After a Windows 10 Update Today Overclocking is lost. WTF Microsoft and Intel???

[Ahhzz]

There are still some benefits to waiting to see what breaks before moving to new advancements....

 

[Static~Charge]

If you are having a problem with that update, uninstall and block it.

Unfortunately, that won't have the desired effect. If that update installed new microcode for the processor, then uninstalling it won't revert the microcode back to the previous version unless a copy of the older microcode was stashed on your system (highly unlikely).

 

[cadaveca]

Real way to fix this is a BIOS update. Your BIOS doesn't like the MEI code, hence the OC problems. Sucks, but what you gonna do? Even my Surface got this update, and it works fine too. Complain to your board makers for not having proper BIOSes ready for this, because its not like they didn't know it was coming.

 

[xkm1948]

Real way to fix this is a BIOS update. Your BIOS doesn't like the MEI code, hence the OC problems. Sucks, but what you gonna do? Even my Surface got this update, and it works fine too. Complain to your board makers for not having proper BIOSes ready for this, because its not like they didn't know it was coming.

X99 is EOL for BIOS support

 

[cadaveca]

X99 is EOL for BIOS support

With Spectre/Meltdown/Foreshadow, it's not, really. That's what is causing this... a spectre/meltdown fix. I've seen both Gigabyte and ASRock releasing updates for X99, but yeah, you're right, this is a problem, and these updates via OS are supposed to counteract that, but the base BIOS code has to be ready for that, and I doubt that the last ASUS X99 update in April was good enough/ready. Gigabyte waited until June...

So those BIOS updates being EOL... are due to the board maker, not Intel or Microsoft...

 

[cucker tarlson]

Using broadwell-c and it's fine, but I use xtu for OC.Try it.

 

[xkm1948]

Looking up tutorials on WinRAID on editing microcode now. Hopefully i can make my own updated UEFI BIOS

 

[MrGenius]

You do that. I'm pretty sure that's not going to solve my problem though. I can still overclock with the BIOS I have from 12/27/13. Overclocking is not the problem. It's the microcode update in the OS that's causing performance loss. And the Spectre mitigation disabling techniques/registry hacks aren't "disabling" it. But I think I know how to do it another way. From what I just read the OS gets the microcode from c:\Windows\System32\mcupdate_GenuineIntel.dll or c:\Windows\System32\mcupdate_AuthenticAMD.dll. So you should be able to replace that file with an earlier version(pre-this fucked update). Which might work. I'm about to find out. Because I do have a an earlier copy of that file for my CPU.

EDIT: Also of note. I've tried 7 restore points dating back to a week and a half ago now. None of them work. The don't make it past the restoring the registry phase. Pretty sure I could still temporarily fix it with an in-place upgrade. Or obviously by using a backup of the entire system. But I can't see how either is more than a temporary solution. The next time it tries to update itself...right back to square one.

 

[Assimilator]

tl;dr If you're running a Broadwell-E system, don't install KB4100347 unless your motherboard BIOS is updated to the absolute latest Intel microcode.

If you've already installed the update and your motherboard doesn't have an updated BIOS, I'd strongly suggest editing your current BIOS to include the latest microcode, then flashing it.

You do that. I'm pretty sure that's not going to solve my problem though. I can still overclock with the BIOS I have from 2014. Overclocking is not the problem. It's the microcode update in the OS that's causing performance loss. And the Spectre mitigation disabling techniques/registry hacks aren't "disabling" it. But I think I know how to do it another way. From what I just read the OS gets the microcode from c:\Windows\System32\mcupdate_GenuineIntel.dll or c:\Windows\System32\mcupdate_AuthenticAMD.dll. So you should be able to replace that file with an earlier version(pre-this fucked update). Which might work. I'm about to find out. Because I do have a an earlier copy of that file for my CPU.

EDIT: Also of note. I've tried 7 restore points dating back to a week and a half ago now. None of them work. The don't make it past the restoring the registry phase. Pretty sure I could still temporarily fix it with an in-place upgrade. Or obviously by using a backup of the entire system. But I can't see how either is more than a temporary solution. The next time it tries to update itself...right back to square one.

You can't just arbitrarily copy old versions of files, Windows Resource Protection will detect that and replace what it sees as a "bad" file with the current one.

 

[RejZoR]

Haswell-E here. Everything seems to work fine. CPU is still overclocked and CPU-Z benchmark scores are the same pre and post update. I do have Spectre protected now (showing in InSpectre) which wasn't before when I downgraded BIOS.

 

[mkdr]

Wasnt the MC update until today optional? Just great...

Did a CrystalDiskMark and I lost about 50% in write speed on my Samsung 970 Evo 500GB. That is an additional loss to previous impacts.

23.08.2018
Sequential Read (Q= 32,T= 1) : 3511.899 MB/s
Sequential Write (Q= 32,T= 1) : 2493.400 MB/s
Random Read 4KiB (Q= 8,T= 8) : 1069.782 MB/s [ 261177.2 IOPS] <- this was already impacted by 30% before
Random Write 4KiB (Q= 8,T= 8) : 1605.116 MB/s [ 391874.0 IOPS]
Random Read 4KiB (Q= 32,T= 1) : 308.759 MB/s [ 75380.6 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 493.769 MB/s [ 120549.1 IOPS]
Random Read 4KiB (Q= 1,T= 1) : 38.656 MB/s [ 9437.5 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 94.370 MB/s [ 23039.6 IOPS]

Today
Sequential Read (Q= 32,T= 1) : 3439.877 MB/s
Sequential Write (Q= 32,T= 1) : 2492.474 MB/s
Random Read 4KiB (Q= 8,T= 8) : 1013.286 MB/s [ 247384.3 IOPS]
Random Write 4KiB (Q= 8,T= 8) : 876.680 MB/s [ 214033.2 IOPS]
Random Read 4KiB (Q= 32,T= 1) : 301.853 MB/s [ 73694.6 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 257.336 MB/s [ 62826.2 IOPS]
Random Read 4KiB (Q= 1,T= 1) : 40.152 MB/s [ 9802.7 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 95.270 MB/s [ 23259.3 IOPS]

 

[MrGenius]

You can't just arbitrarily copy old versions of files, Windows Resource Protection will detect that and replace what it sees as a "bad" file with the current one.

Well, not only have I done it before and had it work(replaced imageres.dll in System32 and SysWOW64 with a hacked version to remove the blue and yellow UAC shield icons from shortcuts), but I just did it again, and it worked again!!! So much for that theory!!! You are half-right though. If you run sfc /scannow the files will be detected as corrupt and replaced. The process is not automatic however.

So yes, as stated...PROBLEM SOLVED. All you need to do is replace mcupdate_GenuineIntel.dll in System32 with an earlier version of the file. The one installed by this update is dated 9/4/18. Which I replaced with the same file dated 4/11/18. Not as easy as it sounds. But totally doable.

First I made sure Spectre protection was disabled. Then copied mcupdate_GenuineIntel.dll from an earlier backup/clone of Windows 10 version 1803. Then booted WinPE with Active@ Boot Disk from a USB drive with a copy of the older version of mcupdate_GenuineIntel.dll. And used that copy to replace the updated version. Easy fricken peasy, lemon fricken squeezy.

Booted up the "updated" Windows 10 to see if it worked. Booted just fine. Noticed the 3-5 second lag in boot time since the update had instantly gone away. Ran a couple benches. FIXED!!! Performance loss GONE!!!

Here's a copy of mcupdate_GenuineIntel.dll from 4/11/18. I don't know what other CPUs it applies to. But if you have a 3770K, or any CPU with CPUID 306A9 I'd imagine, this is the one you need.

 

[xkm1948]

Well, not only have I done it before and had it work(replaced imageres.dll in System32 and SysWOW 64 with a hacked version to remove the blue and yellow UAC shield icons from shortcuts). I just did it again, and it worked again!!! So much for that theory!!! You are half-right though. If you run sfc /scannow the files will be detected as corrupt and replaced. The process is not automatic however.

So yes, as stated...PROBLEM SOLVED. All you need to do is replace mcupdate_GenuineIntel.dll in System32 with an earlier version of the file. The one installed by this update is dated 9/4/18. Which I replaced with the same file dated 4/11/18. Not as easy as it sounds. But totally doable.

First I made sure Spectre protection was disabled. Then copied mcupdate_GenuineIntel.dll from a backup of Windows 10 build 1803. Then booted WinPE with Active@ Boot Disk from a USB drive with a copy of the older version of mcupdate_GenuineIntel.dll. And used that copy to replace the updated version. Easy fricken peasy, lemon fricken squeezy.

Booted up the "updated" Windows 10 to see if it worked. Booted just fine. Noticed the 3-5 second lag in boot time since the update had instantly gone away. Ran a couple benches. FIXED!!! Performance loss GONE!!!

Here's a copy of mcupdate_GenuineIntel.dll from 4/11/18. I don't know what other CPUs it applies to. But if you have a 3770K this is the one you need.

Thank you!!!!

I will try to do this with your file first. That is IF I can get into Windows system first.

Do you think it is possible I can boot off a Linux interactive image from a thumb drive and replace the file using Linux?

 

[MrGenius]

Do you think it is possible I can boot off a Linux interactive image from a thumb drive and replace the file using Linux?

I've never done it. But it sounds like it should work.

 

[Ahhzz]

Thank you!!!!

I will try to do this with your file first. That is IF I can get into Windows system first.

Do you think it is possible I can boot off a Linux interactive image from a thumb drive and replace the file using Linux?

I've been able to in the past, as long as you have a recent enough version to recognize the drive properly. I still remember the first Win8.1 that I had to try to access with my older recovery tool disc...

 

[Punx223]

I have an X99 and X299... both with same KB installed, no issues with OC..

X99 = Strix X99 Gaming & 6950X

X299 = MSI Gaming M7 & 7980XE

I would try reflashing your BIOS... sounds like UEFI screwed the pooch somewhere.

 

[xkm1948]

I have an X99 and X299... both with same KB installed, no issues with OC..

X99 = Strix X99 Gaming & 6950X

X299 = MSI Gaming M7 & 7980XE

I would try reflashing your BIOS... sounds like UEFI screwed the pooch somewhere.

That would makes no sense. I have been using beta 3902 BIOS for a few months now and it has been perfect.

 

[cadaveca]

That would makes no sense. I have been using beta 3902 BIOS for a few months now and it has been perfect.

ASUS has this recurring problem where the BIOS shits the bed if your OC isn't quite stable, so its possible. I've had users report a stable OC not working, when pushing memory, reflash somehow fixes it. What other use is USB BIOS Flashback? Its like they know and then that's the fix...

Wasnt the MC update until today optional? Just great...

Did a CrystalDiskMark and I lost about 50% in write speed on my Samsung 970 Evo 500GB. That is an additional loss to previous impacts.

This is normal when you use a SSD. More data on it = slower drive. You must always test drive speeds with no data and fresh partition to get real potential speeds. You'd be amazed how slow some drives get when they get full.

 

[Punx223]

ASUS has this recurring problem where the BIOS shits the bed if your OC isn't quite stable, so its possible. I've had users report a stable OC not working, when pushing memory, reflash somehow fixes it. What other use is USB BIOS Flashback? Its like they know and then that's the fix...

What he said... I have had BIOS go wonky when overclocked for a long period... try to reflash man...

Why fight or dispute it? you have nothing to lose by trying..I mean worst case it still doesn't OC right?

 

[mkdr]

This is normal when you use a SSD. More data on it = slower drive. You must always test drive speeds with no data and fresh partition to get real potential speeds. You'd be amazed how slow some drives get when they get full.

No, not normal. Free disk space is identical on both tests, around 50GB used of 500GB. TRIM was also used before test.

 

[xkm1948]

OK got off work earlier and used flashback to redo 3902 BIOS. Nope, still BSOD looping. Gonna try to replace the file @MrGenius said earlier. Now I need to use my Mac to make a bootable Linux drive. All it takes is just one f*cking update to f*ck up everything.

 

[Deleted member 178884]

I've been screwed over by stuttering in windows - gotta fix that now, stupid ass nvidia modern driver screwed it up maybe.

 

[Solaris17]

OK got off work earlier and used flashback to redo 3902 BIOS. Nope, still BSOD looping. Gonna try to replace the file @MrGenius said earlier. Now I need to use my Mac to make a bootable Linux drive. All it takes is just one f*cking update to f*ck up everything.

huh? I thought it was inferred that re-flashing might fix your OC issue. Your OS being broken was due to the updates removal. Of course the flash wont fix it.

 

[xkm1948]

YESSS! Swapped the file with Mint on a USB drive and I can get back in system now. Thank you so much @MrGenius
Life Saver.

Backing up my important data now. I will deal with the code/update headache later. Whew.

 

[Solaris17]

YESSS! Swapped the file with Mint on a USB drive and I can get back in system now. Thank you so much @MrGenius
Life Saver.

Backing up my important data now. I will deal with the code/update headache later. Whew.

glad you got it working!