from Hacker News

Ask HN: Why Intel and AMD share the same design flaw?

by fxbois on 1/25/18, 5:23 PM with 6 comments

• by lucozade on 1/25/18, 6:42 PM

  Because they are good ideas if you want your CPU to run fast.

  The combination of features involved: caching, branch prediction and out of order execution are all very valuable. Both individually and in combination.

  What has been discovered is that, under certain circumstances, they combine in ways that can leak sensitive information. This is a very bad thing and needs mitigation/fixing.

  However, it doesn't mean that the underlying features are inherently bad. Just that they have complex interactions that aren't as well understood as, perhaps, was thought.

  When this all settles down, I wouldn't expect these features to be removed. Rather they will be constrained such that they retain as much performance as possible but without the security implications.

• by samfisher83 on 1/26/18, 1:21 AM

  Its how processors are designed. If you take a computer architecture class which I am pretty sure most compsci people have done you learn about most of this stuff. If you don't do caching, branch prediction, pre fetching, out of order execution your processor won't be as fast. Someone just figured a way to exploit the fact that processor are already loading data that it needs to process.

  If you are doing a laundry cycle and you have to wait for the dryer to to finish before you load the washer its going to take lot longer to do the laundry.

• by zelon88 on 1/29/18, 8:04 PM

  I was under the impression that AMD chips weren't vulnerable to Meltdown or variant 1 of Spectre. Only variant 2, Bounds Check Bypass, could be achieved on an AMD chip.

  Is that an incorrect impression?

• by psyc on 1/26/18, 6:19 AM

  Essentially because it isn't a bug in the usual sense. It's a vulnerability that is present even though the processors are doing their primary job correctly.

• by thomas on 1/25/18, 5:28 PM

  Was wondering the same thing! Experts please jump in.