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 Register reports that a 'Kernel memory leaking' Intel processor design flaw will result in a significant hit. A fix could result in an (estimated) slowdown of 5-63% depending on task. AMD processors are NOT affected by this flaw.
This flaw appears to be about a decade old and affects Intel x86-64 architecture. The article claims that a microcode update won't be able to address the issue.
Might be a little premature for the non-Intel folks to be celebrating. It appears 2 problems were found: Meltdown which affects only Intel devices, and Spectre which affects Intel, AMD and ARM devices. Article acknowledges that Meltdown is immediately more serious, but Spectre may have longer term repercussions.
Might be a little premature for the non-Intel folks to be celebrating. It appears 2 problems were found: Meltdown which affects only Intel devices, and Spectre which affects Intel, AMD and ARM devices. Article acknowledges that Meltdown is immediately more serious, but Spectre may have longer term repercussions. Security flaws put virtually all phones, computers at risk | Reuters
And other articles point out that this is largely FUD spread by Intel; the Meltdown bug (Intel, some ARM, no AMD) has been shown to be real and exploitable in many circumstances, the Spectre bug is much more difficult to exploit (either theoretical only or has only been demonstrated within one process) in very limited circumstances. So Intel is saying "these bugs affect all CPUs" to draw attention away from the fact that by far the most serious one only affects theirs. Also don't forget the most serious security and performance impact is in virtualised data centres, which are >99% Intel. So overall the impact on the market from Intel CPUs is at least a couple of orders of magnitude higher than from AMD.
Bad move on his part I would expect All CPU makers shares to rise, well the ones who use speculative branching to as customers scramble to replace affected parts.
Intel should be leading the charge in bringing out parts that are immune to these exploits.
Bad move on his part I would expect All CPU makers shares to rise, well the ones who use speculative branching to as customers scramble to replace affected parts.
Intel should be leading the charge in bringing out parts that are immune to these exploits.
I am wondering about one thing. According to the press, Intel has been informed about the bug in June 2017. However it released new processors on October 2017, knowing that these processors are indeed affected and that fixes are at best semi-possible to fix through firmware and software patches.
I think it goes deeper. INTC management had to know this was possible, but used it as a shortcut to overcome architectural flaws in x86. By running without the seat-belts others provided, they could claim performance advantages and win in the market with no concern about the long term impact on customers. Even if they "get away" with this, they have lost all credibility with smart (cloud) customers who will back away and stop doing business with them. Maybe they will open source their own general CPU design that is more modern than this mess. POWER? Beyond that, imagine using any INTC chip in a car!
From my understanding, this is not an ISA issue (i.e. x86 specific) but an implementation/architectural issue. I am wondering, are all the performance-impacting fixes be applied equally to all processors no matter if they are vulnerable or not? I mean if an AMD processor is not vulnerable, will the patches be applied to it as well as part of the fixes that OS providers will roll out? If this happens then others will be more screwed than Intel.
