Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Starting next week your intel CPU will run 5 to 30% SLOWER...

  • Please log in to reply
21 replies to this topic

#1 DuncanM

DuncanM

    Gemini

  • -----
  • topic starter
  • Posts: 3,104
  • Joined: 03 Nov 2009
  • Loc: Arizona Sky Village or the rain forest

Posted 04 January 2018 - 03:08 AM

See:

 

http://www.cbc.ca/ne...essor-1.4471322

 

 

There's a security flaw in Intel processors and the software patch ( Windows, Linux, Apple) will have a big impact on performance. 



#2 Traveler

Traveler

    Soyuz

  • *****
  • Posts: 3,514
  • Joined: 19 Aug 2007
  • Loc: The Netherlands

Posted 04 January 2018 - 03:53 AM

...and no effect on AMD cpu's ...because the bug isn't in AMD cpu's.


  • happylimpet and tedbiv like this

#3 happylimpet

happylimpet

    Aurora

  • *****
  • Posts: 4,776
  • Joined: 29 Sep 2013
  • Loc: Southampton, UK

Posted 04 January 2018 - 04:14 AM

...and no effect on AMD cpu's ...because the bug isn't in AMD cpu's.

Hooray! I am AMD.



#4 SXBB

SXBB

    Vostok 1

  • -----
  • Posts: 160
  • Joined: 05 Jul 2015

Posted 04 January 2018 - 04:47 AM

Seems that it affects Macs less...



#5 JonnyBravo

JonnyBravo

    Messenger

  • -----
  • Posts: 440
  • Joined: 11 Apr 2017
  • Loc: Dededo, Guam

Posted 04 January 2018 - 06:17 AM

It says processors over the past decade... anyone seen a list they can share?


Sent from my iPhone using Tapatalk

#6 Bean614

Bean614

    Apollo

  • *****
  • Posts: 1,007
  • Joined: 05 Nov 2015
  • Loc: Mass.

Posted 04 January 2018 - 07:10 AM

"Seems that it affects Macs less..."...??? Not sure how you came up with that! If the CPU is an INTEL, it WILL affect your computer, period! Mac, PC, Linux, DIY, "the one your brother-in-law built for you", etc., etc., etc...... I read at least 3 different articles on 'ZDNet' last night, and while I found no 'list', per se, there's enough info out there now for you to determine if you're affected.

#7 Ishtim

Ishtim

    Surveyor 1

  • *****
  • Posts: 1,769
  • Joined: 10 Nov 2007
  • Loc: N. Alabama, USA

Posted 04 January 2018 - 07:43 AM

I started looking into this as well and found that MANY chips may be vulnerable, NOT just Intel... and the "slow down" claims pointed out by the OP is "exaggerated" according to Intel.

"Some researchers have claimed that any fixes could slow down computer systems, possibly by 30%, but Intel believes these claims are exaggerated."
 

"Recent reports that these exploits are caused by a “bug” or a “flaw” and are unique to Intel products are incorrect."

https://newsroom.int...earch-findings/

 

There are two separate security flaws, known as Meltdown and Spectre.
◾Meltdown affects laptops, desktop computers and internet servers with Intel chips.
◾Spectre potentially has a wider reach. It affects chips in smartphones, tablets and computers powered by Intel, ARM and AMD.

http://www.bbc.com/n...nology-42562303

 

 

I guess the PixInsight Batch Preprocessor COULD now take 5.2hrs instead of 4 if you patch your OS. laugh.gif


Edited by Ishtim, 04 January 2018 - 07:43 AM.

  • Oleg Astro, darkstar3d, bsavoie and 1 other like this

#8 Oleg Astro

Oleg Astro

    Astronomy Linux

  • -----
  • Vendors
  • Posts: 452
  • Joined: 12 Aug 2008
  • Loc: Albufeira, Portugal

Posted 04 January 2018 - 08:05 AM

Linux users can install already patched kernel v. 4.14.11.


  • nicknacknock likes this

#9 Oleg Astro

Oleg Astro

    Astronomy Linux

  • -----
  • Vendors
  • Posts: 452
  • Joined: 12 Aug 2008
  • Loc: Albufeira, Portugal

Posted 04 January 2018 - 08:23 AM

Special patch for AMD processors is here.



#10 OleCuss

OleCuss

    Mercury-Atlas

  • *****
  • Posts: 2,655
  • Joined: 22 Nov 2010

Posted 04 January 2018 - 08:58 AM

Yeah, as best I can tell the most easily fixed and most easily exploited flaw is the Meltdown.  We'll soon have it patched and most of us will be fine.  Apparently there is a related issue with AMD processors but the issue appears to be more easily remedied with a minor fix.

 

Spectre may be less easily exploited but may also take as much as a decade to really remedy.  Appears to require changing the architecture of the processors and a substantial redesign of the operating systems.

 

Not claiming to be a guru on any of this but I've read a number of articles on these problems and that is my current take on it.



#11 btschumy

btschumy

    Vendor - Our Galaxy

  • *****
  • Vendors
  • Posts: 3,577
  • Joined: 13 Apr 2004
  • Loc: Longmont, CO, USA

Posted 04 January 2018 - 10:21 AM

"Seems that it affects Macs less..."...??? Not sure how you came up with that! If the CPU is an INTEL, it WILL affect your computer, period! Mac, PC, Linux, DIY, "the one your brother-in-law built for you", etc., etc., etc...... I read at least 3 different articles on 'ZDNet' last night, and while I found no 'list', per se, there's enough info out there now for you to determine if you're affected.

Apple has already fixed the majority of the problem in the macOS 10.13.2 release.

 

"Early indications are that there are no notable slowdowns between a system running macOS High Sierra 10.13.1 and 10.13.2."

 

http://appleinsider....u-security-flaw

 

Of course, that does not mean that whatever changes they plan for 10.13.3 will not cause a slowdown, but at the present time there is no cause for concern.



#12 Traveler

Traveler

    Soyuz

  • *****
  • Posts: 3,514
  • Joined: 19 Aug 2007
  • Loc: The Netherlands

Posted 04 January 2018 - 11:07 AM

To make things clear:

the "fix" (including Mac, Windows and Linux) is a software solution (only) for a hardware issue for at least Intel CPU's....

 

For some history/background/sidenote info about Intel CPU issues:.

This new issue is not new for Intel. In the past Intel had also a hardware issue with the first gen.Pentium cpu's: see this link to FDIV bug



#13 entilza

entilza

    Soyuz

  • *****
  • Posts: 3,774
  • Joined: 06 Oct 2014
  • Loc: Oakville, ON, Canada

Posted 04 January 2018 - 03:06 PM

I'm really curious in the next few weeks how this rolls out!  



#14 lphilpot

lphilpot

    Apollo

  • *****
  • Posts: 1,451
  • Joined: 15 Oct 2005
  • Loc: Central Lousiana, USA

Posted 04 January 2018 - 10:07 PM

Apparently the potential for slowdowns is related to the nature of what the CPU is doing: Many switches between user and kernel mode, more slowdown, fewer switches, less slowdown. I don't know if that maps reasonably well to avoiding filesystem I/O until necessary, etc. The vulnerability is variable by how accurate (or not) speculative execution is. AMD is susceptible to Meltdown but unlikely to be affected by Spectre, I believe. I have both AMD and Intel at home.

 

 

> To make things clear: the "fix" (including Mac, Windows and Linux) is a software solution (only) for a hardware issue for at least Intel CPU's....

 

I'd go even further - It's a software bandaid... a (necessary) workaround, not a solution. And any time hardware ops get moved to software, performance will take a hit.



#15 Oleg Astro

Oleg Astro

    Astronomy Linux

  • -----
  • Vendors
  • Posts: 452
  • Joined: 12 Aug 2008
  • Loc: Albufeira, Portugal

Posted 05 January 2018 - 11:05 AM

...

I guess the PixInsight Batch Preprocessor COULD now take 5.2hrs instead of 4 if you patch your OS. laugh.gif

Look at the my results of PixInsight Benchmark for five Linux kernels on OS Ubuntu Astronomy 16.04 (64-bit):

 

kernel v. 4.4.1     Total: 9858 CPU: 8439 Swap: 32338

kernel v. 4.8.17   Total: 9843 CPU: 8441 Swap; 31396

kernel v. 4.12.14 Total: 9783 CPU: 8361 Swap: 32921

kernel v. 4.14.5   Total: 9786 CPU: 8366 Swap: 32775

 

and last "slow"

kernel v.4.14.11 Total: 9776 CPU: 8343 Swap: 33719

 

Yes, last kernel is more slow, but I don't see minus 5...30%!

 

==========================

kernel v. 4.14.5 vs kernel v.4.14.11

Total: 100% vs 99.898%

CPU: 100% vs 99.725%

Swap: 100% vs 102.880%

 


Edited by Oleg Astro, 05 January 2018 - 11:05 AM.

  • psandelle likes this

#16 entilza

entilza

    Soyuz

  • *****
  • Posts: 3,774
  • Joined: 06 Oct 2014
  • Loc: Oakville, ON, Canada

Posted 05 January 2018 - 11:09 AM

 

...

I guess the PixInsight Batch Preprocessor COULD now take 5.2hrs instead of 4 if you patch your OS. laugh.gif

Look at the my results of PixInsight Benchmark for five Linux kernels on OS Ubuntu Astronomy 16.04 (64-bit):

 

kernel v. 4.4.1     Total: 9858 CPU: 8439 Swap: 32338

kernel v. 4.8.17   Total: 9843 CPU: 8441 Swap; 31396

kernel v. 4.12.14 Total: 9783 CPU: 8361 Swap: 32921

kernel v. 4.14.5   Total: 9786 CPU: 8366 Swap: 32775

 

and last "slow"

kernel v.4.14.11 Total: 9776 CPU: 8343 Swap: 33719

 

Yes, last kernel is more slow, but I don't see minus 5...30%!

 

==========================

kernel v. 4.14.5 vs kernel v.4.14.11

Total: 100% vs 99.898%

CPU: 100% vs 99.725%

Swap: 100% vs 102.880%

 

 

Nice testing!  I've also been reading some computer sites stating it's not that noticeable, perhaps it was exaggerated.  However many of them are using quite fast processors.  Which processor were you using?


  • Oleg Astro likes this

#17 Oleg Astro

Oleg Astro

    Astronomy Linux

  • -----
  • Vendors
  • Posts: 452
  • Joined: 12 Aug 2008
  • Loc: Albufeira, Portugal

Posted 05 January 2018 - 11:24 AM

My other tests were make with y-cruncher v0.7.4.9478-static:

 

------------------------------------------------------------------------------------

Results for "normal" kernel v.Linux 4.14.5-041405-generic x86_64

------------------------------------------------------------------------------------

Constant:              Pi
Algorithm:             Chudnovsky Formula
Decimal Digits:        500,000,000
Hexadecimal Digits:    Disabled
Computation Mode:      Ram Only
Threading Mode:        Push Pool  ->  12 / ?  (randomization on)
Working Memory:        2,748,326,664 (2.56 GiB)
Total Memory:          2,871,790,728 (2.67 GiB)

Start Date:            Fri Jan  5 14:09:51 2018
End Date:              Fri Jan  5 14:11:35 2018

Total Computation Time:    95.085 seconds
Start-to-End Wall Time:    103.945 seconds

CPU Utilization:           1007.95 %  +  0.35 % kernel overhead
Multi-core Efficiency:     84.00 %  +  0.03 % kernel overhead

 

------------------------------------------------------------------------------------

Results for "slow" kernel v.Linux 4.14.11-041411-generic x86_64

------------------------------------------------------------------------------------

Constant:              Pi
Algorithm:             Chudnovsky Formula
Decimal Digits:        500,000,000
Hexadecimal Digits:    Disabled
Computation Mode:      Ram Only
Threading Mode:        Push Pool  ->  12 / ?  (randomization on)
Working Memory:        2,748,326,664 (2.56 GiB)
Total Memory:          2,871,790,728 (2.67 GiB)

Start Date:            Fri Jan  5 14:40:54 2018
End Date:              Fri Jan  5 14:42:49 2018

Total Computation Time:    104.611 seconds
Start-to-End Wall Time:    115.051 seconds

CPU Utilization:           900.69 %  +  0.39 % kernel overhead
Multi-core Efficiency:     75.06 %  +  0.03 % kernel overhead

------------------------------------------------------------------------------------

 

We have these:

 

Total Computation Time:    95.085 seconds vs 104.611 seconds = -10.02%
Start-to-End Wall Time:    103.945 seconds vs 115.051 seconds = -10.68%

 

So, "slow" kernel Linux 4.14.11 is really slower.



#18 Oleg Astro

Oleg Astro

    Astronomy Linux

  • -----
  • Vendors
  • Posts: 452
  • Joined: 12 Aug 2008
  • Loc: Albufeira, Portugal

Posted 05 January 2018 - 11:27 AM

 

 

...

I guess the PixInsight Batch Preprocessor COULD now take 5.2hrs instead of 4 if you patch your OS. laugh.gif

Look at the my results of PixInsight Benchmark for five Linux kernels on OS Ubuntu Astronomy 16.04 (64-bit):

 

kernel v. 4.4.1     Total: 9858 CPU: 8439 Swap: 32338

kernel v. 4.8.17   Total: 9843 CPU: 8441 Swap; 31396

kernel v. 4.12.14 Total: 9783 CPU: 8361 Swap: 32921

kernel v. 4.14.5   Total: 9786 CPU: 8366 Swap: 32775

 

and last "slow"

kernel v.4.14.11 Total: 9776 CPU: 8343 Swap: 33719

 

Yes, last kernel is more slow, but I don't see minus 5...30%!

 

==========================

kernel v. 4.14.5 vs kernel v.4.14.11

Total: 100% vs 99.898%

CPU: 100% vs 99.725%

Swap: 100% vs 102.880%

 

 

Nice testing!  I've also been reading some computer sites stating it's not that noticeable, perhaps it was exaggerated.  However many of them are using quite fast processors.  Which processor were you using?

 

CPU: Intel® Core™ i7-3930K

RAM: 64GB

Swap on RAID0 6x1TB.

 

OS Ubuntu Astronomy 16.04 (64-bit).



#19 entilza

entilza

    Soyuz

  • *****
  • Posts: 3,774
  • Joined: 06 Oct 2014
  • Loc: Oakville, ON, Canada

Posted 05 January 2018 - 12:05 PM

So it seems it should not be a big deal overall.  Perhaps the biggest issue could be for the last 20 years how many programs could have exploited this bug if it was already known to hackers out there :)


  • Oleg Astro and Steve Cox like this

#20 Jon Isaacs

Jon Isaacs

    ISS

  • *****
  • Posts: 85,162
  • Joined: 16 Jun 2004
  • Loc: San Diego and Boulevard, CA

Posted 05 January 2018 - 10:01 PM

So it seems it should not be a big deal overall.  Perhaps the biggest issue could be for the last 20 years how many programs could have exploited this bug if it was already known to hackers out there smile.gif

 

What do you mean no big deal?  For me, it means I have no chance of beating my old Solitaire speed records... :)

 

Jon


  • Traveler, psandelle, bobito and 1 other like this

#21 dmdouglass

dmdouglass

    Surveyor 1

  • *****
  • Posts: 1,924
  • Joined: 23 Dec 2007
  • Loc: Tempe, AZ

Posted 06 January 2018 - 10:36 AM

 

So it seems it should not be a big deal overall.  Perhaps the biggest issue could be for the last 20 years how many programs could have exploited this bug if it was already known to hackers out there smile.gif

 

What do you mean no big deal?  For me, it means I have no chance of beating my old Solitaire speed records... smile.gif

 

Jon

 

Jon...

 

Look at the bright side. This will be a "new beginning", and an opportunity to establish all new "records"....



#22 darkstar3d

darkstar3d

    Messenger

  • *****
  • Posts: 434
  • Joined: 11 Oct 2013
  • Loc: North Metro Atlanta

Posted 06 January 2018 - 01:37 PM

A customer of mine has all hands on deck patching systems. I’m a consultant or I’d be there as well


Sent from my iPhone using Tapatalk


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics