Update microcode for Spectre?

[geneo]

For work you will need this tool: https://www.win-raid.com/t154f16-Too...-quot-UBU.html
There is no performance loss with new microcodes, even with SSD.
Attachment 72058

upd. Asus is behind... Asrock has already begun to spread new bios http://www.asrock.com/support/index.asp?cat=BIOS

I have it too (0x84 with 8700k)and there is a large performance loss with SSD (960 Pro). I would check to make sure you do not have it disabled in the registry.

[pokuly]

For all wanting to test the Spectre performance hit i suggest to download InSpectre https://www.grc.com/inspectre.htm
It checks the patch status and you can disable the Spectre patch, reboot and bench before/after this way.

[RidingTheFlow]

For all wanting to test the Spectre performance hit i suggest to download InSpectre https://www.grc.com/inspectre.htm
It checks the patch status and you can disable the Spectre patch, reboot and bench before/after this way.

inSpectre simply checks the microcode version and assumes that this version (e.g. anything above 0x70) will protect you (along with required registry/OS patches).
It does not actually test that Spectre exploit working or not.

[pokuly]

inSpectre simply checks the microcode version and assumes that this version (e.g. anything above 0x70) will protect you (along with required registry/OS patches).
It does not actually test that Spectre exploit working or not.

It checks if windows is patched AND what the µC does. More can't be done. It will disable the windows part for testing. It shows 'not protected' if your windows status is not patched even if the new µC is there. If your i/o performance gets way better disabling the windows patch it is a good sign it is working,

[RidingTheFlow]

It checks if windows is patched AND what the µC does.

It does not check that uC is "patched". It simply relies on Intel-supplied information that uC version X should give protection against Spectre variants.
It does not check uC more than just checking version number.

More can't be done.

More can be done - just executing test for speculative execution attack scenario (against own test process) and checking if its effective should be sufficient.

[pokuly]

It does not check that uC is "patched". It simply relies on Intel-supplied information that uC version X should give protection against Spectre variants.
It does not check uC more than just checking version number.


More can be done - just executing test for speculative execution attack scenario (against own test process) and checking if its effective should be sufficient.

I still don't get what you are about and let it rest now.
It does check if the OS and Hardware is up-to-date regarding the official recent status.
For me this is also where to check we loose performance and if my install is right.
In theory i can imagine you have to check every single application against Spectre attacks but that is not the point of this thread.

[RidingTheFlow]

I still don't get what you are about and let it rest now.

I am about that having certain microcode version and OS version + registry key may not be enough to actually have *working* side-channel protection, due to complex interplay of different parts required for it to work and large possibility the single faulty part renders whole system non-functional.

Testing every single app is not needed, but at least a single controlled test that branch side channel attack is functioning is.

Due to inSpectre only testing "requirements present" part without actually having a single *functional* test for Spectre variants, may lead to cases where it gives false sense of security by reporting that system is protected, while its actually not.

That's why I was asking if anyone knows for such *functional* test app, instead of a microcode firmware version/OS update checker.

[jjxakergm]

Today a friend discovered the same problem as mine, but already on the Haswell.
0x24 for Haswell - the system shuts down a few seconds after the completion of the computational task.
In turn, I caught this bug on my laptop (Asus X552CL) with 0x1F for Ivy Bridge.
Let Intel update the servers in the Pentagon and the Ministry of Defense as soon as possible, then see what happens
Already many partners have updated their bios, in connection with the recommendations of Intel. I do not envy them...
Intel should be ashamed! let them go in the ass with their microcodes.

[pokuly]

Very interesting Ivy Bridge got the update. I switched from a z77 3570k to my recent system. I don't have it around anymore.
Would be really nice if someone could bench the before/after of such a system with Sofperfect Ramdisk 3.4.8 freeware, NTFS

[jjxakergm]

I do not know how from Ivy Bridge,but performance with Haswell and Kaby Lake increased in many scenarios. this is too high a price, stability suffers.
Clearly, there will be another re-release of microcodes for all processors.
Are already fed up, I hope they will make a mistake in the code. that will allow to remove restrictions on overclocking non-k. as it happened on Skylake.
If they have not already made such a mistake, will need to check periodically
So far only the Skylake non-k can overclocking on z170/z270, It is only necessary to lower the microcode of the processor to version 0x74. Intel allowed once a mistake, can repeat it again. I will not refuse