Iâ€™ve been using a Tesla C870 PCI card with CUDA beta2.0. Iâ€™m very happy with the product and am having excellent results !
Iâ€™m interested in the new Tesla C1060 and S1070 products. Iâ€™ve seen some preliminary product documentation on the Tesla Product Literature page, but havenâ€™t seen much else in terms of news or whitepapers from NVIDIA. Beyond3D has a great article discussing the Tesla 10 Series and the initial performance improvements: Beyond3D: Tesla 10 Series Article
When will the Tesla 10 products be commercially available?
Are there any code examples created explicitly for devices with compute capability 1.2 or 1.3?
In particular, I have a slew of detailed questions about the Tesla S1070 GPU Computing System 1U rack mounted unit with 4 cards. Iâ€™m interested in platform specific topics such as:
Is there a hardware userâ€™s manual or developers guide?
What bandwidths can be expected using different PCI Express interfaces?
How are the individual GPUs tasked when using one verses two host adapters?
How fault tolerant is the system? If one card fails will the otheres continue to opperate? Does a single power supply feed all 4 cards? How is system status monitored by an application?
Any links, articles or wisdom would be helpful.