Incidentally, it's obnoxious that the enthusiast K series CPUs always had them disabled, as well as VT-d (IOMMU) and vPro. While the current implementation of TSX may be immature, and mainly useful 'just' for hardware lock elision, i.e. some percentage boost (although isn't maximizing performance the whole point of K series?)... in principle it could eventually allow new models of programming to be practical, without the huge overhead of STM, so it should be as widely available as possible. And VT-d, while useful for server-level virtualization (as the name suggests), is also required for security against DMA attacks.
I agree, that's really annoying. Thankfully it looks like they may be slowly introducing some of these features to the K line. The 4790k for example is the first k chip that advertises support for VT-d. Still no vPro though.
Everybody still seems to be referencing the original news story, which in turn referenced a tweet. Nobody seems to have a clue what the actual errata is; looking forward to seeing the explanation of the bug and the conditions under which it can occur.
6 comments
[ 2.6 ms ] story [ 23.2 ms ] threadhttps://news.ycombinator.com/item?id=8168569
https://news.ycombinator.com/item?id=8169697
http://ark.intel.com/compare/80806,80807