CPU geekery

[danaeris]

In a critical commentary of Wikipedia, ZDNet's Paul Murphy makes some assertions about the CPU page on Wikipedia:

He writes:

"...what’s been allowed into the Wikipedia’s primary article on CPUs focuses on Intel x86, includes numerous statements offered as fact that are simply not true (everything from the assertion that Intel was the first to create the microprocessor to the claimed design equivalence of Intel’s multi-core packaging to Sun’s CMT/SMP architecture), and essentially denies AMD’s role in repeatedly forcing innovation within the x86 framework.

Exclusions work the same way too: just as the climate change article allows no hint of dissent, the CPU article gives the reader no hints that PPC derivatives dramatically out sell and out perform x86 while using less power and costing significantly less per operation per second."

I'm wondering if any of you can verify or contradict his statements regarding CPUs? Not particularly interested in debating his statements about climate change or Wikipedia just now.

Comments

[nathanjw]

Hmm. That article is definitely heavy on the Intel-based examples; that might or might not be forgiveable in light of the dominance of the Intel x86 architecture in desktop/laptop computers. But the complaints seem a bit out of place; the article isn't (I think) about the complete history of the CPU, or about current market share or comparative performance.

The PPC thing probably refers to the fact that all of the current game consoles use PPC-based CPUs, and sell like hotcakes. It may be right - I don't have numbers - but I'm not sure it's a material fact to that Wikipedia article.

"denies AMD's role in repeatedly forcing innovation..." It's true that the article doesn't mention that, but again, I don't see any reason why it would mention that. A better place for a discussion of that would be a page that was specifically about Intel, or about the Intel x86 architecture - and in fact, it's there.

[hellsop.livejournal.com]

PPC chips are also used in a fair number of other embedded applications, like network hardware and television settop boxes, and Power6 chips power IBM contemporary p- and iSeries servers.

[nathanjw]

Certainly. My last job involved a lot of development on those, especially the Freescale PowerQUICC line. But I think volume isn't that huge, compared to the juggernauts of the gaming consoles, and because the market is more fragmented (among PPC, embedded x86, ARM variants, MIPS variants, and so on).

[geekosaur]

I believe the claim that Intel was the first to create the microprocessor is correct (Intel 4004). Did Murphy give his idea of which was first, and his source?

[hellsop.livejournal.com]

concur essentially with nathanw. Fairchild did have an arithmetic processor on a chip somewhat earlier than Intel's 4004, but it didn't have some important things, like the ability to actually handle memory fetching itself. The 4004 wasn't really very useful as a CPU either. It could manage four bits at a time (Gotta love a computer that can't count past 15) and it took until the 8008 until there was enough capacity to make a real computer out of it. Industry cheerfully went ahead with making Real Computers out of discrete modules and parts rather than the 4004. 8008 got some use with controller modules, but the successor, the 8080, was actually usable to make a general-purpose machine. It also kind of set Intel on a precedent that haunts it today: it used a larger set of the same instruction set that the 8008 did. That "feature" continues on through the 8086, 80386, Pentium, all the way up through the contemporary Pentium M and Pentium-IV chip.

[kallisti.livejournal.com]

Definitely one sided! There is no mention of the Motorola 6800 & 68000 CPUs which evolved to the PowerPC design of today, nor of MOS Tech's 6502.

The 6502 chip that started the Micro-computer revolution, powering the Apple (1, ][+, //e, etc), and the Comodore (PET/VIC-20/C64, etc). Another important but not mentioned chip is the Zilog's Z80, which ran many of the early business micro computers under the CP/M operating system.

It's almost as if an Intel Rep wrote the article...

ttyl

[cjthomas.livejournal.com]

I'd forgotten about these chips, but I'll second the statement that they were important in their times. The 6502 and derivatives, and the Z80, were as-described. The 680x0 series ran the Amiga, which was a competitor to PCs in early days (and did multimedia much better than they did). I don't think the 680x0 was a precursor of the PPC line, though; what I remember of the architectures indicates that they were very different, and the jump from one to the other was a massive undertaking on Apple's part (to port all of their software, produce an emulation layer for 3rd-party backwards compatibility, etc). The PPC was a derivative of IBM's Power line, licensed and adapted by Motorola (who'd made the 680x0 line if memory serves). Apple recently transitioned to x86, which required a similarly huge porting effort (though if I understand correctly some of their lines still use PPC chips, for now).

[cjthomas.livejournal.com]

As far as I know, the answers to the specific objections raised are:

• Intel and TI were both trying to make self-contained CPUs at about the same time. Industrial espionage was involved; I no longer remember who swiped who, but one acquired an interim design for the other's chip, then turned around and patented the concept of the "CPU" based on it, which the other had considered obvious enough to be unpatentable. The espionage was demonstrated by the patented version including various useless/incomplete bits that hadn't been fixed/finished/pruned yet in that version of the design draft.
  
  
• Intel's multi-core chips are indeed CMP-type technology. Intel's "hyperthreading" was likewise re-branded SMT technology. Both of these appeared in academic papers well before anyone put them on silicon to sell. The first CMP chip fabricated was called "Piranha", and was a research chip made by a university if I recall corectly (8 cores, tiny L1 caches per-core, shared (small) L2 cache, directory-based cache coherence on-die to keep them consistent with each other).
  
  
• AMD is indeed noteworthy for fostering competition in the PC CPU market, though they're far from the only ones to do this. First time they did it was in the days of the 386 and 486, when they made clones of Intel's chips that were typically binned at slightly higher clock speeds and sold at a competitive price. Second time they did it was when they absorbed engineers fleeing DEC, and built the Athlon (designed as a "pentium-killer"). Later chips in the Athlon line implemented SIMD floating-point operations, which may have been part of the impetus behind Intel implementing SSE to do this type of operation. Not sure where the PPC line's AltiVec came into the timeline (was after AMD's "3DNow", but I don't remember if it was before, after, or at the same time as SSE). It's important to note that chip design cycles are on the order of a couple of years, so they'd each have learned of the others' intentions before any of these features came to market.
  
  
• The PPC comment sounds at least as biased as what this person is complaining about. x86 is king for desktop systems. PPC derivatives are king for console gaming. Supercomputing tends to vary (used to be split between at least 4 vendors without a clear winner, nowadays tends to be x86 or PPC based clusters, with some exceptions). Embedded systems are sometimes x86 derived, maybe sometimes PPC-derived, sometimes other things like Hitachi's SuperH line, but usually things like ARM for devices where low power consumption matters.
  

That's my two cents worth, at least. I graduated from Comp Eng in 2000, was in industry briefly, have been doing grad-school work since then, and followed the industry in detail up until a couple of years ago, if you're wondering where I pulled all this out of.

[avt-tor.livejournal.com]

I claim no expertise about CPUs, but I do know how Wikipedia works. There is no significant controversy on the talk page of this article. I skimmed quickly through a thousand edits; nothing jumped out looking like an edit war of opinion.

Wikipedia's strongest guidelines demand a non-partisan point of view and that all statements of fact be supportable by external references. This is clearly an area where anyone with an opinion or complaint has the power to fix their own problem. All changes in Wikipedia are recorded, so any assertion of bias could be easily verified and corrected by the people who believe it to be biased.

Wikipedia isn't perfect; it is written and edited by volunteers, so it only as good as the people who decide to contribute to it. That can be anybody, so there is no real grounds for anyone to criticize mistakes that they haven't offered to correct. This article seems a bit light on references, which does suggest room for improvement. However, it is also a bit long and wordy, for which the usual solution is to summarize the main article and put detail into subordinate articles.

But a critique of Wikipedia is just plain disingenuous unless the critic explains what actions they took to fix the articles they had problems with.