Skylake / Kaby Lake Hyper-threading bug

@Vasudev

I haven't check before running the install.bat so is there any changes?

 

Honestly, I don't know. You can uninstall it followed by a reboot then do a check on revision code and proceed to install the patch, okay.

 

It didn't change anything bruh.

Ok, successfully updated by following this guide:

http://forum.notebookreview.com/thr...-laptops-pc-fix-is-here.806451/#post-10560484

 

Now its fixed. Fire up Cinebench r15 using 24 threads, if you any kind of lockup on 4.1GHz post it here.

 

Bruh how do I set it to 24 threads?

 

File > Preferences > Tick Use custom threads with number of threads of your choice.

 

Lol just found out before you responded...firing up now.

@Vasudev here are the results...

41x @24 Threads



41x @12 Threads

 

Good results.

 

It looks like newer microcode for Skylake-X and Kabylake (also KL-X) chips are uploaded by intel.
For those you downloaded the earlier package just download this newer uCode from intel https://downloadcenter.intel.com/download/26925/Linux-Processor-Microcode-Data-File
@hmscott @THEBOSS619 @Papusan @iunlock @Mobius 1 @unclewebb @rinneh @DeeX @Prema
No passwords whatsoever.

 

4c/8t Core i7 can't run more than 8 threads simultaneously

 

Yes you can't but you can see more tiles rendering very fast instead of big tiles.

 

So far, uCode didn't do anything but I feel my PC is running on steroids. Dunno anyone noticed after the latest uCode?

 

The latest ucode includes the "placebo acceleration" update, very highly rated

 

It wasn't in the changelog, one thing is for sure now kabylake cpu are supported including HEDT versions.

 

Updating my guide while we are talking I was gonna post the solution today for kabylake users today but you are faster . Thank you!

I guess since the new microcode update got released... I guess my guide is useless now . but I will post your solution on my guide with only 1 step

 

Um are your processors really faster (the kaby lake ones) after the microcode update?

Seriously......?

 

as a skylake user.. yes I feel difference on smoothness... it became more smoother than before and I'm sure for Kabylake users will have the same effect

Note:-
Guide has been updated for both Skylake user and Kabylake users

Here:-
http://forum.notebookreview.com/thr...s-broken-ht-on-laptops-pc-fix-is-here.806451/

Thanks to @Vasudev for his easy one click solution

 

It would be nice to get some before / after microcode update CPU benchmark results to validate the feeling of improvement.

I think it's a great idea to keep up with your CPU errata fixes through regular Microcode updates, something most people miss out on because MS / vendors don't keep up with them for their customers, and computer owners haven't learned that they need to do this on their own.

But to tell people that they need to upgrade their microcode for performance improvement reasons needs to be backed up with performance benchmark results showing those improvements.

 

Those random stutters are gone but there's minor decrease in CB R15 scores, as usual 10 points lower than my consistent CB score of 682.

 

What were the stutters in that are now gone? That sounds valuable

There are many that complain about stutters with the Skylake / Kabylake CPU's on desktop and in games, some mitigated with OS fixes, some with Nvidia tuning for Optimus, but there are still persistant stuttering reports beyond that.

If there is a real fix for those remaining stuttering issues for Skylake and Kabylake, that would make a better case for issuing recommendations for people to update their microcode on a regular basis - or at least up to the point the stuttering ended.

When did the stuttering stop? With the newest update from July or the May issue?, or?

 

@ THEBOSS619 @hmscott @Vasudev

I ran Rise of The Tomb Raider benchmark today and I must say that it looks smoother now and the stuttering is gone...I'm not sure if it's the effect of the microcode update or because of the monitor currently overclocked to 110Hz?

 

Try backing off to the settings that you had when you last experienced stuttering, and see if it's still gone. Was that the only game where you saw stuttering?

 

Fired up ROTTR again, and it seems smooth/no stuttering even @60FPS with everything set to max, pure hair on (same settings as before having the stuttering issue). I haven't tested other games yet but since this is what I'm currently playing it's where I saw some stuttering during benchmarks and gameplay...especially in the Russian installation map where it's very noticeable.

EDIT:

I will try to revert back to the old microcode and test again...

 

Tested in a short benchmark as with wPRIME v1.5 running 32M ?

 

Its 6.8 sec on 8 threads

 

Try beat your older 6sec 546ms score with the new microcode

 

Like I said, it wasn't on Host Windows 10, I did it using WinPE, no bloat no frills. I might put this in my checklist for Windows 10 Insider feedbacks. One of two features came true but ending up being worse in fact, one being Game Mode and other being Windows Defender w/ Advanced Threat Control & Remediation.

 

where's the test?

 

Impatient? Maybe he is testing it extensively.

 

Well his post was like an hour ago, but fair enough. I'll wait.

 

@Papusan: Here you go, something's off, maybe I need to revert to older microcode, hey it looks like older microcode simply changed the revision code and latest one actually fixes HT issue with some performance loss. Actually reproducible.

I recommend you wait for @Prema's official BIOS. I don't know if @Prema observed it something is not right. Some kind of limit, I could say only 95% of CPU is used even though TS/HWINFO say 100% load. I think 5% loss could be because of error control mechanism double checking every instruction.

 

I applied the previous version, using the one step method, and it appeared that the ucode update version under HWInfo was 39. I downloaded the new version - still 39. So, I ran uninstall (admin, powershell) - still 39!. I did restart between each install or uninstall. My output from CPU-z is identical to that posted here - I cannot see where there is any change. Any thoughts? Thanks for your help, Joe

 

@Jdpurvis
Please turn off the computer, unplug the AC adapter, then unplug the battery connector from the mainboard (if possible). Most models will have an internal battery connector if you unscrew the bottom.
Then wait 30 seconds, reassemble then check the ucode! At the very least, hard AC power cycle the computer.

 

I had been forcing full shutdown with each reboot. However, I tried "one more time", and it succeeded. Thanks!! Joe

 

You have to unplug AC.
Windows 10 doesn't do full shutdowns anymore either unless you hold down shift key while clicking the shutdown selection. And all these newer laptops have fast UEFI boot.
What microcode revision is it now?

 

Fast boot can usually be disabled in the BIOS. Windows quick start can be turned off in the system settings part of the power plan control panel.

 

I was forcing full shutdown with each reboot by holding down the shift key. It didn't seem to take the first couple of times - even when I tried an uninstall and reinstall. However, it finally did take, and revision moved from 39 to BA, and it shows correctly in HWINFO and in RW-Everything, so I think it is correct. Many thanks! Best, Joe

 

Bro @Phoenix Where are yoo? Please, run a lots of bench. Then download this bug fix, install and run the same benchmarks all over again Maybe yoo already have installed this?

 

Got a skylake ucode 8A, no issues. Perhaps MS should provide microcode updates regularly. Their last one just from memory was like 2 years ago.


Sent from my iPhone using Tapatalk

 

I prefer to cut off it's head instead, and disable hibernation - Fast Startup and Fast Shutdown depend on it. Side benefit is you get back 1x memory size in storage space from hiberfil.sys hidden file on C partition.

Start As Administrator cmd window

powercfg -h off

Reboot, enjoy

 

Why would people bother with this fake shutdown today? Most part of people use ssd's and not old spinning disk in computers nowadays.

 

Hibernation, Fast Startup / Shutdown keeps data state alive even after power off, for easy "data forensics".

 

I'm sure M$ want your data Faster the better, bruh Or they can have access to your data with telemetry in fake shutdown

 

Not surprisingly, even with earlier Windows versions, I find if I disable the ethernet just before I shutdown, it shuts down much faster...

 

@hmscott @Papusan @Mr. Fox @iunlock @Mobius 1 @judal57 @rinneh @THEBOSS619 @unclewebb @Phoenix
I noticed strange CPU boost in frequency from 800Mhz to decrease in stock clocks by 1.92MHz.
I will show before and after uCode results on HWBOT prime:
Before:

After:

So the reason for increased system responsiveness is because of increased clocks even on locked mobile chips, bclk simply increased by 0.2-0.8 MHz. Also, CPU performance isn't consistent i.e in some apps the performance is quite good whereas in some cpu benchmark performance is average.
@Mr. Fox @Papusan and @Phoenix: You guys just be able to OC higher by 0.4-0.8MHz on KBL chips. The voltage spiked at 1.1v on HQ chips.

 

Does it bother your OCD when the blck isn't a perfect 100?

 

Partly yes. Then again for me blck was 99.76MHz now its 99.78 or 99.87MHz after the microcode update. Though with a small performance loss in some benchmarks.

 

It was already 1 AM during that time so I hit the sack.

@hmscott @Papusan @Vasudev

Anyways, I reverted back to the original microcode and ran some benchmark with ROTTR but it would seems that the micro-stuttering is back...also Cinebench R15 result is 892 (original microcode) but installing back to the newest microcode gives me 899 (highest I can get) in CB R15:



You guys should really try tests on your own too to confirm if this new microcode update really does something good...I'm also currently downloading 3DMark and going to run some more tests.

 

Try with speedshift in TS. No need for HWINFO temp sensors window

 

Will do!

How much for the min, max, and range?