Anyone remember the 487DX? or 486 SX?

[DocClock aka MadClocker]

Like the P4, way back when, intel unleshed their latest and greatest...the 486DX, Well some math proffessor found a flaw in the 486 math co-processor, it was off after I think 6 significant digits, and when Intel found out, they implemented a "fix" by asking manufacturers to add an extra socket for their new math co-processor on upcoming boards, and in the mean time, Intel disabled the integar (math co-proc), and labled it 486 SX, and you had to buy the math chip to fit in the other socket, if you wanted higher math computations.
Here's the kicker...the math co-processor was actualy a 486DX! and when pluged in, it took over all functions, and the SX chip was esentialy turned off, however you still needed the original chip for the board to function, which by the way is why and when they came up with their slogan "Intel inside", and the original comercial showed an empty extra socket on a mobo.
And at the time, the 386 DX 33 would smoke the 486SX, because of the "working" math chip on the 386DX.
I don't think they will get away with that trick again, but it shows intels quick thinking to avoid a real marketing disaster, and I am sure intel will come out with an even better integar setup for the next gen P4.
So in closing Intel's prob with the P4 is nothing new.
That ends my history lesson for the younger clockers in our crowd. :-)

[Phil]

the integer unit does whole number sums, the maths co-processor was an fpu for floating point sums, just though I'd point that out as you wrote that they disabled the ineteger unit when it was the fpu (maths co-processor)
I think intels worse problem was the floating point unit on the pentium 60/66 processors that got some calculations wrong, something that put many academics and people using the chips for scientific calculations off intel

[Lancelot]

I remember the possibility to upgrade a 486SX by adding a 487 co-processor. However I still have a 486 rig that I only use as a 'typewriter' for email and stuff when my 6 year old son is gaming on my main rig. It has a real Vesa Local Bus mobo, woohoo! and just one CPU socket. When I noticed the mobo supports CPU's upto the 486DX2-66, I got one at a PC fair for $4(!) put it in the same socket as the 33SX was in, adjusted some jumpers and all of a sudden it became reasonably useable under Win95. So how come at first you needed a second socket, while now I got double speed, more cache and a co-processor?! Seems that Intel has always been upto their tricks with 'intermediate chips', like the first P3's (katmai) before the 'real' coppermine's came out, and now the first P4's (willamette) that will be followed up shortly with the more definate version...

[Phil]

at least the katmai's got a reasonable life span and coppermine chips worked in their motherboards, heck I run a coppermine in my 2 1/2 year old motherboard that I had with my p3 233, and the katmai wasn't a complete botch like the p4, the new version due out may as well be a new chip if it gets all the improvements it needs

[Newbie_Doo]

My recollection of the 586 launch related to floating point processing errors. If I recall correctly, the 586 (original Pentium) was derisively referred to in academic circles as the "585.99999986975" in honor of the defect.

:)

[Phil]

thats what I was getting at Newbie_Doo, I just found the magazine I had with the article in it, it used to come up with the answer 256 instead of 0 for some equations,
an equation that invokes it is

z=x-(x/y)*y
correct solution for x and y is z=0
when x=419835 and y=3145727

pentiums used to answer it with 256

(just checked it on my celeron 600, comes back at z=0)