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!
The new Intel processors have proven vulnerable to attack by malware called Spoiler, while AMD processors have proved immune. Intel says a software patch may repair the flaw, but it looks like it might take a redesign of their processors. If this is the case Intel has serious problems that has been reflected in the price of the stock and also in the price rise of AMD's stock. If Intel looses the competition for processors, where will they be? If Intel has to do a redesign of the chip, how difficult, costly and time consuming will this task be, any thoughts or comments on this appreciated.
Arthur, last I heard, Spectre also can't be fully fixed with a SW patch either. The problem lies in speculative execution support, fundamental to all high-speed processors. I'm sure more exploits around speculative execution will be found for years to come. I have no skin in this game, but seems to me this report has a heavy dose of marketing spin on the latest such discovery. I see no reason why AMD should be particularly immune from future discoveries.
Arthur, last I heard, Spectre also can't be fully fixed with a SW patch either. The problem lies in speculative execution support, fundamental to all high-speed processors. I'm sure more exploits around speculative execution will be found for years to come. I have no skin in this game, but seems to me this report has a heavy dose of marketing spin on the latest such discovery. I see no reason why AMD should be particularly immune from future discoveries.
That's an interesting question Arthur. One solution, to stop using speculative execution, isn't practical given our expectations on performance. There are various suggestions to refine architectures to mitigate risk. But at the end of the day complex systems are impossible to test exhaustively across all possible sequences so defenses of any kind just keep pushing out the boundaries of what you "know" is safe. I doubt we will ever be able to say that a system is immune to any possible attack (and I guarantee if we do, the hackers will get busy to find a way to prove us wrong).