Target PC :: Coppermine Slot 1 650 MHz Processor

Why do I run lousy DOS tests? I have a theory and it's proving to be very useful. Looking back to the Combo of the Year article, I achieved 616 MHz (5.5x112) in DOS and this translated to a stable 550 MHz (5.5x100) at default (2.0) voltage. My theory is that when you shove your processor as high as it can go, booting to a floppy and not corrupting your hard drive Windows install, you should be able to shave off the upper 10% and wind up with a stable 2D/3D speed. For example, if the max speed is 616 MHz at maximum voltage (i.e. 2.3 for the Celeron), taking 90% of that, we arrive at 554 MHz for a default voltage. The numbers quoted below for Vcore are the minimum, not maximum required to obtain stable operation.

*FSB (MHz)*	*Speed (MHz)*	*Vcore (Volts)*	*Chip Temp (F)*	*Case Temp (F)*
100	650	1.65	89	77
124	806	1.65	95	77
133	867	1.65	98	77
137	891	1.65	98	78

Several things should be readily apparent. The first being that I could test no higher than 137 MHz FSB. This was not due to a video card or SDRAM limitation as I used an old 1MB ISA video card and two different RAM strips. The 137 MHz FSB using factory cooling hit the limit of the L2 cache most likely and did not hit the 75C (167F) internal temperature limit. The second is the core voltage. Moving the voltage to 1.7 or even to the limit of 1.8 volts had no effect on maximum speed.

As I had no Voodoo 3 PCI video card available, I had to use an old 1MB PCI card to verify stable 2D/3D speeds. As this isn't the optimum scenario, caution must be exercised when pushing an AGP to >66 MHz speeds. My 32MB TNT2 Ultra couldn't withstand 3D testing for more than 15 minutes at the 133 MHz FSB (89 MHz AGP) speed. This was a limitation of the 440BX chipset, not a limit of the chip.

*FSB (MHz)*	*Speed (MHz)*	*Vcore (Volts)*	*Chip Temp (F)*	*Case Temp (F)*
100	650	1.65	95	80
116	754	1.65	110	80
124	806	1.65	111	82
133	867	1.65	112	85

Again, what's of particular interest is that the core voltage doesn't need a bump beyond default to crank it up. This means that motherboards without voltage tweaks that possess good FSB selection should have similar results to ones that do. Because this is a factory heatsink/fan setup, the interior case temperature doesn't reflect the rising internal chip temperature as in higher CFM cooling scenarios. This shows that the factory heatsink does a relatively good job of cooling the chip without blowing all the heated air around the case, thusly heating everything else up with it.

In the case of the P3 Coppermine, my 90% Windows/DOS theory proves to be false with anything less than radical cooling. Working backwards from the Win98SE tests, the maxmum FSB is DOS should be 146 MHz (133x1.1) instead of 137 MHz FSB. This isn't entirely bad though. For this particular sample, it appears to run stable in Windows to within 3% of it's maximum DOS speed. The Celeron 366 required a lowering of a full 10% to achieve the same stability. This could be a testimony to the greater stability of the 256 KB Coppermine.

Speeds in the 800+ MHz class are darn fast--for any application. The VIA 133A chipset has the magic 1/2 AGP multiplier, so the 650@866 could easily become a reality with a few other well matched components. At the time of this writing, the 1 GHz chips are all but unobtainable and the 866 MHz version costs approximately triple of the 650 MHz chip. If you can't wait for dedicated Coppermine boards and the overclockability results of the Coppermine128 (Celeron PPGA) 566 MHz wonder, the Coppermine 650 MHz is a very good buy indeed. Even if you don't possess a motherboard with the 133 MHz FSB, as long as it can support Coppermines, the 650 makes a great upgrade from an old P2. As the prices drop to the $200 range in a few months, this P3 will become the overclocker of choice for those requiring 256 KB of full speed cache. Highly recommended for standard clockers and overclockers.

Related Reviews: The Intel Pentium III Coppermine FC-PGA