What controls Turbo Core in Xeons?

[escksu]

It was known issue for xeon v4 and corona.
And with unlocked turbo it is now affecting v3, i guess
https://corona-renderer.com/forum/index.php/topic,12018.msg80645.html#msg80645
TDP limitation?

@The Stilt
Could you run coronabench test and say what`s you turbo
https://corona-renderer.com/stuff/Corona1.3.Benchmark.zip
And please someone with 2683 v3 unlocked test it also


dual cpu system unlocked?

Hi yes,its dual. It has to be either 2696v4 or 2699v4. O

 

[Ignis Aeternum]

So, I have kind of succeeded with my Asrock X99 WS and E5-2686V3 (QS).
I only get 28 multiplier max on all cores.
Any ideas what I might check or try to improve this?

edit: HWiNFO64 was showing the multipliers all over. CPU-Z shows stable 28x during rendering both in Cinebench and Corona Render.

Is it even possible to lock all 18 cores on max turbo? 'Cause there's a E5-2696V3 (QS) with a max turbo of 3.8 GHz...
Will this work for ASRock X99 WS-E & WS-E 10G or Asus X99-E WS & WS 10G ? 'Cause these are probably the best X99 motherboards on the market, Asus' actually the best, having two 8-pin CPU power connectors.

 

[fromax]

Hi yes,its dual. It has to be either 2696v4 or 2699v4.

Could you post a link on that forum?

 

[Jann]

Is it even possible to lock all 18 cores on max turbo? 'Cause there's a E5-2696V3 (QS) with a max turbo of 3.8 GHz...
Will this work for ASRock X99 WS-E & WS-E 10G or Asus X99-E WS & WS 10G ? 'Cause these are probably the best X99 motherboards on the market, Asus' actually the best, having two 8-pin CPU power connectors.

From what I gather, any Asrock board work, as their bios is easier to edit. On Asus you would have to do it manually as described previously in this topic.
I don't know what's possible, for now I've seen 12-14 core cpus get their max 3-3.3GHz turbo on all cores. Only one other 2686V3 owner replied with results. He got 2.9GHz, but that was with increased BCLK.
Waiting for someone to figure out if this is the max we get

 

[KAL2017]

I received my ram Friday, the motherboard is supposed to be here Monday. I ordered two E5-2667 V3 ES cpu's and they are also supposed to be here Monday. I started to have second thoughts on doing this mod on a new dual cpu motherboard so I went with the E5-2667 V3 instead of the E5-2683 V3.

 

[fromax]

do it manually as described previously in this topic

Well it`s not that staightforward because offsets differ in each bios.

 

[Ignis Aeternum]

From what I gather, any Asrock board work, as their bios is easier to edit. On Asus you would have to do it manually as described previously in this topic.
I don't know what's possible, for now I've seen 12-14 core cpus get their max 3-3.3GHz turbo on all cores. Only one other 2686V3 owner replied with results. He got 2.9GHz, but that was with increased BCLK.
Waiting for someone to figure out if this is the max we get

I'm not talking about 2686V3, it's about 2696V3 (it is 2699V3 OEM), the 2696V3 QS has a max turbo of 3.8 GHz, so I'm just wondering if The Stilt got his 2699V3 with a max turbo of 3.6 GHz to 3.1 for 18c, then what having a max TB of 3.8 will get us to...

 

[Dufus]

@vadimkara That is correct, the V3_1 driver was written only to make adjustments that were required before a microcode update took place. If the microcode you have allows full max turbo then adjustments can be made in BIOS and / or OS as usual, no need for the V3_1 driver. The V3_1 driver will (hopefully) not change anything if it detects overclocking is locked, turbo ratios are not adjustable, CPU isn't of CPUID 306f?, microcode is present or is an unlocked CPU.


@Jahn Sorry if I wasn't clear but throttling was with reference to power and current, not temperature. For instance running Linpack even with AVX2 down clocking to 27x multi with AVX2 draws 175W package power, nearly 50% more than TDP.

1.044V seems high for 28x ratio unless using AVX2 and would have thought more likely a VID for 35x ratio so this may be holding the CPU back some. Maximum ratio's can be affected by VID and this seems to be why my own ratio drops to 27 when using AVX2. I don't know if there is a way to bypass this but it certainly would be nice if it could be done.


@escksu, 50k for 44 cores doesn't seem that high. 14 cores @ 3.15GHz scores 21500, 44 cores is over 3x as many and the V4 probably has a 5% better IPC improvement over the V3 so all 44 cores would likely be averaging somewhere around 2.2 - 2.3GHz.


@fromax Some AVX2 / non AVX turbo ratio's can be found here http://www.intel.com/content/dam/ww...cification-updates/xeon-e5-v3-spec-update.pdf

As already mentioned full core utilization drops my 2683v3 to 27x ratio from 30x. Even with a simple test program that uses an AVX2 instruction once every millisecond or less does the same even though power, current and temperatures are low. If less cores are active the ratio can be higher.


@jet1981 If you follow and go through the same steps as fromax you should get an idea of how to do it. Make sure you have a way to recover if it fails to post such as dual BIOS chips. I have to catch up on some other stuff so might no be around so often but I think there's enough info in the thread already just maybe a little spread around. Maybe someone would like to start another thread with everything consolidated into one post and if we get some improvements even better.

 

[mrpiggy]

Had an odd thing happen with the EFI bootup. Everything was working fine with the auto load of the V3.EFI file and decided to drop in a 2nd 980GTX in SLI for some benchmark fun. Everything was still working, but then I removed the 2nd 980GTX and the PC would not auto load the V3.EFI file anymore.. Had to remove and then re-install using the UEFI shell before it would work again.. Just weird. Maybe normal behavior? I don't know enough about how the UEFI boot works. Just thought I'd mention it.

 

[escksu]

@Dufus, are you help me to point out what I need to change for 2696v4 CPU? I know it probably won't work but still want to give it a try.

Also, where can I read up more about microcode hacking? Seems that info is quite rare.

Thanks.

 

[katlen]

Had an odd thing happen with the EFI bootup. Everything was working fine with the auto load of the V3.EFI file and decided to drop in a 2nd 980GTX in SLI for some benchmark fun. Everything was still working, but then I removed the 2nd 980GTX and the PC would not auto load the V3.EFI file anymore.. Had to remove and then re-install using the UEFI shell before it would work again.. Just weird. Maybe normal behavior? I don't know enough about how the UEFI boot works. Just thought I'd mention it.

Hi, it happend to me once. The cause was that I added a new hdd and the hdd order was changed from fs1:\ to fs2:\. If you switch sata ports it might happen too.

 

[Murilo900]

Hello guys, just registered here because of this topic, I am amazed of what you guys reached here, this kind of bios mod is mindblowing, especially when we are talking about xeon and high frequencies
So, I am thinking of getting one V3 just because of this glitch, it's a game changer for me, because I need high frequencies on my tasks, and seeing one 2683v3 running at 3ghz is amazing.
My question is, do you guys this microcode modding works in a X99 SOC CHAMPION? thanks for all

 

[TLN]

Hello guys, just registered here because of this topic, I am amazed of what you guys reached here, this kind of bios mod is mindblowing, especially when we are talking about xeon and high frequencies
So, I am thinking of getting one V3 just because of this glitch, it's a game changer for me, because I need high frequencies on my tasks, and seeing one 2683v3 running at 3ghz is amazing.
My question is, do you guys this microcode modding works in a X99 SOC CHAMPION? thanks for all

If you're already have this mobo - might be worth trying to. If you're buying a new one - I'd get a Taichi instead.

 

[Murilo900]

Yeah, this is my favorite one, I had alot of x99's and only the SOC CHAMPION gave to me 100% stability at everything I thrown at her, although I didn't tested any asrock. I was trying to mod the bios and I noticed that UBU have problems with GIGABYTE bios, so reading more here I saw that it could be modded by HxD too, I think it will work for me. I already bought one 2683v3, I am hoping that will work.

 

[spronkin1]

Hello! I have an engineering sample processor e5-2678 v3 and I could only get 3.2 GHz instead of 3.3 GHz (0x27 microcode)

 

[Murilo900]

I've researched alot about flashing the GIGABYTE X99 SOC CHAMPION and I gave up because there is no information about it and I don't want to screw up the mobo, any one could mod for me the a bios without the microcodes? I have here a X99 DELUXE too, if anyone could mod for me the bios of any of those mobos I would be foverer grateful, I can't do these things, too much insecure about it

 

[escksu]

I've researched alot about flashing the GIGABYTE X99 SOC CHAMPION and I gave up because there is no information about it and I don't want to screw up the mobo, any one could mod for me the a bios without the microcodes? I have here a X99 DELUXE too, if anyone could mod for me the bios of any of those mobos I would be foverer grateful, I can't do these things, too much insecure about it

Actually you not need to be afraid. Your board has dual BIOS. If you screw up the main BIOS, you still can boot from the backup bios. Then just simply flash back the main BIOS.

 

[custom90gt]

This is an awesome thread! I'm going to give it a shot on a e5 2695 V3 ES when I have some free time. It should be a good replacement to my E5-2630 V4 that I currently have. I hope I can get it to work on Server 2016...

 

[TLN]

Actually you not need to be afraid. Your board has dual BIOS. If you screw up the main BIOS, you still can boot from the backup bios. Then just simply flash back the main BIOS.

Is that that easy to flip BIOS and edit it from UEFI console nowadays?
I got ASUS mobo, and will give it a try, once I get CPU's. But I never tried dual bios boards, and got no idea how fast I can switch between BIOSes.

 

[Murilo900]

Actually you not need to be afraid. Your board has dual BIOS. If you screw up the main BIOS, you still can boot from the backup bios. Then just simply flash back the main BIOS.

The problem is how to do it, how to mod the bios. I didn't find nothing about this on GIGABYTE, UBU can't flash it because incompatibility, it remains for me MMTool or UEFITool, or even HxD, I can't find the offsets of the microcodes in there on HxD

 

[vercety]

Hi I am new to the forum and I am really excited about this topic, but I am a bit lost with all the process I have a Xeon E5-2686 V3 paired with an Asus X99 Deluxe motherboard with bios 3101 (+32gb ramm hyperx 2666, 980ti hybrid and Windows 10 in legacy mode). I have read the process of page 3 many times, but I am not too sure about the Bios editing part and how to install everything (as someone said with Asus bios might be different.) Which steps should I follow then? Many thanks to all!

 

[mr.hat]

Hi I am new to the forum and I am really excited about this topic, but I am a bit lost with all the process I have a Xeon E5-2686 V3 paired with an Asus X99 Deluxe motherboard with bios 3101 (+32gb ramm hyperx 2666, 980ti hybrid and Windows 10 in legacy mode). I have read the process of page 3 many times, but I am not too sure about the Bios editing part and how to install everything (as someone said with Asus bios might be different.) Which steps should I follow then? Many thanks to all!

use UBU for asus, and "biosflashback" for flashing

 

[Ionstream]

Things aren't looking too good on my end. I've tried flashing the edited BIOS (using the UBU toolkit) on my SM DP board, but the system had trouble booting up. I ended up spending a few hours restoring the system to its original state.

Not sure if it's because I deleted an extra file (containing mcupdate_GenuineIntel in it's name), or it's because UBU really has issues with AMI Aptio V BIOSes. I'll give it a shot again, once I learn how to use UEFITool.

@Dufus Could you explain how this hex-editing magic works? I know we're trying to remove code from a BIOS, but how do I pinpoint the section of code to remove? Also, is it as simple as deleting 500 lines from the BIOS?

On a side note, the guys at win-raid have received take-down notices for their utility from AMI, so I'd be careful about posting any links here. We don't want the Anandtech forums attracting the wrong kind of attention.

 

[tentoo]

Hi! Much respect to @Dufus, @C-Power/Tw0tch.
I have MSI X99A Raider and xeon 2650 v3 and I would like to try this mod. Sadly UBU doesn`t support msi bios. So could someone give me a proper instruction how to earse ucode manually or make such modification for me and share bios https://msi.com/Motherboard/support/X99A-RAIDER.html#down-bios
I would appreciate any help)

For MSI X99A SLI PLUS and xeon 2673 v3 use next:

F0C69E 03 -> 05
F0C6E6 03 -> 01

F1C29E 03 -> 05
F1C2E6 03 -> 01

What i must change for MSI x99a raider and E5-2650v3?
Thanks.

 

[Ionstream]

You can try poking around these 3 sites:

http://www.overclock.net/t/1577530/guide-how-to-update-the-cpu-microcode-on-the-x99-ami-bios
http://forum.notebookreview.com/thr...-intel-haswell-cpu-microcode-bug-hack.790177/
http://www.win-raid.com/t18f16-Guide-Manual-AMI-UEFI-BIOS-Modding.html

I'm still figuring this out, but the gist of it involves finding the start and end of the ucode, because they are stored sequentially.