pro core failings" -- despite having half a dozen or more articles on the subject of varying details -- which they are too lazy to read, or too stupid to figure out how to click a link to.
Let's contrast this to what they have. One article where they do some benchmarks and agree with all my technical points. But the technical similarities don't seem to matter, they didn't like my tone.
I also nearly busted a gut laughing at that crack about "solid and innovative work done by the P6 team". Was the innovation where they borrowed from RISC, or where they borrowed SIMD? Was the "solid" work where they had the overflow bug, or where they had the Orion chipset bug (that caused lousy I/O performance for all the early P6 adopters)? Sometimes people are most funny when they aren't trying to be. They went on and on.
My big question is, if the x86 ISA is such an insurmountable hindrance to performance, then how is it that we've seen such performance increases in the x86 line? How is it that the P6 in particular has done so well? Apparently, the situation is not as deplorable as some would have us believe.
The thing I find least appealing about this site is their continued use of strawman arguments. They make this stupid inferences. Just because the P6 core is successful does not mean that it is RISC. I've never implied that it is unusable -- and there are huge legacy, business, and personal bias that go in to choosing a processor or computer. I've stated many times that despite the Pentium being a radically inferior design, that if you throw 10 or 20 times as many resources at it, you can overcome most of the flaws -- which is precisely what Intel has done. I never said the situation was "deplorable" -- only the design is. In quite a few of my articles I've mentioned that implementation is good and effective DESPITE the design. In fact that is the most impressive thing Intel has -- good hackers that can get around the really bad design. (I'm sorry I can't call what they do real engineering -- but engineering requires forethought and planning). The whole ending of that paragraph is nothing but a strawman attack, filled in with the usual uninformed assumptions. They imply that I said something I didn't, then attack that. This isn't journalism -- it is like trying to debate with an over-hormonal uninformed teen (1) about things that they don't understand, and are too arrogant and naive to understand that they don't understand.
(1) My apologies to teens. I've dealt with many that behave quite maturely, and can question and disagree without snottiness -- but they aren't over at Arse-Technica. I also know that their name isn't Arse Technica -- but I figure if their editors can't tell the difference between "Mac kiddo" and MacKiDo (despite a whole article explaining what it means), then they likely have their heads in their arse.
The strawman arguments go on and on. Saying things like:
...it's hard for me to accept that my processor is so outdated that it's practically criminal to own it. The simple fact of the mater is that the P6 core, the mind-parent of the Pentium Pro, Pentium II, Celeron and Pentium III processors, is not dead.
I agree -- but I am left to ponder under what drug-induced delusion makes them think I've said anything of that kind? It is an outdated design! DESIGN! The ARCHITECTURE SUCKS! But the implementation is not THAT bad. If you want to run WindowsNT or PC software then the P6-Core is completely usable (and arguably the top of their performance heap). But if you are looking at the design it stinks on ice. A Yugo or a Pinto will get many where they want to go -- but that doesn't mean it is a well designed car. Despite being a crappy design, the PentiumIII is still pretty quick (though overpriced) and the best architecture they have (until the K7 comes out later this year). I don't understand why some mental-teens (who see the world through polarized black-and-white glasses) can't understand the differentiation, or the facts.
The article goes on (and on, and on)...
...and all of you who are in the middle of sending me mail claiming that I'm worshiping Intel might was well stop now. The media's fundamental assertion, heck, Ars' assertion, is right. The PIII is just a PII with a change or three (SSE), redressed in marketing hoopla.
So in other words -- I was right all along with my technical assertions. They just didn't like the tone (and so had to misrepresent what I was saying). But they go on to try make some weird point to make it look like what I said was something else.
Yet there's a second part to this equation. Part I is only true if you acknowledge that the PIII as it stands now isn't quite the processor its 800 MHz brethren will be.
"Part I is only true if"? What the hell is that? Sorry, Part I is true on its own whether Part II ever happens or not. Considering Intel's history of late (and over promising), it is very risky and naive to assuming that they are going to deliver on their promises in a timely fashion. When the 800, or 8000 MHz version of the Pentium III exist, we can talk about it. The fact is that Intel said the PentiumIII was going to be shipping (initially) in a .18µ process, and be 500 - 600+ MHz and so on (their usual hype and over promising). They delivered something that was far short of their early hype -- and now we get to wait another 6 months or a year for them to deliver on their initial promises -- and once again PC advocates are blindly saying that we should ignore the now, and look pretend tomorrow's promises exist today. When they deliver, we will discuss it again.
Here's where people like Every constantly get burned. Someone rips Pentium X, says it's the end of the road, etc., etc., and a year from then the Pentium x+1 is puttin' smack down.
The article then continues its inevitable and unmerciful descent into the kingdom of the lame, with claims about how "people like Every" get burned. I never said the PentiumX is at the end of its road. That is yet another strawman -- do you notice the pattern? There could be a Pentium27, all based on the P6 core, and Intel could still sell a bazillion of them -- so I make no such predictions of doom. Wintel people have proven they will buy whatever Intel or Microsoft sells them. The fact is that many PC buyers are stupid enough that it doesn't matter what garbage Intel or Microsoft feed them, they will gladly chow it down, and tell each other how good it tastes. That only reflects on them, not me. My point is that design wise, it keeps taking Intel more and more money and time to get each new "version" out -- and each version is delivering less and less relative growth. EVENTUALLY that will bite them in the ass, despite what ignorant advocates claim, and despite the inane claims that they are still selling so they must be good. The next big delivery isn't scheduled until 2001 (Willamette), which is 4 years late!
The facts are that since the release of the Pentium, Intel has been able to grow performance at a rate of 58% per year -- keeping with Moore's law (barely) -- but that rate of growth is averaged, and of late it has been slowing down. Yet, Motorola and IBM have been growing in performance (not MHz, but performance) at a rate of 68% per year -- and that 10% per year difference is starting to show. PowerPC has had more evolutions and revolutions, and has some far bigger jumps in store in the near future (with MultiCore processors in the loop). This does not mean that Intel is doomed -- there are plenty of sheep, suckers, and poor bastards dependent on their stuff (and that guarantees revenues for years and possibly decades) -- but it also doesn't mean that because lots of people drive Yugo's that I should give up my BMW. They can have their crap, and I'll continue to advocate the superiority of a well thought out design.
Even their rant against me almost admits these things (in a watered down way) --
Intel may be averaging a moderately slow 3.5 year generation period between significant core changes. So what? There's one thing that's very important to remember: Intel's development cycles are dictated more by profit margins than by development per se. Intel milks the P6 because it can, it has to (CISC dev cycles are longer than RISC), and it should (from a business perceptive).
In other words. Intel is a monopoly that will only grow as fast as competition drives it. For once I agree with them. For now the RISC and PowerPC advantage hasn't been enough of a threat to them (mainly because PC users are afraid of changing, and IS doesn't want to give up the fatted calf of high support costs and a huge demand for IS). So Intel's falling behind hasn't been hurting them. But at this rate (or an accelerated one), how long until even the stupid open their eyes (and minds)?
Arse-Technica's conclusion seems to be that we should pretend that the emperor has great clothes (because everyone else says they do), and that Intel's mode of business has been successful in the past, so it will continue to be in the future. The problem with "driving by watching the rearview mirror" (and looking at marketshare and past performance as an indicator of future trends) is that it fails to anticipate anything. While it is true that things don't change often, they do change. If you assume that because there hasn't been a turn in the last 10 miles, that things won't change for the next 10, then you are foolish. Everyone that has made future predictions based on past results eventually gets bit (hard). Should we all be lemmings, rushing towards the cliff edge and assume everything is great, just because we haven't fallen off yet? I don't think so.
For the record, the tales of lemmings committing suicide by running off cliffs is fallacy. I use it as an ironic (and popular) example, and also as a way to ridicule "common beliefs". Lemmings run off cliffs just like Intel is good because they are big, and things are going to be good in the future because they are great right now. All equally inane claims.
The article is littered with other cases of either ignorance or intentional misinformation. Like I had said in my article:
"The PowerPC is going on its 4th generation and was released in 1993, which means it takes 2 years (max) per PowerPC generation. RISC matters--those that don't think it matters can't do the math."
To which their response was:
Maybe I missed something in computer architecture class. Is there some sort of Zen RISC design technique that allows you to channel your chi energy in such a way that your processors evolve quickly? ... No technology evolves itself, and Intel has shown, even with limited tech, that they can produce an amazingly powerful platform. I don't care what a textbook says. Reality is right in front of our faces.
Well DUH! That was the whole flippin' point of RISC. So yes, he did miss something -- the whole point of his computer architecture class! The point of RISC was to simplify the architecture in instruction set complexity (not count), so that you could use that space for other things, and evolve processes and architectures more quickly. Since time-to-market on processes means performance, and evolution can mean performance (or achieving design goals) this is very critical stuff. Yes, Intel has done an amazing job of implementation and hacking, and yes the textbooks do say that design matters -- but as the author points out "he missed something". The reality is right in front of our faces -- Intel has been losing ground. Their embedded market has all but evaporated. Almost all other architectures have gone to RISC (or RISC / VLIW / EPIC which are all RISC derivatives). Even Intel is trying to be as RISC like as they can be -- with a RISC-Like back end, and making IA64 into a RISC chip, and so on. So the point of this guys article seemed to be an excuse to attack me for stating the same things he would say (if he had a clue) -- but at least he told me what I already knew, he missed something (like the point) while sleeping through his design class, just as he did while scanning my article.
The author concludes with:
Perhaps the saddest part of all of this is that, for most commentators, the CPU is secondary.
I don't understand why that is sad -- the truth is the CPU is only one variable in the complex "what is performance" equation. Systems design matters. In fact, because memory performance has not been keeping up with processor performance, I suspect that memory speed (and design) will be far more important in the next 5 - 10 years, and as much a differentiating factor (in performance) as the processor itself. So he doesn't seem to understand that obvious fact either. (More napping through the import parts I suspect). Just to make sure we know he is a half-informed PC-biased buffoon, he ends with the following:
The only stake these dolts have in the PPC is the fact that they like the GUI that runs on top of it. If Apple said tomorrow, "we're dropping PPC and moving to Alpha" you'd never hear of PPC again, just a bunch of spooging about how the Alpha is the "most advanced processor on the planet"...except this time they'd be closer to the truth. What can you do when your OS only runs on Motorola?
<Sigh> Some people just don't get it. The Alpha is/was a great processor in performance for scientific applications -- but it is actually a lousy general purpose processor. Why? It costs too much (power, heat, money, memory, and so on). It makes a great server and scientific machine where you can ignore those things -- but that is its "niche". But some clueless PC types only understand "big" and "fast" without understanding the underlying design tradeoffs, and why for many things it is not the ultimate machine -- despite MHz and lots of execution units and big FP results.
As for "these dolts in the PPC", I could try to explain to this person why the Mac is not just a GUI and is instead a whole OS (unlike Window9x which is a GUI on top of DOS), or I could try to explain to him about user experience and how people want to use tools (and not waste time having to fix the tools or try to keep them running), and so on. I could go on about how the Mac is actually the usability of the entire System (hardware and software), and that yes, Apple could put the Mac User Experience on a different platform, and if it was as tightly integrated and as good a solution, I would advocate those advantages as well. But the sad fact of the matter is that I don't think he has the mental capacity to understand what I would be saying. To some it will always be about measuring horsepower, RPMs, toque and wheel size -- and never about style, usability, handling, comfort or reliability. Which is really sad when you think about it.
So go over there, and explore their site, read their articles, see if what they have to say is of value... and if you don't think it is then don't waste your time responding on their little BBS (that generates "hits" and ad revenue for them) -- just leave and don't return. I know I frequent them as little as possible, and when I do visit, I keep finding reasons to avoid them in the future. Don't get mad at these inflammatory malinformed pathetic post-teens -- feel the pity for them that they are "too smart to learn anything new" and "too dumb to really get it".