Intel Pentium III 500E CPU and 550E FC-PGA Review

An anonymous reader says "This article on the Intel Pentium III "Flip Chip" 500/550E shows some great overclocking potential for this CPU. " Its a fairly technical article, but a nice one.

Comments

Re:Is overclocking really that important?

[Mr. Flibble]

I am currently running a Celeron 300A @ 450 Mhz.
I can certanly testify that the speed increase in games (and kernel compiling!) is signifigant. I have not benchmarked it I admit, but I can certanly feel the difference when it is at 300 vs. 450.

SMP won't help with Quake 3 Arena (It will under NT, Linux SMP for Q3A is not yet avalible) Or Half-Life, or many other games.

Half-Life for example is a very processor-dependant game. Many games currently are. Although graphics card technology is growing rapidly, few current games fully exploit the cards abilites. So, in short a high clock speed is VERY important.

SMP might help if you are running BeOS, as it can force multi-threading, but the current state of gameing does not support SMP, hence there will be no bonuses associated with it.

Having said that, My next system will be a Dual Celeron 366 PPGA on a Abit BP6 motherboard. I will clock the chips to 500 (possibly 550 with water cooling, just because its geeky). I am doing this because:

1) Linux will FLY on that system.

2) Quake 3 Arena should have Linux SMP by the time I get the Motherboard.

I personally am not all that interested in overclocking my Graphics Card, I dont think that the yield in framerates is enough. The CPU is a different story IMO.

Re:Is overclocking really that important?

[Tackhead]

> I seriously question the sanity of anyone who overclocks a production system.

I'd echo that - but with one very important qualifier that completely inverts the intent of your original post, namely "without knowing what they're doing".

The rules for successful overclocking are actually pretty simple - know the relationship between FSB speed and the multipliers and dividers that turn that FSB speed into AGP and PCI speeds, and how that relationship varies as a function of motherboard vintage.

Would I run a C366 (5.5 x 66 MHz) at 75 MHz FSB, to get 412.5 MHz? Hell no:

• AGP: 2/3 of 75 is 50 MHz, and I'm (wastefully) underclocking the AGP bus, or if it runs at 1 * 75 is 75 MHz, I'm (far-beyond-dangerously!) overclocking my AGP bus.
• PCI: 1/2 of 75 is 37.5 MHz, and I'm overclocking my PCI bus by maybe 10%, which most hard drives can handle, but perhaps not all.

That "hell no" argument goes double if you ask me to run it at 83 MHz FSB * 5.5 for 458 MHz.

But if I crank that puppy up to 100 MHz FSB, I have - modulo heat - rock-solid stability:

• AGP: The mobo should detect 100 MHz FSB and switch to 2/3 of 100 MHz, or 66 MHz. You're running at spec.
• PCI: The mobo should detect 100 MHz FSB and switch to 1/3 of 100 MHz, or 33 MHz. You're running at spec.

That's right - as long as it can be cooled effectively, a C366 running at 550 MHz is more stable than one running at 412 or 458.

A similar set of calculations can be performed for other CPUs running at other core multiplier frequencies, and will reveal different "sweet spots" where the PCI and AGP busses run in spec. (In an ironic twist, buying a "faster" CPU of the PII and Celeron vintages actually makes things worse :)

So would I allow some random kiddie who says "0vercl0cking iz k00l 4 gamez, d00d!" crank a production system of unknown processor vintage and motherboard capability to "as fast as you think you can get it, kid"? No, that's madness. He might have a CPU that'll run fine at 550 MHz at 2.1V, but not at 2.0V. Instead of turning the voltage to 2.1V and testing heat dissipation, he leaves it at 2.0V and "steps back" to run the FSB at 83. Two weeks later, the hard drive explodes and everyone blames overclocking. It wasn't the fault of overclocking, it was the fault of one particular overclocker.

But would I, if starved for speed (and cash - in that I couldn't afford to buy the speed off the shelf), allow an overcautious doomsayer dissuade me from applying overclocking in such a way as to get speed without sacrificing stability? No again - that would be just as insane.

Like everything else in computing - blind adoption and blind rejection is madness. A clued admin will realize that there's a middle ground.

If an admin has the power to select quality equipment, and the clue to tweak said equipment without risking stability, the right answer to "Admin, we want a fast server but are on a low budget" is "yes, I can do that, but only if you let me select every component and not quibble if the motherboard I choose costs $10 more than the cheapest one on the market."