News Intel processors affected by a Kernel Memory Bug!!!

Intel CPU have "Kernel Memory Leak" bug, which passes the data, without checking if it is secure and by doing that they are getting more perfromance.

Fix may decrease performance by 5-30%

Source

 

[MarcParis]

Guru3D benchmarked the pre and post benchmarks. Gaming performanve are afected only slightly.
http://www.guru3d.com/articles_pages/windows_vulnerability_cpu_meltdown_patch_benchmarked,1.html

Interesting result was on 3d mark timespy cpu result...patch is clearly impacting result.
So physics in game could be also impacted in game.
Let’s see..

 

[GuilleAcoustic]

All repetitive algorithms like physic calculation, encryption, compression, etc. will be subject to slowdowns.

From a programming perspective, this is more than likely.

 

[EdZ]

Interesting result was on 3d mark timespy cpu result...patch is clearly impacting result.

Yeah, but the performance improved post-patch.

My guess is that because the patch's exploit mitigation is on the 'edges' of syscalls (extra delay introduced when switching between user and kernel pages), part of the mitigation for the impact of that extra delay would be to batch syscalls to execute together when possible, so you only get hit by one set of switch penalties per batch rather than one set per op. Batching can be an inherent efficiency improvement (when batching is possible) and in something like Timespy where Futuremark made a benchmark to deliberately hammer draw calls and do other "low level API-y things" might be getting a boost just because it was deliberately inefficient to start with.

 

[MarcParis]

Yeah, but the performance improved post-patch.

My guess is that because the patch's exploit mitigation is on the 'edges' of syscalls (extra delay introduced when switching between user and kernel pages), part of the mitigation for the impact of that extra delay would be to batch syscalls to execute together when possible, so you only get hit by one set of switch penalties per batch rather than one set per op. Batching can be an inherent efficiency improvement (when batching is possible) and in something like Timespy where Futuremark made a benchmark to deliberately hammer draw calls and do other "low level API-y things" might be getting a boost just because it was deliberately inefficient to start with.

You are right i’ve mixed pre post on cpu timespy..

 

[AleksandarK]

Aaaaaaand let the games begin
https://arstechnica.com/gadgets/201...tion-lawsuits-regarding-meltdown-and-spectre/
Three lawsuits vs Intel

 

[MarcParis]

Here are some results including firmware update to counter specter.
No major impact on gaming and content creator softwares...but this time there is a big decrease on write speed on ssd (especially nvme ssd)...to be confirmed.

 

[bledha]

All repetitive algorithms like physic calculation, encryption, compression, etc. will be subject to slowdowns.

From a programming perspective, this is more than likely.

My wife is pretty bummed about just buying a relatively high end Intel equipped laptop for her MATLAB calculations which span over major topographic areas and sometimes take hours as it is - we are expecting to see even more time tacked onto that...not sure though. Looking into whether or not MathWorks has something released for Meltdown.

 

[AleksandarK]

My wife is pretty bummed about just buying a relatively high end Intel equipped laptop for her MATLAB calculations which span over major topographic areas and sometimes take hours as it is - we are expecting to see even more time tacked onto that...not sure though. Looking into whether or not MathWorks has something released for Meltdown.

I use MATLAB too with i3. Didnt notice anything past few days.

 

[bledha]

I use MATLAB too with i3. Didnt notice anything past few days.

Thanks @AleksandarK, for the input! Did you run any major programs or calculations? We have an older version of MATLAB running on a 7th Gen HQ i7. Do you know if the issue with this was solved OS side or should we be on the lookout for an update from MathWorks?

 

[EdZ]

No major impact on gaming and content creator softwares...but this time there is a big decrease on write speed on ssd (especially nvme ssd)...to be confirmed.

Looks like the 960 Evo that everyone is using to test is affected, while other SSDs can even see performance improvements, e.g. Optane. Could just be Samsung being slow on rolling out updates.

 

[AleksandarK]

Thanks @AleksandarK, for the input! Did you run any major programs or calculations? We have an older version of MATLAB running on a 7th Gen HQ i7. Do you know if the issue with this was solved OS side or should we be on the lookout for an update from MathWorks?

I just ran a few calculations for distacnce mesuring and a few logic simulations. Mostly simulation of wind turbine under different air directions. I stole that idea from their website lol.
I also did a fun project where i programed arduino based robot to learn to avoid objects. Not my original idea either. Jist saw a YT video and wanted to try to replicate it. I use matlab just for fun.

I have newest version running on 3rd gen i3(slooooooooooooooow). I didnt check for any updates either on OS side or program side, but i believe it wont be impacted too much

 

[jØrd]

IEEE has a good write up w/ some good links at the end here:
https://spectrum.ieee.org/tech-talk...processor-meltdown-vulnerability-was-thwarted

 

[AleksandarK]

Red Hat mesured the impact of the patches themself
https://www.servethehome.com/red-hat-outlines-meltdown-spectre-patch-performance-impacts/
Biggest reported impact is 19%.