Conclusion
I'm getting sick of saying "it's all about the features" these days. I seem to say that an awful lot when looking at derivative boards based on long established chipsets, or graphics cards based on identical reference designs where manufacturers can't even be arsed changing the cooler.However, that mantra does apply here. We've discussed the main reasons why; Athlon 64's implementation of the memory controller on the processor takes away a lot of the core logic guesswork we're used to doing with other chipsets.
With NVIDIA giving the 250-series chips a feature upgrade, a 'fixed' HyperTransport implementation, some new disk based ability with the changed ATA and SATA controller, 1000Mbit/sec Ethernet capability to let marketroids tick a box, along with an all round spit and polish, nForce3 250 is rather good. The hardware firewall for example is a fine chipset level feature upgrade, scoring NVIDIA good brownie points.
I still have reservations about the single chip design, but it does have its advantages in certain scenarios. Plus, I still have reservations over the ability of board makers to implement designs that can lock the AGP and PCI bus, appeasing the enthusiast, but that will come in reviews of boards that claim to implement it.
In some respects, the most important nForce3 250 designs will be the versions implementing onboard graphics cores. They weren't on test here, examination of which will also have to wait.
It definitely feels like it's just a bootstrapped product in a market where the core logic doesn't dictate performance that much, where "it's all about the features" is something you'll have to keep reading for a while. In those respects, NVIDIA have done well.
Other reservations lie with NVIDIA's lost market share to K8T800 designs in the initial run of boards to support the new CPUs. However, with the Athlon 64 market still small, I doubt they care too much about that. The questions lies with board parners still seeing a market for nForce3 250 designs, given poor sales of 150-based boards. It's something like a 90/10 split in VIA's favour, at the time of writing.
Encouraging, especially with the new SATA controller and GigE MAC with firewall. They do add value to the chipset as a base, providing per-bridge costs aren't too high. I'm hedging my bets, but I slightly prefer K8T800 Pro on paper, at least as a basis for great enthusiast motherboards, going by existing K8T800 designs. It remains to be see what non-reference boards will do with the bridge, but here's hoping its good stuff and they don't just drop it in to exsting 150-based designs and change scant little else.
The jury is out. The chipset itself is good, we just need good boards based on it in return.
