What controls Turbo Core in Xeons?

[itou31]

Hi,
I just test with a E5-2696V3 and asus X99-E WS/USB3.1 bios 3502 mod and V3.EFI (Dufus) and µ39 in Win10. All seems OK (all core at x31 and 1 core at x38) I'am able to BCLK to 103.5

 

[Young-Chap]

I have the same motherboard as you with 2696v3 after upgrade to bios 3402 or 3502 the maximum single thread turbo x28 instead of x38 but with bios 3302 or before max single thread turbo x38.

My question in you setup with bios 3405 dose the cpu in single thread turbo to the max at x35 or it max at x31?

XMA my max multiplier is still at x35 in my bios. I did notice that under load it would drop to 3100 from the 3200

 

[Young-Chap]

Hi,
I just test with a E5-2696V3 and asus X99-E WS/USB3.1 bios 3502 mod and V3.EFI (Dufus) and µ39 in Win10. All seems OK (all core at x31 and 1 core at x38) I'am able to BCLK to 103.5

itou31 can you tell me the difference in clock speeds before and after the bios mod? Thanks

 

[sciff]

Hello sciff

Can you please make a mode for bios 3302 because it the only bios which can turbo to x38 in single thread. Maybe with the modded 3302 it get more than x28 on all core and still has x38 in single thread.

http://dlcdnet.asus.com/pub/ASUS/mb...302.zip?_ga=1.106753475.1564184574.1458152219
still can’t boot to Uefi shell
Thank you

https://yadi.sk/d/OyQ4WTzx3HFBQN

 

[itou31]

itou31 can you tell me the difference in clock speeds before and after the bios mod? Thanks

before mod : x27 max allcore and x38 one core

 

[xma]

https://yadi.sk/d/OyQ4WTzx3HFBQN

Thank you very much

How you did it? I tried but I could not keep getting Error in Replacing File

 

[itou31]

sciff explains somewhere in the thread by replacing directly in CAP file (2 instances of 14 AA 03 0C to 18 AA 03 08). Only valid for asus x99-e ws/USB3.1 bios 3402 and 3502 and perhaps 3302 (not checked). This hex mod is to made UBU support aptio V for asus and get rid of this "error in replacing file"

 

[sciff]

itou31, exactly. I used HxD hex editor and I searched for 2E 06 A0, right below which I could find 14 and 0C.
In case of 3302 BIOS I just skipped first instance because there were three search results, and the first one was in the irrelevant part of the BIOS.

I came across this method here: http://www.win-raid.com/t18f16-Guide-Manual-AMI-UEFI-BIOS-Modding-4.html#msg21755

First value (95) is different in case of that BIOS (for an MSI motherboard), as you can see. But what you need to understand, is when you change 0C to 08 directly by editing the .CAP file itself (subtracting 4 effectively), you have to compensate for this change by adding 4 somewhere, otherwise the .CAP file's checksum will change and the BIOS Flashback won't work (I tried). The same location as lordkag suggests, seems safe, so that's where I added 4 by changing 14 to 18.
But don't rely on offset address values (00DFE730 and the like). Use search function.

 

[xma]

itou31, exactly. I used HxD hex editor and I searched for 2E 06 A0, right below which I could find 14 and 0C.
In case of 3302 BIOS I just skipped first instance because there were three search results, and the first one was in the irrelevant part of the BIOS.

I came across this method here: http://www.win-raid.com/t18f16-Guide-Manual-AMI-UEFI-BIOS-Modding-4.html#msg21755

First value (95) is different in case of that BIOS (for an MSI motherboard), as you can see. But what you need to understand, is when you change 0C to 08 directly by editing the .CAP file itself (subtracting 4 effectively), you have to compensate for this change by adding 4 somewhere, otherwise the .CAP file's checksum will change and the BIOS Flashback won't work (I tried). The same location as lordkag suggests, seems safe, so that's where I added 4 by changing 14 to 18.
But don't rely on offset address values (00DFE730 and the like). Use search function.

thanks


Something wrong with my cpu or the MB. Nothing working with it. The modded bios 3302 behave the same as bios 3403 and 3502 max turbo x28 single core and all cores. I gave up on this system.

 

[ajschot]

The 2900 score was using the same motherboard (ASRock Extreme 4) as you.

You have a bottleneck somewhere. Try running the original V3 driver Dufus uploaded, that is the one I am using.

edit:
Turn off SVID telemetry in the BIOS, that will allow your CPU to go beyond the 240 Watt limit.

Hey thanks for your input, i tried with Original V3.EFI and VMWare microcode 39 and same thing happens only x31 on all core's also if i don't load VMWare microcode the machine takes the x36 standard in Windows 10. I renamed the file and then i can load x27 but it does not make any difference. Also if i don't load a microcode in windows 10 the machine freezes.
I also disabled SVID (it is left out in 3.60, but if i ut modded 3.40 in i get the even worse results...

For finding a right solution i bumped the bLCK back to 100. But i still have no idea why i can not get higher on Windows. Maybe i need to do a fresh install and look if that is giving me higher numbers.
There must be way to get it at least at x33.
For testing i also disabled the overheat protection for the CPU but also this did not change anything.

Or maybe something wrong with my mod? however than it should not load V3.EFI and i should not see in the bios 0 at the microcode, right
EDIT: Did try bios 3.30, 3.40 and did not make any change.... still not better then x31

Otherthing tested out... when using the 2683v3, windows does not use the x36 microcode.... and i can easily load x27 microcode but with the 2696v3 i need to rename the windows dll file.

 

[sciff]

I have the same motherboard and it can sustain an all core turbo of 3.3 and score over 2900 on cinebench. He has a bottleneck somewhere.

What about AVX applications? Does it run at 3.3GHz in those also?

Guys, can anyone re-upload V3x2 EFI driver variations that Welsper published previously? The link has expired

Haha the smile above is way too dramatic

Use original V3.efi (V3_1.efi) for PC with single CPU and V3x2_cup.efi (from attached link) for multi CPU PC. And remove microcode update from OS.

V3x2 with variuos configurations: https://ufile.io/b7d061
cup - with uncore max (max perfomance, but high power drain on idle)
cp - without uncore modification
none/27/39 - microcode version

I've made small test (Power profile: ballanced):

R15 @ 7Zip@8Th --- max result from 8x run. Corona - single run.
cp27 & cp39 --- with C3 enabled.
7Zip@8Th - simulates the loading of not all available cores.

 

[itou31]

Hi sciff,
Thanks for all your Bios settings and all advices you done on the asus x99-E WS/USB3.1.
My E5-2696V3 is a R2 also, have also plug 2 12V connector on the board.
Cannot go up to x33. I load the V3 original. Did you go with V3_1.EFI ?

Here is the v3_2 : https://ufile.io/gmjtf

 

[ajschot]

Hi sciff,
Thanks for all your Bios settings and all advices you done on the asus x99-E WS/USB3.1.
My E5-2696V3 is a R2 also, have also plug 2 12V connector on the board.
Cannot go up to x33. I load the V3 original. Did you go with V3_1.EFI ?

Here is the v3_2 : https://ufile.io/gmjtf

Well better then mine i can't pass x31 and i also have 2 cpu connectors connected.
Also on microcode 27 i can not go x38 on single core it does not higher then x32 which is strange, with x39 it does. I have same stepping as you have only i use a Asrock board and use a Corsair RM750x PSU (that should be enough power)

 

[xma]

Why i can’t mode the bios i did as Sciff wrote but still get Error in Replacing File

Even I tried the already modded bios by Sciff still getting Error in Replacing File

What I am doing wrong ?

 

[sciff]

Hi sciff,
Thanks for all your Bios settings and all advices you done on the asus x99-E WS/USB3.1.
My E5-2696V3 is a R2 also, have also plug 2 12V connector on the board.
Cannot go up to x33. I load the V3 original. Did you go with V3_1.EFI ?

Here is the v3_2 : https://ufile.io/gmjtf

You're welcome
Well, I get x33 no matter which EFI-driver I use. I can only suggest you to double-check the BIOS settings.

It's just that I prefer to use the ones without microcode (and I'm not loading any through VMware driver), then Corona renderer also works at CPU's x33 (whereas with µcodes ver.27 and ver.39 the all-core multiplier doesn't go higher than x27 and x28 respectively). My favourite is V3x2_cp.efi, still good performance, but lower power consumption in idle (my system takes 97W with EIST enabled).

 

[ajschot]

You're welcome
Well, I get x33 no matter which EFI-driver I use. I can only suggest you to double-check the BIOS settings.

It's just that I prefer to use the ones without microcode (and I'm not loading any through VMware driver), then Corona renderer also works at CPU's x33 (whereas with µcodes ver.27 and ver.39 the all-core multiplier doesn't go higher than x27 and x28 respectively). My favourite is V3x2_cp.efi, still good performance, but lower power consumption in idle (my system takes 97W with EIST enabled).

well on OSX I get all core's on x35 - x37! But not on Windows it won't get above x31 but I will try to reinstall windows maybe it installed something strange....

 

[itou31]

Hi, just test with V3x2_cp.efi and no microcode in windows 10 : x32 only . Score in Cinebench R15 also go back from 2700 to 2500 even I'm at x32 !? but other bench seems to be better.
and when load vmware drivers 0x39 multiplier go back to x31
I'm with BIOS 3502

 

[itou31]

Hi all,
Just take some time to test all possibility with the Asus X99-E WS/USB3.1 with E5-2696V3 R2
With BIOS 3502 and BIOS settings provide by sciff.
[URL=http://www.hostingpics.net/viewer.php?id=469467Capture.jpg][/URL]

So as sciff said, the best seems to be V3_2_cp.EFI with no microcode driver. I can get x32 (with bclk at 104). But score in Cinebench r15 is strangely low.

 

[ajschot]

Hi all,
Just take some time to test all possibility with the Asus X99-E WS/USB3.1 with E5-2696V3 R2
With BIOS 3502 and BIOS settings provide by sciff.


So as sciff said, the best seems to be V3_2_cp.EFI with no microcode driver. I can get x32 (with bclk at 104). But score in Cinebench r15 is strangely low.

Does your machine works with cp? Mine freezes, it also freezes when i don't have a ucode loded i can only use cup27 and cup39 otherwise my machine won't work how it should be...

Also it is something to see than all benchmarks gives different numers, one is better without other with x27 or x39

 

[xma]

Today I flash again the modded 3205 bios by sciff and played with the bios setting. I don’t have V3.EFI hack installed because I can’t boot UEFI shell.

I found out by changing the the cpu voltage form auto to offset + 0.001 (or any value) and core ratio from auto to x38 I could get x31 in windows without V3.EFI hack and with microcode 36 ( windows default driver ) when running aida64 stability test or HandBrake it will drop to x27 . Deleting the microcode got x32 , running aida64 stabilty test or HandBrake x30-x31. as itou31 found Cinebench r15 score lower with no microcode.

 

[txgy]

Today I flash again the modded 3205 bios by sciff and played with the bios setting. I don’t have V3.EFI hack installed because I can’t boot UEFI shell.

I found out by changing the the cpu voltage form auto to offset + 0.001 (or any value) and core ratio from auto to x38 I could get x31 in windows without V3.EFI hack and with microcode 36 ( windows default driver ) when running aida64 stability test or HandBrake it will drop to x27 . Deleting the microcode got x32 , running aida64 stabilty test or HandBrake x30-x31. as itou31 found Cinebench r15 score lower with no microcode.

I want to know how to delete Windows microcode
thanks

 

[traderjay]

Thanks so much for this thread! Just wondering anyway to modify the turbo for my system:

Motherboard: X10SRL-F (latest bios)
CPU: Intel XEON E5 2696 v3

 

[GTZ]

Today I flash again the modded 3205 bios by sciff and played with the bios setting. I don’t have V3.EFI hack installed because I can’t boot UEFI shell.

I found out by changing the the cpu voltage form auto to offset + 0.001 (or any value) and core ratio from auto to x38 I could get x31 in windows without V3.EFI hack and with microcode 36 ( windows default driver ) when running aida64 stability test or HandBrake it will drop to x27 . Deleting the microcode got x32 , running aida64 stabilty test or HandBrake x30-x31. as itou31 found Cinebench r15 score lower with no microcode.

Hi all,
Just take some time to test all possibility with the Asus X99-E WS/USB3.1 with E5-2696V3 R2
With BIOS 3502 and BIOS settings provide by sciff.


So as sciff said, the best seems to be V3_2_cp.EFI with no microcode driver. I can get x32 (with bclk at 104). But score in Cinebench r15 is strangely low.

With no microcode your cache will not go any higher than 2300 under heavy load. CB R15 shows this anomaly. With a microcode it will boost to the maximum (my case 3000). This has been discussed a few pages back.

Example
My 2696 will score 2925 with ucode 27 (any ucode really) with an all core turbo of 3300mhz and cache at 3100mhz. Without a ucode it will run at 3400mhz but cache will only be running at 2400mhz and score only 2800. Both examples running an 104.5 BCLK overclock.

My take out of this is:
If your workload does not use AVX instructions set run a ucode. However if it does and is stable without a ucode run without one. If a program fully takes advantage of cache, in my findings for every 500-600mhz is equal to 100-150mhz clock speed. Since running without a ucode under AVX loads it will only throttle by 100mhz as apposed 500mhz running a ucode. Making the cache deficit up.

 

[xma]

I want to know how to delete Windows microcode
thanks

In windows directory search for “mcupdate_GenuineIntel.dll" you will found two rename or delete

Then use VMWare tool: https://labs.vmware.com/flings/vmware-cpu-microcode-update-driver and click uninstall to remove the microcode

 

[itou31]

Thanks so much for this thread! Just wondering anyway to modify the turbo for my system:

Motherboard: X10SRL-F (latest bios)
CPU: Intel XEON E5 2696 v3

Hi,
if you don't scare to try, you could mod the bios with UBU (see page 3). There is no " Error replacing file" with the X10SRL6.801 bios. So you can try, even UBU has a warning about AMI APTIO V. But use at your own risk. My motherboard has bios Flashback so, if it fails, I can reflash with no problem.