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

[remixedcat]

This is why the Windows Update Management Tool is such a good little program. It allows you to block updates just like previous Windows versions.

does it work on home editions? I found a tool like that but it won't work on home editions. My tablet has home. This is why I can't use the registry or the GPO.

 

[rtwjunkie]

does it work on home editions? I found a tool like that but it won't work on home editions. My tablet has home. This is why I can't use the registry or the GPO.

That I do not know the answer to at this moment. @Mussels has the link in his signature. Maybe he knows or the link has info. Probably not, though.

I can test it out on my HTPC this evening, since it is W10 home, and it hasn’t been turned on since this month’s updates came out.

 

[remixedcat]

That I do not know the answer to at this moment. @Mussels has the link in his signature. Maybe he knows or the link has info. Probably not, though.

I can test it out on my HTPC this evening, since it is W10 home, and it hasn’t been turned on since this month’s updates came out.

well the app works itself. le's see if keeping that service disabled works the trick. Hope so!!

 

[freeagent]

Well, I spoke too soon, I found out yesterday that my oc with ht is not stable. If I turn ht off then its fine. That blows. I swear it was good to go when I ditched 10 and put 7 back on, but the next day it was back to its shenanigans. Not sure what to say! Havent had a real good look at AMD in about 10 years.. Looks like I maybe picking one up soon. At least the guys running newer Intel hardware have some hope left.

Edit:

I've got 4.4 nailed down with ht, 4500 and up is pretty difficult, it seems like it needs more voltage, but when it gets it, it becomes unstable. So, I've got 4.4 nailed down with ht, and with ht off I can do 4700. Not the doom and gloom I thought. But its not all sunshine and lollipops either. My x5690 is purring away just another day.

 

[fullinfusion]

Download this stand alone tool and run it before windows updates.. http://download.microsoft.com/download/f/2/2/f22d5fdb-59cd-4275-8c95-1be17bf70b21/wushowhide.diagcab

Google how to use it properly.

Also you need to be aware the little tool will sometimes let M$ sneak a hidden update through so to have better control set your network connection to metered, that way it will prompt you about the update and you can then go and run the tool and block it again.. Trust me I have my system locked down from M$ bullshit firmware patches..

 

[agent_x007]

After update, interesting thing happend : AIDA64 cache benchmark isn't detecting OC... UNLESS Windows is put into "High Performance" power plan (I made 100% core clock as minimum) :

^"Balanced"

^"High Performance", OC was done from BIOS level (FYI).
I was using latest microcode for IB-E : 0x42D (thank you R-T-B !), before update got downloaded/installed.

 

[lexluthermiester]

After update, interesting thing happend : AIDA64 cache benchmark isn't detecting OC... UNLESS Windows is put into "High Performance" power plan (I made 100% core clock as minimum) :
View attachment 107337
^"Balanced"
View attachment 107338
^"High Performance", OC was done from BIOS level (FYI).
I was using latest microcode for IB-E : 0x42D (thank you R-T-B !), before update got downloaded/installed.

It's kinda sad that an OC is needed just to get back up to the level of performance it had before a software update.

 

[agent_x007]

WELL... I tested this on my new CPU
"RTX Off" :

"RTX On" :

Basicly, it's 75-100MHz perf. drop in my case (for Cinebench R15 multi score).
Interesting bit is that it made this CPU stable on a bit lower Vcore (which is good, when cooling high Vcore is a problem). I think reason for this is CPU does less stuff per cycle, so it's easier to stabilise.
Additional testing must be done to confirm that...

 

[hat]

lol @ RTX off/on

Though, again, these patches mostly impact performance in very specific applications. Cinebench isn't one of them.

 

[agent_x007]

Catch is, Ivy and earlier CPUs don't have the tech needed for low perf. impact of these patches (invpcid instruction if I remember correctly). That's why I get "Slower" perf. in InSpectre program, and why Cinebench score is indeed affected in my case (small impact of ~1,5%, but it's there).

 

[lexluthermiester]

Basicly, it's 75-100MHz perf. drop in my case (for Cinebench R15 multi score).
Interesting bit is that it made this CPU stable on a bit lower Vcore (which is good, when cooling high Vcore is a problem). I think reason for this is CPU does less stuff per cycle, so it's easier to stabilise.
Additional testing must be done to confirm that...

Have you also tested drive access performance?

 

[agent_x007]

Have you also tested drive access performance?

AS SSD : ~7% slower read time, and ~5% slower write time.

If anyone wants to test this :
To disable fix, use this command in cmd prompt (with admin rights) :

reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v FeatureSettingsOverride /t REG_DWORD /d 3 /f

reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v FeatureSettingsOverrideMask /t REG_DWORD /d 3 /f

To enable it, use this (cmd with admin is again required) :

reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v FeatureSettingsOverride /t REG_DWORD /d 0 /f

reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v FeatureSettingsOverrideMask /t REG_DWORD /d 3 /f

In both cases a RESTART is required for changes to take effect ("shuting down" and pressing power button, usually doesn't count as restart in Win 10).

 

[Ruru]

How windows can disable something what is set in bios/uefi? Windows should have no access to bios settings.

Pentium G3258.. Even with Z87 board it forced itself to stock clocks, had to delete the microcode update file manually.

I just installed this update with an i7-5820K on Asus ROG Strix X99 Gaming.

The update had no effect my overclock of 4.5ghz.

Same here, 5820K @ 4.5GHz & Rampage V Extreme.

 

[Zyll Goliat]

lol @ RTX off/on

Though, again, these patches mostly impact performance in very specific applications. Cinebench isn't one of them.

Yep...agree,I noticed that on my old X58 platform with updated microcode performance lost in cinebench is almost negligible but for example in 3d mark Firestrike physics test performance lost is around 10%.....
Here I just do the Firestrike testing to make sure what is the exact lost in performance

 

[cucker tarlson]

Contrary to what I wrote earlier,I do have some spectre/meltdown impact on my pc,so let me break it down.
My method of comparing it to older results was flawed,and since inspectre can enable or disable them at request,I could do some relevant testing.

so first let's start with cdm ssd perfromance,4K r/w is the only one I test.
on my adata 128gb: 30/85 with patch on, 34/118 patch off
on samsung 850s:32-34/88 patch on,35-39/122 patch off

difference is not noticeable on windows boot,but games do load faster and I'm not dreaming it.

watch dogs 2 - my go to game for cpu testing. note that what I achieved was in an extremely cpu heavy game with low settings and temporal filtering on,a scenario that is likely never actually used for gaming,but relevant for cpu testing

patch on: 100-102 fps patch off: 112-116 fps, 13% performance lost.

so great job intel,you done fucked up. I'll be running with patches off since I don't really care and there's nothing for me to lose except for performance,but AMD got themselves a new customer when 7nm Ryzen drops. New intel cpus like 9700k will probably take less (if any) of a hit since they're newer and bios support is better too, but my points stands unless they offer me a free replacement.
From e6300,through q6600,2500k,3570k,4790k to 5775c,thank you for your service intel,you are dismissed.Time to go back to AMD

 

[agent_x007]

Yep...agree,I noticed that on my old X58 platform with updated microcode performance lost in cinebench is almost negligible but for example in 3d mark Firestrike physics test performance lost is around 10%.....
Here I just do the Firestrike testing to make sure what is the exact lost in performance

View attachment 107405

I got +3,4% with my setup without Meltdown update active :


Comparison : https://www.3dmark.com/compare/fs/16451213/fs/16450669

 

[Zyll Goliat]

I got +3,4% with my setup without Meltdown update active :
View attachment 107409
View attachment 107410
Comparison : https://www.3dmark.com/compare/fs/16451213/fs/16450669

Well....then good seems like X79 handling better changes in microcode at least when it comes to this specific apps(FirestrikePhysics)

 

[Gripen90]

I have an MSI X99S Gaming 7 (BIOS from 25th of may 2016) and a Core i7 5960X@4.1GHz and it isn't affected by this update.
However my main rig with MSI X99S Krait Edition (BIOS from 7th of july 2016) with a Core i7 6900K@4.2GHz is affected by this update. Installing it and I'm down to 3.7GHz. I have blocked the windows update with the MS tool and the CPU is running 4.2GHz again.

 

[lexluthermiester]

AS SSD : ~7% slower read time, and ~5% slower write time.

Ok, not as some. I've seen as bad as 20% degradation.

 

[agent_x007]

Maybe updated microcode in UEFI makes drop less apparent ?
I did not checked how my disk behaves with Windows before I updated the UEFI...

 

[R-T-B]

Ok, not as some. I've seen as bad as 20% degradation.

Nehalem and chips on lga1366 seem hit the worst, which makes sense I suppose.

 

[Zyll Goliat]

 

[cadaveca]

After update, interesting thing happend : AIDA64 cache benchmark isn't detecting OC... UNLESS Windows is put into "High Performance" power plan (I made 100% core clock as minimum) :
View attachment 107337
^"Balanced"
View attachment 107338
^"High Performance", OC was done from BIOS level (FYI).
I was using latest microcode for IB-E : 0x42D (thank you R-T-B !), before update got downloaded/installed.

Oh, so power plans affect OC now... so the OS offer dynamic overclocking. Interesting. Funny how no one noticed this, and everyone thought things we just broken... and see to be ignoring the fact you said as much. I mentioned I wasn't having any problems myself...and I'm pretty sure I have high performance already selected. So this is all a bunch of drama for no reason...? Well, I guess it's not "no reason"... people simply didn't understand what actually changed. Myself included.

 

[hat]

You shouldn't need a high performance power plan to OC. If you do that you throw idle power saving features out the window. Most of us probably don't want our over clocked chips running north of 5GHz 24/7. Smells like a bug to me if that is indeed the case.

 

[cadaveca]

You shouldn't need a high performance power plan to OC. If you do that you throw idle power saving features out the window. Most of us probably don't want our over clocked chips running north of 5GHz 24/7. Smells like a bug to me if that is indeed the case.

Yeah, it does, but it might also explain why some are seeing issues, and others are not. It'd also be nice if there was simpler ways to change these power plans... but in the grand scheme of things, this is something I would like to see developed further.