Windows Patch KB4497165 - Nerfs Intel Compute
Posted: Thu Jun 04, 2020 1:03 am
Posting here at this is specific to certain Intel CPUs running various Windows variants - If there is a better home for this please feel free to move !!
There is a Windows Patch with new Intel microcode aimed at addressing a number of security threats which may well appear in your Windows Update list at some point … and you may want to be aware of the potential impact of installing it prior to doing so and make your own decision as to the cost/benefit of doing so
I am usually very supportive of auto updating systems as soon as patches arrive - all my kit does so by default - and I tend to keep my builds a bit ahead of the curve, so it is with a bit of irony that I am posting this "heads up" about https://support.microsoft.com/en-us/hel ... de-updates
Now the patch is for all the right reasons - information security is important for the most part - but not specifically on this hardware with its current software build as I am in an unusual position of actually not gaining much (if any) benefit from the patch as the build is totally dedicated to FAH and will be blown away/rebuilt when I want to use the server for another task … but I patched anyway out of habit
Two days later having confirmed exactly what the cost of the patch is from my logs that I keep for Beta Team duties I am uninstalling the patch as I type this
Consistently over the last 48hrs my 32/56 and 24/56 slots that I run which usually push out between 250k and 300k PPD have been pushing through less than 200k PPD … What I am seeing is a reduction in WUs throughput that reduces PPD by over 25%
… and Yes I have checked normal variances, yes I have compared WUs from same projects, yes I am sure the impact is this great, yes etc.
I will be doing some work to actually look into the raw processing hit on TPF so as to unpick the QRB and see how much it has actually slowed the compute down by ... various posts indicate possibly 10% or more - and it is by all accounts the compute intensive multi-threaded processes that are getting hit the worst so two Core_a7 (avx_256) on a twin 14core Xeon utilising all 56 threads may be close to a worst case scenario
Now please don't take this post as myself saying you shouldn't install/update this patch … There will be many situations where installing this is absolutely the right thing to do … I am simply pointing out that at least in my case there is a significant degradation of processing throughput - and that there may be people who like myself have the luxury of being able to not care about this level of security threat and may wish to consider carefully how much it might "impair" their system before installing.
Oh and for anyone running Intel CPUs who has been pulling their hair out trying to spot what has caused that noticeable drop in PPD - you may want to check if you received this patch about the time your reduced output started
For a non Microsoft perspective of the patch (imo actually quite well balanced) … https://meterpreter.org/microsoft-reiss ... indows-10/
There is a Windows Patch with new Intel microcode aimed at addressing a number of security threats which may well appear in your Windows Update list at some point … and you may want to be aware of the potential impact of installing it prior to doing so and make your own decision as to the cost/benefit of doing so
I am usually very supportive of auto updating systems as soon as patches arrive - all my kit does so by default - and I tend to keep my builds a bit ahead of the curve, so it is with a bit of irony that I am posting this "heads up" about https://support.microsoft.com/en-us/hel ... de-updates
Now the patch is for all the right reasons - information security is important for the most part - but not specifically on this hardware with its current software build as I am in an unusual position of actually not gaining much (if any) benefit from the patch as the build is totally dedicated to FAH and will be blown away/rebuilt when I want to use the server for another task … but I patched anyway out of habit
Two days later having confirmed exactly what the cost of the patch is from my logs that I keep for Beta Team duties I am uninstalling the patch as I type this
Consistently over the last 48hrs my 32/56 and 24/56 slots that I run which usually push out between 250k and 300k PPD have been pushing through less than 200k PPD … What I am seeing is a reduction in WUs throughput that reduces PPD by over 25%
… and Yes I have checked normal variances, yes I have compared WUs from same projects, yes I am sure the impact is this great, yes etc. I will be doing some work to actually look into the raw processing hit on TPF so as to unpick the QRB and see how much it has actually slowed the compute down by ... various posts indicate possibly 10% or more - and it is by all accounts the compute intensive multi-threaded processes that are getting hit the worst so two Core_a7 (avx_256) on a twin 14core Xeon utilising all 56 threads may be close to a worst case scenario
Now please don't take this post as myself saying you shouldn't install/update this patch … There will be many situations where installing this is absolutely the right thing to do … I am simply pointing out that at least in my case there is a significant degradation of processing throughput - and that there may be people who like myself have the luxury of being able to not care about this level of security threat and may wish to consider carefully how much it might "impair" their system before installing.
Oh and for anyone running Intel CPUs who has been pulling their hair out trying to spot what has caused that noticeable drop in PPD - you may want to check if you received this patch about the time your reduced output started
For a non Microsoft perspective of the patch (imo actually quite well balanced) … https://meterpreter.org/microsoft-reiss ... indows-10/
