Talk:Intel 80486

From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Computing (Rated C-class, High-importance)
WikiProject icon This article is within the scope of WikiProject Computing, a collaborative effort to improve the coverage of computers, computing, and information technology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
C-Class article C  This article has been rated as C-Class on the project's quality scale.
 High  This article has been rated as High-importance on the project's importance scale.
 

Comments[edit]

can you come on internet and surf well with a 80486 computer?

Totally. My first internet experience was actually on a 386 (monochrome screen, slow at 19200 baud, and accessed via AOL - but still the actual internet...). Much of my formative internet learning was on various SX/25s, DX/33s and DX2/50~66s, DX4/75~100s at school and home (with 14.4, then 33.6kbit modems... and an awesome 5x86/133 at 56kbit :-D). 16 bit bus, no FPU, 4mb RAM, 25mhz and a 512kb ISA video card, and you can still make a reasonable fist of getting online --- though it may have a bit of trouble these days. DX2/66 with 16mb and a PCI graphics card will, and did, do better. However I think for surviving a lot of the modern internet, that's broken a lot of the original concepts of universal usability etc with its embracing of flash and other things - that were originally supposed to be optional add-ons to easily portable HTML code - as the backbone of major important sites, you may need at least a 64mb P133 (which my dad did until a couple years ago - everything parsed pretty slowly, but it worked), or a 128mb P200 MMX (my brother's emergency machine, which is a fair bit more tolerable). You WOULD be able to come online and surf successfully with an old 486, if needs be ... but it would be hella slow and possibly fall over completely at the more complex sites. 193.63.174.210 (talk) 14:53, 24 October 2008 (UTC)
A 50mhz 486 or faster with 16mb of ram is suitable for surfing the internet. Just you'll need to use a lightweight browser, as most browsers such as Internet Explorer and Mozilla are very bloated. K-Melon and Netscape 3.04 come to mind. A software modem would also be a no-no.
The internet is usable on a 486SX/25 (the slowest 486 ever made) running Linux and the Dillo browser. However, a lot of web pages can't be accessed (such as Hotmail and Yahoo web mail accounts), and others look strange in this browser. Samboy 04:16, 23 Apr 2005 (UTC)
I think the SX/16 would take the prize as the slowest 486, if you can find one - and it would have a hard fought battle against the SLC/33 I had to endure in an old monochrome laptop I picked up almost literally for buttons at college. A 486 only in name, that one, I think. As noted much further below, just doing something like loading a VGA resolution JPG off disk and downconverting it for display was hilariously slow - like 10+ seconds, and larger ones could take a minute or more. A complex webpage would make it break down in tears... so, that or an equivalent performance chip is probably your lower limit for even vaguely tolerable graphical browsing, because with a 33k modem it'd be decoding and displaying pictures at about the same speed as they came down the line with 100% CPU utilisation. Still, it ran Word and Excel perfectly well, which is what I "bought" (er, bartered :D) it for... can't really ask more than that. It does work as a baseline however - take that (probably cacheless and FPU-less) 33mhz part, or the certainly FPU-less 16/25mhz one, and note that a DX2/66 is at least 4x faster than them in real world use... 193.63.174.11 (talk) 11:29, 27 April 2011 (UTC)

Curio: why is the "Intel Overdrive" specified as being a DX2? There were numerous Intel Overdrives, not all of them were DX2s.

It shouldn't be. The "Overdrive" 486 chips were available in both DX/2 and DX/4 form.

WHat socket did the 486 use? --Tarquin 15:38 Apr 28, 2003 (UTC)

A few different ones, Tarquin. Cris Hare is the best source for this stuff. Tannin

For a quick rundown - some socket 2, a lot of socket 3 (the first ZIF type IIRC?), and quite a lot were either soldered or in 286/386 style static sockets where you needed special tools (or at least a couple of screwdrivers and good co-ordination) to safely pull them out. My 1994 vintage board was sock 3 (and 72 pin ram with PCI, thanks) 193.63.174.210 (talk) 14:57, 24 October 2008 (UTC)

Regarding the new image of two SX 486 chips, could it be moved to the Intel 80486SX article? It is just that SX chips were not 'normal' 486s, and putting them in this article is analogious to placing a picture of a Celeron at the front of an article about the Pentium 4. Also (a tip), images this wide should be put below the first piece of text because people on lower resolution screens (such as the 800x600 I'm on at the moment) will see an image taking up all the screenspace above the first text of the article. Crusadeonilliteracy 16:08, 26 Nov 2003 (UTC)

Uh, they're "486"es aren't they? I've seen a lot of them in use, as low-end "486" machines as they were budget parts which still gave good office performance and were compatible with 486-accelerated code. Think of it less as "Pentium 4" and more whatever chip-family codename the related P4s and Celerons belonged to. 193.63.174.11 (talk) 11:29, 27 April 2011 (UTC)

I thought Intel always referred to this chip as the i486, never the 80486. I could be wrong. --Furrykef 06:39, 24 May 2004 (UTC)

I believe title "80486" is incorrect. 386 was officially "80386", but in 486 Intel dropped 80-prefix, and in Pentium numbers altogether. See pictures of prosessors, name "Intel i486" is printed on them, nothing says "80486". -- Mikko Paananen 14:28, 12 Jun 2004 (UTC)
I concur. 80486 was still a common layman's designation. The article should be renamed to Intel i486 and a redirect added from 80486. -- Jherico 00:55, 26 May 2005 (UTC)
You shouldn't confuse the technical identifier i.e., the chip product ID 80486 with the shorter (simpler, nicer) names used for marketing or by mere users like i486, 486. Also many terms used nowadays (like x86, it was either IA-32, i386 or 80x86) were never used at that time. --195.62.99.203 04:00, 15 Jun 2005 (UTC)
Here's a little bit of proof "http://logout.sh.cvut.cz/~astro/pictures/trade/a80486sx25.jpg". In case that link doesn't work, just use Google's image search for "80486". You'll see that at least some models of Intel (but also AMD) have 80486 printed on it. --195.62.99.203 04:06, 15 Jun 2005 (UTC)
By the way, you can clearly read "80486" on the photo at the top of the article. --195.62.99.203 04:53, 19 Jun 2005 (UTC)

FSB speed in GHz??[edit]

Anyone else notice that FSB is given in GHz and seems to be irreversible? If not, please edit and inform! Lahs0n (talk) 11:56, 2 August 2008 (UTC)

Clock Multiplier[edit]

I think I don't quite understand that clock multiplier thing. Could you, please, explain it using eg. 486 DX4 processor @ 100 MHz? does it mean that processor runs @ 100 MHz, while system bus is clocked @ 33MHz, ie. three times less?

--89.102.25.151 17:30, 5 June 2006 (UTC)

Yes, a 100 MHz Intel DX4 uses a 3x multi on a 33 MHz bus. It's a clock tripled CPU. A DX2 50 runs on a 25 MHz bus. A AMD 5x86 133MHz is actually a 486 with a 4X multiplier on a 33 MHz bus (lots of people put the bus at 40 MHz and get a 160 MHz 486!) --Swaaye 18:32, 5 June 2006 (UTC)
*sticks hand up* oh yeah, I was there, I was that soldier :D ... 160mhz 5x86, 24mb Fast-Page RAM, a 2mb PCI graphics card, ESS688 audiodrive, 1gb hard disc... the world was my lamb pasanda kebab. And with "recent" (up to about 18 months ago) experience of using an old pentium laptop in the modern world, I'd like to think it'd still be partway practical with the right software. We just don't get that kind of massive overclocking any more, I don't know why... +20% cpu speed and it still ran acceptably cool and absolutely stable. Just wish I had some kind of suitable ultra-cooling rig at the time (for the enhanced conductivity...) so I could make it work at my motherboard's 50mhz speed :D 200mhz enhanced high-cache 486 core, anyone? 193.63.174.210 (talk) —Preceding undated comment was added at 14:10, 24 October 2008 (UTC).
Cuh, I go and say that, and then what? I get news that a friend has just overclocked a Core Duo he bought around the time I wrote that comment ... by 21%. Sod. 2.3Ghz to 2.8Ghz, and running stable. Seems you CAN still do it. Well done that chap :p 193.63.174.10 (talk) 11:57, 18 May 2010 (UTC)

DX2/DX3/DX4[edit]

Regarding the comment "DX4 runs at triple the clock speed because DX3 (supposed to run at 2.5x the clock speed) never hit the market".. This is conjecture, and is absolutely false.

The reason the DX4 was called such was due to the lawsuit in the mid 90s that decided that Intel could not use a number as a trademarked name. They could not call their chips "486" anymore; this is why they called it "DX4". "DX" for their naming system that had been in use, and "4" as a reference to the 486.

You'll note that 486 DX4 chips are not called "486" at all anywhere in Intel literature or on the packaging. They are simply called "Intel DX4". The 4x/3x multiplier "confusion" is because of Intel's use of "DX2" for their clock-doubled processors. Although Intel's intent at the time with "DX2" was to use it as a suffix that denoted a doubled clock speed, the use of "DX4" cannot be compared as such.

Intel never planned to release a "DX3", and there were no 2.5x 486-based chips from Intel.

Is it not just an early stab at their now common strategy of naming things such as to disguise their true performance level (e.g. the "Pentium M 740" in my laptop, that's actually a third-gen 1.73ghz example), or appear faster than they are? In this case suggesting to anyone who doesn't read the small print that's it's 2x faster than an equivalent DX2. Intel would probably have argued that a DX4 is more than 3x faster than an equivalent DX("1") because of the larger and more sophisticated cache, but it's still misleading. 193.63.174.210 (talk) 14:45, 24 October 2008 (UTC)


What does "dx" stand for?

My own hunch is that DX, SX etc don't stand for anything. They're like the SL, SR, RX, LX letters used to name cars. The X looks futuristic and distinctive, and it sounds snappy when you say it. D has one syllable. Neither DX nor SX are a word. Other combinations of letters might have been undesirable - EX would have people calling it the "486-ecks", FX looks rude, WX is a mouthful to say, AX is "axe", and so on. Lupine Proletariat 15:09, 18 May 2006 (UTC)
Both these theories sound possible. However neither is sourced... Nil Einne 15:17, 3 July 2006 (UTC)
This is unsourced as well, but might they not have roots both in "simplex" and "duplex" transmission modes (as the DX has twice the bus width of the SX ... was there mention on the main page that the SX was limited to 16mb RAM?), and in that the actual bus width for the SX is a Single word, and the DX can transmit a Double word with each clock cycle? With the X coming both as a common ending letter for tightly reduced american acronyms as a whole (eg Px for pharmacies), but also the TX/RX pairing of Transmit and Receive in radio & electronics - and the next letter of the alphabet from W, because SW & DW may have looked wierd. Also bear in mind that the SX & DX originally came from the 386 days.... and the first 386 SXs slotted into slightly modified AT-286 boards where the "single" 16 bit channel would have been the standard, and 32 bit would be double the norm. There's all manner of sensible, but almost unprovable (without having access to their committee meeting minutes) reasons for choosing these. I don't think there was ever a clock doubled 386 (they maxed at 40mhz, probably on a slightly modified *486* board, as far as I was aware), but it probably would have got the DX2 designation itself - it makes sense. Not that the things that make the most sense usually get chosen, but sometimes they do. ((And... if my dim recollections of Railroad Tycoon are correct... a DX Goods loco was one with twin cylinders for improved horsepower...?)) —Preceding unsigned comment added by 193.63.174.210 (talk) 14:19, 24 October 2008 (UTC)
(Extra to my own post there --- this deviates a bit from the 386 reasoning, as all their FPUs were external... but, the SXs also had a Single processor onboard, or at least claimed to, whereas the DXs had two... the main core (central processor), and the FPU (maths processor)... the original dual-core intel chips, in fact - more reasoning for the choices? :-) 193.63.174.210 (talk) 14:42, 24 October 2008 (UTC)

More on DX3/DX4

I was the original commenter objecting to idea that the "DX3" was reserved for a "2.5x multiplier". Although there are references to this fact (which I believe is false) on the web, I recall specifically by reading news articles at the time that the name "DX4" was chosen because of the "DX" tag already being used on chips and "4" being added to it to keep the name in line with their 4th generation (486) chips, after lawsuits decided that Intel could not copyright a number, as stated in the article. Also note that the clock-tripled DX4 is simply called the "Intel DX4", which leaves more credence to the fact of not being able / wanting to use the "486" name anymore. Additionally, the "4" in DX4 does make some sense as a name, whereas the "3" meaning 2.5x makes none. A good explanation of this (and agreement with this thesis) can be seen at Karbosguide, a PC reference site which pre-dates Wikipedia and the other articles claiming "DX3=2.5x" nomenclature on the web. To wit: (Note, the site's author, Michael B. Karbo, is not a native english speaker, and the site is translated by staff - hence the grammar errors. This does not detract from Mr. Karbo's extensive PC experience.)

"Contrary to what you might think, the DX4 were not named for a quadrupling. They were named this way because of the registry of Intel's 80486 and 80586 names. The DX4 name is separated from that context, so it could be patented. If DX3 referred to a tripling, this would not work. The same type of problem caused the next generation chip to be named Pentium, rather than 80586." Tcp100 06:12, 3 August 2006 (UTC)

WP bit in cr0[edit]

I think it ought to be mentioned that the 386 had a design flaw in its paging system. It did not enforce write protection at CPL 0. This makes it very inconvenient to implement copy-on-write in 386-compatible operating systems, because the kernel could not simply access user memory directly expecting an exception if a page were copy-on-write. Modern Windows NT in particular wouldn't be able to handle this.

The 486 solves this by enforcing write protection at CPL 0. However, to remain compatible with the 386, it has to be disabled by default. Bit 16 of cr0 was defined as the WP bit, enabling this feature.

The other big feature of the 486 from a systems programming perspective was the addition of important atomic primitives: cmpxchg (really cemented by Pentium's cmpxchg8b), xadd, bt, bts, btr. cmpxchg in particular, because all atomic primitives can be implemented based on it. (These need "lock" to be atomic; don't use them in non-atomic situations because they're slow.)

bswap turned out to be nice for the SHA hash functions but clearly wasn't an important improvement.

-- Myria 10:09, 9 August 2006 (UTC)

486SL -- DX not SX[edit]

From memory, I seem to remember that the 486SL was a variant of the 486DX rather than the 486SX as noted in this article. We had a Zenith Data Systems 486-SL25 notebook (4MB RAM, 25MHz CPU) back in 1993-4. According to http://everything2.org/?node=486sl, it is a variant of the -DX chip as well. Is this article a suitable source, or do we need to find a better one?

--193.120.178.201 15:56, 10 August 2006 (UTC)

What about the SLC? I'd like to know what the difference is there. I had a laptop with one of these in (obviously a very low power chip, as in energy consumption, as it didn't even have a heatsink (and good thing too, as the battery was useless, just six nicad C-cells soldered together in a plastic shrinkwrap)... which made reading the code easy when the lid was lifted), but I don't see it listed here. If it was DX based, I'll eat some suitable piece of clothing - it was ok for general low-end windows stuff, including typing stuff in Word 6 (where it could just-about keep up, so long as you weren't scrolling past clipart), but anything more than that absolutely slayed it. At 25mhz it took about a minute to decode a VGA-rez JPG photo that our DX2/66 desktop could manage in under 10 seconds. An achingly under-capable bit of silicon. 193.63.174.210 (talk) 14:32, 24 October 2008 (UTC)

Dates?[edit]

When was the 486 introduced?

-- It was introduced in 1989, became obsolete for new PCs in the mid 90s. It is still used in embedded systems, but Intel will stop making it in 2007.

You're right. I have an old 486SX-25 dated back to 1989.

AppleMacReporter 15:57, 29 December 2006 (UTC)

I have recently heard about a multiprocessor-capable 486, but the article does not mention it. Can someone confirm its existence or nonexistence? - Anonymous 19:14, 17 May 2007 (UTC)

Look at Sequent Computer Systems; Alecv 12:07, 18 May 2007 (UTC)
Thanks. So, the CPU didn't itself provide support for SMP operation, but engineers at Sequent had implemented it "around" the stock processor with their own supporting hardware, if I understand correctly? Also, I believe those Sequent's SMP systems should be (briefly) mentioned here, in the 486 article. - Anonymous 01:04, 19 May 2007 (UTC)
OTOH, when I think about it, in times of sequential, in-order execution CPUs, not much support for SMP should be needed on the CPU side besides simple, bus-locking memory access primitives like XCHG (and in 486 even more powerful LOCK XADD and LOCK CMPXCHG). - Anonymous 01:10, 19 May 2007 (UTC)
I wouldn't be surprised at seeing an old multi-CPU 486 server somewhere. Remember it was Intel's flagship CPU for at least 4 years. If you wanted a high-end, i386 compatible machine, that's what you would have been using. 193.63.174.11 (talk) 11:29, 27 April 2011 (UTC)

Differences between the 386 and 486[edit]

Can anyone put a full command set difference list ? Sometimes it called 486+. AFIAR, 486 command set contains BSWAP command to translate a big endian "network order" data. Alecv 20:28, 23 May 2007 (UTC)

Correct. It also added some instructions very useful for synchronizing concurrent computation (f.e. SMP) like XADD and CMPXCHG. Later generations of 486 also added the CPUID instruction. There were some other additions, but not many, and I don't remember them all. I'll try to find out the full list and add it here. - Anonymous 14:47, 30 May 2007 (UTC)
Alright, seems my memory is better than I suspected :). I forgot only three instructions, all of them for low-level operating system use. That would be INVD, INVLPG, WBINVD. Hope the list helps you. - Anonymous 15:53, 30 May 2007 (UTC)

'Obsolescence' section (and bus speeds thing)[edit]

I have some problems with this ... I think it's been written from the personal standpoint of someone who had a fairly old 486 that was killed by the factors mentioned in this and has expanded it to "all" machines. I'd like to offer a competing one which comes from the other end of the spectrum, as this chip actually lasted quite a long time in the marketplace and saw several shifts in the underlying architecture - 30 to 72pin SIMMs (an essential change for anyone wanting a serious DX board, as the chips had to otherwise be installed in matching QUADS unless clever motherboard chipsets were used), FPM to EDO, common integration of reasonably quick IDE drive, floppy & port controllers, shifts from E/ISA thru VESA/MCI to PCI, etc etc etc.

I make no apologies for length, there's a bit of exposition to be done on this and we've plenty of space.

486s were not killed by Windows 95 and its high memory requirements. Let me spin you a tale of a reasonably mid-range DX2/66 bought in early 1994 (for a HUGE amount of cash compared to a modern PC, but no-where near as much as was being asked for the fledgling Pentium-60, 66 & 75 systems), by no means an end-of-the-line machine... 66s thru 100s were still on sale as new PCs for a couple of years afterwards, even after the P60~75 were killed off.

CPU was a typical bread-and-butter AMD job, perfectly respectable by the standards of the time, and for quite a while after purchase capable of handling anything we could throw at it with rarely a whiff of slow-down. Even Quake did quite well, if kept at low-rez VGA that was then par for the gaming course. Living in an easily upgradable ZIF socket that would support everything up to the (unfortunately mythical) DX4/120, and later a 5x86/133 (...160).

It had 3 built in PCI slots, VESA not yet gone the way of the dinosaur but rapidly and obviously heading for it, boards and cards supporting it mainly being around to offer continued compatibility for anyone having to replace one or the other, and for budget systems, rather than "for serious"; installed, a 2mb PCI SVGA card that could do SXGA (in 256 colours) or SVGA in true-colour, which was plenty enough colour and rez for someone moving up from an Atari ST; and a CD-quality sound card in one of the ISA slots.

It had built-in IDE that would do PIO mode 4 and some limited DMA (ATA-33 not yet standardised) and handle discs up to 4Gb; attached to which was a 540mb Seagate drive that I still have, and still works well and as fast as you'd expect, and a double-speed ATAPI CDROM drive that still made a decent stab at serving up Final Fantasy 7 when it survived into a later machine.

It had four 72-pin SIMM slots, and the unusual ability to function with only a single module installed (presumably at a performance hit as the memory bus was reduced to that of an SX), and was sold with 8mb onboard in two 4mb sticks of fast-page. The manual provided with it suggested it could support up to 64mb by using 16mb modules, which I have tested... can't remember if I ever tried 32mb ones in it though. Also between the RAM and the CPU was 128kb of SRAM cache, good for the entire 64mb spread.

It quite happily endured the upgrade to Windows 95, with a bigger hard disc and a memory improvement soon after the installation to smooth things along (a "whole new PC" would still have been a similar cost to a decent second-hand hatchback, so although these upgrades were still costly, they were "no contest" compared to the alternative). It probably would have made it to Win 98SE, which it would have been more than capable of supporting, if not instead killed by GAMES. We found that some titles (e.g. C&C Red Alert) were a bit much for the 66mhz, so moved on to an interim 133mhz Evergreen overdrive, with all other components - bar a more compatible but otherwise near-identical 486 motherboard - remaining the same...

...then my brother bought Final Fantasy 7 having looked at the box and seen "133mhz" (and a 3D card, or 166 without) as the minimum requirement. 133mhz Pentium that is. It played, but almost unusably slowly. The choice was to get a 3D card, or upgrade the CPU & motherboard. We did the latter, as it would better futureproof the machine against other non-game applications as well, and so the 486 was finally consigned to history so far as our family was concerned. However I took it for my own and made a frankenstein's monster out of all manner of discarded parts from home, school, neighbours' 286s, etc, and happily used it for schoolwork until it was time to go to University (in late 2000). Whereupon I inherited the Cyrix (now with 3D card, too) - and a friend of mine took an ultimate-level spare-parts modernisation of the 5x86 off my hands to do wordprocessing with.

So unless you were using a 486 of type that ALREADY looked out of date by the time we took ownership of that beast (e.g. the SX-25s that school were using, which DID only have ISA/VLB and 30-pin SIMMs) because they were still literally using previous-generation, high-end-386 technology, Windows 95 wasn't that bad news. There were Pentiums that suffered as badly with it, or worse, such as a P100 laptop of my mother's that was horrendously slow... because it had a hard-pressed 8mb and was very difficult and expensive to upgrade (and awful integrated graphics that couldn't be upgraded). Just as it's nonsense that you need an extra powerful Core 2 Duo CPU with multi gigabytes of memory to run XP & office and "multitask" effectively. A sub-Ghz Pentium III with a half decent motherboard, a non-embarrassing flavour of PCI graphics and 512mb does the job admirably (Celerons, the "SX" chips of the modern world, less so), despite a casual observer thinking such a rig would be well over the hill - until you ask it to run an up-to-date game of course.

(I'll reserve extensive comment on ME & Vista, they're exceptions to the general rule... get as powerful a PC as you can, and pray)

I could probably get 2000 running on the thing by maxing the RAM... pop in a PCI USB card, an ATA-133 controller with a fast hard disc or even an SSD, a DVDRW and the last of the PCI-classic graphics cards, and it'll probably not seem too different overall from the Celeron I'm typing from that's crippled by the actual slowness XP has and a rubbish hard disc :D ... so long as I don't get ambitious for anything more than Office and Email of course. Being a home cinema is probably beyond it.

Feeling tempted now actually. That's within my grasp using the next pay cheque and eBay ;) Been feeling like I should do something to rival those guys who got XP "working" within 20mb and 8mhz (took over 24 hours to boot...)

Oh yeah - and the "bus speeds" bit of the title... eh, that kind of got integrated into all of the above. It would have been possible probably to do 50 and 66mhz bus speeds, and had the Pentium not come along and made it all a bit moot by launching with 50, 60 and 66mhz speeds bus speeds, it might have been tried. VLB and probably a lack of suitably stable northbridge technology put the dampers on it originally, but as noted we did make the jump to PCI before too long (and the board I donat... sorry, SOLD :D to my pal in 2000 did support 50mhz WITH PCI), it was merely the pointlessness of it, and a lack of any suitable chips because of development switching to the P5 generation (which would have happened probably around 92-93, still during the VLB period, as development and early adoption of such things tended to be some years ahead of widespread desktop use at the time, rather than now where they're rushed to market ASAP to catch a march on the opposition), that stifled it. 50mhz itself maybe wouldn't have caught on too well itself for the same reason DX2/66s were actually more popular than DX4/75s, and the DX4/100s weren't quadruplers... the PCI and some other buses had to be clocked down to 25mhz in order to cope with it stably (a smarter move however than having it ALL run at 50, as for VLB), which would have caused quite a performance hit for graphics and disc transfers. Not sure if 40 had to go down to 20... it might have been switchable, or fractionally divided to 30. Leaping right to 60 or 66mhz bus speed would probably have been a more popular move. (Hmm... a quad-clock, 66mhz 486 chip - 266mhz?)

.... which I guess isn't really a strong argument against the Vesa bus keeping 486 bus, and loosely therefore the core cpu speeds down. Just the way that bit of the article looked suggested that all/most 486 systems were lumped with this rather ill-conceived piecemeal replacement of the ISA bus, and that's why they died out --- when actually its shortcomings (and that of its MCA rival) were identified fairly quickly and it died a swift death in the face of the lower-peak-speed but more versatile, reliable, simple & compact (a big plus point for motherboard and add-in card designers) and user friendly PCI, and their fade-out was ultimately more due to the increasing demand of first-person / general 3D (Quake, NFS2, Flight Sim) and complex 2D (C&C, GTA) games and other graphics/multimedia applications that couldn't entirely be alleviated by add-in cards (which helped for 3D games and MPEG video, but little else). After all PCIs speed didn't go up (except in industrial/professional level designs) over the entire course of its existence, from the early-mid 90s to its replacement by PCI-x, so that wasn't actually too much of a limiting factor (the real problem with VLB was not slowness, but its direct coupling to the CPU's front-side bus, which was too variable in speed and too weak of signal); even AGP was mainly only useful for games, and as a way for integrated cards to steal system RAM.

There's a point in there somewhere. Bite me, it's late on Friday afternoon and I should be on my way home rather than looking for answers for some work problem online and getting distracted. Just be glad I'm merely making a point for now, and if I go to edit the article after maybe looking some stuff up, I should be more compos mentis than compost mental :) 193.63.174.210 (talk) 16:47, 24 October 2008 (UTC)

original: 193.63.174.210 (talk) 16:27, 24 October 2008 (UTC)

OK, I have rewritten this paragraph. - Yuhong (talk) 00:47, 28 November 2009 (UTC)
Thanks ... looks like a similar thing is in effect on the Am586 page though. I'll have a go at that myself. And actually... sorry about the sprawl. It's even making my eyes hurt. Looks like it was from a period when I was lone working in a rather dank, cramped office and living off coffee. 193.63.174.10 (talk) 12:11, 18 May 2010 (UTC)

486 on Hubble[edit]

http://www.popularmechanics.com/science/air_space/4288705.html http://www.theregister.co.uk/2008/10/16/hubble_back_up/ http://sm3a.gsfc.nasa.gov/downloads/sm3a_fact_sheets/advanced_computer.pdf

These mention that a 486 chip is on board Hubble Space Telescope. (The last also mentions there was a 386 earlier and some other architecture I'm not familiar with). Hoemaco (talk) 19:49, 16 October 2009 (UTC)


Quake Pentium-only code?[edit]

Um, Quake II maybe? Which not only gasped for air on anything below a P166 MMX, but really only began to shine with a 3D card (ANY 3D card... even a horrid S3 ViRGE) ... but as I've already noted, Quake itself runs just dandy on a DX2/66 if you resign yourself to ModeX VGA (320x200/240)... which was the default gaming resolution of the day anyhow, because high-rez VESA standards were still in their infancy. There may have been some code in there which auto-detected a Pentium and used some more efficient routines for improved framerates at "high" (i.e. 640x480 VGA) resolution (I remember an actually sluggish combat flight sim explicitly featuring such a user-switchable option, and the editor of PC Gamer magazine mentioning people worshipping at the altar of his top-of-the-line P200 when the full version of Quake was sent for review... but note that it's already 3x faster anyway!)... but it didn't necessarily mean the game either could not be played, or ran unplayably slow, on the at-the-time very considerable installed userbase of lower performance machines. 193.63.174.11 (talk) 11:29, 27 April 2011 (UTC)

Gaming and FPU[edit]

I think that the importance of a (powerful) FPU for 3D gaming was grossly overestimated during the lifetime of the 486 - probably due to marketing efforts of Intel.

In 1997 the german computer magazine ct tested/measured (using the performance counter registers of the Pentium) the time an application spent in FPU code. They tested several apps including some 3D games of that time. The max usage was in povray (chess.pov) with around 24% CPU time. The games (POD, Descent 2, Afterlife) spent 6.7, 1.7 and 1.5% in FPU routines. This means that even a very drastical improvement of the FPU wouldn't have resulted in a high speedup of any game.

They concluded (ct 9/1997, p.256): "Bei Spielen unter Windows 95 liegen die Gleitkommaanteile besonders niedrig, nur POD bildet die Ausnahme. Spieleprogrammierer versuchen natürlich, jeden FPU-Aufruf zu vermeiden und nehmen alle notwendigen Berechnungen viel lieber in Integer-Arithmetik vor. Große Performance-Vorteile der Pentium-Prozessoren bei etlichen populären Spielen sind also eher Folge einer speziell auf die Architektur dieser CPUs optimierten Programmierweise."

(My) Translation In games under Win95 the floating point part is remarkably low, just POD is an exception. Game programmers of course try to avoid any FPU call and prefer using integer arithmetics. Big performance advantages of Pentium CPUs in many popular games are therefore rather due to optimizations for this CPU architecture.

So, I think, the reference to FPU performance in the gaming section should be removed.

The cited article is available for download - for a fee ;) http://www.heise.de/artikel-archiv/ct/1997/9/256_kiosk

79.204.143.250 (talk) 10:47, 24 May 2011 (UTC)

That's correct. Many games in that era often used FIXED POINT integer math back in that era. It was much, much faster than floating point, and the accuracy of floating point math wasn't necessary for the typical game. It's still faster to this day, but the difference isn't as big as it once, and neither is really used heavily anymore, thanks to GPU-oriented design. Renegrade (talk) 01:59, 2 October 2011 (UTC)

How can be 486DX2 if it don't have cache at all?[edit]

So 486 processor runs twice faster than bus 50 MHz, so gains speed 100 MHz. RAM speed also 50 MHz. So that point in DX2 (double bus clock) if 16KB L1 cache apears only with Pentium microprocessor?
Or there was faster memory at that time (like 80 MHz, 100 MHz)?
"Since the introduction of its first microprocessor in 1971, Intel has steadily increased

the number of general purpose registers in its microprocessors. The 4004, Intel’s first microprocessor, had no general purpose registers per se, although a complete 4-chip computer, consisting of the 4001, 4002, 4003, and 4004 chips, included 16 RAM locations that were used as registers. Its successors, the 8008, 8080, and 8085, incorporated six general purpose registers, as well as an accumulator, within the processor chip itself. The 8086 microprocessor has eight general purpose registers, as do the 80286, 80386, and 80486 microprocessors. The Pentium microprocessor also has 8 internal general purpose registers, but they are 32 bits wide, as opposed to the 16 bits of its predecessors. Intel’s most recent microprocessor (as of this writing), the Itanium microprocessor, has 128 general purpose integer registers and an additional 128 general purpose floating point registers. Intel first introduced cache memory into its Pentium microprocessor, starting with 16K of cache memory. It soon increased this to 32K, and further increased the amount in later processors. The Itanium microprocessor contains three levels of cache with over 4 MB of cache memory."

Chip vs. systems[edit]

Shouldn't all that stuff about exciting new misfeatures of personal computers be someowhere *else*, not in this article about a particular microprocessor? Or, at most, a small pointer placed saying something like "Mass market personal computer design evolved new features to take advantage of the 486 processor."--Wtshymanski (talk) 13:39, 23 August 2012 (UTC)

intel Quark[edit]

The upcoming Quark CPU is more or less a 486 CPU, as pointed out in the german c't mag in http://www.heise.de/ct/artikel/Prozessorgefluester-1979287.html and if one compares the "i486 Hardware Reference Manual" with the "intel® Quark SoC X1000 Core Hardware Reference Manual" (just google them), they are right. Might be something to add to this article. — Preceding unsigned comment added by 146.52.29.42 (talk) 15:39, 21 October 2013 (UTC)

040-based Macintosh[edit]

"For a time Apple attempted to compete with Intel's clock doubling 486DX2 systems by publicising misleading doubled clock rates for its '040-based Macintosh Performa systems, despite the lack of any clock doubling." - here is complete analyse of 040 clocks: http://milan.kovac.cc/atari/040CLOCK.TXT - MC68040 does have PCLK (Processor CLocK) and BCLK (Bus CLocK), first usually two times higher than BCLK but it DOES NOT affect performance of 040 so yes, Apple did mislead customers citing clock that does not affect performance of 040 (but 68040 DOES HAVE two clocks (not in sense of intel "clock doubling")) — Preceding unsigned comment added by Calimero (talkcontribs) 09:12, 13 December 2013 (UTC)

List of Intel i486 Microprocessor[edit]

Should we create a separate page for list of Intel i486 microprocessors? Like their newer brother do have list of microprocessor of their own. Can the new list be included the sSpec numbers list as well? After all, some of the i486 microprocessor did includes these sSpec numbers.

Also, I like the pictures included in these list here as well. We can use the information from Wikimedia Commons page of Category:Intel i486 listed there. Rjluna2 (talk) 23:23, 12 May 2014 (UTC)