Let’s take a quick look at Tualatin.
Essentially, it’s a PIII dragged through the ringer one more time in .13 micron.
It’s really meant to eventually become Celeron III (though cache will no doubt be sliced a bit for that incarnation).
An Upgrade Chip?
I asked you yesterday about this, and the primary interest in this chip seems to be as an upgrade chip to current systems (a few were
interested in SMP setups).
Tualatin isn’t supposed to work on current Intel motherboards. However, the change doesn’t appear to be that major, since a revision of the Intel 815 boards is supposed to accommodate it.
One website stated that their correspondent in Taiwan heard rumors that while Tualatin doesn’t work on Intel 815 or Via boards, it does on BX boards. Mind you, that’s not “some people
have got one, and it works for them.”
This really doesn’t make a whole lot of sense. The VRMs on BX boards are marginal at best for Tualatin (Digit-Life reports voltage of 1.1-1.15V for later Tualatin Xeons, and BX VRMs go down to at best 1.3V).
True, it’s possible this alleged chips are just being run at 1.3V, but you could do that just as well on an 815 or Via board (matter of fact, a few of the later ones may have VRM 9.0 compliant devices, which get down to around 1.1V.
Here’s my suspicion about what’s going on:
Intel’s going to be releasing a bunch of processors in the next couple months. There will be server Tualatin with 512Kb cache. There will be
desktop Tualatins with 256Kb cache running at 133Mhz FSB. Finally, there will be a 1Ghz and 1.1Ghz 100Mhz PIII introduced.
The roadmap information, while it indicates elsewhere where the PIII is Tualatin or Coppermine, doesn’t do so for these chips. We’re not sure which they are, and here’s the possible key to the puzzle. I think there’s a good chance those chips running on BX boards are those 1.1/1 Ghz CPUs.
They may be just regular PIIIs, in which case, they would have no problem running on older boards. Or they might be Tualatins modified to run at higher voltages.
Intel had planned to come out with a “Coppermine-T” to cover the upgrade needs of current Intel platform users but those plans have gone by the wayside. Either Intel expects everybody to buy a new mobo (not too wise), or they’ve put the “bridge” chip under the Tualatin effort.
After all this time being focussed on AMD chips, we’d probably would remind old-timers or inform newcomers that Intel chips are multiplier-locked. They have been since August, 1998, and nobody’s been able to break that.
That means you can only overclock by increasing the FSB clock over default. For those of you unaware of the ramifications of FSB overclocking, please go here for a quick lesson.
The “regular” Tualatins will be set to run at 133Mhz. Depending on your equipment, some of you may not be able to hit 150Mhz (especially with older mobos and/or older/cheaper RAM). If you have relatively new equipment, you might reasonably expect something between 150-160Mhz. Anything above that is pretty much luck.
So the “regular” Tualatins will probably not be too overclockable due to system bottlenecks. (If those 1.1/1Ghz new PIIIs are really Tualatins, the 1Ghz will probably be a good candidate for easy overclocking, but see below).
The CPU itself? In the past, the fastest chip Intel plans to release in a series has been a very base point upon which to calculate how far the chip will initially go (using high-end air cooling). Add 10% to that figure, and you have an estimate that’s been very accurate in the past.
The quickest Tualatin planned at the moment is 1.26Ghz. Add 10% to that, and we get a figure just shy of 1.4Ghz. That’s about where I would bet overclocked Tualatins will mostly end up.
Some of you may say, “but the chip should run cool at that speed, surely it can do more.” Coolness isn’t everything. The initial Coppermines ran cool, too, up to about 800Mhz, but cool or not, they usually went no further.
Eventually, the Tualatins may do better as they get modified and tweaked, just don’t expect to buy one next month with a fan and expect to get close to 2Ghz.
Or they may not: the range of frequency on these is pretty narrow for a new line of CPU. Intel didn’t build Willy just for the hell of it; they built it because the PIII architecture was becoming too limiting.
Just don’t see Tualatins as competitive with Palominos on performance grounds. Probably will cost a good deal more than equivalent TBirds for quite some time. But for those who have decided for whatever reason that they don’t want AMD, or don’t want Willy, either, they may be worth taking a look into.