I never knew the P2-450MHz was uncommon. I have one of those but it came from an IBM workstation machine, so I guess that just reinforces the point. I haven't actually ever seen a P3-450 but I ended up with a bunch of 500MHz versions. Out of those, 2 out of 10 had the 300MHz cache chips in them (suitable for 600MHz core).

I'd love to overclock a 300A someday, just because I've never done it. I never had a Voodoo either, because they were too expensive. So if the parts were handy, that's what I'd build but that's only because of my own personal angle.
I scorned the Celeron name back then, quite foolishly, and opted for a K6-3 system instead. Budget was a factor, but I really can't remember how much more, if any, it would have cost me to build a Celeron if I had taken that idea seriously. I was also uninformed/fearful of overclocking.
In retrospect, a good 440BX board with a 300A would have been awesome, and very upgradable later on with slockets as well.
I do wonder how honestly stable many of those 300A@450MHz builds actually were. Gamers didn't have the highest standards in the era of Windows 98. Instability was easily blamed on software, and I don't think objective stress testing was as common.

The 300A's were extremely stable. Consider this. The 300A wasn't really a 300 at all. It was kind of like what Nvidia did with Maxwell. It was a next generation test of sorts which later became the P3 Coppermine series. The original Celerons sucked ass, a lot. Intel had to add cache but how, while keeping cost down? They integrated it into the CPU die, which allowed it to run at the CPU speed and not some divider. Now that wasn't a P2 300 die they took... it was the 450 die. So it was a 450 die with 128kb integrated cache. It was extremely rare to find a 300A that wouldn't run stable at 450. By that point Intel had their stuff down pat with that fab process. I ran them in SMP systems which would greatly magnify any flaws, not in a gaming system but in a production system running NT 4.0. Do not think of the 300A at 450 as overclocked but instead think of the 300A at 300 as an underclocked 450. We will probably never see such a CPU again.

swaaye wrote:

Of course it makes little sense to mess with overlooking today. Other than to relive it.

At the same clock speed, 512K half speed vs 128K full speed really isn't a tangible difference.

Yeah, if you ditch the P2 450 and move up to a 450 or 500mhz Pentium III you've basically wiped out any advantage the 300A at 450mhz had over the P2 450 and a slow P-III can be had for just a few dollars so not worth playing around with overclocking anymore.

sliderider wrote:

swaaye wrote:

Of course it makes little sense to mess with overlooking today. Other than to relive it.

At the same clock speed, 512K half speed vs 128K full speed really isn't a tangible difference.

Yeah, if you ditch the P2 450 and move up to a 450 or 500mhz Pentium III you've basically wiped out any advantage the 300A at 450mhz had over the P2 450 and a slow P-III can be had for just a few dollars so not worth playing around with overclocking anymore.

Pentium III 450 mhz and 500mhz still had the 512 cache running at half speed,so they weren't faster then a p2 450mhz,only improvement were the sse instructions.

Nahkri wrote:

sliderider wrote:

swaaye wrote:

Of course it makes little sense to mess with overlooking today. Other than to relive it.

At the same clock speed, 512K half speed vs 128K full speed really isn't a tangible difference.

Yeah, if you ditch the P2 450 and move up to a 450 or 500mhz Pentium III you've basically wiped out any advantage the 300A at 450mhz had over the P2 450 and a slow P-III can be had for just a few dollars so not worth playing around with overclocking anymore.

Pentium III 450 mhz and 500mhz still had the 512 cache running at half speed,so they weren't faster then a p2 450mhz,only improvement were the sse instructions.

The Pentium III had more improvements than just the addition of SSE. 2 million transistors were added, execution units to improve pipeline performance, and an improved L1 cache controller. A 300A only just outperforms a P-II 450. The collective changes to the P-III were more than enough to make up for it.

No. The Katmai performs exactly the same as a Deschutes at the same clock-speed, unless SSE is used. The Coppermine was the true Pentium 3.

When it comes to a voodoo1: both cpu`s are extremely fast and completely equal (since no voodoo1 games support sse)

The voodoo1 will be out of breath way before the cpu does. It is maxxed out around a 333mhz p2, imo there isn't much to gain beyond that. It can even get in trouble around 450mhz because the cpu is in some cases too fast and some games can get graphical corruptions because of that... Most games run fine.

The absolute safe max for a cpu for the voodoo1 is a p2 400mhz imho.

I use it with a p1-233mmx. If a game needs a faster cpu it often needs more graphics power too so the game is much better off in my v2 system or higher anyway.

The best match for a p2-450 is a single voodoo2. But it is fast enough to enjoy v2 sli or a voodoo3 with higher resolutions.