What controls Turbo Core in Xeons?

[xeon_fan]

In general does it make any sense to buy E5-2698 v3 if max turbo for all cores will be significantly lower than on E5-2697 v3?

Someone with E5-2698 v3 should provide their max frequency speeds but from the max turbos perspective both looks the same yet 2698 v3 has two more cores 🙂

 

[m1919]

This result is amazing. I was considering buying these two chips: https://www.ebay.es/itm/Intel-Xeon-...=item2604adc264:g:FzIAAOSwDYxbsOFO:rk:14:pf:0 and possibly OC them to 3.6GHz. Are these QS chips a good idea? I was looking for 18 core xeon, but non-ES chips are still quite expensive.

QS chips are usually production stepping. Make sure never to get anything that is QExx.

QFxx, QGxx appear to be same stepping as retail usually.

What is the max turbo for all cores that someone managed to achieve using E5-2698 v3 or E5-2697 v3?

What you see in the screenshot is the absolute max overclock I was able to get. That was actually not stable and I had to back off to 105.3 MHz in bios, was getting some Nvidia driver crashes and some PCIE related errors in event viewer.

I tried for the full 3.9 GHz, but one of the procs failed to initialize at 105.8 MHz bus.

I'm satisfied with the result, turned out a lot better than I thought it would.

Hi one additional question ...
what this mean ? where did you left the microcode update ? in Windows ?


Thanks !

Would be great to run e5 2697 v3 [QS] at x37

@ziollos - I have QS version of CPU and it does work flawlessly !

Yes, I left the latest Intel microcode update that Microsoft pushed out to address some of the Spectre exploits. Apparently Microsoft will be releasing mitigation for performance issues related to Spectre, so I don't really feel the need to play around with the older microcodes.

I had for e5 2697 v3[QS] :

• 1-11 cores - turbo 3.6 ghz (x36)
• 12-13 cores - turbo 3.5 ghz (x35)
• 14 cores - turbo 3.4 ghz (x34)

But now I see the x37 can be even achieved ... need to check that 🙂

I think the turbo boost you will see are dependent on your board and possibly how good your procs are. Also different workloads will produce different all-core turbo.

Also, if you're running a Z10PE-D8 WS or the -D16, you probably will have to get a bios programmer and a separate bios chip to program with the modded bios.

I was completely unable to use bios flashback to get the modded bios in place.

 

[ziollos]

What I wanna now before making a purchase is basically is there significant difference between E5-2698 v3 and E5-2697 v3 in terms of max turbo boost on all cores during 100% load. My workload will gain more from fewer cores at higher clock speed so if there is like 0.2GHz difference and more I will go for 14 core version.

 

[m1919]

What I wanna now before making a purchase is basically is there significant difference between E5-2698 v3 and E5-2697 v3 in terms of max turbo boost on all cores during 100% load. My workload will gain more from fewer cores at higher clock speed so if there is like 0.2GHz difference and more I will go for 14 core version.

E5-2698 v3 be seems to have a lower TDP than the E5-2697 v3. 135W vs 145W.

I would assume the 2698 would boost lower on all cores as its max multiplier appears to be 36x and TDP is lower. Maybe someone else can comment on their experience with the 2698.

If the 2698 is significantly more expensive than the 2697, choice should be clear.

My 2697s are actually unstable at close to 3.9 GHz. Too much negative voltage offset on core. Going to have to compile some custom EFI drivers that are slightly less aggressive. It felt stable for a couple days usage, so I think I can probably still try for that 37x multiplier.

 

[m1919]

Also noticed that 37x multiplier application seems to require cold boot + PSU power cycle because I was getting inconsistent result. I would switch EFI driver, reboot and still get 36x multiplier, even on drivers where I was sure I had gotten 37x multiplier before.

 

[kjboughton]

E5-2696v3 is the top dog. 38x top multi and 18 cores. 145W TDP

 

[ziollos]

Ok, given that 2696 is still quite expensive I decided to go for E5-2697 v3 (it has 0.3GHz higher all cores turbo than 2698 and I'm assuming that this difference is hard to equalize even after OC). My question is: is cpu from this link legit? https://www.ebay.es/itm/Intel-Xeon-...h=item23ad1a047b:g:EOUAAOSwYHtbsN0a:rk:1:pf:0
I know it is QS, but it is still significantly cheaper (400€ per piece) than other QS chips from ebay auctions (which people are even more eagerly buying than this cheaper version). What do you think could be the catch?

 

[dj_adam21]

E5-2696v3 is the top dog. 38x top multi and 18 cores. 145W TDP

Can you post your bios settings for 3.8 ghz turbo on all cores and also which efi do you use?
Thanks

 

[BeZnojeK]

Hello to all. Help please with bios for asus rampage v extreme / u3.1 and .efi under e5-2630l v3 QS.
I tried to do it myself, but in the new UBU_v1_70_rc16_1 I didn’t find an item to remove the microcode, but I don’t get through the old UBU_v1_65_7 one (with the firmware it gives an error "selected file is not a proper bios")
bios v.3801

 

[xeon_fan]

E5-2696v3 is the top dog. 38x top multi and 18 cores. 145W TDP

Is this 18 cores at 38x ?
Just used your undervolting profile (v3x4-0.10b-i306f2-rc9_90_50_50.efi, thanks !) and was able to achieve 13 at x36 and 14 at x35 (at 101 mhz bclk). My e5 2697 v3 [qs] also has 145 TDP ... that's why I'm asking if it's max turbo on all cores 🙂

 

[kjboughton]

Is this 18 cores at 38x ?
Just used your undervolting profile (v3x4-0.10b-i306f2-rc9_90_50_50.efi, thanks !) and was able to achieve 13 at x36 and 14 at x35 (at 101 mhz bclk). My e5 2697 v3 [qs] also has 145 TDP ... that's why I'm asking if it's max turbo on all cores 🙂

Yes, all cores will turbo to 3936MHz (103.6 BCK is current highest stable for me) for 2x CPUs. As loading increases the multi drops.

 

[imimixa]

Yes, all cores will turbo to 3936MHz (103.6 BCK is current highest stable for me) for 2x CPUs. As loading increases the multi drops.

And you load the microcode?
If so, how?

 

[kjboughton]

I load 0x1F at Windows init

 

[imimixa]

I load 0x1F at Windows init

Thank.
Than the microcode 3C is worse than 1F, what is the difference?
What are the values of the Intel power gadget during idle and load?

 

[xeon_fan]

Yes, all cores will turbo to 3936MHz (103.6 BCK is current highest stable for me) for 2x CPUs. As loading increases the multi drops.

Will try then different versions of your EFIs, will try also to search 1F microcode 🙂
Still ... unbelievable result 🙂

 

[BeZnojeK]

Hello to all. Help please with bios for asus rampage v extreme / u3.1 and .efi under e5-2630l v3 QS.
I tried to do it myself, but in the new UBU_v1_70_rc16_1 I didn’t find an item to remove the microcode, but I don’t get through the old UBU_v1_65_7 one (with the firmware it gives an error "selected file is not a proper bios")
bios v.3801

Help me plz 🙁

 

[xeon_fan]

I load 0x1F at Windows init

Hi 🙂 could you share 0x1F microcode or point where it can be obtained ?

Thank you!

 

[kjboughton]

Hi 🙂 could you share 0x1F microcode or point where it can be obtained ?

Thank you!

http://www.kristech.net/haswell/0x1F.zip

 

[raun0]

Thank you! When I have time I look into it.

I will follow your work with great interest 🙂

 

[foureight84]

Got the hack applied on my Z10PE-D8 WS with dual E5-2697 v3s.

Sadly, I couldn't get to 3.9 GHz, but it's close.

Running the 3407 bios with microcode removed and the v3x2_80-50-50_39_vcc1.8.efi driver MOF posted a while back.

I left the latest microcode update from Intel in place. It seems to boost to 3.8 in most situations.

Try using a lower voltage offset like -20 or use this https://github.com/freecableguy/v3x4/releases/tag/v3x4-0.10b-i306f2-rc9 the one without voltage offset. Also in your bios overclock settings, change loadline calibration to level 9 for both CPUs and also voltage capacity to 140%. Also for the ram, voltage capacity to 140% while manually set ram voltage to 1.2v. I have the same motherboard and 105.6 seems stable for me.

Running 2667v3 @ 3.8ghz

 

[xeon_fan]

http://www.kristech.net/haswell/0x1F.zip

@kjboughton - could I ask you which EFI configuration are you using ?

I have tested several and I cannot go any better than 13c - x36, 14c - x35.

Currently I'm using yours v3x4-0.10b-i306f2-rc9_90_50_50.efi (but tested a lot of them) and basically for all of them I'm getting the same results for 0x1F and 0x27 and very bad results for 0x39 (CPU stuck at x35 even for one core).

Thank you!

 

[kjboughton]

@kjboughton - could I ask you which EFI configuration are you using ?

I have tested several and I cannot go any better than 13c - x36, 14c - x35.

Currently I'm using yours v3x4-0.10b-i306f2-rc9_90_50_50.efi (but tested a lot of them) and basically for all of them I'm getting the same results for 0x1F and 0x27 and very bad results for 0x39 (CPU stuck at x35 even for one core).

Thank you!

I run a custom EFI.
-70mV for CORE
-30mv for CLR
104.8 BCLK (105 BCLK or greater downgrades QPI from 4.8GHz to 4.0GHz)

Like I said, lighter loading will boost all cores to 38x (~3982MHz)
Heaver loading and you will see somewhere around 34x, 35x

EDIT: Watercooling is a MUST for these speeds

 

[ziollos]

Hello again,
I bought these 2687 v3 QS and they seem to work perfectly fine. What for overclocking I searched this thread, but it is already too big to effectively use it as a source of knowledge (a lot of posts from the beginning of the thread were helpful, but I have no idea if that's current best solution).
I am using ASUS Z10PE-D16 motherboard.
Could someone post their modded bios to this motherboard or briefly explain step by step what to do to get these CPUs to 3.5-3.6 on all cores?
If this knowledge can be found in this thread I'm sorry, but I was unable to find it 😉

 

[ziollos]

Got the hack applied on my Z10PE-D8 WS with dual E5-2697 v3s.

Sadly, I couldn't get to 3.9 GHz, but it's close.

Running the 3407 bios with microcode removed and the v3x2_80-50-50_39_vcc1.8.efi driver MOF posted a while back.

I left the latest microcode update from Intel in place. It seems to boost to 3.8 in most situations.

To flash bios on your motherboard did you use something like this: https://www.amazon.es/gp/product/B07DRCVZB3/ref=oh_aui_detailpage_o01_s00?ie=UTF8&psc=1 ?
I have almost the same setup (d16 instead of d8) and am wondering what coolers are you using? Right now I have 2x nh-u12s in nzxt s340 case and in aida64 stress test cpu temp maxes out at 67C (stock without OC). I am going to add nf-a12x25 for both coolers as push and put current nf-f12 as pull which should lower the temps another 5C.
I am afraid that it won't have enough overclocking headroom and the only option left will be custom water loop (nh-u14s won't fit in s340, but comparing it to nh-u12s it would be give me 5C lower temps which isn't worth switching to another case).
What do you think. Is 62C(after adding 2x nf-a12x25) during 100% load at stock enough to overclock cpus to 3.7GHz? Maybe I should look at another cooler brands?

 

[xeon_fan]

I run a custom EFI.
-70mV for CORE
-30mv for CLR
104.8 BCLK (105 BCLK or greater downgrades QPI from 4.8GHz to 4.0GHz)

Like I said, lighter loading will boost all cores to 38x (~3982MHz)
Heaver loading and you will see somewhere around 34x, 35x

EDIT: Watercooling is a MUST for these speeds

That's unbelievable result!

For me so far the best results for me are -
-70-50-50 @ 1.9VCCIN with 104 mhz BCLK(over this my nvme driver throws randomly BSOD]
14 cores x35
13 or less at x36
[tested in CINBENCHMARK - 2475 max reached]
with AVX @ x33 for all cores

I have watercooling but the temperatures are under 50C - For now I guess the main problem is that my CPU goes max 130-140W and this is restriction which does not allow me to work with all cores at x36. Even "normal" cooling would handle my CPU now.