Meatball wrote on 2022-02-22, 19:08:
Pentium III 850/100/256 Slot 1 MISB
I bought this with the intention of upgrading a Slot 1 800MHz P3, but I have since changed my mind and I will keep this sealed. I think what changed my mind was looking at a Quake III benchmark and noticing a P3 677/133 pulls ahead of the 850/100 and a 733 beats it soundly. Better to keep this 850 as a historical piece and switch to another system if I need 4fps more (compared to an 800) that badly…heh…
In general, it's not worth spending "significant" amounts of money, (and I peg that at probably more than $20 depending on how broke I'm feeling and whether it's right in front of me or not) to do a CPU upgrade that's less than another 50% of the one you're running... because 15-20% is the you can barely feel it level, and have to run benches to make sure you didn't get ripped off. Then 20-33% is you can tell after you use it a bit, 33-50 is a definite yes this is faster, but you're not really blown away.
But also the bus speed boost can be quite noticeable because of effect on entire system, everything gets the bump, not just the number crunching.
You'll notice the "classic" x86 speeds 16,20,25,33,40,50 are all kinda rounded off 25% increments... but later range planning might have had binning for lack of yield performance, and matching competitor model speeds, and marketing fussing about having a "gap" and round numbers selling better and all sorts of nonsense polluting the ideal of putting 'em far enough apart so you can at least tell.
As you get near the top of the range, or "fastest on X speed bus" etc, you start paying out the ass for the couple of notches nearest the top. So when it comes to range topping performance, I usually shoot for being in sight of it, rather than being at it, because it's almost double the price. For myself, sub 20% bumps don't usually feel worth opening the case for when I've got the CPU sitting there. Only time that happens really is "this CPU is a bit faster than that CPU, but the thing that needs A CPU of this type next doesn't need much performance, so it's a waste for it to have the slightly faster CPU, so I guess I'll bother swapping them over and giving it the slower one and the other system the faster one. " 🤣 .... then you've just seen by purchases I might go for something in the 33% bump area if it's dirt cheap. 1.6 CellyM to 1.8 PentM, where the Celly probably has a 15% celeron tax on the performance. 1.5 Willamette to 2.0 P4.. ... ... ... but if I really got a passion for X machine being as fast as possible, the possible would likely mean a limit of $50 worth of possible, unless it was a very modern box.
So your PIII there, I'd be inclined to see if it liked 124Mhz bus, the 800.. though if it's an older stepping a B stepping, not sure it will like 133, I think it has the ghz bug which shot intel in the foot when they tried to break 1Ghz ... then they fixed it.. and I think it's fixed for all CPUs of that stepping, which I'm not recalling if it's C or D now... anyway, some will clock fine past 1Ghz, some will seem to but have strange errors. (Differently strange than a CPU pushed to limit of it's clock speed envelope I mean)
Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.