Haswell Microcode 23 released. Spectre vulnerability patched, 9% slowdown :-(

Hi All,

Looks like Dell won't be officially releasing any updated system bioses for the M17x range nor the AW17 ranger to fix spectre http://www.dell.com/support/article...e-2017-5754---impact-on-dell-products?lang=en

Intel have released microcode 23 for Haswell, so I can include that in an update for the aw17 unlocked bios but early indications from the desktop platform is that we are likely to see a 9% slowdown and less overclocking potential :-(

http://www.overclock.net/t/1645289/...spectre-performance-and-stability-differences

Javers

 

Where can we get microcode updates for bios ? I have a GT72 from 2014 with 4th Gen core I7 and intend to mod my bios to include mitigation to Spectre.

 

It's not just a drop in the performance it's also throwing errors and instability if you read the OCN thread. I think they went too aggressive and did a botched job the microcode patch v23 on this situation as the fallout is exponentially growing + the 10nm fiasco, at CES zero info on the 10nm node and usual PR stuff. Considering how bad the CEO's stock sale news and the 11bn USD loss in share value & the future due to the Zen+ on track and the aggressive pricing advantage loss, how the CFL was pushed and how it's affecting them from the Class Action Lawsuits.

I suggest people not to go haywire on patching their systems left and right and end up getting BSODs and weird issues with this super urgent patch which requires understanding of the situation of the three CVEs (CVE-2017-5754 - Meltdown, CVE-2017-5753 and CVE-2017-5715 - Spectre) & How its related to you on Windows unless you don't know what you are doing while going OL or your browsing habits + your data security needs & how not so stable it is on the avg-consumer end. IF you cannot analyze the situation like here, go ahead and patch it with whatever you can.

This news has drawn a lot of speculation and articles before the embargo lift on Jan 9th. Which is today, one week before it this happened, now its catastrophic to Intel & everyone involved.

Today is patch Tuesday and there are issues already on that end with AMD platforms + M$ on screwing it up earlier with the blacklisting on the AVs & the Windows crashes now, M$ halting the updates on AMD platform & the corporate are affected badly too.

Plus Google and Mozilla are on track with updating their browsers for mitigating the risks by the end of this month, well Chrome update 64 is coming on Jan 23. And Nvidia too.

 

Microsoft Says Windows 10 PCs Running Haswell Or Older Intel CPUs Can Expect Significant Slowdowns Post-Spectre Patch-hothardware.com

Now for some good news and bad news. We'll get the good news out of the way first and tell you that Variant 1 and Variant 3 will have "minimal performance impact" for users. However, bad news comes with the revelation that the mitigation protocols put in place with Variant 2 can have a profound effect on system performance, especially for users running Haswell (or older) processors on Windows 10 and Windows Server customers (regardless of what processor being used).

• Windows 10 PCs with Skylake, Kaby Lake or anything newer may see "single-digit slowdowns", but for most users the impact will be minimal.
• Windows 10 PCs with Haswell or older processors will see "more significant slowdowns" and Microsoft notes that a segment of customers may "notice a decrease in system performance”.
• Windows 7 and Windows 8 PCs powered by Haswell or older processors will see a "decrease in system performance" for "most users".

As for customers running Windows Server, "Hold on to your butts":

Windows Server on any silicon, especially in any IO-intensive application, shows a more significant performance impact when you enable the mitigations to isolate untrusted code within a Windows Server instance. This is why you want to be careful to evaluate the risk of untrusted code for each Windows Server instance, and balance the security versus performance tradeoff for your environment.

 

I already patched my haswell to microcode v23 and had zero performance impact in gaming, gaming benchmarks, numerical simulations and running Linux virtual machine and running simulations on that VM.

 

And you get same 4K scores in ASSSD and CrystalDiskmark?

 

Believe me I hate Intel as much as the next guy ... But I am being objective, the numbers don't lie.

 

You say you have patched with new microcode and the patches from M$ Morons and still get exactly same performance in IO-intensive applications? Gaming and similar tasks won't show decrease in performance. If you have previous bench scores from mentioned software above... Maybe re-run the tests and compare results.

 

Already ran my Master thesis simulation that used to finish at best in 776.69 secs and it went up to 807.06 secs which is around 3% difference and this is within margin of normal computational variance. As for normal usage scenarios I have one cold boot 3Dmark CPU score higher than my average before the update.

EDIT: Quite honestly I am not sure if it can be considered as an IO operations intensive simulation, but for the usage scenarios I stated earlier, quoting myself " gaming, gaming benchmarks, numerical simulations and running Linux virtual machine and running simulations on that VM" I have no impact.

 

People stating errors even at stock so I am waiting a bit to see more feedback

 

As of yesterday Intel are now recommending everyone roll back to microcode v22 so fingers crossed for less performance drop / better reliability with the next microcode release.

https://newsroom.intel.com/wp-content/uploads/sites/11/2018/01/microcode-update-guidance.pdf

Javers