Yeah, it's hard to say because it would be very application-specific. An application that makes many system calls to the kernel is going to be more affected than one that makes fewer such calls. It would also depend on how well developers have optimized a given app--may only experience inconsequential longer app opening times if things such as "warming" the cache at startup are done to make an app more responsive during actual usage.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Intel Chips Have Memory Access Design Flaw and Fix Could Lead to Performance Drop
- Thread starter MacRumors
- Start date
- Sort by reaction score
No, it is VERY significant. If there is a Safari bug, for example, they can read kernel memory - so malicious Javascript or Java or HTML5 (or heaven forbid, Flash) could all potentially trigger it. Once you can read kernel memory, you can obtain security keys. With the keys, you can then use them to increase your privileges to, say, root or any other process and then do whatever you want.
This is a huge bug. Like the Intel FDIV, only much, much worse because it isn't just returning incorrect floating point values, but giving essentially everyone access to everything. (The Intel F00F bug was mild in comparison too.)
Yes, the reaction of some people is laughable, but at the same time this is inexcusable by Intel to go 10 years without picking this flaw up. You could almost accept it affecting one generation of CPU's, but 10 years worth??
It also shows how poor Intel have become as a technology company as the last 10 years has essentially been a slow trickle of repackaging the same products over and over.
Also interesting to learn that the Intel CEO sold all of his stock that he was allowed to sell in November.
https://www.fool.com/investing/2017/12/19/intels-ceo-just-sold-a-lot-of-stock.aspx
The little weasel knew what was coming.
If the Intel problem could be fixed by a software updated with no impact "damage" to the end user (i.e., no speed penalty) then it's one of those no harm no foul issues and most everyone will move on. But if the software fix results in a speed reduction (a form a damage to the end user), get ready for a class action law suit against Intel. Look at how fast the lawsuit was filed against Apple for their IOS speed throttling to deal with older batteries. This is a MAJOR screw up by Intel.
There is a possibility that macOS isn’t affected due to Mach kernel design. The macOS kernel does not map its own memory into each user processes address space.
Citation:
http://flylib.com/books/en/3.126.1.91/1/
Citation:
http://flylib.com/books/en/3.126.1.91/1/
The question is why this has come out in the open? Of course the NSA knew about it and exploited it.
I was referring to the macOS security issue, not the Intel one.
I've reviewed the information Intel has provided to Microsoft and Apple and am well aware of the implications of the flaw.
We don't know that. We only know that it took decades for it to become known outside of Intel and the NSA.
If intel wanted to give the nsa a back door there are a lot more direct ways they could do it. And for what it’s worth, I don’t believe intel would ever agree to do such a thing. Slightly more likely would be an nsa employee embedded in intel unbeknownst to them, but I doubt that, too.
Intel is working to patch this on Linux. Intel just tried to cripple ALL processors, not just their own. AMD guy jumps in to say "Our processors aren't impacted by this issue and don't need to be slowed down too." Shady stuff on Intel's part.
https://lkml.org/lkml/2017/12/27/2
Edit: Upon further consideration, this is actually the way to do it. You know there's a large issue so you patch across the whole. Allow exceptions to whitelist out of it, rather than attempt to blacklist only the flawed processors as you don't know for sure if it impacts others too.
https://lkml.org/lkml/2017/12/27/2
Edit: Upon further consideration, this is actually the way to do it. You know there's a large issue so you patch across the whole. Allow exceptions to whitelist out of it, rather than attempt to blacklist only the flawed processors as you don't know for sure if it impacts others too.
Last edited:
I actually just bought one last month too with the upgraded i7 spec bump. Got dayamn!
Ditto. Here's hoping my 3 month old $2500 machine doesn't slow down to the speeds of the 10 year old one it replaced.....
The solution should be to force intel to fix the problem, in hardware, in a timely manner, with no performance degradation, and replace every single chip for free regardless of computer brand or age, and regardless of the magnitude of the cost to intel.
Until the collective "we" begin to hold corporations responsible, in meaningful ways, the continued screwing of the average individual for the sake of the few at the top will continue.
In my case, I'm literally (no, not figuratively, literally) on the ragged edge of a 1 year old gaming rig - liquid cooled and overclocked, because I game stereoscopically in true 3D modes almost exclusively, and have for many years (generally this means I get 1/3 the frame rate I would get without stereoscopic mode enabled). I absolutely can't lose even 10% of my CPU performance (the conversion to stereoscopic is not just GPU intensive, it is also very CPU intensive) since that would drop my performance below acceptable levels in many recent games.
So yeah, I guess if the fix really does slow down my system I'll be joining a class-action, for which I'll receive a free mostly-dead hamster and Intel will get a tiny little slap on the wrist...
Until the collective "we" begin to hold corporations responsible, in meaningful ways, the continued screwing of the average individual for the sake of the few at the top will continue.
In my case, I'm literally (no, not figuratively, literally) on the ragged edge of a 1 year old gaming rig - liquid cooled and overclocked, because I game stereoscopically in true 3D modes almost exclusively, and have for many years (generally this means I get 1/3 the frame rate I would get without stereoscopic mode enabled). I absolutely can't lose even 10% of my CPU performance (the conversion to stereoscopic is not just GPU intensive, it is also very CPU intensive) since that would drop my performance below acceptable levels in many recent games.
So yeah, I guess if the fix really does slow down my system I'll be joining a class-action, for which I'll receive a free mostly-dead hamster and Intel will get a tiny little slap on the wrist...
Because you aren't a very critical thinker.
[doublepost=1514995497][/doublepost]
That fix isn't practical, even if they offered it.
Plus, you got full use out of it until now, so they only owe you some depreciated value of your current chip.
Some of you seem to think companies owe you the world because you spent a few hundred bucks on their product 4 years ago.
When clamoring for class action suits or that Intel should do XYZ, remember where those costs will land... on the consumers who buy future Intel products. Sure, the shareholders will hurt some as well... which again is pretty much everyone with a retirement account / 401k since many will likely include mutual funds that also have some Intel stock.
I'm just glad I haven't gotten around to buying that MBP15 I'd been thinking about. Guess I'll sit on the sidelines a bit longer.
I'm just glad I haven't gotten around to buying that MBP15 I'd been thinking about. Guess I'll sit on the sidelines a bit longer.
Intel is used all across the government. If the government wants it, Intel will have to do it.
But, I also don't think this was intentional to create, but I do think that it is likely that people at the NSA and Intel knew about this and let it go on. We know the reason for the NSA to act like this, they are anti-American. What we don't know is how involved Intel was in covering this up.
A guy has been making benchmarks in Linux and it's not as bad as it may seem. Gaming and video encoding seem unaffected. Disk IO with SSDs is severely affected though.
https://www.phoronix.com/scan.php?page=article&item=linux-415-x86pti&num=2
https://www.phoronix.com/scan.php?page=news_item&px=x86-PTI-Initial-Gaming-Tests
https://www.phoronix.com/scan.php?page=article&item=linux-415-x86pti&num=2
https://www.phoronix.com/scan.php?page=news_item&px=x86-PTI-Initial-Gaming-Tests
That is outdated information, your source talks about MacOS 10.4, which allowed some 64-bit apps but still had a 32-bit kernel and it had to flush the TLB with every syscall to give apps the full 4GB address space. Since MacOS X 10.6 Snow Leopard, it has a fully 64-bit kernel and doesn't flush the full TLB when making syscalls. It was one of the most well-known performance improvements that were introduced with Snow Leopard.