Intel, AMD, VIA & Freescale CPU Microcode Repositories Discussion

Special Topics CPU Microcodes
441

Intel cpu50654_platB7_ver02006906_2020-04-24_PRD_B163D4FA
Intel cpu50657_platBF_ver05002F01_2020-04-23_PRD_8026BDDF

ucodes.rar (85.2 KB)

442

Intel cpu50656_platBF_ver04002F01_2020-04-23_PRD_FF889A84

cpu50656_platBF_ver04002F01_2020-04-23_PRD_FF889A84.rar (51.2 KB)

443

Nothing new in Windows 10 build 19640.

444

!New_cpu406E3_platC0_ver000000DC_2020-04-27_PRD_459D4483.bin
!New_cpu50653_plat97_ver01000157_2020-04-24_PRD_D44BC2EB.bin
!New_cpu506E3_plat36_ver000000DC_2020-04-27_PRD_65F0B5D1.bin
!New_cpu806E9_plat10_ver000000D6_2020-04-27_PRD_9BEC442B.bin
!New_cpu806E9_platC0_ver000000D6_2020-04-27_PRD_C979B5EA.bin
!New_cpu806EA_platC0_ver000000D6_2020-04-27_PRD_D9DD064F.bin
!New_cpu806EB_platD0_ver000000D6_2020-04-27_PRD_21854E7E.bin
!New_cpu806EC_plat94_ver000000D6_2020-04-23_PRD_5B2D4D96.bin
!New_cpu906E9_plat2A_ver000000D6_2020-04-23_PRD_741B08C6.bin

https://github.com/intel/Intel-Linux-Proā€¦rocode-20200609

20200609.zip (844 KB)

445

Special Register Buffer Data Sampling Advisory
-https://www.intel.com/content/www/us/en/ā€¦l-sa-00320.html

CVEID: CVE-2020-0543

Description: Incomplete cleanup from specific special register read operations in some Intel(R) Processors may allow an authenticated user to potentially enable information disclosure via local access.

CVSS Base Score: 6.5 Medium


HSW C0 6-3c-3/32 00000027->00000028 Core Gen4
BDW-U/Y E0/F0 6-3d-4/c0 0000002e->0000002f Core Gen5
HSW-U C0/D0 6-45-1/72 00000025->00000026 Core Gen4
HSW-H C0 6-46-1/32 0000001b->0000001c Core Gen4
BDW-H/E3 E0/G0 6-47-1/22 00000021->00000022 Core Gen5
SKL-U/Y D0 6-4e-3/c0 000000d6->000000dc Core Gen6 Mobile
SKL-U23e K1 6-4e-3/c0 000000d6->000000dc Core Gen6 Mobile
SKX-SP B1 6-55-3/97 01000151->01000157 Xeon Scalable
SKX-SP H0/M0/U0 6-55-4/b7 02000065->02006906 Xeon Scalable
SKX-D M1 6-55-4/b7 02000065->02006906 Xeon D-21xx
CLX-SP B0 6-55-6/bf 0400002c->04002f01 Xeon Scalable Gen2
CLX-SP B1 6-55-7/bf 0500002c->05002f01 Xeon Scalable Gen2
SKL-H/S R0/N0 6-5e-3/36 000000d6->000000dc Core Gen6; Xeon E3 v5
AML-Y22 H0 6-8e-9/10 000000ca->000000d6 Core Gen8 Mobile
KBL-U/Y H0 6-8e-9/c0 000000ca->000000d6 Core Gen7 Mobile
CFL-U43e D0 6-8e-a/c0 000000ca->000000d6 Core Gen8 Mobile
WHL-U W0 6-8e-b/d0 000000ca->000000d6 Core Gen8 Mobile
AML-Y42 V0 6-8e-c/94 000000ca->000000d6 Core Gen10 Mobile
CML-Y42 V0 6-8e-c/94 000000ca->000000d6 Core Gen10 Mobile
WHL-U V0 6-8e-c/94 000000ca->000000d6 Core Gen8 Mobile
KBL-G/H/S/E3 B0 6-9e-9/2a 000000ca->000000d6 Core Gen7; Xeon E3 v6
CFL-H/S/E3 U0 6-9e-a/22 000000ca->000000d6 Core Gen8 Desktop, Mobile, Xeon E
CFL-S B0 6-9e-b/02 000000ca->000000d6 Core Gen8
CFL-H/S P0 6-9e-c/22 000000ca->000000d6 Core Gen9
CFL-H R0 6-9e-d/22 000000ca->000000d6 Core Gen9 Mobile


@Marvin that was the reason for microcode 28 haswell finally!

446

On the LGA2066 (X299) platform with each new microcode, the performance collapses
I thought Intel had made an effort on that sideā€¦

447

https://software.intel.com/security-softā€¦r-data-sampling

So we finally know what bit 9 is:
RNGDS_MITG_DIS support is enumerated by CPUID.(EAX=07H,ECX=0).EDX[SRBDS_CTRL = 9]==1


448



What are your numbers to back that up? Running 7800X @ X299 myself, and i didnā€™t noticed any perf problems after inserting 0x2006906 into my bios via UBU.

449

@javanse

With 10980XE
Old microcode is better

450

Nothing new in Windows 10 build 19645.

451



What are your numbers to back that up? Running 7800X @ X299 myself, and i didnā€™t noticed any perf problems after inserting 0x2006906 into my bios via UBU.



Performance has fallen across everything since these garbage "security issues" started years ago. Voltage requirements also shot way up for the same settingsā€¦ +40mv is now +80mv, cute. Only once in a while a good mc comes like 0xCA for 8700k but then gets quickly pushed back by more garbage like 0xD2 and D6.

Intel is doing a fascinating jobā€¦ and no its not "cpu degradation", verified this multiple times by putting back old mc + equal version of windows at the time so its not overridden, back to the usual much lower voltage for the same settings.

452

For my 8700k D6 must be the Special Register Buffer Data Sampling and CrossTalk mitigation. The first time since a while i see flac encoding with my setting became faster after D2! From ~133x to ~138x speed. Before all mitigations it was above 150x speed.
No real benchmarks, sorry.

453

Iā€™ve stopped updating at B4 for my 8700K. The CA microcode is not good at all, and at default settings, for me and a few other MSI users, it would drop the max uncore/ring multi from 44x to 43x. I just cannot see how Intel thinks this sort of stuff is legitimate vulnerability patching, there have been so many holes by now that they should refund the CPU value IMO.
Oh, and the overvolt/undervolt settings might be related to other BIOS settings, not specifically the microcode. For example, MSI confirmed that they changed a ā€œCPU Lite Loadā€, i.e. they raised it so CPUs with worse silicon quality can overclock, at the price of a higher stock voltage. That setting alone, buried in the BIOS, made me have to go from -0.080V to -0.150V to achieve the same-ish under load voltages.
But, yeah, the new microcodes, at least for CFL-S, are quite the horror. B4 is OK and offers protection up to MDS or so, without much of an impact outside NVMe benchmarking.

454

Update from lastest Windows 10 Insider 19645.1 Build .

+ Icelake U/Y 706E5 Microsoft team rolled back from 00000046 microcode to 00000032 microcode.

455

Bugs are reported in the latest batch of updates (April to June 2020).

Test with care. Shambolic!

https://tracker.debian.org/news/1152789/ā€¦o-oldoldstable/

456

More Official Info here
1. https://software.intel.com/security-softā€¦oduct-cpu-model

@tistou77 @vmanuelgm

457



Interesting table, these models are marked retired and no new mc is expected for them:
Sandy Bridge: 206A7 206D6 206D7
Ivy Bridge: 306A9 306E4 306E7

So for this moment that means:
cpu206A7_plat12_ver0000002F_2019-02-17_PRD_8D15DF31.bin
cpu206D6_plat6D_ver00000621_2020-03-04_PRD_7ED51E79.bin
cpu206D7_plat6D_ver0000071A_2020-03-24_PRD_AC8779F8.bin
cpu306A9_plat12_ver00000021_2019-02-13_PRD_A8DF9213.bin
cpu306E4_platED_ver0000042E_2019-03-14_PRD_3C222CE1.bin
cpu306E7_platED_ver00000715_2019-03-14_PRD_AD89AC0D.bin

458

Yes, anything before Haswell (4th Gen) is EOL nowadays.

459

Sandy bridge is not affected by Special Register Buffer Data Sampling vulnerability. So it not need microcode anyway.

460

Warning regarding:

cpu406E3_platC0_ver000000DC_2020-04-27_PRD_459D4483
cpu506E3_plat36_ver000000DC_2020-04-27_PRD_65F0B5D1


Source: https://github.com/intel/Intel-Linux-Proā€¦mment-644885826