sunaiac wrote:

Samir wrote:

boxpressed wrote:

I just purchased a large lot of 70+ 386/486/586 class CPUs, mostly Intel, but some Cyrix and AMD.

Are some Intel 486 SX/DX/DX2 models more rare/collectable than others?

The 486dx-50 (not dx2-50) was quite rare. This is because the clock ran at 50Mhz rather than being 25Mhz and then doubled internally. It was on par with the performance of a 486dx2-66.

That's actually an urban legend.
I'll post my numbers this evening.

DX2-66 should be around 10% faster because in addition to a faster internal clock, Intel also added some additional cache that the DX-50 doesn't have so it doesn't have to communicate with main memory as often, nullifying the DX-50's bus speed advantage. (At least I remember reading that someplace a while back).

Sorry I'm late 😀 :

MB-1433/50 UIV-A rev5, 256Kb, 32MB, Tekram DC680-T 4MB, ET4000/W32p

Cache is slower on the 50MHz FSB, and there's still 33% internal frequency difference, that's a lot.
I do not know of any cache difference between the two.

sunaiac wrote:

Sorry I'm late :) : […]
Show full quote

Sorry I'm late 😀 :

MB-1433/50 UIV-A rev5, 256Kb, 32MB, Tekram DC680-T 4MB, ET4000/W32p

Cache is slower on the 50MHz FSB, and there's still 33% internal frequency difference, that's a lot.
I do not know of any cache difference between the two.

Intel rated the DX-50 at 41 MIPS (less than half a MIP higher than a DX2-50) and the DX2-66 at 54 MIPS and that's a bigger difference than raw clock speed alone would seem to imply.

sliderider wrote:

Intel rated the DX-50 at 41 MIPS (less than half a MIP higher than a DX2-50) and the DX2-66 at 54 MIPS and that's a bigger difference than raw clock speed alone would seem to imply.

Mmm, 54/66 gives 0.8181 IPC for the DX2-66 and 41/50 gives 0.82 IPC for the DX50.
Why do you say it's a bigger difference than clock rate would imply ?

I'm curious: while synthetic benchmarks may indeed indicate no instruction-level benefit from a 50Mhz FSB, isn't that a bit of a crude measurement? Surely a more rounded, real world test would show clear benefits from components such as a cached controller, graphics card and RAM running faster in combination (assuming the system can be made to run stable)?