As commented on by Joe Landman at his blog, Tesla C1060s that plug into desktops and workstations and form the cornerstone of the Personal Supercomputing initiative have been hard to come by so far this summer (note that this is a different unit than the rack-mountable S1070). Last year the company was announcing partnerships around the C1060 to promote the Personal Supercomputing idea, including this announcement with Cray for the CX-1, and as recently as May announced that Dell would sell the cards in their high-end products.
What hasn’t been as widely discussed is the fact that the units can be hard to come by. From Joe’s post in mid-June
If you haven’t heard, Tesla’s are hard to come by. We have several Pegasus systems that customers have purchased, that we can’t get the units for. All of the distributors and resellers we have spoken to indicate that they are getting a small fraction of their orders filled. We have had units on order over a month. Several more orders, and a hard deadline to get units filled.
Joe recently updated his blog to report that things were loosening up a little
They are getting backlog serviced as fast as they can. Looks like they are clearing it out pretty quickly. This is good.
I reached out to NVIDIA’s Andrew Humber early this week to get the official word on what’s going on with these units. Here’s what he had to say
Due to the large demand we are seeing across the board — a number of large cluster installations, additional demand from OEMs, and the success of a developer promotion we are currently running, we are left in a position of ramping up production to meet demand. This is naturally a good problem for NVIDIA to have, but we understand that it might leave our customers waiting in some cases. Please note that product is flowing through at high volume and customer backlog is being fulfilled worldwide.
As Andrew points out, this is a high class problem to have, and one shared by (for example) Apple with its iPhone and iPod lines. I don’t run a manufacturing business, but I imagine it’s pretty hard to forecast demand for a new category of product, and guessing too high is probably very expensive.