Windows 10: New Intel Microcode for Spectre V3a, V4 & L1TF Gets Released

#5611530

fantaskarsef

2018-11-28 17:19

Supporting Broadwell but not Haswell? F U Intel, F U so much.

#5611534

Fox2232

2018-11-28 17:37

It is that time of the year again... It's Bench Time!

#5611535

nevcairiel

2018-11-28 17:39

fantaskarsef:

Supporting Broadwell but not Haswell? F U Intel, F U so much.

Intel has released updates for L1TF and Spectre v3a all the way back to Nehalem. Microsoft usually rolls those out in waves however, since older systems may require more validation. You can always inquire with your motherboard vendor about a new BIOS with new Microcode if you need it in a more timely manner.

#5611541

scoter man1

2018-11-28 17:54

Better hope it doesn't wipe C:\

#5611561

Venix

2018-11-28 19:13

Any eta about the rollback before they get it right ? 😛

#5611562

Exodite

2018-11-28 19:14

With these patches generally not being deployed back to Sandy Bridge it sometimes feels like my 2600K is getting better and better, relatively speaking. :P

#5611564

tsunami231

2018-11-28 19:23

Eh I tired of this, by the time this all said and done, all the performance hits will add up

#5611568

yeeeman

2018-11-28 19:31

Let the "performance decreased significantly with the new ucode updates from Intel" news begin...

#5611579

Robbo9999

2018-11-28 19:42

I've got a 6700K Skylake CPU and it's been on the latest C6 Microcode since July this year (from motherboard BIOS), previous to that I was on C2 microcode which was the previous Spectre protected version released through Microsoft, and there has been zero performance hit with transferring to this latest C6 microcode - in fact it's marginally faster, but hardly, splitting hairs, but it's not slower. So I think folks will be fine with this latest microcode update through Windows 10 - the Spectre Variant 4 protection is not enabled by default so there's no performance loss. And I'm guessing that Spectre Variant 4 protection is not enabled by default because it's not deemed a high risk and also combined with the high performance hit if it were to be enabled.

#5611588

tsunami231

2018-11-28 20:06

Robbo9999:

I've got a 6700K Skylake CPU and it's been on the latest C6 Microcode since July this year (from motherboard BIOS), previous to that I was on C2 microcode which was the previous Spectre protected version released through Microsoft, and there has been zero performance hit with transferring to this latest C6 microcode - in fact it's marginally faster, but hardly, splitting hairs, but it's not slower. So I think folks will be fine with this latest microcode update through Windows 10 - the Spectre Variant 4 protection is not enabled by default so there's no performance loss. And I'm guessing that Spectre Variant 4 protection is not enabled by default because it's not deemed a high risk and also combined with the high performance hit if it were to be enabled.

last update my motherboard got was for C2 microcode that was back in earlier 2018 I dont expect there to be update anytime soon from asrock, then again I also dont care to do the updates via motherboard. I dont see or feel any performance hits though. Im sure they are there if you looking at benchmarks, but real world performance for me I dont see any hits

#5611591

schmidtbag

2018-11-28 20:20

If any of these patches did some of the same stuff that almost went through with the Linux 4.20 kernel, the performance results are going to scare some people. The performance hit was so bad that the patch was pulled and needed to be tweaked. To my understanding, if you want full security, you have to disable hyper threading.

#5611595

Robbo9999

2018-11-28 20:29

tsunami231:

last update my motherboard got was for C2 microcode that was back in earlier 2018 I dont expect there to be update anytime soon from asrock, then again I also dont care to do the updates via motherboard. I dont see or feel any performance hits though. Im sure they are there if you looking at benchmarks, but real world performance for me I dont see any hits

Yeah, there's definitely been a performance hit (in the region of 3%) if you compare my 6700K with zero Spectre protections versus where it's at now with C6 microcode, but there's been no loss for me from going from C2 to C6 microcode though, which is what this latest Microsoft patch is all about.

#5611599

Fender178

2018-11-28 20:40

tsunami231:

Eh I tired of this, by the time this all said and done, all the performance hits will add up

I am with you there. With my x58 xeon rig. I disabled the Spectre and Meltdown patches for better performance since it was causing issues when I was testing PS3 emulation with it enabled. I feel that with older rigs below 4th generation Core i series I would disable the patches for better performance because the performance hit is greater for the older CPUs.

schmidtbag:

If any of these patches did some of the same stuff that almost went through with the Linux 4.20 kernel, the performance results are going to scare some people. The performance hit was so bad that the patch was pulled and needed to be tweaked. To my understanding, if you want full security, you have to disable hyper threading.

Also to my understanding the performance hit effected Virtualization and Servers as well worse than games did. If what you are saying is true about disabling hyperthreading... I am wiling to bet that most of us will not do that because it will add only more to the performance hit. Especially with games that take advantage of the extra threads.

#5611615

schmidtbag

2018-11-28 21:45

Fender178:

Also to my understanding the performance hit effected Virtualization and Servers as well worse than games did. If what you are saying is true about disabling hyperthreading... I am wiling to bet that most of us will not do that because it will add only more to the performance hit. Especially with games that take advantage of the extra threads.

Yes, games for the most part don't really get affected, but, they're not that demanding of CPUs in the first place, especially if you have at least 4 physical cores. The main reason why disabling HT was suggested was actually because those patches were basically just doing exactly that, except through software rather than BIOS. Since that was so counter-productive, Linus and other kernel devs had the same argument as you, which is why they retracted the patch and instead adjusted it to be a little more "open minded" about what gets to use hyper-threaded threads. As a result, some tasks still have a performance hit but it's not as dire as it was just a week ago.

#5611616

HeavyHemi

2018-11-28 21:51

I guess I will be the test victim and see if this screws up overclocking on X99 for a bunch of users. I should be okay as my BIOS is already updated with this ucode version. Thanks EVGA! Edit.... KB4465065 is not compatible with your computer... Well running 18282.1000 probably explains that.... 😛

#5611621

lucidus

2018-11-28 22:31

HeavyHemi:

I guess I will be the test victim and see if this screws up overclocking on X99 for a bunch of users. I should be okay as my BIOS is already updated with this ucode version. Thanks EVGA! Edit.... KB4465065 is not compatible with your computer... Well running 18282.1000 probably explains that.... 😛

It might already be a part of the OS, double check with hwinfo or whatever you prefer.

#5611646

rl66

2018-11-29 00:22

fantaskarsef:

Supporting Broadwell but not Haswell? F U Intel, F U so much.

read it twice... and keep cool 🙂 it's not YET but they are working on it. Broadwell is on 1150 but it is more Skylake than Haswell, and so it seem more easy to release the patch with the 1st wave.

#5611654

Irenicus

2018-11-29 01:15

nevcairiel:

Intel has released updates for L1TF and Spectre v3a all the way back to Nehalem. Microsoft usually rolls those out in waves however, since older systems may require more validation. You can always inquire with your motherboard vendor about a new BIOS with new Microcode if you need it in a more timely manner.

Uhhhh correct me if I'm wrong, but I do believe BIOS microcode is different to OS microcode.

schmidtbag:

Yes, games for the most part don't really get affected, but, they're not that demanding of CPUs in the first place, especially if you have at least 4 physical cores.

What planet are you living on and what games do they play there? Games can be very demanding of CPU, even with hex cores and above. Maybe you only play old games? Regardless, you're incorrect

#5611656

TheDeeGee

2018-11-29 01:22

By the time they roll out Haswell i'm probably on Zen2.

nevcairiel:

Intel has released updates for L1TF and Spectre v3a all the way back to Nehalem. Microsoft usually rolls those out in waves however, since older systems may require more validation. You can always inquire with your motherboard vendor about a new BIOS with new Microcode if you need it in a more timely manner.

That's why i will never ever buy Gigabyte again, they havn't even bothered releasing Spectre BIOS for Z87.

#5611659

user1

2018-11-29 01:32

Irenicus:

Uhhhh correct me if I'm wrong, but I do believe BIOS microcode is different to OS microcode.

FYI the os microcode isn't different from the bios microcode, its just a binary blob that gets loaded , can be done by the bios or the os. you could insert this new microcode into the bios if you wanted too.