3 ms·
There's no conceptual vulnerability if the system the chip is being integrated into isn't multi-tenant (where a "tenant" can be a literal time-share user, or so
by derefr 8y ago
There's no conceptual vulnerability if the system the chip is being integrated into isn't multi-tenant (where a "tenant" can be a literal time-share user, or something like a website's Javascript code executing in a web browser.)
So: cloud computing, and consumers that might be running dumb code from the internet (= i{3,5,7} and Xeon = volume production) should get chips with no hyperthreading; while HPC, locked-down workstations, and gaming PCs that probably don't have any important data on them (= i9 = low-batch-size production) should get chips with hyperthreading.
- Coding_Cat 8y agoUnfortunately, it's the multi-tenant applications that tend to gain the most from hyperthreading. For HPC it's usually turned off, and I think for gaming and workstations it also has little if any use.
- derefr 8y agoI think I've seen either Intel or AMD explain hyperthreads as being useful in the case of a gamer who is also using a software encoder to stream their gameplay. Also, HPC doesn't necessarily imply a uniform workload; it just implies that all workloads are trusted. Really, modern HPC clusters are still "multi-tenant"; they're just tenants who are all e.g. employees of the same company, or researchers at the same university. So you can describe them, from a security perspective, as one "tenant." None of them would have anything to gain by doing a side-channel attack against one-another.