Windows: How to get latest CPU microcode without modding the BIOS

Discussion in 'Operating Systems' started by mbk1969, Jan 5, 2018.

  1. tsunami231

    tsunami231 Ancient Guru

    Messages:
    14,725
    Likes Received:
    1,854
    GPU:
    EVGA 1070Ti Black
    newer cpu would "see" less of hit is what was said by intel, seen reports of people with my cpu that say they have no hit other that said they do have hit. i keep puting off the updates cause i dont want to know if there is no hit my ocd will be fine, if the is and i see it my ocd will be triggered >>

    Which why atm i waiting for MS to push this update threw WU to do the test that would be much easier to revert
     
    Last edited: Mar 18, 2018
  2. EdKiefer

    EdKiefer Ancient Guru

    Messages:
    3,127
    Likes Received:
    394
    GPU:
    ASUS TUF 3060ti
    what CPU are you running now?
     
  3. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    Newer CPUs have a feature Process Context ID which can help in (Spectre?) mitigation which protects user space from kernel space by clearing whole TLB cache at any transition from user space to kernel one (and back), thus affecting all processes in system. The presence of PCID feature allows to clear only TLB cache belonging to current process (one transiting between spaces).

    But Google came with much more lightweight solution - retpoline, so for example Linux devs decided to use it instead of heavy TLB cache cleaning.
    (If I am not mistaken. I have read all this one time several months ago.)
     
    Dragondale13 and CaptaPraelium like this.
  4. CaptaPraelium

    CaptaPraelium Guest

    Messages:
    229
    Likes Received:
    63
    GPU:
    1070
    "best" for performance, but not for security. Just enable it all, and deal with the bad performance. It sucks. Deal with it.
     
    Dragondale13 likes this.

  5. EdKiefer

    EdKiefer Ancient Guru

    Messages:
    3,127
    Likes Received:
    394
    GPU:
    ASUS TUF 3060ti
    Yes, I read about the Google solution and was assuming Intel/MS would also adopt this method after they had problems with first stability wise.
     
  6. tsunami231

    tsunami231 Ancient Guru

    Messages:
    14,725
    Likes Received:
    1,854
    GPU:
    EVGA 1070Ti Black
    yah my process supports that, according to aida64 anyway, but it doesnt say its enabled

    HWinfo64

    I cant find anymention of it, I found line that say pcommit intructions ( same thing?) that says not present
     
    Last edited: Mar 19, 2018
  7. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
  8. tsunami231

    tsunami231 Ancient Guru

    Messages:
    14,725
    Likes Received:
    1,854
    GPU:
    EVGA 1070Ti Black
    yah your tool says
    Windows OS support for PCID performance optimization is enabled: True [not required for security]

    Which i assume means it "supported" which is what Aida64 says but Aida64 also says enabled next to things that are supported and enabled, supported is not same as enabled so question becomes what turns it on? is it some bios setting?

    I know intel trusted platform protection and SGW are disable in my bios which why those are not supported accord and reported as not supported in Aida64 and Hwinfo64
     
    Last edited: Mar 20, 2018
  9. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    No. BIOS has nothing to do with that. It is OS kernel thing. I don`t know how devs of Aida64 implemented this feature query. Since I only translated original Microsoft PowerShell script to C# (to build an executable app version) I can guess that MS devs know what to do. If they state "support is enabled" I tend to believe that the feature is used.
    Feature by itself is supported by patched version of Windows and is not supported by unpatched version of Windows.
    So patched Windows supports the PCID feature but feature is used (enabled) only in the presence of CPU with PCID.

    Update: I have checked the PowerShell script and I can say - it is not clear whether PCID feature is reported as being enabled or as being capable due to presence of proper hardware.
     
    Last edited: Mar 20, 2018
  10. CaptaPraelium

    CaptaPraelium Guest

    Messages:
    229
    Likes Received:
    63
    GPU:
    1070
    Apparently it refers to the presence of the hardware support, but I'm not so sure. We need to find a rather OLD CPU which says the hardware support is not there, and then run the powershell script and see if the results are different. We can use sysinternals coreinfo to find the PCID support presence in hardware
     

  11. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    I tend to believe that if PCID feature is present then it is used by patched Windows10. We don`t need to search OLD CPU because on mine CPU script clearly states "Windows OS support for PCID performance optimization is enabled: False".
     
  12. tsunami231

    tsunami231 Ancient Guru

    Messages:
    14,725
    Likes Received:
    1,854
    GPU:
    EVGA 1070Ti Black
    i my my old 8400e system my dad is using, say it PCID is false and that cpu sure dont support it, If i had my i7 920 it would say same thing i believe
     
  13. CaptaPraelium

    CaptaPraelium Guest

    Messages:
    229
    Likes Received:
    63
    GPU:
    1070
    Then you must have an old CPU because all the new ones support it.... Either that, or you don't have the patches installed for software support? And that's the point of checking an old cpu, with the patches installed. A process of elimination.

    Also, note that the process I described, was to check for the presence of PCID support in hardware (using coreinfo), THEN run the script. Not, to run the script to check for support since we don't know what the script is actually reporting; hardware, software, or both.
     
  14. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    Of course I have old CPU - IvyBridge-E 4930K. According to Wikipedia PCID feature was introduced in Westmere architecture in 2010 but not in all models. And according to this article Linux could utilize this feature back in 2012.

    As for the script and unpatched Windows, when WIndows is unpatched script doesn`t print "Windows OS support for PCID performance optimization is enabled:" at all.

    It can be that Windows kernel devs could avoid using that PCID feature due to not 100% coverage.
     
    CaptaPraelium likes this.
  15. dr_rus

    dr_rus Ancient Guru

    Messages:
    3,882
    Likes Received:
    1,015
    GPU:
    RTX 4090
    My Ivy Bridge machine is running rather badly after all the updates (Windows + new BIOS with updated microcode). I mean, there are slowdowns which can be seen when just doing stuff on the desktop which weren't there previously.
     

  16. Dragondale13

    Dragondale13 Ancient Guru

    Messages:
    1,527
    Likes Received:
    244
    GPU:
    GTX 1070 AMP! • H75
    I've had to drop back to my old OC of 4.5ghz at 1.36v, ddr3 2133 oc to 2400 to kind of make up for the loss.
    Overclock at 4.6ghz is no longer stable at 1.39v, guessing I'm needing at least 1.41v which is too much for my cooler H100i V2.
    It's new microcode "patches" so chances are the cpu now has to work just a lil bit harder at same frequencies, driving temps up.

    i7 3930k HT - On
    VTT, IMC - 1.2v
     
    Last edited: Mar 22, 2018
  17. EdKiefer

    EdKiefer Ancient Guru

    Messages:
    3,127
    Likes Received:
    394
    GPU:
    ASUS TUF 3060ti
  18. EdKiefer

    EdKiefer Ancient Guru

    Messages:
    3,127
    Likes Received:
    394
    GPU:
    ASUS TUF 3060ti
    Looking at your voltages, I think your 4.6 Vcore wasn't really enough, fully stable.
    I am no expert but "generally" once you get above 4.4ghz, each step needs "about" 0.040-0.050 extra.
    So I agree with your statement of needing around 1.41v if 1.36 was/is stable at 4.5.

    On memory, do you really see improvement other than in BM.I never really mess with memory other than CM=1.
    How loose did you have to make timings to get 2133 memory to 2400?
     
  19. Vidik

    Vidik Master Guru

    Messages:
    619
    Likes Received:
    220
    GPU:
    MSI 1070 Gaming Z
  20. Dragondale13

    Dragondale13 Ancient Guru

    Messages:
    1,527
    Likes Received:
    244
    GPU:
    GTX 1070 AMP! • H75
    Went messing with it again and I'm game benchmark, game, general use stable so far..at 4.6ghz 1.39v..2133mhz..CL9 - CR1
    Disabled C3 / C6, left C1E enabled, EIST disabled.
    VTT / IMC - 1.1v

    About the mem oc, no it makes very little difference in game benchmarks, in gaming...cpu bound areas in Watch Dogs 2 for example in the night with rain or fog is heavy on the system at 4k, ultra, textures very high, shadows low, headlight shadows - my car, distance/detail slider off...it can make those instances "more acceptable" vs 2133.

    Can't do 4.6ghz "and" oc mem, have to compromise.So far going with cpu oc seems best but I'll only know for sure with continued use.
    Ran Firestrike and got 17556..physics test 50 avg...combined 39 avg...but that could change after couple days...but it's what I was getting regularly pre-patch.
    Since patching and running at 4.5ghz 2400mhz CL10 CR2, it was 17138 - 17320 range..physics 48..combined 37.

    If I run memory at 2400mhz, timings I use 10-12-12-31-2T or 11-12-11-31-2T or 11-13-13-31-2T
    1.2v VTT / IMC
    Cas latency 10 is what I aim for.

    EDIT 1: Nope, FFXV and Stormblood benchmarks keep freezing with the system unresponsive.
    That's it 4.5ghz 2400mhz...will leave power states as is with 1.37v.

    EDIT 2: Ok I'm hard headed..now at 1.395v..topping out at full load 1.4v...just passed Stormblood bench at 4.6ghz.

    EDIT 3: Fail! Lol!!

    EDIT 4: Add 2400mhz at 11-11-11-30-2T best so far for gaming.WD2 running better than before the security patches and cpu at 4.6ghz 2133mhz CL 9 1T.
    Happy gamer! :) ....and secure. :cool:

    EDIT 5: Timings were unstable in WD2, ...TRCD couldn't remain on 11 :( guessing it's the board defaulting to it's own profile for 2400 which is 12, so with that in mind gave 10-12-12-36-2T a try..upped the TRAS to 36 as that's the "round trip" value and giving leeway to the memory not being 2400 sticks, 31 might have been too tight resulting in a drop in performance.
     
    Last edited: Mar 27, 2018

Share This Page