Celeron FCPGA 533

UPDATE 4/28/00

I purchased a CeleMine 533A from PC NUT and received it this AM – readers are asking how the Celeron IIs compare to the original Celerons, and so I compared it to a C433 running at 542 MHz (6.5 x 83) and also to a PII 333 @ 515 MHz using an ABIT BX6-2 as the test platform.

A quick look at the back of the C533A reveals a very sparse landscape – compare this to the 600E below – talk about circumcision!

Interestingly, the BIOS on the BX6-2 is the first one and I deliberately did not update to see what would happen. I used an ABIT Slotket III (courtesy of NATA Computers), setting the voltage on the Slotket at 1.5 volts. Fired it up and it booted fine; the BX6-2 identified it as a Pentium III. I then went into BIOS to adjust settings and then off and running. If you have any hesitation about CeleMines and older BX boards, with the proper adapter you should not have an issue.

TEST SYSTEM

Nothing sexy here:

• ABIT BX6-2
• 128 MB Samsung GH
• IBM 10 G 7,000 rpm
• Matrox G400 16 MB
• Ensoniq PCI
• Intel NIC
• CD ROM and HP Burner

THE TEST

I ran Si Sandra 2000 PRO benchmarks on the three CPUs:

• Celeron 533A @ 533 MHz (66 MHz FSB);
• Celeron 433 @ 542 MHz (83 MHz FSB);
• Pentium II 333 @ 515 MHz (103 MHz).

RESULTS

The major difference is in Multi-Media; this is because the C533A includes SSE capabilities while the C433 and PII 333 do not. The differences in Memory Performance can be attributed to the increase in SDRAM speeds due to higher FSBs – 26% greater at 83 MHz and 59% greater at 105 MHz. I have not yet tested cache-intensive apps but would expect to see some differences there also.

BTW: I ran the C533A at 800 MHz, 1.5 volts, no problem.

Original Review 3/31/00

Many many thanks to PC NUT for spending the time with me to test this CPU. It’s not for sale yet but PC NUT will be one of the first to carry it. We had a very limited time to play with this but I’ll try to answer as many questions as I can about it.

SUMMARY:

• Overclocking Nostalgia March 2000: Remember the good old C300A? Did 450 out of the box.
• Overclocking Nostalgia March 2002: Remember the good old FCPGA C 533? Did 800 out of the box.

And that about says it. By sheer luck, we got to play with an engineering sample of the about to be released Celeron FCPGA 533 MHz CPU. It is sweet. This baby runs at 1.5 volts because it is a 66 MHz part. We found that to run it stable at 100 MHz, you have to increase the voltage to at least 1.65 volts – coincidentally (?) the same voltage for the PIII 600E running at 800 MHz. Note closely the multiplier – 8x.

The C533 is a PIII FCPGA so it has the same feature set – SSE and ATC. It also means that it will only work in FCPGA supported boards and slotkets.

Depending on the motherboard, voltage may vary – we had to increase it to 1.9 volts on a slotket with the SOYO +IV. On the MSI 6309, I could run at 800 MHz at 1.65 volts.

The pics below show something interesting:

Not much news here – except for the labeling, it looks like a PIII.

Hello! What have we here? Look closely at the back – there’s some missing parts! It would appear that the circumcision takes place after birth, not during. The pic below clearly shows “the scene of the crime”:

Pure speculation on my part, but if I were Intel and I make all my chips exactly the same (as they have done in the past), then making Celerons out of PIIIs happens on the flip chip circuitry, NOT on the die. The so-what is this is truly a PIII with half its cache disabled…the same PIIIs that overclock so well.

In our testing, the Celeron performs pretty much like a PIII except for lower scores for CPU Mark – a clear indication of the half-cache penalty. However, SI-Sandra benches were almost indistinguishable from a PIII running at 800 MHz. The following table outlines what we found:

All tests with PC133, CAS2 RAM, CPUs running at 800 MHz.

The half-cache penalty shows clearly in CPU Mark32 results on the SOYO+IV, both chips running at 800 MHz:

PIII 600 E: 2260
FCPGA  533: 1330

These test used an ALPHA PEP66 for cooling – the Celeron ran at about 100F under stress, so cooling is going to be extremely important in achieving stability at 800MHz and beyond.

CONCLUSION:

Engineering samples are not production run CPUs, but they are not off the mark either. Everything we know to date indicates that the Celeron FCPGAs are going to breathe a new life into the zillions of BX boards that are ripe for CPU Upgrades. However, the high multipliers on these Celerons (533MHz = 8x) could mean that the Celeron FCPGA 533 is the “chip of choice”. In addition, super-cooling a C533 may push this chip into the 900s. This is going to be fun!