We’re now just about at the point where realistic 166MHz + overclocking is becoming feasible to those less than extreme hardcore.
Yes, I’m well aware people have been doing this for months already, but a lot of it hasn’t yielded much or indeed proved counterproductive because a few pesky details have often been ignored.
Memory Fast, Memory Right
Sure, pioneers broke the 166MHz barrier a long time ago, but at first, they had to downshift their memory settings to get it to work. This often left them no better off than if they had stayed at 133MHz with fast settings.
Only now are we seeing memory modules at least somewhat capable of 166MHz at fast settings under vaguely normal conditions.
The Corsair XMS3000 is the only module which is getting a good deal above 166MHz under less than harrowing conditions.
The Crucial PC2700 seems to do alright at 166MHz, but not much more.
True Samsung PC2700 is currently the overclockers’ choice for its price, but when pioneers talk about the need to “treat it” with around 3V for a while before it will perform great feats, this is what I’d call harrowing and not something to be recommended for the less-than-Captain Courageouses out there.
Joe already has some Crucial. I’m going to get some Corsair, and maybe some Samsung, too.
You Can’t Put A Size 12 Foot In A Size 10 Shoe
Unfortunately, this has been what many have been unknowingly doing in pursuit of memory MHz on Athlon boards, and here’s why that is.
At 133MHz/266MHz FSB, the bandwidth of an Athlon motherboard is 2.1Gb/sec. At 150MHz; it’s 2.4Gb/sec. At 166MHz; it’s 2.7Gb/sec. The Front Side Bus not only handles memory transfers; it also normally handles all the other communications between other devices in your computer like hard drives and sound cards and network cards. They all have to fit into that FSB space.
What people have been doing to jerking up memory speed so high by using asynchronous buses that they appear to be jamming the FSB with more bandwidth than the FSB can handle. This appears to be why folks don’t do as well as they expect when they benchmark real activities rather than some memory score.
We’re going to have to experiment a little, but we suspect that if you have some superduper RAM; you’re going to be better off running FSB and memory synchronously at 175/175 than something like 142/190.
(In theory, this is not supposed to be a problem with PIV boards, because they theoretically have at least 3.2GHz bandwidth at 100Mhz; 4.2GHz at 133MHz. However, again in theory, we don’t exactly understand how a quad-pumped FSB can handle double-pumped memory any better than a double-pumped FSB. We have a few questions here, too.)
Of course, there’s a little problem with running 175/175, and that is . . . .
The Forgotten PCI Divisor
A lot of people have forgotten (or never knew) that this little thingamajig can stop your overclocking fantasies stone cold. It has not helped that many if not most motherboard manufacturers nowadays don’t indicate this in the BIOS settings or even in their manuals.
If you need to get up to speed on this subject, read this.
How important is this? Let me put it to you this way. I plan on doing a lot of 166Mhz+ testing. If I can’t find out for sure (not maybe, not assume) the mobo has a /5 PCI divisor, that mobo gets stricken off the list. Period.
And if you plan on doing the same thing, you’d be crazy not to do otherwise.
My Summer Project
We hope to do a lot of high-speed FSB testing this summer on both the Athlon and PIV platforms.
We say “hope” because while we don’t think Athlons will be any problem (we’ll start off with an XP and a KT333, then go to Thoroughbred and a KT400 later on), getting a PIV to run an FSB at 166MHz is pushing it. Not ridiculous currently, but no sure thing, either.
Right this moment, the 1.6A is the only candidate with a decent fighting chance to do this. Sometime late this fall, the lowest-speed 133Mhz FSB PIVs should have a good shot at it (along with the 1.8A if it’s still around).
Other equipment will be tested as appropriate as time goes by, but our priority is going to be to see how high and how stably products will work in this environment.
We’ve been really quiet on the hardware front the last few months because we knew this had to be the next frontier, and we also knew all the pieces weren’t in place yet for all but the most hardcore wandering into this wilderness, the fur trappers of this new frontier, so to speak.
It’s still pretty early, and it probably won’t become mundane until near the end of the year, but it’s now time for at least the pioneering homesteaders to show up.