Crosshair VII Hero Essential Info Thread

[gupsterg]

Last update: 17/11/18

Last 3 updates to OP:-

i) Added section Is the lack of ASUS T-Topology an issue on C7H?

ii) Added section To use manual OC or Precision Boost Override? (see post 3)

iii) ASUS PE section has new link to post by The Stilt on his tweak in PE3/4.

ROG Crosshair VII Hero
Essential Info Thread



by Elmor

Ryzen: Strictly technical - Chapter: Pinnacle Ridge By The Stilt

ZEN+ Delidded - Before/After Temperatures - [Ryzen 5, 7, 2600, 2700X] by Der8auer

ASUS CROSSHAIR VII HERO - Everything You Need To Know! X470 In Depth Review by Der8auer

Elmor's C7H UEFI folder link

PSA regarding C7H:-

Early batch of motherboards will exhibit an issue where [Auto] or manually set 1.05V for 1.05V SB voltage is shown as ~+50mV in UEFI readback box, SW monitoring in OS and on ProbeIt points. See the post by Elmor on OCN regarding this, link.

Do not use HWMonitor, it shows monitoring data which is not actually occurring, excessive CPU/RAM MHz, etc.

Preface

A C7H has been supplied to me by ASUS 😄 , courtesy of ASUS and Elmor :cool: . I do not work for ASUS or have any committent to them, I am just a "user" who has received a sample and doing what I would do if I had bought a C7H.

My plan is to expand OP to collate essential information for using this board. It will be gathered from web/user shares/own meddling. It will emulate some threads I did on OCN for Ryzen+C6H & TR+ZE; so do check back 🙂 .

Ryzen 2nd Gen "Pinnacle Ridge"





The C7H review by The Stilt

Original link, translated link.

• RAM Slots are daisy chain topology instead of ASUS T-Topology on C6H. Daisy chain favours 2 dimm slots/1DPC config.
  
• VRM 10+2 phases, 10 for VCORE, 2 for SOC, the mosfets have integrated temperature sensors.
  
• Six layer PCB.
  

Besides other information within review, I really liked the data on VRM information.



Is the lack of ASUS T-Topology an issue on C7H?

From my experience in some ways yes and no. I have found same CPU/RAM used on C6H you will have an easier ride with 4 dimms than C7H. I used Samsung B die single sided/ranked 8GB dimms.

A recent post by The Stilt highlighted how motherboard vendors can not tune firmware as they can on Intel platform (source link).

The worst memory related issue on this platform is that there is no way to tune the memory controller parameters. They are hard coded into the memory controller firmware (PMU) and cannot be changed by anyone but AMD. This prevents the manufacturers from optimizing the parameters specifically for their designs.

From what I have read in some detailed reviews is that AMD guidance on RAM tracing is daisy chain, so I would assume the firmware would be better tuned by AMD for that topology. Which favours the C7H as it uses daisy chain topology.

ASUS Performance Enhancer

Options within UEFI page Advanced > AMD CBS > NBIO Common Options > Precision Boost Override Configuration (Changed to XFR Enhancement in UEFI 1101). Precision Boost Override term is Precision Boost Overdrive. Some useful links to reference:-

• What each preset level of PE tweaks, link.
  
• Some further information from The Stilt on PE levels, link.
  
• Some further information from The Stilt on voltages seen when using PE, link.
  
• This linked post is also interesting on what the tweak within PE3/4 does.
  
  

  Even if you run 2700X with LN2 and disable all of the power / current limits, it will not boost higher than e.g. 4.35GHz for the best two cores of the CPU.
  
  4.35GHz for the best two cores of the CPU (marked with a golden and silver star in Ryzen Master), 4.2GHz for the rest (1-2C load).
  4.075GHz for all cores, unless limited by PPT, TDC, EDC, thermal or reliability (FIT). Clock reductions starts at 85°C (95°C tCTL), unless configured to a lower value.
  
  The power management must be reconfigured in order to allow higher frequencies, my "eXFR" ("Performance Enhancer" on ASUS boards) does just that.

  
  
  

ASUS Performance Enhancer/Precision Boost Override/Precision Boost Overdrive simply put allows a user of Ryzen to retain single core/thread boosting, but tweak nThread/Core boosting (ref above image).

You can also setup your own Performance Enhancer/Precision Boost Override Configuration (Changed to XFR Enhancement in UEFI 1101).

Note: Not available on Non X CPU.



PPT Limit - Package Power Tracking, watts limit CPU can use.

TDC Limit -Thermal Design Current, maximum current the VRM can supply to CPU, based on thermal aspect of VRM.

EDC Limit - Electrical Design Current, maximum peak current the VRM can supply to CPU, based on VRM components.

Precision Boost Overide Scalar - Increased scalar value relaxes voltage limiting rules of FIT monitoring.

These settings info is based on previous linked post plus quote below from The Stilt Ryzen: Strictly technical thread.

The "Precision Boost Override" feature available on 400-series motherboards allows increasing the physical limiters mentioned earlier. On SKUs belonging to the 105W TDP infrastructure group, the default limiters are following: PPT 141.75W, TDC 95A, EDC 140A and tJMax of 85°C (absolute, excl. offset).

When "Precision Boost Override" mode is enabled (AGESA default), PPT becomes essentially unrestricted (1000W), TDC is set to 114A and EDC to 168A. These limits can be customized by the ODM so that the new limits will comply with the electrical characteristics of the motherboard design in question.

Here is a link to a thread that people tinkering with PBO may find handy, link. Also below is images from Threadripper gen 2 launch slide deck with good info on PBO.



Which core is best?

As of HWINFO v5.83-3435 "Core Quality" is being shown, link to author's post. Perf #0 being best and highest # being worst. Latest Ryzen Master on Pinnacle Ridge only shows best core (denoted by gold star ) and second best (denoted by silver star).



BCLK Fluctuations?

Ryzen has no hardware to readback BCLK correctly. Now some owners of Ryzen/C7H may say I've never noted a fluctuation, there could be a few reasons to this.

• They have not captured it, as they did not have monitoring running at the time of event.
  
  
• Like HW, but differing samples can exhibit difference in fluctuation of BCLK readback. This is not limited to AMD CPUs without hardware to readback BCLK, but even Intel CPUs can exhibit this.
  
  
• They have correctly set up HWINFO for BCLK monitoring on CPUs without hardware for BCLK readback. When launching HWINFO pick settings, then untick option in below image highlighted in red.
  

So all in all what you set as BCLK in UEFI is what Ryzen/ASUS board will use and it is a steady frequency. This was confirmed in C6H thread at launch by Elmor who works in ASUS ROG MB R&D.

Which Prime95 to use?

The Stilt has advised v28.10 build 1 is best to use (link to app in this post), using custom setup of 128K 128K in place FFT will enable user to get correct CPU voltage setup. Then I use also custom setup of 8K 4096K with ~13GB RAM used out of 16GB (ie ~80% installed RAM); this will exercise the CPU/IMC/RAM.

How to test single or multiple cores?

Use HWINFO to identify core "Quality", perf #0 is best core and 7 is worst. Then ref below image on setting affinity and setup of P95. If using W10, in Task Manager click "Details" tab, find prime95.exe, right click and select "Set affinity" and you will see similar as screenie below.



Load Line Calibration!?

This cropped ups pretty much daily. The problem I think is there are far too many youtube video where you have x person setting y voltage on a board and then saying ahh look it drooped to z. What they over look saying is there is a load line specification. Even a video I saw around Ryzen launch where Robert Hallock from AMD is feature skimmed over the subject with the usual nonsense. At least at the time we had The Stilt and Raja@ASUS to guide us 😉 .

Any how here we go with the info:-

i) Der8auer has a great video on the subject, link.

ii) AM4 LL spec by AMD and info on dLDO by The Stilt.

iii) LLC on C6H, link 2 by The Stilt (still relevant for C7H).

iv) How to calculate VDROOP by The Stilt.

So I set a 4.0GHz PState 0 and VID of 1.281V, stock LLC (ie [Auto]). In UEFI, which creates load on CPU (you'll see even when CPU default it will PB/XFR in UEFI), I saw ~1.274V steady on DMM applied to ProbeIt point. UEFI readback box was bouncing, at and ever so slightly below VID set.

In OS whilst I had P95 loading CPU with custom 128K 128K in place FFT I saw ~1.208V on DMM applied to ProbeIt point.



Light load (ie UEFI) VDROOP was less, as when CPU come off load it is not gonna overshoot wildly. Heavy load led to more VDROOP, which again makes sense. As when using stock LLC it should be this way, so I get less of an overshoot when CPU come off load vs increased LLC.

Below is a simple graphic to show load line effect and calibration.



What about CPU voltage!?

Many I time since moving to Ryzen I will see discussions based on when CPU is stock that (x)V is observed in usage, so CPU must be able to sustain that when OC'ing!?

To see what the actual maximum voltage FIT allows the CPU to run at in various different scenarios is, I disabled all of the other limiters and safe guards. With every other limiter / safe guard disabled, the reliability (FIT) becomes the only restrain. The voltage command which the CPU sends to the VRM regulator via the SVI2 interface and the actual effective voltage were then recorded in various scenarios. In stock configuration the sustained maximum effective voltage during all-core stress allowed by FIT was =< 1.330V. Meanwhile, in single core workloads the sustained maximum was =< 1.425V. When the “FIT� parameters were adjusted by increasing the scalar value from the default 1x to the maximum allowed value of 10x, the maximum all-core voltage became 1.380V, while the maximum single core voltage increased to 1.480V. The recorded figures appear to fall very well in line with the seen and known behavior, frequency, power and thermal scaling wise.

The seen behaviour suggests that the full silicon reliability can be maintained up to around 1.330V in all-core workloads (i.e. high current) and up to 1.425V in single core workloads (i.e. low current). Use of higher voltages is definitely possible (as FIT will allow up to 1.380V / 1.480V when scalar is increased by 10x), but it more than likely results in reduced silicon lifetime / reliability. By how much? Only the good folks at AMD who have access to the simulation data will know for sure.

Quote source link

What should SOC voltage be set to!?

The C7H increases SOC when left on [Auto] based on "auto rule" not an "auto determination" of silicon characteristics. Increasing RAM past official speeds makes this occur. Best is to tune SOC as required. The Stilt experience quoted below.

There are clear differences in how the memory controller behaves on the different CPU specimens. The majority of the CPUs will do 3466MHz or higher at 1.050V SoC voltage, however the difference lies in how the different specimens react to the voltage. Some of the specimens seem scale with the increased SoC voltage, while the others simply refuse to scale at all or in some cases even illustrate negative scaling. All of the tested samples illustrated negative scaling (i.e. more errors or failures to train) when higher than 1.150V SoC was used. In all cases the maximum memory frequency was achieved at =< 1.100V SoC voltage.

I have found my R7 2700X (Batch: UA 1805SUS) scales this way for SOC:-

3200MHz set to 0.900V in UEFI (Tested using The Stilt Safe 3200MHz timings preset but TRC 44 & TRFC 256)
3333MHz set to 0.912V in UEFI (Tested using The Stilt Fast 3333MHz timings preset)
3400MHz set to 0.925V in UEFI (Tested using The Stilt Fast 3466MHz timings preset)
3466MHz set to 0.956V in UEFI (Tested using The Stilt Fast 3466MHz timings preset) Note: If CPU stock
3466MHz set to 0.968V in UEFI (Tested using The Stilt Fast 3466MHz timings preset) Note: CPU PState 0 OC 4.1GHz

All testing above so far on UEFI 0601, RAM used F4-3200C14D-16GTZ, CPU was stock. I found at UEFI defaults the SOC was ~0.800V, I just picked 0.900V for my starting point. Perhaps 3200MHz could be attained with less SOC on my CPU, I may test this at later stage.

Is RAM MHz king?

As with Summit Ridge, Pinnacle Ridge also at >3200MHz RAM timings play a bigger role than frequency. This article pits The Stilt's 3466MHz preset found in "DRAM Timings Control" of UEFI against upto 3600MHz C16.

What RAM to get?

Samsung B Die single rank/sided is the most favoured RAM IC for Ryzen, in useful links is a thread with helpful info when making a purchase. In regard to QVL I share The Stilt's opinion, link. I have used non QVL RAM with C6H, ZE and plan to do so with C7H. There are many user shares of having bought QVL listed RAM and not gaining higher speeds. Reference below image to know what is OC RAM speed on Ryzen.

Ryzen gen 1 officially supported max 2666MHz for single rank/sided in 1 dimm per channel configuration, gen 2 is 2933MHz. This increase may have been gained by tweaks to silicon and or AGESA, or perhaps eaten into OC headroom, I have no idea. In a nutshell I'd say keep expectations realistic, higher density RAM usage equals lower attainable MHz, generally speaking. Be prepared to put in some time and effort to gain higher speed RAM. Below is a table from Ryzen gen 1 reviewers guide, but with added gen 2 officially supported RAM MHz info.



Note: Only SR config has gained increased RAM MHz officially, this information was confirmed via another mobo maker and consulted The Stilt.

Testing RAM stability?

HCI Memtest and GSAT have been mainly what I have used. Little while back a new app has hit the scene, RAM Test by Karhu Software. This app was developed with the support of Ryzen owners on a Finnish forum. When I read The Stilt approved of this app that was all I needed to know to recommend it also. There is also a thread on OCN regarding this app, link.

Below listed sections moved to post 3:-

RAM Timings Info

How to use Flashback (ie updating motherboard UEFI)

How many ways does the board post!?

Some of what I do with OC'ing

Creating Window 7 ISO with drivers for usage on C7H

FAQ

Useful Links

[gupsterg]

Some timings I've used so far on the Micron E 2x16GB, link.

I managed to get SOC down to 1.025V as well, in this ZIP is most recent testing, 3x AIDA64 link.

Previously on 4x8GB (single sided/rank) when using 3800MHz, I had seen I could set tRTP 5, but it would still be 6 (left screenie). On 2x16GB (dual sided/rank) you can set below 10, but it does not take (right screenie).



*** edit ***

On OCN where I also posted this information a member pointed out that tRTP needs to be even when using gear dowm mode, link.

This prompted me to read a Micron DDR4 white paper and found at end of page 93/top page 94:-

When operating in 2N gear-down mode, the following MR settings apply:

• CAS latency (MR0[6:4,2]): Even number of clocks
• Write recovery and read to precharge (MR0[11:9]): Even number of clocks
• Additive latency (MR1[4:3]): CL - 2
• CAS WRITE latency (MR2 A[5:3]): Even number of clocks
• CS to command/address latency mode (MR4[8:6]): Even number of clocks
• CA parity latency mode (MR5[2:0]): Even number of clocks

So with GDME below must be even:-

• tCL
• tWR
• tCWL
• tRTP

Now I have tRTP of 8 with 3800MHz.

*** edit ***

A Samsung DDR4 PDF (page 82/83) also has same information as Micron.

For the operation of geardown mode in 1/4 rate, the following MR settings should be applied.
CAS Latency (MR0 A[6:4,2]) : Even number of clocks
Write Recovery and Read to Precharge (MR0 A[11:9]) : Even number of clocks
Additive Latency (MR1 A[4:3]) : 0, CL -2
CAS Write Latency (MR2 A[5:3]) : Even number of clocks
CS to Command/Address Latency Mode (MR4 A[8:6]) : Even number of clocks
CA Parity Latency Mode (MR5 A[2:0]) : Even number of clocks

*** edit 2 ***
Gear down mode is part of JEDEC DDR4 Rev B specification, so it's not a manufacturer specific implementation.

[Athalayogi]

anyone can help me.
what about temperature pch crosshair vii hero with 9 3900x?
i don't want to build with x570 because that have fan on pch.

[gupsterg]

RAM Kit: BLS2K16G4D32AESB

DRAM CAS# Latency [16]
DRAM RAS# to CAS# Read Delay [19]
DRAM RAS# to CAS# Write Delay [19]
DRAM RAS# PRE Time [17]
DRAM RAS# ACT Time [43]
Trc [60]
TrrdL [9]
Tfaw [36]
Twr [16]
TrdrdScl [4]
TwrwrScl [4]
Trfc [580]
Trtp [8]

3x AIDA64 result link.

~10hrs stability testing link.

Athalayogi wrote:
anyone can help me.
what about temperature pch crosshair vii hero with 9 3900x?
i don't want to build with x570 because that have fan on pch.

As I have water cooled GPU tends to be ~40-50C depending on room ambient, how I'm using PC, etc. I have seen some at ~70C where a GPU is say air cooled and dropping warmed air in area. Even those temperatures are not an issue for PCH.

[madzh]

Friends, please tell me. Can I put two m. 2 drives (SATA or pci-e) so that I don't lose 16x for the video card on the Crosshair VII motherboard? For example, install nvme ssd in m. 2_2, and m. 2_1 sata ssd in m. 2. In this combination of drives, how many pci-e lines will the video card have left? Will anything change on the pci-e lines if you put the Risen 3900x processor (for example).

[bluestang]

I wouldn't worry about losing the x16 for the GPU. You're not going to saturate x8 let alone x16 quite honestly.

[gupsterg]

Just wanted to share this ZIP, order files by time, ~49hrs back to back mixed testing of 3800C16@1.35V.

I have used several kits of G.Skill, all Samsung B die, bins 3200C14, 3600C15, 4000C18. None of those have given me the OC buzz that the Crucial Ballistix Sport LT 2x16GB kit (BLS2K16G4D32AESB) have. Main reason being is for ~£100 I got 32GB of DDR4 with nice OC headroom and not lacking in performance vs Samsung B die.

Phenomenal value for sure.

madzh wrote:
Friends, please tell me. Can I put two m. 2 drives (SATA or pci-e) so that I don't lose 16x for the video card on the Crosshair VII motherboard? For example, install nvme ssd in m. 2_2, and m. 2_1 sata ssd in m. 2. In this combination of drives, how many pci-e lines will the video card have left? Will anything change on the pci-e lines if you put the Risen 3900x processor (for example).

bluestang wrote:
I wouldn't worry about losing the x16 for the GPU. You're not going to saturate x8 let alone x16 quite honestly.

+1 to bluestang comment.

I use 3x NVMe on C7HWIFI. One is fitted to M.2_1 motherboard slot, then I use the UEFI bifurcation setting to change a PCI-E slot to 4x/4x, using ASUS Hyper M.2 x16 Card I have 2x NVMe on there. My GPU only gets 8x and performs as it did with 16x.

There is ~2-3% on average performance difference on RTX 2080 Ti by going 8x vs 16x, link.

[leo5111]

is there a later bios then 3004? not sure what issue is im on win 10 64 and sometimes it seems to hang for a few seconds opening a hard drive i ran every spy/virus checker im clean that way

[liakou]

leo5111 wrote:
is there a later bios then 3004? not sure what issue is im on win 10 64 and sometimes it seems to hang for a few seconds opening a hard drive i ran every spy/virus checker im clean that way

Go to your Power plan settings, Advanced, then under Hard drives, change the "turn off hard drives after" to Never.

[leo5111]

liakou wrote:
Go to your Power plan settings, Advanced, then under Hard drives, change the "turn off hard drives after" to Never.

its been set that way, also cpu is 3900x

[Gundam0081]

man i really hope they give good bios updates for this board when ryzen 4000 series hits,

also what are some good overclockable memory kits for this board??