PCI Express Gen2 Support

In our briefing call with NVIDIA, it was mentioned that the new Quadro FX boards announced today would all support PCI Express Gen2 link and transfer speeds when plugged into a Gen2 host. The PCI Express controller in G84 is able to work at Gen2 speeds, but the spec isn't completely supported in hardware, so there's some specific software (driver and VGA BIOS level) to ensure it works.

Regardless, the new boards will support Gen2 when hosts appear, with NVIDIA working to make sure the supports works properly in upcoming non-NVIDIA core logic that supports Gen2 with AMD and Intel processors. Full hardware support for the Gen2 spec will have to wait for a future round of Quadro models.

The higher bandwidth available with Gen2 will be well supported by certain pro-level applications that rely on high GPU-to-host readback performance, to get data off the board for further processing. The higher available bandwidth and support by the new G84-based Quadro products means that they could be an attractive buy if your app is readback limited in that fashion.

Full hardware support for Gen2 in Quadro products will come in future products.

Thoughts

We mentioned in our look at the Quadro FX 5600, and earlier in this article, that there are certain inherent advantages to a unified, load-balanced and mostly orthogonal architecture when it comes to pro-level acceleration.

The new Quadro products announced today are all designed to take advantage of those inherent properties, to varying degrees of course, along with the latest in display output technology and bus connection that G8-series products provide. Each new Quadro FX is an improvement on the G7-based equivalent it replaces, often in key areas, and with the two lower-end announced models NVIDIA are looking to hit some new key price points and drive down the barrier to entry they perceive as an issue.

$99 for the Quadro FX 370 should help NVIDIA pick up a new swathe of pro-level customers put off by the high price of a Quadro compared to the equivalent GeForce model. At $199, the Quadro FX 570 should be a new volume winner -- at least based on the on-paper specs for perf/price/watt -- and we assume NVIDIA are looking to it to improve market share over the levels they already enjoy.

The $699 Quadro FX 1700 undercuts the existing Quadro FX 3500 as mentioned, and should do fine business in the mid-range pro space, and that just leaves a hole for NVIDIA to fill between $699 and $1999. We assume they'll be looking to do that sooner rather than later.

Lastly, AMD's recent announcement of new R6-based FireGL products will have NVIDIA on their toes. We get the sense that NVIDIA waited to see how that roll-out would go, before pricing and positioning these new Quadros to be competitive. With ~$500 FireGL V5600 looking like good competition for the Quadro FX 1700, we look forward to that matchup in particular when we get round to pitting to two together.

We've got an extended look at the FX 5600 coming up, hopefully versus the AMD FireGL V8650 (R600, 2GiB framebuffer, similar unified shading architecture), and then we'll take a closer look at the new Quadros as they come on stream in October, versus their new FireGL counterparts.

To sum up, the on-paper specs for the new Quadro products, including the lowly NVS 290, look appealing from a price/perf/watt standpoint, and the features inherent to the new G8-series architecture will do the products additional favours. NVIDIA enjoy a strong showing in the professional space, and these new products should help that endeavour along nicely for them.