I don’t think the US gov’t is taking kindly to Nvidia desperately trying to stay in the Chinese market.
When they tried to reroute all shipments ahead of the deadline the US moved it to be effective immediately.
All it takes is one small change on the sanctions and their new chips go kaput.
These sanctions will negatively impact the next generation GPUs from all vendors, as they will be designed for China first.
Suppose one looks closely at Nvidia’s alleged data center product lineup for China. In that case, they will notice that the family is meticulously designed to avoid any possible violations of the latest U.S. export rules concerning AI and HPC GPUs. All the new AI and HPC offerings from Nvidia are designed to fit into the green zone in the chart.
How much does having the best performing hardware actually matter in AI?
When you are able to choose between 1 rack of latest Nvidia products vs 3 racks of last gen to get results in the same time frame, you find that the performance matters a lot
I work in the field and I don’t know what the others here are saying. In practice the difference in performance doesn’t matter, we literally just launch a bunch of jobs on the cluster, and if they finish faster it doesn’t matter because I either haven’t checked my jobs, or I’m not ready to collect the results yet, or I’m just dicking around with some experiments or different parameters so the actual speed of completion doesn’t matter. A bunch of weaker GPUs can do the same task as a stronger one, only memory really matters. Doubly true if the company is big enough that power consumption is a drop in the bucket in terms of operational costs.
What actually matters is the overall workflow, stuff like the cluster having downtime is way more impactful to my work than the performance of a GPU, or the ease of designing and scheduling jobs/experiments on the cluster.
Also in the end all of this is moot, this type of training is probably the wrong approach to AI. Note that a child does not need a million images of a ball to recognize a ball, and it would instantly be able to recognize soccer balls, basket balls, etc as all balls after learning one. The way we train our AIs cannot do this, our current approach to AI is just brute force.
not much, garbage (quickly) in, garbage (quickly) out, the training data set and novel techniques matter more than how quickly it’s processed after a certain point.
These sanctions indirectly make Intel and AMD even more competitive in HPC in China due to the performance cap. Take for example Gaudi2, its significantly cheaper than Nvidia’s offerings but also comes up a bit short on performance (ends up being better price to performance than Nvidia). These sanctions would affect Gaudi2 too but not the Chinese variant Intel is making, and with the U.S. government setting a performance cap suddenly the other metrics become far more important, like pricing and availability.