What controls Turbo Core in Xeons?

[Cata40]

E5-2683 v3 BIOS settings

OC Tweaker

CPU Config

'Max Core Ratio' needs removing as it is irrelevant for this CPU. Long duration still reverts to 120W but use false reading to bypass this otherwise limited to a RAPL of 40 seconds. Use to run BCLK at 105.1MHZ but saw I was getting a lot of corrected WHEA errors on the PCIe bus with NVMe at of 103MHz and GTX1080 at ~102MHz. Can reduce with lower link speed but just doesn't seem worth it for a few percent.


FIVR Config

Can hit max turbo -1 with AVX2 and -20mV offsets but seems to ramp up over a few seconds from max turbo -2 to -1. -30mV gives much faster response. And yes, VSA is positive.

Voltage Config

Most likely could lower DRAM voltage but haven't bothered. Timings reduced.


V3 Config

OC lock is being used for power reading and not OC lock. Need to change if I get some enthusiasm back. Tjmax doesn't do anything here and some other settings such as "Perf_Plimit_Differential" which if I guess correctly is actual for dual socket and P-State differential between sockets. Maybe I'm guessing wrong.


Advanced CPU

Could add package C2 here but PC2 doesn't seem to make a lot of difference.

Anyway that's what I'm running and works for me but might not work so well for someone else. Needs a fair bit more work yet, maybe some time later on.

Edit: Thanks Cata40, hope imgur works better than Photobucket did

I have an Asrock x99 oc formula , my bios is same to yours.
Why you have v3 configuration and me not in bios?

 

[wiseguyuk]

-80 Core, -50 Uncore, PowerCut (VCCIN = 1.900V) [be sure to set your BIOS to Auto]
download here

Sorry to bother you again. where is VCCIN in Z10PE-D8 bios ?

I just loaded the driver after switching it off for more than 10 seconds. I think it worked ok.

Something is not right. The CPUs run at 3400 MHz, but my cinebench score is around 3880.

 

[MOF]

Looks like only server motherboards needs to use powercut enabled drivers. If you are using OC capable x99 or WS motherboard you are free to use non-powercut drivers.

 

[Dufus]

I have an Asrock x99 oc formula , my bios is same to yours.
Why you have v3 configuration and me not in bios?

Similar, not exactly the same.

Because it's part of the modifications I was doing which really need starting from scratch again.

 

[Cata40]

Similar, not exactly the same.

Because it's part of the modifications I was doing which really need starting from scratch again.

I forgot, I asked about V3 in the integrated bios, just the v3 master maker.
Very interesting what I saw there, really great

 

[wiseguyuk]

Looks like only server motherboards needs to use powercut enabled drivers. If you are using OC capable x99 or WS motherboard you are free to use non-powercut drivers.

@kjboughton , is it possible to compile a 80,50 non-powercut one ?
what should be my cinebench score considering we have the same motherboard & CPUs ?

 

[MOF]

@kjboughton

With rc5 "1pt650pc" powercut works on all contition. After the restart gives error massage on post screen but still works.

With rc8 drivers powercut only works after soft resets(restart).

 

[kjboughton]

Sorry to bother you again. where is VCCIN in Z10PE-D8 bios ?

Well, you're down 4 whole cores from before. You may find you want them enabled as this config is really specific to my uses and was more for comparison purposes while we got you going. Assume you have NUMA enabled and Home Snoop set. Cinebench gets a big boost from HT. CPU Input Voltage on the first BIOS page is VCCIN.

 

[kjboughton]

@kjboughton

With rc5 "1pt650pc" powercut works on all contition. After the restart gives error massage on post screen but still works.

With rc8 drivers powercut only works after soft resets(restart).

I have engineered the driver to function this way. On first cold boot with any PC driver set, a reboot is required. Your system does it if you have CPU Input Voltage set to anything other than Auto (if that option is available), it's just really fast that you might not even notice it. A reboot is required once those settings are programmed. The exception being warm restarts where it sticks. Cold boot always means immediate reboot to take effect. The new driver will warn if reboot required and stay silent when good.

 

[wiseguyuk]

Well, you're down 4 whole cores from before. You may find you want them enabled as this config is really specific to my uses and was more for comparison purposes while we got you going. Assume you have NUMA enabled and Home Snoop set. Cinebench gets a big boost from HT. CPU Input Voltage on the first BIOS page is VCCIN.

I can't thank you enough for your help.

I had disabled NUMA & it was Early snoop. should I change them to get better performance ?

 

[MOF]

I couldnt test it enough but apparently minus uncore voltage offsets makes performance drops.

Edit: false alarm.

 

[kjboughton]

I can't thank you enough for your help.

I had disabled NUMA & it was Early snoop. should I change them to get better performance ?

I would recommend enable NUMA, set Home Snoop Mode, disable Hyperthreading. Enable C3

And... as stated previously... you're running Dual Channel. You need to be Quad Channel to feed all those cores at those speeds....

 

[wiseguyuk]

I would recommend enable NUMA, set Home Snoop Mode, disable Hyperthreading. Enable C3

And... as stated previously... you're running Dual Channel. You need to be Quad Channel to feed all those cores at those speeds....

I made the setting & it is better performance. Individual cores run at 3800 Mhz. Under load they all run at 3400 Mhz.

I enabled all the 36 cores. Under load all the 36 cores are running at 3300 Mhz

My Cinebench looks much better

I have 2 sets of dimms .. 16gb x 8 and 32 gb x 4. I will change it this weekend as it is fiddly to change them with the huge radiators with push & pull fans. I only need 128gb memory.

 

[MOF]

(Modified versions of v3x2_50_vcc.efi)

1.7V vccin
v3x2_20-50-50_vcc.efi
v3x2_30-50-50_vcc.efi
v3x2_40-50-50_vcc.efi
v3x2_60-50-50_vcc.efi
v3x2_70-50-50_vcc.efi
v3x2_80-30-50_vcc.efi
v3x2_80-50-50_vcc.efi
v3x2_90-50-50_vcc.efi

1.8V vccin
v3x2_20-50-50_vcc1.8.efi
v3x2_30-50-50_vcc1.8.efi
v3x2_40-50-50_vcc1.8.efi
v3x2_60-50-50_vcc1.8.efi
v3x2_70-50-50_vcc1.8.efi
v3x2_80-50-50_vcc1.8.efi
v3x2_90-50-50_vcc1.8.efi


(Modified versions of v3x2_50_39_vcc.efi, ucode 39 included)

1.7V vccin
v3x2_39_vcc.efi
v3x2_20-50-50_39_vcc.efi
v3x2_30-50-50_39_vcc.efi
v3x2_40-50-50_39_vcc.efi
v3x2_60-50-50_39_vcc.efi
v3x2_70-50-50_39_vcc.efi
v3x2_80-30-50_39_vcc.efi
v3x2_80-50-50_39_vcc.efi
v3x2_90-50-50_39_vcc.efi

1.8V vccin
v3x2_20-50-50_39_vcc1.8.efi
v3x2_30-50-50_39_vcc1.8.efi
v3x2_40-50-50_39_vcc1.8.efi
v3x2_60-50-50_39_vcc1.8.efi




-80mV vcore, -50mV cache, -50mV system agent, 1.8V vccin, Powercut + ucode39
v3x2_80-50-50_39_vcc1.8.efi

-90mV vcore, -50mV cache, -50mV system agent, 1.8V vccin, Powercut + ucode39
v3x2_90-50-50_39_vcc1.8.efi

-90mV vcore, -60mV cache, -50mV system agent, 1.8V vccin, Powercut + ucode39
v3x2_90-60-50_39_vcc1.8.efi

 

[CANONKONG]

Similar, not exactly the same.

Because it's part of the modifications I was doing which really need starting from scratch again.

Can you share your MOD BIOS？Thank you.

 

[Dufus]

Can you share your MOD BIOS？Thank you.

Once it's completed, sure. Might be a while though.

 

[Methanoid]

I had uploaded,you can find it in the default link.

Default link? Where is that.. I did try searching through your posts but didnt find....

 

[glitchman]

Once it's completed, sure. Might be a while though.

Your custom UEFI looks incredible!

 

[magicmanred]

Sweet mother of Xeon Juice....

I came across all of this info (no pun), and boy oh boy this is an AMAZING find!

I have an Intel Xeon E5 2696 (aka 2699) v3. Stepping 2, Revision R2. 18c 36t Haswell-E
It's in an ASRock X99E-ITX/ac motherboard with bios 3.6 (8/9/2016).

Naturally, as any sane person may assume... I didn't read through all 68 pages of this thread...
But I did do some reading, and found a walk-through by a user Ace123 on the XS Forums.
After the reading, and some googling, I successfully converted my boot drive to GPT from MBR (Now UEFI in Windows 10 instead of Legacy Bios) so that I could get started.

I downloaded the most recent BIOS for my MB (in 3 flavors: dos, windows, and raw)
and generated mod files with UBU. It DOES say something about my board/bios having "known issues" for the mod.

Has anyone done this before on this motherboard?
And are the "known issues" that UBU says when I'm generating the mod bios something to be concerned about?

When I open the bios in UBU, I get the following:
<--------------------------
Scanning BIOS... Please wait...
Platform BIOS AMI Aptio 5
Brand ASRock
Remove ASRock Protection
Remove ASRock new protection...
Found Option ROM in GUID A0327FE0-1FDA-4E5B-905D-B510C45A61D0 20FEEBDE-E739-420E-AE31-77E2876508C0
Found Option ROM in GUID A0327FE0-1FDA-4E5B-905D-B510C45A61D0 E095AFFE-D4CD-4289-9B48-28F64E3D781D
Found Option ROM in GUID A0327FE0-1FDA-4E5B-905D-B510C45A61D0 15DAA290-F7D3-4CEF-9527-EBFF67A02182
Found Option ROM in GUID A0327FE0-1FDA-4E5B-905D-B510C45A61D0 15DAA290-F7D3-4CEF-9527-EBFF67A02182
Found EFI Intel Raid Controller GUID 91B4D9C1-141C-4824-8D02-3C298E36EB3F
Found EFI AMI NVMe Driver GUID 634E8DB5-C432-43BE-A653-9CA2922CC458
Found EFI Intel LAN PRO/1000 Undi GUID 9006872D-3A86-4BAE-A2F0-E527B9D7119E

This BIOS on Aptio 5 platform, known issues:
- It is not supported by the update files in the GUID A0327FE0-1FDA-4E5B-905D-B510C45A61D0
- There may be problems with updating the CPU microcode

Press any key to continue . . .
---------------------------------------->

I went ahead and selected the appropriate options for Haswell-E.
Selected most recent (39) version and now have modded bios files.
(I did this for all 3 flavors of bios files with success).

I have a USB 2.0 flash disk formatted and ready with the V3.EFI in root.
I have the EFI folder with the Shellx64.efi file in it, as well as a Boot folder in that with the Bootx64.efi file in there.
I put UBU in a folder on there as well just in case, as well as my factory bios files and modded bios files.

I'm pretty much ready to give it a whirl. I have all of the instructions for when I boot with the USB flash disk in UEFI, and I also have the VMWare files ready for when I get into windows with the modded bios.

Just wanted to see if anyone knows about these "known issues" and if I should just pull the trigger.

 

[wiseguyuk]

I did a lot of testing

Here is what I am going to keep for now.

Best driver for my set up
v3x4-0.10b-i306f2-rc8_90_70_50.efi
Microcode 39 (with 1F, I had a lot of WHEA errors)

This is the final result

Thanks to All the guys who helped me, especially @kjboughton.

 

[Micrornd]

I did a lot of testing

Could you possibly post your final Cinebench scores for 36 cores 36 threads and 72 threads please?
Or your best, if it is some other combination of cores and threads.
I'm try to see how much difference this mod makes

 

[wiseguyuk]

Could you possibly post your final Cinebench scores for 36 cores 36 threads and 72 threads please?
Or your best, if it is some other combination of cores and threads.
I'm try to see how much difference this mod makes

There you go. I never ran it with hyperthreading ON because I never enabled it.

1st is the last score for all 36 cores.
2nd one (I think) is for 32 cores on microcode 27
3rd one (I think) is for 32 cores on microcode 39
4th is for 28 cores on microcode 39.

Without this exploit, I think I had around 2650.

 

[Micrornd]

There you go. I never ran it with hyperthreading ON because I never enabled it.

1st is the last score for all 36 cores.
2nd one (I think) is for 32 cores on microcode 27
3rd one (I think) is for 32 cores on microcode 39
4th is for 28 cores on microcode 39.

Without this exploit, I think I had around 2650.

Well I hope you had more than that originally, if not that was just terrible
I'm seeing only 3730 with 36c/36T with the 3A microcode without any mods, so that is quite an increase.
If you decide to enable hyperthreading, for comparison I'm seeing 4563 @36C/72T

 

[wiseguyuk]

Well I hope you had more than that originally, if not that was just terrible
I'm seeing only 3730 with 36c/36T with the 3A microcode without any mods, so that is quite an increase.
If you decide to enable hyperthreading, for comparison I'm seeing 4563 @36C/72T

I ran it in a Dell T7910 motherboard . Even the other benchmarks were quite pathetic.. Probably the bios was throttled for noise & air cooling.

with 72 threads I get the below scores

 

[Cata40]

Once it's completed, sure. Might be a while though.

I guess it works on all Asrock boards. Looks incredible!