If you look at the benchmarks for the Ti200, they trail the “old” GF3 card.
You have two potential reasons for that:
Why did nVidia come up with a “new” card that’s worse than the “old” card?
Well, this one’s supposed to be a lot cheaper than the high-end Ti500. If this performed as well as the “old” GF3, relatively few people would pay an extra
$150-200 for the Ti500 for just a 10% improvement. More would pay that much for a 20% improvement.
Sounds like a potential overclocking situation to me.
Unfortunately, those relatively few places with both cards didn’t look at this.
Let’s see what the prospects are:
Core Speed: The cores of all these Titanium chips are being made using a 0.15 (rather than the old 0.18) micron process. This looks to be
a pure shrink; NVidia didn’t even change the revision number. Nonetheless, the shrink led to enough of an improvement to let nVidia market the Ti500 with a 240MHz core and it seems capable of 265-270MHz without too much fuss.
It could be that these 240MHz chips are hand-picked. It could be that nVidia’s yields are erratic enough that functional chips have a wide spectrum of maximum speeds.
However, I find it very doubtful that chips using an improved process aren’t going to normally do at least as well as those made with the old process. I would be very surprised if Ti200s couldn’t normally run at a speed of at least 200MHz.
The performance of a GF3 is much more dependent on memory bandwidth than core MHz, so memory speed is really what we have to look at.
At least according to this picture from XBit Labs, the Ti500 uses 3.5ns DDR memory. The old GF3 uses 3.8ns DDR memory.
So what does the Ti200 use?
None of the reviews I saw indicated that, and since the GF3 Titaniums have a heatspreader on the memory chips, looking at pictures doesn’t help.
Nor are the manufacturers a lot of help yet. There’s two cards available for at least pre-sale at the moment: the Visiontek Xtasy 6564, and the Leadtek WinFast GeForce3 Titanium 200 TDH.
Visiontek isn’t admitting to making the product yet on its website, while Leadtek’s information looks inaccurate (it shows faster memory for the Ti200 model than the Ti500, which makes no sense).
In any case, current prices are skyhigh anyway, no better than “old” GF3 prices. You shouldn’t be in any rush to buy.
What you need to do first before buying is to find out what kind of memory is going into the Ti200 card you want to buy. If it turns out to be 3.8ns RAM, for instance, there’s no reason why you shouldn’t be able to run the same memory as an old GF3 at least as fast as on an old GF3.
What Do You Call Overclocking an Underclocked Card?
Those who like to push the envelope are probably scoffing at these timid estimates, but this article is meant for those who don’t want to push the envelope.
There’s a very good chance the Ti200 is a deliberately underclocked (as opposed to one with a reasonable safety margin) card. In other words, rather than a 15% margin to the ragged edge, we may well have a 30% margin instead.
The memory is a big question mark which needs to be answered, but if the answer is satisfactory, we’re most probably looking at a risk-free overclock that would half the performance difference between the Ti200 and Ti500.
Given that the price difference between the two is $150-$200, getting a couple answers first is well worth the time and effort.