[Various] Overclocker discovers Xeon E5 V3 Errata, Engineers exploit to unlock Turbo

[dingmel]

you have launch the uefi shell (the elf file.)

finally managed to get it to load manually, (Many thanks to StyM and zanzabar)

I ran into a new problem with getting it to auto load however.

I used the bcfg command as suggested in the guide, which adds a new entry to the RvE's EFI menu which is to load the V3 driver automatically.

When i enable that, it defaults to the 25x max on all cores... weird.

When i manually load the driver from the shell, I get 30x. XD

Somehow V3.EFI isn't autoloading even though "bcfg drivers dump" shows its set

[Saren]

Here is my workaround for auto load in case efi driver settings does not work.

Put a startup.nsh inside efi partition's root, content as follows:
load fs0:\EFI\Apricity\v3x2_faster.efi
exit

The drive specified and path to the efi file must be configuration to your own settings.
___________
BTW I found v3x2_mc39.efi has equal or better performance than v3x2.efi by simple ffmpeg benchmark despite CPU clocked lower.
I need more benchmark to conclude that.

[Ace123]

Man I like what's happening in here.

[dingmel]

Here is my workaround for auto load in case efi driver settings does not work.

Put a startup.nsh inside efi partition's root, content as follows:
load fs0:\EFI\Apricity\v3x2_faster.efi
exit

The drive specified and path to the efi file must be configuration to your own settings.
___________
BTW I found v3x2_mc39.efi has equal or better performance than v3x2.efi by simple ffmpeg benchmark despite CPU clocked lower.
I need more benchmark to conclude that.

Thanks Saren. I'll look into it. I believe my problem may be due to secure boot or one of the related settings within the Asus EFI. Still tinkering.

Edit : I managed to figure out why it wasn't loading each time. Secureboot was in the way XD. Anyone managed to get Secureboot up with this running?

[dingmel]

Thanks Saren. I'll look into it. I believe my problem may be due to secure boot or one of the related settings within the Asus EFI. Still tinkering.

Edit : I managed to figure out why it wasn't loading each time. Secureboot was in the way XD. Anyone managed to get Secureboot up with this running?

Edit : Looks like i spoke too soon. After i was trying out different bclk settings, the system crashed, and i needed to clear the CMOS.

Now, after adding the bcfg entry, and a startup.nsh entry (which intially worked), on top of disabling Secure Boot, it still doesn't auto load.

Any ideas?

[StyM]

i'm going manual for now, it's annoying though...

[dingmel]

i'm going manual for now, it's annoying though...

Ditto...

The frustrating thing for me though is it randomly autoloads, and I can't seem to figure out the causative variable. Secureboot is completely off, though I could boot into the EFI shell with it enabled. do correct me on this, V3.EFI is an unsigned driver, so secureboot could've been theoretically blocking it from loading, but if the EFI shell loads, why wouldn't V3 load. I'm preplexed XD

[StyM]

not totally sure on that.
noticed that the max OC isn't stable on BOINC 24 hrs i've gotten two restarts already.
now going back to 2.7ghz OC see if it's stable for a 24 hrs full load operation.

[BenchZowner]

Why not increase the Vcore ?

[VietOZ]

I dont think there's an option for that as Stym is using SM 2p board

[BenchZowner]

Ah SuperMicro.

Too bad.
I'd easily swap to an Asus WS board which features a more powerful BIOS and of course Voltage options to stabilize the Xeons at 3GHz.

[Saren]

not totally sure on that.
noticed that the max OC isn't stable on BOINC 24 hrs i've gotten two restarts already.
now going back to 2.7ghz OC see if it's stable for a 24 hrs full load operation.

Were you using the 0x27 microcode patch or the 0x39 one until you found out the system was unstable? I am fine with 0x39 one so far.

[Ace123]

My system is slightly faster with 27

[dingmel]

Ah SuperMicro.

Too bad.
I'd easily swap to an Asus WS board which features a more powerful BIOS and of course Voltage options to stabilize the Xeons at 3GHz.

Aren't the Xeon E5's Vcores hard locked? I have the setting in my bios, but changing it doesn't do anything apparently.

[zanzabar]

Aren't the Xeon E5's Vcores hard locked? I have the setting in my bios, but changing it doesn't do anything apparently.

it depends if your using a linked or unlinked vcore on your board. intel wants you to use the linked set up so the cpu controls it, but some boards (i think asus sells one and tyan has one) will run unlinked like an overclocking 1p board will do where the board does what it wants with the cpu only requesting a voltage so you can offset it.

[Saren]

My system is slightly faster with 27

Have you compared it with 39 using benchmarks? What are the results?

[Ace123]

I ran both cinebench repeatedly with bios microcode loaded. 27 repeatedly just edged out 39 every time. Also w prime is slightly faster with 27. If you check my hwbot all my 2683 submissions were 27.

[Ace123]

not totally sure on that.
noticed that the max OC isn't stable on BOINC 24 hrs i've gotten two restarts already.
now going back to 2.7ghz OC see if it's stable for a 24 hrs full load operation.

What do you mean max oc?

[Ace123]

i'm not using that file, i believe that file is for single CPU only.
there are two files for use on 2P system, you may look for v3x2.efi (3.0ghz) or v3x2_mc39.efi (2.7ghz) , i'm currently using v3x2.efi for full turbo.
went into the efi shell and loaded the v3x2.efi file, copied it to the efi\boot folder.
unfortunately supermicro efi shell doesn't have bcfg, so not sure how to make the efi driver stick permanently.
so for now, everytime i reboot i have to load manually the driver...

Wait a minute, I'm confused. V3x2 runs your 2683 at 3ghz fully loaded?

[StyM]

Wait a minute, I'm confused. V3x2 runs your 2683 at 3ghz fully loaded?

it runs them at 2.8ghz and 2.9ghz , different speed , i've gotten two restarts so far on this config when running boinc 24hrs.
if i go with v3_mc39 it runs them at 2.7ghz , no restarts so far.

i'm using 212 evo for cooling, max temp is 65 deg on one cpu.
don't think the restart is related to heat.

[StyM]

Were you using the 0x27 microcode patch or the 0x39 one until you found out the system was unstable? I am fine with 0x39 one so far.

i'm running v3x2_mc39 with 39 vmware microcode, so far no restarts.
they're running @ 2.7ghz

[StyM]

What do you mean max oc?

different efi driver different speed, v3x2 runs my cpu at 2.8 & 2.9ghz
while v3x2_mc39 runs my cpu at 2.7ghz.

[Saren]

I ran both cinebench repeatedly with bios microcode loaded. 27 repeatedly just edged out 39 every time. Also w prime is slightly faster with 27. If you check my hwbot all my 2683 submissions were 27.

I see. I would like to use 27 too, but StyM reported it being unstable. I should stay on 39.

[Saren]

different efi driver different speed, v3x2 runs my cpu at 2.8 & 2.9ghz
while v3x2_mc39 runs my cpu at 2.7ghz.

Hi, are you currently following this guide? https://www.supermicro.com/support/f....cfm?faq=21555
I am using the config stated above.
However, somebody said while using microcode exploit, following the guide above gives you a more stable system.
Can you use the 27 one and follow the guide to give it a try?

[StyM]

thanks, will try that..