You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Amazon is now supporting something they call an EC2 F1 instance (cryptic name?) to support FPGA-based accelerations in the cloud. You can build FPGA-based accelerators, uploads your design to F1 instances in the cloud, then use those instances to accelerate compute. I blogged about this earlier, in that case for Intel/Altera being used to accelerate Microsoft Azure compute. This is looking like an interesting trend for the FPGA business.
Good question Dan. In the Microsoft Azure Altera/INTC case, they are being used for reconfigurable network processing. The Amazon case seems to be a broader play targetting any form of acceleration (for those knowledgeable enough to take advantage of the capability). Deep learning for vision recognition etc would be a possibility, but perhaps NVDA has too much of a head start in that area. But there are presumably plenty of other areas where accelerators are needed but the DSP structure isn't the best platform.
Machine learning encompasses a fairly broad set of algorithms. Some of these algorithms may run better on GPUs and others may run better on FPGAs. So the answer is they don't compete that much directly, it depends on what algorithms are being used.
Small business or large retailers are only as good as their servers. Companies that know what they're doing online have better products and more sales than companies that don't. I've seen it with Columbia Sportwear struggles as well as Nike.
Webhostings and servers played a huge part in electing Donald Trump. Talking to that industry should be the first priority for the democratic party. It won't happen because it makes too much sense.
