Intel Skylake processors crash at specific Prime number calculations
Click here to post a comment for Intel Skylake processors crash at specific Prime number calculations on our message forum
Fox2232
As much as AMD Phenom bug. Nearly impossible to happen, normal users did not have to care, and yet BIOS fix was issued. And in process it degraded performance a bit.
GeniusPr0
If it has to compute the prime specified then the system will freeze.
Octopuss
How can this be fixed with a BIOS update, when apparently something is done wrong on the hardware level (like, some messed up instructions or something)?
Undying
Thats what i thought. They said its a Skylake architecture hardware error. No bios can fix that.
Kaarme
There must be millions of Skylake systems running already. If it was a common problem during running common applications, like games, it would be a huge issue by now, with PCs crashing regularly left and right.
Denial
Huh? Intel pushes out Microcode bios updates with nearly every processor release to fix various errata. This isn't anything new aside from the fact that it's a bigger flaw than you usually see.
Warrax
That's what firmware do, to an extend.
FerCam™
So this is the new 586 aka Pentium?
nexxusting
Definitely "Huh?". Lol.
Obviously somebody doesn't even know what a microcode update even is. Doesn't just allow newer CPU'd on a mobo.
Neo Cyrus
For most users this means nothing unless an updated BIOS to work around this causes a performance hit like with the Phenoms. For those who are doing intense calculations, a defect like this could be a real problem.
Reminds me of the Pentium FDIV bug, which resulted in some calculations being wrong.
Razoola
Could get real messy if intel also cripple the overclock hack in the same fix
haz_mat
Some things can be fixed in BIOS, happens all the time in fact (as mentioned, errata is not uncommon). I think its too early to tell if there will be a performance hit from the fix, but that is certainly possible.
This particular issue is different than the infamous FDIV bug - there was no system hang there, instead it was yielding incorrect computation when dividing floats.
In that link to the intel forums they mention the issue is only reproducible when hyperthreading is used with AVX instructions on a skylake chip - and they're able to trigger it when computing on a specific large-exponent. A very specific set of conditions to say the least.
From this alone, there is no reason to conclude that this issue is only able to be triggered under these exact circumstances. Someone in that thread even mentioned an ongoing discussion at tomshardware about skylake systems freezing up while idle for no reason, even after RMA'ing multiple components.
I'm sure more information about the issue will be revealed after they have a fix ready, as disclosing too much at this stage is a definite security risk.
WareTernal
So your insane OC may be stable after all?
"There's also FPGAs and other ICs" - that's interesting...What the last Intel CPU that was field programmable?
EspHack
I wouldn't be surprised, but mobo makers can enable it back right? I'm not exactly sure how the process goes, intel cant offer a bios for every mobo from every oem, all they can offer is a fixed code, which each manufacturer then fits into their bios, so intel would have to hide their malicious code inside it or something
framesx2x
Seriously people there is no way that you cloud fix this, This is IC level problem. A Microcode bios can’t fix or reprogram a processor, which is impossible do in a user level. the only way to fix this, Intel need to do a recall on these processor i7-6700k OR being to start reimbursing everyone who has this processor in hand. it sounds more like they made a mistake with the transistors layout.
Neo Cyrus
Of course they don't fix the problem through BIOS tweaks, they work around it if possible, hence why people are worried about a performance hit.
A mistake with the transistor layout? Why would you think that would be the cause of it crashing when doing calculations with a specific number?
Denial
You work around the bug. It's done every single processor generation.
http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/desktop-6th-gen-core-family-spec-update.pdf
Skylake has 53 known errata issues, 54 if you count this one. Haswell had 108:
http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/5th-gen-core-family-spec-update.pdf
I'm sure they'll end up finding more as Skylake matures. But regardless Intel says it has a fix. Yeah sure, if the Fix ends up crippling Skylake performance by a huge amount then Intel should recall the processor. But there is zero indication of that being the case.
Loobyluggs
That's binary for you. Everysoften someone tries to divide by zero, or calculate how many prime numbers there are.
Silly billy's
rflair
Moderator
What prime does it fail on 2 or 3?
dwiewolverine
stick with '-E' system..never doubt it:)
