Some serious security bug in INTEL CPUs?? Since Westmere possibly

[RJARRRPCGP]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

Possibly off-topic: Some recent windows 7 update keeps making my wife's twin Opteron X2 BSOD. For now she's on my Dual Xeon Westmere rig until I get back from a long company trip. Possibly related to the Intel patch reboot issue?

It's probably the known bug that was reported recently!

[Topcat]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

^
yay for running unpatched for ages...

[ratdude747]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

Possibly off-topic: Some recent windows 7 update keeps making my wife's twin Opteron X2 BSOD. For now she's on my Dual Xeon Westmere rig until I get back from a long company trip. Possibly related to the Intel patch reboot issue?

[RJARRRPCGP]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

It has a fucking BSOD issue

For 99 percent of the time, QFT!

[Per Hansson]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

now intel says the meltdown and spectre patches can cause reboot problems in older chips!

Source: Reuters

says it all, my AMD systems dont reboot - ever.

so for the sake of completeness, can intel tell us how often their cpu's reboot normally?

I understand that there isn't a clear cut answer, but Intel is really getting on my nerves with the information manipulation. Feels like everything Intel says is a lie now.

I agree hannah, the above quote is a nice example.
No where does Intel mention BSOD (Blue Screen) in their public communication.
But that is exactly what it is:
Just because the default in Windows is to restart _after_ a BSOD does not mean the CPU has a restart issue.
It has a fucking BSOD issue Intel!

[hannah]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

That's the age-old argument in the PC world...much like Ford vs Chevy in the car world....or coke vs pepsi....republican vs democrat...its an endless argument that no winner will ever emerge from.

I understand that there isn't a clear cut answer, but Intel is really getting on my nerves with the information manipulation. Feels like everything Intel says is a lie now.

[Curious.George]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

Stupid question perhaps, but I have for some time been planning to deploy linux, will it help with this problem or make no difference since this is a hardware issue.

The first question you have to ask is how "exposed" the machine in question is to the introduction of "foreign" code (directly or indirectly). An attacker has to get code to execute ON your machine in order to take advantage of any of these exploits.

So, does your OS (or any applications) allow for the entry (deliberate or clandestine) of "instructions" (which may or may not be direct machine code) from an external/remote, untrusted source?

A machine that doesn't have a network connection is immediately less vulnerable. Likewise, the absence of a (privileged) "console" or other mechanisms for introducing new media to the system (autoplay, executables invoked from shells, etc.)

Likewise, an OS that requires signed executables (presumably, the CA vouches for the binaries that it signs) is largely immune as there's an intermediary required to get the code to execute.

Your first order of business is thus to nail down any paths into the system that can be leveraged to inject the sort of code that can tickle these particular exploits.

[Topcat]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

I didn't see 5400-series Xeons (harpertown) in the master list from Intel....which would also cover yorkfield and those similar (C2Q)....either they're not susceptible or they didn't go back that far....but LGA771 was the predecessor to LGA1366 (nehalem & westmere, 5500 and 5600 series), which is on the naughty list... I figured they would go back that far, there still a lot of these in operation.

[retiredcaps]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

Stupid question perhaps, but I have for some time been planning to deploy linux, will it help with this problem or make no difference since this is a hardware issue.

Most linux distros have deployed mitigation against Meltdown using KPTI. The big name distros like Ubuntu, Redhat, Suse, etc all have it patched. Some of the smaller, less frequently updated distros may not have KPTI.

Spectre is a much longer process to fix.

[llonen]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

Stupid question perhaps, but I have for some time been planning to deploy linux, will it help with this problem or make no difference since this is a hardware issue.

[RJARRRPCGP]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

??
the original xbox chip was a 700MHz pentium

P III 733 Mhz. (Probably a Coppermine)

[Curious.George]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

No, its not a fundamental x86 Problem, its a Problem with all CPUs.

No, it isn't. And, it isn't a natural consequence of speculative execution -- you can design a processor to do speculative execution and still preserve the protections that are being subverted. But, it costs more silicon.

Security exploits are ALWAYS a matter of economics -- someone, somewhere, doesn't want to "pay" the price of maintaining the security that has been designed into the system (whether it is a piece of hardware, software or human policies).

People want to be able to get into their homes if they've misplaced their keys so leave a spare "under the mat" -- so they aren't inconvenienced (in that RARE event) or to avoid the cost of calling a locksmith.

Folks want easy to remember passwords so even if you require a 14 character password (to defy rainbow table cracking), they'll pick from a much smaller subset of the 14 character passwords that are potentially available -- cuz "pDiUoETDK9fwJl" is a lot harder to remember than "my name is bob" (both passwords being 14 characters).

Software assumes input from the user will fit in the space set aside by the software developer and saves the cost of limiting the data to that space in the name of saving a few instructions per iteration.

Projecting the access control bits forward through the memory accesses proposed by speculative execution costs more than simply ignoring them and hoping the actual bus cycle that is run is not contrary to their protections. Or, inhibiting the caching of their results (and thus taking the performance hit, there). (some of the fancier ARMs carry this information forward as they speculate)

It doesn't because speculative excecution is inevetable if you want performance.

The question that should be asked is why do you need this performance? Why is the software so sluggish and bloated that it needs so many trillions of clock cycles to do "whatever". Even when I'm rendering a 3D CAD model, I don't need that performance -- cuz I can always find something else to do while waiting for the render to finish (and, I can do it on the same computer!)

Its interesting to drag out old software and run it on newer hardware and see how blazingly fast it is! Then, ask yourself what you've "gained" by running the newer (slower) software on that hardware.

[diif]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

??
the original xbox chip was a 700MHz pentium

Yeah, the 360 is the one I meant. Forgotten, because I was busy PC gaming.
https://hackaday.com/2018/01/08/spec...-for-xbox-360/

[stj]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

??
the original xbox chip was a 700MHz pentium

[diif]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

The original Xbox chip, made by IBM used speculative execution.

[Behemot]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

What the heck are you talking about? Their site is not screwed by some ovelaying menu craps like intels for them mobile craps which I always have to delete away in dragonfly…

[RJARRRPCGP]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

For all the Intel bashers (me included), I just want to remind them all that AMD isn't immune to this either, and isn't the holy grail savior in this mess. Just sayin'. Several of the articles have mentioned that AMD has been less than forthcoming with information for patches, etc. The hypocrisy runs deep.

I don't like the AMD driver web site and server(s)! (At least for their Windows software!) Makes Intel look like a total angel! LOL!

First off the AMD driver web site design, reminds me of a web site that's ran by radical Euro (or Australia) (or NZ) fascists...

[Topcat]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

For all the Intel bashers (me included), I just want to remind them all that AMD isn't immune to this either, and isn't the holy grail savior in this mess. Just sayin'. Several of the articles have mentioned that AMD has been less than forthcoming with information for patches, etc. The hypocrisy runs deep.

[Stefan Payne]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

Sure but the problem is that now we're finding it's a fundamental problem with x86 (well, superscalar execution), and switching from superscalar to VLIW was a solution.

No, its not a fundamental x86 Problem, its a Problem with all CPUs.
And that nobody thought about attacking it until now, so no security features to prevent that were implemented - well at Intel at Least, at AMD it seems people thought a bit more but sacrificed performance for that...

Define nonsense if it solves the problem at hand.

It doesn't because speculative excecution is inevetable if you want performance.

And AFAIR either the last Itanics had something like that or it was planned. So no, id didn't really solve the problem.


As for VLIW: AMD had something like that from the 2000 up to the 6k Series - wasn't that great even in GPUs...

[hikaruichijo]

Re: Some serious security bug in INTEL CPUs?? Since Westmere possibly

Intel lied about the maximum TDP of their cpus multiple times since the Pentium 4 and again with the first gen core i7, I've tested and i7 920 that said 130W TDP and it reached more than 200W with prime95, the heat sink that came with it couldn't hold the heat for five minutes. But here people always said that AMD cpus were hotter because of the famous silly Tom's hardware video that removed the heat sink while computer was running, who is so stupid to do that? well Tom's hardware reviewers were. Sure Intel paid god money to do that video.