Clover General discussion

February 12, 2018

I see now you edit and this for me is a correct way to answer (also with a noob I can be here)

your statement about the equivalence is not correct at all

My cpu and also other "unsupported" cpu have all the steps it can allow from 12 to 37

in the past with null cpu kext only 3 steps 12 22 28

Power management is fine and time ago we have discussed with piker alpha to find a different solution because EIST enabled was fine but it has only that problem I have said before

Then, 500 point less in benchmark is correct because cpu without patch and with EISt bounch from x22 to x 25 states during load

without EIST when system is loaded it stays for all 44 Threads @x26 so if you do some calculation...

However thank you for your time

Yes, so as I said this patch is disabling speedstep, which is equivalent to NullCpuPowerManagement.kext, which is terrible and causes your CPU to run full speed all the time. Maybe a better solution would be to edit the p state data for the SMBIOS you are trying to use so that it reaches more and higher states since they are setup for the CPU models apple uses???

apianti · February 12, 2018

I see now you edit and this for me is a correct way to answer (also with a noob I can be here)

your statement about the equivalence is not correct at all

My cpu and also other "unsupported" cpu have all the steps it can allow from 12 to 37

in the past with null cpu kext only 3 steps 12 22 28

Power management is fine and time ago we have discussed with piker alpha to find a different solution because EIST enabled was fine but it has only that problem I have said before

Then, 500 point less in benchmark is correct because cpu without patch and with EISt bounch from x22 to x 25 states during load

without EIST when system is loaded it stays for all 44 Threads @x26 so if you do some calculation...

However thank you for your time

I did do the calculations and made a perfect equivalency. You are completely disabling speedstep so your CPU is running full speed all the time, 100%. It is never reducing speed or thermal overhead. NullCpu never gave you any steps, it is exactly that, no steps. It is bouncing around between those states during load because it knows how much energy it is using and the p states that have been configured for it say that at that energy level you need to be at multiplier state whatever or you are using too much energy. If you edit these states to represent the power energy level that your CPU needs for its states then they will happily be reached. Currently you are basically performing a constant burn-in test.

February 12, 2018

apianti, it is not correct your statement here "Currently you are basically performing a constant burn-in test"

With that patch I am forcing to achieve higher states when cpu need of them

otherwise it is not possible (many tries with frequency vector and other things done)

however thank you for your point of view

AppleIntelInfo.kext v2.9 Copyright © 2012-2017 Pike R. Alpha. All rights reserved.

Settings:
------------------------------------------
logMSRs..................................: 1
logIGPU..................................: 0
logCStates...............................: 1
logIPGStyle..............................: 1
InitialTSC...............................: 0xa62fd12f682 (519 MHz)
MWAIT C-States...........................: 8480

Processor Brandstring....................: Intel(R) Xeon(R) CPU E5-2696 v4 @ 2.20GHz

Processor Signature..................... : 0x406F1
------------------------------------------
 - Family............................... : 6
 - Stepping............................. : 1
 - Model................................ : 0x4F (79)

Model Specific Registers (MSRs)
------------------------------------------

MSR_IA32_PLATFORM_ID.............(0x17)  : 0x0
------------------------------------------
 - Processor Flags...................... : 0

MSR_CORE_THREAD_COUNT............(0x35)  : 0x16002C
------------------------------------------
 - Core Count........................... : 22
 - Thread Count......................... : 44

MSR_PLATFORM_INFO................(0xCE)  : 0x20080C3BF2811600
------------------------------------------
 - Maximum Non-Turbo Ratio.............. : 0x16 (2200 MHz)
 - Ratio Limit for Turbo Mode........... : 1 (programmable)
 - TDP Limit for Turbo Mode............. : 1 (programmable)
 - Low Power Mode Support............... : 1 (LPM supported)
 - Number of ConfigTDP Levels........... : 1 (additional TDP level(s) available)
 - Maximum Efficiency Ratio............. : 12
 - Minimum Operating Ratio.............. : 8

MSR_PMG_CST_CONFIG_CONTROL.......(0xE2)  : 0x8402
------------------------------------------
 - I/O MWAIT Redirection Enable......... : 1 (enabled, IO read of MSR(0xE4) mapped to MWAIT)
 - CFG Lock............................. : 1 (MSR locked until next reset)
 - C3 State Auto Demotion............... : 0 (disabled/unsupported)
 - C1 State Auto Demotion............... : 0 (disabled/unsupported)
 - C3 State Undemotion.................. : 0 (disabled/unsupported)
 - C1 State Undemotion.................. : 0 (disabled/unsupported)
 - Package C-State Auto Demotion........ : 0 (disabled/unsupported)
 - Package C-State Undemotion........... : 0 (disabled/unsupported)

MSR_PMG_IO_CAPTURE_BASE..........(0xE4)  : 0x10414
------------------------------------------
 - LVL_2 Base Address................... : 0x414
 - C-state Range........................ : 1 (C6 is the max C-State to include)

IA32_MPERF.......................(0xE7)  : 0x2D11193B58
IA32_APERF.......................(0xE8)  : 0x3F22E36604
MSR_0x150........................(0x150) : 0x1F00000000

MSR_FLEX_RATIO...................(0x194) : 0x0
------------------------------------------

MSR_IA32_PERF_STATUS.............(0x198) : 0x263500002500
------------------------------------------
 - Current Performance State Value...... : 0x2500 (3700 MHz)

MSR_IA32_PERF_CONTROL............(0x199) : 0xFF00
------------------------------------------
 - Target performance State Value....... : 0xFF00 (25500 MHz)
 - Intel Dynamic Acceleration........... : 0 (IDA engaged)

IA32_CLOCK_MODULATION............(0x19A) : 0x0

IA32_THERM_INTERRUPT.............(0x19B) : 0x0

IA32_THERM_STATUS................(0x19C) : 0x884F0000
------------------------------------------
 - Thermal Status....................... : 0
 - Thermal Log.......................... : 0
 - PROCHOT # or FORCEPR# event.......... : 0
 - PROCHOT # or FORCEPR# log............ : 0
 - Critical Temperature Status.......... : 0
 - Critical Temperature log............. : 0
 - Thermal Threshold #1 Status.......... : 0
 - Thermal Threshold #1 log............. : 0
 - Thermal Threshold #2 Status.......... : 0
 - Thermal Threshold #2 log............. : 0
 - Power Limitation Status.............. : 0
 - Power Limitation log................. : 0
 - Current Limit Status................. : 0
 - Current Limit log.................... : 0
 - Cross Domain Limit Status............ : 0
 - Cross Domain Limit log............... : 0
 - Digital Readout...................... : 79
 - Resolution in Degrees Celsius........ : 1
 - Reading Valid........................ : 1 (valid)

MSR_THERM2_CTL...................(0x19D) : 0x0

IA32_MISC_ENABLES................(0x1A0) : 0x850089
------------------------------------------
 - Fast-Strings......................... : 1 (enabled)
 - FOPCODE compatibility mode Enable.... : 0
 - Automatic Thermal Control Circuit.... : 1 (enabled)
 - Split-lock Disable................... : 0
 - Performance Monitoring............... : 1 (available)
 - Bus Lock On Cache Line Splits Disable : 0
 - Hardware prefetch Disable............ : 0
 - Processor Event Based Sampling....... : 0 (PEBS supported)
 - GV1/2 legacy Enable.................. : 0
 - Enhanced Intel SpeedStep Technology.. : 1 (enabled)
 - MONITOR FSM.......................... : 1 (MONITOR/MWAIT supported)
 - Adjacent sector prefetch Disable..... : 0
 - CFG Lock............................. : 0 (MSR not locked)
 - xTPR Message Disable................. : 1 (disabled)

MSR_TEMPERATURE_TARGET...........(0x1A2) : 0x640200
------------------------------------------
 - Turbo Attenuation Units.............. : 0 
 - Temperature Target................... : 100
 - TCC Activation Offset................ : 0

MSR_MISC_PWR_MGMT................(0x1AA) : 0x402000
------------------------------------------
 - EIST Hardware Coordination........... : 0 (hardware coordination enabled)
 - Energy/Performance Bias support...... : 1
 - Energy/Performance Bias.............. : 0 (disabled/MSR not visible to software)
 - Thermal Interrupt Coordination Enable : 1 (thermal interrupt routed to all cores)
 - SpeedShift Technology Enable......... : 0 (disabled)
 - SpeedShift Interrupt Coordination.... : 0 (disabled)
 - SpeedShift Energy Efficient Perf..... : 0 (disabled)
 - SpeedShift Technology Setup for HWP.. : No (not setup for HWP)

MSR_TURBO_RATIO_LIMIT............(0x1AD) : 0x1E1F202122232525
------------------------------------------
 - Maximum Ratio Limit for C01.......... : 25 (3700 MHz) 
 - Maximum Ratio Limit for C02.......... : 25 (3700 MHz) 
 - Maximum Ratio Limit for C03.......... : 23 (3500 MHz) 
 - Maximum Ratio Limit for C04.......... : 22 (3400 MHz) 
 - Maximum Ratio Limit for C05.......... : 21 (3300 MHz) 
 - Maximum Ratio Limit for C06.......... : 20 (3200 MHz) 
 - Maximum Ratio Limit for C07.......... : 1F (3100 MHz) 
 - Maximum Ratio Limit for C08.......... : 1E (3000 MHz) 

MSR_TURBO_RATIO_LIMIT1...........(0x1AE) : 0x1C1C1C1C1C1C1C1D
------------------------------------------
 - Maximum Ratio Limit for C09.......... : 1D (2900 MHz) 
 - Maximum Ratio Limit for C10.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C11.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C12.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C13.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C14.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C15.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C16.......... : 1C (2800 MHz) 

MSR_TURBO_RATIO_LIMIT2...........(0x1AF) : 0x1C1C1C1C1C1C1C1C
------------------------------------------
 - Maximum Ratio Limit for C17.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C18.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C19.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C20.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C21.......... : 1C (2800 MHz) 
 - Maximum Ratio Limit for C22.......... : 1C (2800 MHz) 

IA32_ENERGY_PERF_BIAS............(0x1B0) : 0x0

MSR_POWER_CTL....................(0x1FC) : 0x2904005B
------------------------------------------
 - Bi-Directional Processor Hot..........: 1 (enabled)
 - C1E Enable............................: 1 (enabled)

MSR_RAPL_POWER_UNIT..............(0x606) : 0xA0E03
------------------------------------------
 - Power Units.......................... : 3 (1/8 Watt)
 - Energy Status Units.................. : 14 (61 micro-Joules)
 - Time Units .......................... : 10 (976.6 micro-Seconds)

MSR_PKG_POWER_LIMIT..............(0x610) : 0x785A0001584B0
------------------------------------------
 - Package Power Limit #1............... : 150 Watt
 - Enable Power Limit #1................ : 1 (enabled)
 - Package Clamping Limitation #1....... : 1 (allow going below OS-requested P/T state during Time Window for Power Limit #1)
 - Time Window for Power Limit #1....... : 10 (2560 milli-Seconds)
 - Package Power Limit #2............... : 180 Watt
 - Enable Power Limit #2................ : 1 (enabled)
 - Package Clamping Limitation #2....... : 1 (allow going below OS-requested P/T state setting Time Window for Power Limit #2)
 - Time Window for Power Limit #2....... : 3 (20 milli-Seconds)
 - Lock................................. : 0 (MSR not locked)

MSR_PKG_ENERGY_STATUS............(0x611) : 0x68134D8D
------------------------------------------
 - Total Energy Consumed................ : 106573 Joules (Watt = Joules / seconds)

MSR_PKGC3_IRTL...................(0x60a) : 0x0
MSR_PKGC6_IRTL...................(0x60b) : 0x0
MSR_PKGC7_IRTL...................(0x60c) : 0x0
MSR_PKG_C2_RESIDENCY.............(0x60d) : 0x22D7C756F7A
MSR_PKG_C3_RESIDENCY.............(0x3f8) : 0x0
MSR_PKG_C2_RESIDENCY.............(0x60d) : 0x22D7C756F7A
MSR_PKG_C6_RESIDENCY.............(0x3f9) : 0x5194C0CC5CA
MSR_PKG_C7_RESIDENCY.............(0x3fa) : 0x0

IA32_TSC_DEADLINE................(0x6E0) : 0xA62FFC74C35

CPU Ratio Info:
------------------------------------------
Base Clock Frequency (BLCK)............. : 100 MHz
Maximum Efficiency Ratio/Frequency.......: 12 (1200 MHz)
Maximum non-Turbo Ratio/Frequency........: 22 (2200 MHz)
Maximum Turbo Ratio/Frequency............: 37 (3700 MHz)
P-State ratio * 100 = Frequency in MHz
------------------------------------------
CPU P-States [ (12) 30 32 ]
CPU C6-Cores [ 0 2 3 4 6 9 10 11 13 15 17 19 23 25 26 28 31 32 33 34 35 36 37 38 39 40 41 42 43 ]
CPU C6-Cores [ 0 1 2 3 4 5 6 8 9 10 11 12 13 15 16 17 19 20 22 23 25 26 27 28 30 31 32 33 34 35 36 37 38 39 40 41 42 43 ]
CPU P-States [ 12 30 32 (37) ]
CPU C6-Cores [ 0 1 2 3 4 5 6 8 9 10 11 12 13 14 15 16 17 19 20 21 22 23 25 26 27 28 30 31 32 33 34 35 36 37 38 39 40 41 42 43 ]
CPU P-States [ 12 30 32 33 (35) 37 ]
CPU C6-Cores [ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 25 26 27 28 30 31 32 33 34 35 36 37 38 39 40 41 42 43 ]
CPU C6-Cores [ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 ]
CPU P-States [ 12 30 31 32 33 35 (37) ]
CPU P-States [ 12 30 31 32 33 35 (36) 37 ]
CPU P-States [ 12 30 31 32 33 34 35 36 (37) ]
CPU P-States [ 12 26 30 31 32 33 34 35 36 (37) ]
CPU P-States [ (12) 26 28 30 31 32 33 34 35 36 37 ]
CPU P-States [ 12 25 26 28 30 31 32 33 34 35 36 (37) ]
CPU P-States [ 12 25 26 27 28 30 31 32 33 34 35 36 (37) ]
CPU P-States [ (12) 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 ]
CPU P-States [ (12) 22 23 24 24 25 26 27 28 29 30 31 32 33 34 35 36 37 ]
CPU P-States [ (12) 18 19 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 ]
CPU P-States [ (12) 15 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 ]
fabios-iMac-Pro:desktop fabio$

I did do the calculations and made a perfect equivalency. You are completely disabling speedstep so your CPU is running full speed all the time, 100%. It is never reducing speed or thermal overhead. NullCpu never gave you any steps, it is exactly that, no steps. It is bouncing around between those states during load because it knows how much energy it is using and the p states that have been configured for it say that at that energy level you need to be at multiplier state whatever or you are using too much energy. If you edit these states to represent the power energy level that your CPU needs for its states then they will happily be reached. Currently you are basically performing a constant burn-in test.

Edited February 12, 2018 by Guest

Trung_Nguyen · February 12, 2018

Oh god, no. I told you how to properly determine if there is a key stroke, inspect the key aggregator protocol.

So It kind of like this?:

Clover->Aggregator(eatin' keystrokes):GM, any keystrokes for t'day

Aggregator: yes->Clover invoke GUI/startloader with argument else start normal

or

Clover->Aggregator(eatin' keystrokes):GM, alt key pressed?

Aggregator: yes->Clover invoke GUI

boot.efi->Aggregator: receive key then parse args?

apianti · February 12, 2018

apianti, it is not correct your statement here "Currently you are basically performing a constant burn-in test"

With that patch I am forcing to achieve higher states when cpu need of them

otherwise it is not possible (many tries with frequency vector and other things done)

however thank you for your point of view

First, I can only go on the information that your have given me and you still have yet to say WHAT that patch is doing, or even where it came from. You said that you were only getting x26 now on all 44 threads. Under 100% from a multithreaded test, it is IMPOSSIBLE to have all cores at max turbo, this is a burn-in, you do not want this at all. The whole point of turbo is that it uses energy from threads/cores that are not activated or using all of the available thermal overhead and diverts that energy to a different thread/core to TEMPORARILY increase the performance state. If it was supposed to run at that speed then that would be the nominal frequency of the CPU and there would be no such thing or need as turbo states. ~~Just to give you an example, your CPU has a TDP of 85W but under full load during your test there, your CPU was well into 130W.~~

EDIT: Also your frequency is going well into 3GHz+ while not under load, what's up with that?

EDIT2: Nevermind. I see, I was looking at the v3 of that chip, apparently the v4 is 2.2Ghz and 3.2 GHz turbo, with 145W TDP. So either way it's still not correct.

EDIT3: It's impossible to have any turbo under 100% load on every thread/core. So what is this patch doing?

February 12, 2018

You are continuing to say not exact things I don't understand why

My cpu has a TDP of 150W and could achieve a maximum of 180 w

It is a Xeon 2696 V4

In osx during full load it reach a maximum of 140 with some stress test

In real life a maximum of 125w

In windows it pass over 150 W in stress test also, in some benchmark it reaches for all cores/thread (44) 2.8 Ghz (cinebencz) in OSX only 2,6Ghz

also my wattage usage in idle is about 10w or less (it depends from services I have loaded)

the point of my first message about this patches is:

Now, thanks to the finding of PMHeart we (all) can use previously named Okrasit Performances fix patch also in 10.13.4.beta

No more of this

Then you are saying or understanding other things

Ie:

"You said that you were only getting x26 now on all 44 threads"

Why?

this is not enough for you?

X 28 is the max I can achieve for all cores and due some limitation in osx or AVX benchmark test it is limited for all cores to X26

2 cores limited to x37 and so on till x28 for all cores

First, I can only go on the information that your have given me and you still have yet to say WHAT that patch is doing, or even where it came from. You said that you were only getting x26 now on all 44 threads. Under 100% from a multithreaded test, it is IMPOSSIBLE to have all cores at max turbo, this is a burn-in, you do not want this at all. The whole point of turbo is that it uses energy from threads/cores that are not activated or using all of the available thermal overhead and diverts that energy to a different thread/core to TEMPORARILY increase the performance state. If it was supposed to run at that speed then that would be the nominal frequency of the CPU and there would be no such thing or need as turbo states. Just to give you an example, your CPU has a TDP of 85W but under full load during your test there, your CPU was well into 130W.

Edited February 12, 2018 by Guest

apianti · February 12, 2018

So It kind of like this?:

Clover->Aggregator(eatin' keystrokes):GM, any keystrokes for t'day

Aggregator: yes->Clover invoke GUI/startloader with argument else start normal

or

Clover->Aggregator(eatin' keystrokes):GM, alt key pressed?

Aggregator: yes->Clover invoke GUI

boot.efi->Aggregator: receive key then parse args?

I don't understand this at all. Instead of discarding all the keystrokes by reading them all before loading drivers, after loading the drivers, locate the aggregator protocol and inspect it's key buffer..........

You are continuing to say not exact things I don't understand why

My cpu has a TDP of 150W and could achieve a maximum of 180 w

It is a Xeon 2696 V4

In osx during full load it reach a maximum of 140 with some stress test

In real life a maximum of 125w

In windows it pass over 150 W in stress test also, in some benchmark it reaches for all cores/thread (44) 2.8 Ghz (cinebencz) in OSX only 2,6Ghz

also my wattage usage in idle is about 10w or less (it depends from services I have loaded)

the point of my first message about this patches is:

Now, thanks to the finding of PMHeart we (all) can use previously named Okrasit Performances fix patch also in 10.13.4.beta

No more of this

Then you are saying or understanding other things

Ie:

"You said that you were only getting x26 now on all 44 threads"

Why?

this is not enough for you?

X 28 is the max I can achieve for all cores and due some limitation in osx or AVX benchmark test it is limited for all cores to X26

2 cores limited to x37 and so on till x28 for all cores

I am reading intel documents. Yes, I mistook that you had the v3, not v4, but either way, the numbers you are saying do not match up to what intel says. First the CPU is 2.2GHz so that is all the higher you should be hitting on a 100% multithreaded test, second 2.6GHz is not even one of the turbo multipliers the CPU can reach! They are 37/37/35/34/33/32/31/30/29/28.... And nominal is 22. How are you even reaching 26?

EDIT: DUDE, there is a thread on github with you and Pike participating, where he clearly tells you not to use that patch because it defeats the purpose of XCPM, as I was suspecting that it was.

February 12, 2018

Edit2 " is not correct

2696 V4 22 core/44 threads

TDP 150 w (145 is for 2699 V4)

max turbo is 3700Mhz for two cores (for all core/threads is 2800 Mhz)

this patch cuts the benefit od EIST enabled which in OSX produce a lower frequency for Higher load states and frequency with EIST bounches from 2.2Ghz to 2.4Ghz or so during load test

with this patch during load (only during load test ) frequency is locked to 2600 GHz

then when load is finished system have lower frequency as steps as usually

---

EDIT: Also your frequency is going well into 3GHz+ while not under load, what's up with that?

EDIT2: Nevermind. I see, I was looking at the v3 of that chip, apparently the v4 is 2.2Ghz and 3.2 GHz turbo, with 145W TDP. So either way it's still not correct.

EDIT3: It's impossible to have any turbo under 100% load on every thread/core. So what is this patch doing?

I see now you have studied (thank you for dude)

Pike says this not having as you a similar cpu to test

https://ark.intel.com/products/91317/Intel-Xeon-Processor-E5-2699-v4-55M-Cache-2_20-GHz

so you can study better (mine is a OEM version of 2699 V4 with some different specs (150 w and x37 for max turbo instead of 145 and x36)

forgive me but it is very ridiculous to follow how you edit your posts to find some prove I am saying {censored}

If you read and understand well that thread first EIST was advised to be disabled..then after all my testing on my system Pike advice to enabled it because from appleintelinfo output data it was supported well

And the I have said of bad performance in OSX with EIST enabled

thats all

I don't understand this at all. Instead of discarding all the keystrokes by reading them all before loading drivers, after loading the drivers, locate the aggregator protocol and inspect it's key buffer..........

I am reading intel documents. Yes, I mistook that you had the v3, not v4, but either way, the numbers you are saying do not match up to what intel says. First the CPU is 2.2GHz so that is all the higher you should be hitting on a 100% multithreaded test, second 2.6GHz is not even one of the turbo multipliers the CPU can reach! They are 37/37/35/34/33/32/31/30/29/28.... And nominal is 22. How are you even reaching 26?

EDIT: DUDE, there is a thread on github with you and Pike participating, where he clearly tells you not to use that patch because it defeats the purpose of XCPM, as I was suspecting that it was.

apianti · February 12, 2018

Edit2 " is not correct

2696 V4 22 core/44 threads

TDP 150 w (145 is for 2699 V4)

max turbo is 3700Mhz for two cores (for all core/threads is 2800 Mhz)

this patch cuts the benefit od EIST enabled which in OSX produce a lower frequency for Higher load states and frequency with EIST bounches from 2.2Ghz to 2.4Ghz or so during load test

with this patch during load (only during load test ) frequency is locked to 2600 GHz

then when load is finished system have lower frequency as steps as usually

Actually that is EIST working exactly as it should. It should be bouncing around the nominal frequency when around 100% multithreaded load, there's not much left over energy to enter turbo states in this situation. And I made a typo, I meant 3.7GHz, evidenced by my following comment where I wrote out the turbos.

Trung_Nguyen · February 12, 2018

I don't understand this at all. Instead of discarding all the keystrokes by reading them all before loading drivers, after loading the drivers, locate the aggregator protocol and inspect it's key buffer..........

I am reading intel documents. Yes, I mistook that you had the v3, not v4, but either way, the numbers you are saying do not match up to what intel says. First the CPU is 2.2GHz so that is all the higher you should be hitting on a 100% multithreaded test, second 2.6GHz is not even one of the turbo multipliers the CPU can reach! They are 37/37/35/34/33/32/31/30/29/28.... And nominal is 22. How are you even reaching 26?

EDIT: DUDE, there is a thread on github with you and Pike participating, where he clearly tells you not to use that patch because it defeats the purpose of XCPM, as I was suspecting that it was.

Do we need to handle apple boot key combo, or just check if there is keystrokes to invoke the GUI?

I meant if boot.efi's job including read from the protocol?

apianti · February 12, 2018

Do we need to handle apple boot key combo, or just check if there is keystrokes to invoke the GUI?

I meant if boot.efi's job including read from the protocol?

Well should the GUI start if you know the key combination is meant for boot.efi? Probably not. It should probably just boot, but that's only if boot.efi is the selected bootloader, otherwise it should enter the GUI because those key combinations don't mean anything for other OSes and you can get to all that stuff from the GUI in that case.

I see now you have studied (thank you for dude)

Pike says this not having as you a similar cpu to test

https://ark.intel.com/products/91317/Intel-Xeon-Processor-E5-2699-v4-55M-Cache-2_20-GHz

so you can study better (mine is a OEM version of 2699 V4 with some different specs (150 w and x37 for max turbo instead of 145 and x36)

forgive me but it is very ridiculous to follow how you edit your posts to find some prove I am saying {censored}

If you read and understand well that thread first EIST was advised to be disabled..then after all my testing on my system Pike advice to enabled it because from appleintelinfo output data it was supported well

And the I have said of bad performance in OSX with EIST enabled

thats all

I am looking at an Intel document for Xeon E5-2696 v4. I edited my posts before you had responded. I am just still trying to figure out why you need this patch when you could change the p state information. Maybe I just don't understand sacrificing a ton of power management for a small increase in output.

February 12, 2018

first three benchmark are with patch enabled (or without EIST enabled in bios)

last two benchmark with EIST enabled in bios and without patch in clover

edit

if you can share Intel 2696 V4 document I will be grateful (no need for 2699 V4 but exactly mine)

I am looking at an Intel document for Xeon E5-2696 v4. I edited my posts before you had responded. I am just still trying to figure out why you need this patch when you could change the p state information. Maybe I just don't understand sacrificing a ton of power management for a small increase in output.

znmacc · February 12, 2018

@slice @PMHeart

I suggest to optimize Chinese descriptions in Clover Install GUI such as Driver64UEFI.

If descriptions have some mistake please fix it.

Thanks.

AptioInputFix

Vit9696编写的针对FileVault2启动界面的键盘和鼠标输入设备的支持

AptioMemoryFix

Vit9696自OsxAptioFix2驱动优化而来的新驱动，支持更多新特性，例如：

自动为boot.efi寻找最适合的内存地址，避免启动错误

当slide值不能被使用的时候提供对KASLR的支持

新增系统在低内存地址下的安全模式的支持

确保系统不会出现slide溢出的问题

尝试修复更多的内存分布问题

支持硬件NVRAM

优化一些休眠的问题（暂时不稳定）

（请勿与其余内存修复驱动同时使用）

CsmVideoDxe-64

提供对CSM模块的支持，某些主板若要安装非UEFI系统或者非安全启动的系统，例如Windows7，Linux等设备，需要开启CSM模块，此时需要加入该驱动以修复CloverGUI的显示问题

Clover推荐关闭CSM模块启用原生GOP显示模块

EmuVariableUefi-64

macOS使用NVRAM存储一些系统变量，大部分的UEFI主板在配合合适的Aptio驱动后支持原生硬件NVRAM，但是少部分主板不支持NVRAM或者NVRAM的支持有问题，此时建议加入该驱动，该驱动通过开机时加载位于EFI分区内的nvram.plist内容到nvram中，以模拟NVRAM支持

需要注意的是，是用此驱动，需要勾选“安装RC scripts到目标磁区”选项才有效

Fat-64

可选择的64位FAT文件系统的支持

OsxAptioFixDrv-64

Dmazar编写的针对UEFI固件的内存问题修复的驱动，对休眠支持不完善

（请勿与其余内存修复驱动同时使用）

OsxAptioFix2Drv-64

Dmazar编写的针对UEFI固件的内存问题修复的驱动，在1代基础之上完善了休眠等高级功能的支持，部分机型需要手动设置slide值

（请勿与其余内存修复驱动同时使用）

OsxAptioFix3Drv-64

Vit9696等作者在OsxAptioFix2Drv-64的基础之上进行了优化，修复了大多数新设备的NVRAM支持，该驱动在部分机型依然需要手动设置slide值

（请勿与其余内存修复驱动同时使用）

OsxFatBinaryDrv-64

可选择的64位FAT文件系统的支持

OsxLowMemFixDrv-64

针对UEFI固件内存问题修复的驱动简化版

（请勿与其余内存修复驱动同时使用）

PartitionDxe-64

支持非常态的分区配置，如苹果混合分区或Apple分区图。

UsbkbDxe-64

FileVault2启动界面的键盘输入设备的支持

UsbMouseDxe-64

FileVault2启动界面的鼠标输入设备的支持

Hi,

請問如果之前安裝了OsxAptioFix2Drv-64 我可以直接update lover時更換為第三版OsxAptioFix3Drv-64？有何好處？

apianti · February 12, 2018

first three benchmark are with patch enabled (or without EIST enabled in bios)

last two benchmark with EIST enabled in bios and without patch in clover

Please give a geekbench result.

EDIT: ~~I don't see how you could be getting any states when this is the status of the CPUs configuration~~, EIST is clearly disabled:

EDIT2: I see now, it's just disabling EIST and HWP, so basically just using speedstep always to the complete maximum..... Which means its not using XCPM.

EDIT3: I never explain enough, I don't think and then it's confusing or something. Basically, this patch appears to always set the target performance state to the maximum value which means that nothing is really working because it's not changing that value to get better performance/power ratio, it's just going all out performance. So even though EIST is enabled, it's really not, if that value is never changed then it will always choose the highest state available, not the most appropriate.

MSR_IA32_PERF_CONTROL............(0x199) : 0xFF00
------------------------------------------
 - Target performance State Value....... : 0xFF00 (25500 MHz)
 - Intel Dynamic Acceleration........... : 0 (IDA engaged)

...

MSR_MISC_PWR_MGMT................(0x1AA) : 0x402000
------------------------------------------
- EIST Hardware Coordination........... : 0 (hardware coordination enabled)
- Energy/Performance Bias support...... : 1
- Energy/Performance Bias.............. : 0 (disabled/MSR not visible to software)
- Thermal Interrupt Coordination Enable : 1 (thermal interrupt routed to all cores)
- SpeedShift Technology Enable......... : 0 (disabled)
- SpeedShift Interrupt Coordination.... : 0 (disabled)
- SpeedShift Energy Efficient Perf..... : 0 (disabled)
- SpeedShift Technology Setup for HWP.. : No (not setup for HWP)

Here are your turbos:

MSR_TURBO_RATIO_LIMIT............(0x1AD) : 0x1E1F202122232525
------------------------------------------
- Maximum Ratio Limit for C01.......... : 25 (3700 MHz) 
- Maximum Ratio Limit for C02.......... : 25 (3700 MHz) 
- Maximum Ratio Limit for C03.......... : 23 (3500 MHz) 
- Maximum Ratio Limit for C04.......... : 22 (3400 MHz) 
- Maximum Ratio Limit for C05.......... : 21 (3300 MHz) 
- Maximum Ratio Limit for C06.......... : 20 (3200 MHz) 
- Maximum Ratio Limit for C07.......... : 1F (3100 MHz) 
- Maximum Ratio Limit for C08.......... : 1E (3000 MHz)

MSR_TURBO_RATIO_LIMIT1...........(0x1AE) : 0x1C1C1C1C1C1C1C1D
------------------------------------------
- Maximum Ratio Limit for C09.......... : 1D (2900 MHz) 
- Maximum Ratio Limit for C10.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C11.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C12.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C13.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C14.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C15.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C16.......... : 1C (2800 MHz)

MSR_TURBO_RATIO_LIMIT2...........(0x1AF) : 0x1C1C1C1C1C1C1C1C
------------------------------------------
- Maximum Ratio Limit for C17.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C18.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C19.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C20.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C21.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C22.......... : 1C (2800 MHz)

Funky frank · February 12, 2018

Will it have any influence on performance, if I change the auto-clover-calculated c3-latency value from 0x00FA to 0X03E9? Or is that outdated anyway?

apianti · February 12, 2018

Will it have any influence on performance, if I change the auto-clover-calculated c3-latency value from 0x00FA to 0X03E9? Or is that outdated anyway?

Ugh, I would just get those from your original tables. Most of the generated CPU stuff is outdated, my advice is never use any of the generated stuff and just patch your tables.

February 12, 2018

Attached a Geekbench test

up without patch

down with patch

I don't want open another page about your XCPM assertion..but I can only say with or without patch I have always same appleintelcpuinfo output , temperatures, steps..only performances is lower

PS.

Pikeralpha opened a case with Geekbench software because How can also see from this partial IPG graphics during test, Geekbench do not push my xeon at its power and frequency limits

Please give a geekbench result.

EDIT: ~~I don't see how you could be getting any states when this is the status of the CPUs configuration~~, EIST is clearly disabled:

EDIT2: I see now, it's just disabling EIST and HWP, so basically just using speedstep always to the complete maximum..... Which means its not using XCPM.

MSR_IA32_PERF_CONTROL............(0x199) : 0xFF00
------------------------------------------
 - Target performance State Value....... : 0xFF00 (25500 MHz)
 - Intel Dynamic Acceleration........... : 0 (IDA engaged)

...

MSR_MISC_PWR_MGMT................(0x1AA) : 0x402000
------------------------------------------
- EIST Hardware Coordination........... : 0 (hardware coordination enabled)
- Energy/Performance Bias support...... : 1
- Energy/Performance Bias.............. : 0 (disabled/MSR not visible to software)
- Thermal Interrupt Coordination Enable : 1 (thermal interrupt routed to all cores)
- SpeedShift Technology Enable......... : 0 (disabled)
- SpeedShift Interrupt Coordination.... : 0 (disabled)
- SpeedShift Energy Efficient Perf..... : 0 (disabled)
- SpeedShift Technology Setup for HWP.. : No (not setup for HWP)

Here are your turbos:

MSR_TURBO_RATIO_LIMIT............(0x1AD) : 0x1E1F202122232525
------------------------------------------
- Maximum Ratio Limit for C01.......... : 25 (3700 MHz) 
- Maximum Ratio Limit for C02.......... : 25 (3700 MHz) 
- Maximum Ratio Limit for C03.......... : 23 (3500 MHz) 
- Maximum Ratio Limit for C04.......... : 22 (3400 MHz) 
- Maximum Ratio Limit for C05.......... : 21 (3300 MHz) 
- Maximum Ratio Limit for C06.......... : 20 (3200 MHz) 
- Maximum Ratio Limit for C07.......... : 1F (3100 MHz) 
- Maximum Ratio Limit for C08.......... : 1E (3000 MHz)

MSR_TURBO_RATIO_LIMIT1...........(0x1AE) : 0x1C1C1C1C1C1C1C1D
------------------------------------------
- Maximum Ratio Limit for C09.......... : 1D (2900 MHz) 
- Maximum Ratio Limit for C10.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C11.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C12.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C13.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C14.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C15.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C16.......... : 1C (2800 MHz)

MSR_TURBO_RATIO_LIMIT2...........(0x1AF) : 0x1C1C1C1C1C1C1C1C
------------------------------------------
- Maximum Ratio Limit for C17.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C18.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C19.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C20.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C21.......... : 1C (2800 MHz) 
- Maximum Ratio Limit for C22.......... : 1C (2800 MHz)

ohhhh

with edit2 you now understand the meaning of that patch!

edit

added sysctl machdep.xcpm output

Last login: Mon Feb 12 18:32:48 on ttys000
fabios-iMac-Pro:~ fabio$ sysctl machdep.xcpm output
machdep.xcpm.mode: 1
machdep.xcpm.hard_plimit_max_100mhz_ratio: 37
machdep.xcpm.hard_plimit_min_100mhz_ratio: 8
machdep.xcpm.soft_plimit_max_100mhz_ratio: 37
machdep.xcpm.soft_plimit_min_100mhz_ratio: 8
machdep.xcpm.tuib_plimit_max_100mhz_ratio: 37
machdep.xcpm.tuib_plimit_min_100mhz_ratio: 8
machdep.xcpm.tuib_enabled: 0
machdep.xcpm.power_source: 0
machdep.xcpm.bootplim: 0
machdep.xcpm.bootpst: 37
machdep.xcpm.tuib_ns: 0
machdep.xcpm.vectors_loaded_count: 1
machdep.xcpm.ratio_change_ratelimit_ns: 500000
machdep.xcpm.ratio_changes_total: 368499
machdep.xcpm.maxbusdelay: 0
machdep.xcpm.maxintdelay: 0
machdep.xcpm.mbd_mode: 1
machdep.xcpm.mbd_applications: 0
machdep.xcpm.mbd_relaxations: 0
machdep.xcpm.forced_idle_ratio: 100
machdep.xcpm.forced_idle_period: 30000000
machdep.xcpm.deep_idle_log: 0
machdep.xcpm.qos_txfr: 1
machdep.xcpm.deep_idle_count: 0
machdep.xcpm.deep_idle_last_stats: n/a
machdep.xcpm.deep_idle_total_stats: n/a
machdep.xcpm.cpu_thermal_level: 0
machdep.xcpm.gpu_thermal_level: 0
machdep.xcpm.io_thermal_level: 0
machdep.xcpm.io_control_engages: 0
machdep.xcpm.io_control_disengages: 1
machdep.xcpm.io_filtered_reads: 0
machdep.xcpm.io_cst_control_enabled: 0
machdep.xcpm.ring_boost_enabled: 0
machdep.xcpm.io_epp_boost_enabled: 0
machdep.xcpm.epp_override: 0
fabios-iMac-Pro:~ fabio$

apianti · February 12, 2018

Attached a Geekbench test

up without patch

down with patch

I don't want open another page about your XCPM assertion..but I can only say with or without patch I have always same appleintelcpuinfo output , temperatures, steps..only performances is lower

The performance is lower because you are trading off power management for unbridled speed. Geekbench runs actual algorithms so the utilization changes as it is used, which is a good thing because it gives a more real time usage result. The scores are marginally different, the utilization is pretty much similar, I'm sure the statistical variance would be small but that power consumption is absolutely more, a lot more. The performance is around 9% gain but how much more energy is that taking? In the first one, you don't hit 120W at all, in the second you hit it seven times, and a very extended eighth time. Just those spots each look around about ~20W more, most of the spots look about ~20W more. So since it was hitting 100W but now hits 120W in the same place, that is a 20% increase in power usage. You are trading a ~20% increase in power consumption for a ~9% increase in output.

EDIT: Also 4000 is geekbench's your CPU is great score.

ohhhh

with edit2 you now understand the meaning of that patch!

Does not mean I think it is a good patch. Also that was what I was asking you to explain to me but you weren't.

edit

added sysctl machdep.xcpm output

Last login: Mon Feb 12 18:32:48 on ttys000
fabios-iMac-Pro:~ fabio$ sysctl machdep.xcpm output
machdep.xcpm.mode: 1
machdep.xcpm.hard_plimit_max_100mhz_ratio: 37
machdep.xcpm.hard_plimit_min_100mhz_ratio: 8
machdep.xcpm.soft_plimit_max_100mhz_ratio: 37
machdep.xcpm.soft_plimit_min_100mhz_ratio: 8
machdep.xcpm.tuib_plimit_max_100mhz_ratio: 37
machdep.xcpm.tuib_plimit_min_100mhz_ratio: 8
machdep.xcpm.tuib_enabled: 0
machdep.xcpm.power_source: 0
machdep.xcpm.bootplim: 0
machdep.xcpm.bootpst: 37
machdep.xcpm.tuib_ns: 0
machdep.xcpm.vectors_loaded_count: 1
machdep.xcpm.ratio_change_ratelimit_ns: 500000
machdep.xcpm.ratio_changes_total: 368499
machdep.xcpm.maxbusdelay: 0
machdep.xcpm.maxintdelay: 0
machdep.xcpm.mbd_mode: 1
machdep.xcpm.mbd_applications: 0
machdep.xcpm.mbd_relaxations: 0
machdep.xcpm.forced_idle_ratio: 100
machdep.xcpm.forced_idle_period: 30000000
machdep.xcpm.deep_idle_log: 0
machdep.xcpm.qos_txfr: 1
machdep.xcpm.deep_idle_count: 0
machdep.xcpm.deep_idle_last_stats: n/a
machdep.xcpm.deep_idle_total_stats: n/a
machdep.xcpm.cpu_thermal_level: 0
machdep.xcpm.gpu_thermal_level: 0
machdep.xcpm.io_thermal_level: 0
machdep.xcpm.io_control_engages: 0
machdep.xcpm.io_control_disengages: 1
machdep.xcpm.io_filtered_reads: 0
machdep.xcpm.io_cst_control_enabled: 0
machdep.xcpm.ring_boost_enabled: 0
machdep.xcpm.io_epp_boost_enabled: 0
machdep.xcpm.epp_override: 0
fabios-iMac-Pro:~ fabio$

This is meaningless if you patch the place where it writes the target performance state to always have the highest value.

February 12, 2018

The performance is lower because you are trading off power management for unbridled speed. Geekbench runs actual algorithms so the utilization changes as it is used, which is a good thing because it gives a more real time usage result. The scores are marginally different, the utilization is pretty much similar, I'm sure the statistical variance would be small but that power consumption is absolutely more, a lot more. The performance is around 9% gain but how much more energy is that taking? In the first one, you don't hit 120W at all, in the second you hit it seven times, and a very extended eighth time. Just those spots each look around about ~20W more, most of the spots look about ~20W more. So since it was hitting 100W but now hits 120W in the same place, that is a 20% increase in power usage. You are trading a ~20% increase in power consumption for a ~9% increase in output.

EDIT: Also 4000 is geekbench's your CPU is great score.

Does not mean I think it is a good patch. Also that was what I was asking you to explain to me but you weren't.

This is meaningless if you patch the place where it writes the target performance state to always have the highest value.

Uhmm

Maybe is my today english

ok

I use some 3d software for my job

they use intensively cpu when I do some rendering

with patch 20 % faster and cpu locked @2600 Mhz

without patch no

In windows instead no differences

watt consuming is another things and I would not like to argument because n my rig I have had also two TitanZ Graphic card (maybe you know how much power they need in the past)

Now

This patch is useful for me and for people who have (for various reasons EIST enabled in bios)

And the reason is always the same OSX or some tool I use to boot inside (like bootloader or other) with EIST enable produce for unsupported CPU worst performance

Now if it is possible to use this "boost" only when I need (during rendering IE) I will be happy

but for now it wasn't possible and for this fact I and other people are using that patch from Sierra

PS

thank again to PMHeart to discover it for 10.13.4 beta

and for the sake of a right answer I have answered here to your first request

#16745

Funky frank · February 12, 2018

Ugh, I would just get those from your original tables. Most of the generated CPU stuff is outdated, my advice is never use any of the generated stuff and just patch your tables.

I don't have tables anymore, I am now member of the exclusive xcpm club.

apianti · February 12, 2018

I don't have tables anymore, I am now member of the exclusive xcpm club.

You still have tables, lol. That even requires tables.... But I meant if you want to know your c states information get it from your original table. Also how does XCPM handle c states? Like p states?

Funky frank · February 12, 2018

XCPM seems to use secret tech, and it is described at the page I linked and you blocked. What is this, a kindergarten? Did t0nym4c also block links to you? LOL

apianti · February 13, 2018

XCPM seems to use secret tech, and it is described at the page I linked and you blocked. What is this, a kindergarten? Did t0nym4c also block links to you? LOL

I didn't block anything. lol, I'm pretty sure the link that it redirects to instead describes why it is blocked. It doesn't use secret tech, it just uses intel hardware features, there's a whole part of the developer's manual just for power management. It almost certainly uses monitor and mwait instructions.

Funky frank · February 13, 2018

It was ironically meant. Not you in person, but you as insanely-team. Well, I find that blocking kindergarten like. Together you could reach much more , e.g. a proper wiki with always state of the art info. Whatever. Surely u know about that, only wondering that this is not the default clover setting for haswell... Since it seems to be faster, SSD seems to be snappier. Or I am just tired.

xcpm works here with and with cpu ssdts the same. And dual opencl is a Bliss.

Trung_Nguyen · February 13, 2018

GOT IT WORKING:Apple boot key.

Just test it with CMD+V

Currently trying to get the alt key to invoke GUI

Clover General discussion

Recommended Posts

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites