You could always try a PCI IDE/SATA controller and see if the performance changes. Either way, that Dell seems great. Congratulations!
Btw, is a lot of stuff in that Dell propertary? Like PSU and connectors?

Thanks! The PSU is. It has an AT-style connector with 3 blue and 3 black wires. The connector doesn't conform to any of my other AT boards or boards with supplemental connectors like it. It also has a proprietary front end connector (power on, reset button, leds etc). It's also got this huge passive heatsink for the processor. Even with the case open, it's whisper quiet. Everything else is standard. That's a good idea on the controller. I really want to see if I can isolate the issue and report back. I'm looking at doing a couple of game benchmarks (quake in software etc) and when I'm done I'm gonna pry off the PcChips heatsink (prominently labeled Intel i440EX/LX) and see if the print on the ceramic jives.

It's kind of neat running games on a top-end OEM gaming machine from the time. This thing probably cost someone $3k or more new and I found it by accident making a craigslist buy, I went there for another system just to pick up a horizontal ATX case for an HTPC build.

I can't resist expressing my surprize for your results!

These two CPU's are far more apart that the small difference your numbers indicate. Maybe it's the type of benchmarks that don't show this clearly? I don't know, but there's something wrong here. I mean, Sandra even says the celeron is faster...

In case you want to play a little more (since you're interested in numbers and benchmarks), I'd run again the same and a couple of other benchmarks with the CPU's on the same pc.

A specific disk test might show something interesting... buffered speed, seq speed, and CPU utilization

Ok, I've popped the Celeron into the Dell system and run the same benchmarks.

Dell XPS R400 with Celeron 400 / 66MHz FSB / 128k L2
Scisoft Sandra 2001:
CPU: 1051 / 525 (-2.2% / -1.7%)
Multimedia: 1418 / 569 (-2.8% / -1.5%)
Memory: 195 / 228 (-19% / -38%)

PCMark2002:
CPU / Memory / HDD
1015 / 804 / 434
3% lower / 22% lower / 6% lower - variance with Pentium II 400

Based on these scores, it looks like the board doesn't feature independent bus speeds for CPU and Memory. At any rate, this is more in line with what one would expect for the two processors in comparison. Looking at the HD scores, keep in mind it is the same HD swapped to the different systems. PC Chips must be engaged in shenanigans with the IDE controller and/or bus. Setting memory aside, however, the two processors are very comparable. It would have been worth the cost savings back in 1998 to buy a Celeron instead of a Pentium II with these performance numbers.

I'm working on a Quake benchmark and I'm adding 3dMark2001 to the original post above.

konc wrote:

These two CPU's are far more apart that the small difference your numbers indicate.

I really wonder where you got that from.
400MHz Celeron has always been pretty much as fast as 400MHz PII.

I think this only applied, when a Celeron was overclocked and ran at 100 MHz bus speed.

Hey OP I have a Dell Dimension 4100 with the same proprietary pinout. The half AT plug is a non-stadard varient of a auxillary power connector often found on Pentium 3 and Pentium 4 computers. You can find adapters to use standard power supplies with the non-standard Dell motherboards for under $5 online 😀

sunaiac wrote:

konc wrote:

These two CPU's are far more apart that the small difference your numbers indicate.

I really wonder where you got that from.
400MHz Celeron has always been pretty much as fast as 400MHz PII.

Maybe I didn't explain it properly, I should have said that a system with a PII is much faster that the same system with a same speed Celeron.

Which is logical if you think it in many ways. OK a benchmark relying on CPU MHz's doing for example math calculations, will show similar results.
But how can a system that has a Celeron with a slower FSB, smaller (half) cache and running the memory at lower speed (because of CPU FSB) stand close to the PII when its only advantage is running the cache at full clock speed?

Results from real scenarios, such as the last benchmark added by the OP, are in my opinion indicative of the difference of the two CPU's when used on the same system. I'm not discussing whether it was a better decision or not to buy a Celeron instead of a PII back then, that's another thing, although in my opinion Mendocino Celerons were a smart buy. Just that the actual difference of these two CPU's in real scenarios is much more than what a CPU benchmark shows. Hope this clarifies everything 😀

I understood correctly, but I disagree.
I'm pretty sure putting the celeron on the 440BX at 66MHz bus speed would give pretty much the same 3Dmark score as the P2.

Ok, I did a 3DMark2001 run with the Celeron on the BX board. After booting into XP, I verified through Scisoft Sandra that the Celeron and the memory was running at 66MHz bus speed.

Dell XPS (440BX), Pentium II 400, 100MHz FSB
3DMark2001 SE, 640x480x16: 532

Dell XPS (440BX), Celeron 400, 66MHz FSB
3DMark2001 SE, 640x480x16: 471

The Celeron with the same board and lower memory clock is 13% slower than the Pentium II. It's important to keep in mind that this is a game environment test. The synthetic benchmarks for arithmetic from Sandra indicate the processors themselves are nearly equal. Intel's decision to allow the L2 cache on the Celeron to run at full speed is a great equalizer. From a practical standpoint, however, the package of a Celeron with 66MHz bus is slower than the Pentium II.

PC Chips (440EX/LX), Celeron 400, 66MHz FSB
3DMark2001 SE, 640x480x16: 438

What's further interesting is that the same processor and same speed memory on a different board is 7% slower than the BX.

I'm very surprised.
All my tests proved them equal in real games (It's burned in my memory, because I paid for the Pentium 2 at that time 😁)
Is 3Dmark 2001 testing things memory-BW heavy ?

Tests from that time confirm my memories : http://www.hardware.fr/articles/98-1/nouveau-celeron.html (see Quake 2 and redline racer at the end)
No game should be depending this much on the 66-100MHz bus difference.

ODwilly wrote:

Hey OP I have a Dell Dimension 4100 with the same proprietary pinout. The half AT plug is a non-stadard varient of a auxillary power connector often found on Pentium 3 and Pentium 4 computers. You can find adapters to use standard power supplies with the non-standard Dell motherboards for under $5 online 😀

They are only sporadically available, though. I got two last year and haven't seen any for sale since then.