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]

mikev190 wrote:
Yeah I've noticed that too. My cpu doesn't go above 1.45v and all core voltage is more like 1.337v instead of 1.325v. I will post some screenshots when I get back to my computer for you!

I am using a BeQuiet! Dark Rock Pro 4. I only got it the other day and sadly the paste that came with it had a air bubble in so there was hardly any in the tube and the paste I had from awhile back has dried up. I'm getting temps of 56-60c playing WoW classic which only uses 10% of the cpu and 85c+ just running AIDA64 stress test. Hoping with some decent paste and a good 'X' method will sort my temps out as they don't match the reviews of the cooler. I think if I wasn't thermally limited it would be better for sure.

I will check out eBay and see what I can do. I need to start saving for Xmas for the little ones so need to work it all out. My dad has a 3600X and his boosts to 4.3Ghz stock in games. It's a shame mine doesn't. I blew my luck on my old 2700X which was a good sample.

Yeah ~1.45-1.46V sporadic peak when I use a PBO OC is max I've seen. Usual for say all cores boost average of ~4.3-4.34GHz with PBO OC, averages ~1.425V give or take few mV. Usual for stock is ~1.325V for all cores average boost of ~4.2GHz, only if I use unmodded AGESA ComboPi-AM4 1.0.0.3AB/ABB UEFI do I see ~50mV on top of that case test and then a PBO OC does not break past ~4.275GHz, but still uses ~1.45V+.

When I ran AIDA64 stress test on 2700X I saw some nasty temps even on water, I can't recall what version it was, wasn't old as always use what is latest at the time. It was when I picked a specific CPU test. For me usually it's Kahru RAM Test, P95, RealBench, Y-Cruncher, some encoding and just general benching, etc. I've stuck to spreading TIM using a strip of plastic off say an amazon gift card. Works well for me, used on Intel/AMD/Threadripper.

[mikev190]

gupsterg wrote:
Yeah ~1.45-1.46V sporadic peak when I use a PBO OC is max I've seen. Usual for say all cores boost average of ~4.3-4.34GHz with PBO OC, averages ~1.425V give or take few mV. Usual for stock is ~1.325V for all cores average boost of ~4.2GHz, only if I use unmodded AGESA ComboPi-AM4 1.0.0.3AB/ABB UEFI do I see ~50mV on top of that case test and then a PBO OC does not break past ~4.275GHz, but still uses ~1.45V+.

When I ran AIDA64 stress test on 2700X I saw some nasty temps even on water, I can't recall what version it was, wasn't old as always use what is latest at the time. It was when I picked a specific CPU test. For me usually it's Kahru RAM Test, P95, RealBench, Y-Cruncher, some encoding and just general benching, etc. I've stuck to spreading TIM using a strip of plastic off say an amazon gift card. Works well for me, used on Intel/AMD/Threadripper.

I ran 12 instances of HCI with hwInfo running. I didn't run long as I don't have much time on tonight. The allcore speeds were higher than normal. Generally all core speed is a lousy 3950Mhz. Should be better tomorrow so will repost then when I have sorted my cooling etc. This is at stock with no PBO + 3600Mhz ram @ 1.4v SoC @ 1.05v cLDO VDDG @ 0.95v

[mikev190]

gupsterg wrote:
NP, just a small fish in a big pond 😉 .

Same drill as C7H UEFIs to flash. As I hadn't downed the 0002 versions for C6H, C6HWIFI & C6E, before links went down from high traffic, I could only do "official" 7403 for C6H, C6H WIFI & C6E.

Download link.



Cheers for info, not had the experience of one 🙂 .

Flashed perfectly thanks!

So far from minimal testing I am getting spikes of 4150 and rarely 4200. In games it is around 4100 but think it will be more like 4125 when this hot weather passes. It is still some ways off 4.2Ghz but it is an improvement nonetheless. PBO and AutoOC still don't work and actually lower boosting but we can't have it all! Cinebench score seemed the same as before

Thanks again!

[madamaou]

simple question, am I not seeing AMD CBS options because I'm using a x470 (crosshair vii hero) with a ryzen gen 1 (1700x) or are they indeed missing from the 2703 bios? I mean, I do have a perfectly functional, stable, cool 'n quiet (^^), p-State OC that i'm carrying from an old saved profile but if I wanted or needed to tweak something there that would be outright impossible, unless if I downgraded to a pre 1.0.0.0 agesa -- which doesn't sound like a compelling proposition.

[Vali_Pali]

gupsterg wrote:

Reverts it to what you would see on an AGESA ComboPi-AM4 1.0.0.2 UEFI.

Someone used it with a 3900X on reddit, requested/shared experience only in my inbox so far.

Confirming this takes the max boost clock to 4.625 GHz. Thanks for this.

[xeizo]

Vali Pali wrote:
Is this suppose to fix the boost behaviour?

That would be the entire point of the mod AFAIK 🙂

[WheelC]

gupsterg wrote:
Thanks to The Stilt's info I have managed to mod SMU FW 46.34.00 into C7H & C7HWIFI UEFI 2703 & 0002.

2703 is current "Official" and 0002 is one with ASUS WMI possibly fixed.

PBO+150MHz set on my R5 3600 on UEFI 2501, 2703 and modded 0002, can be seen in WMVs in this ZIP.

Disclaimer

Download & use UEFIs at your own risk, I experienced no adverse effects so far, but only limited testing done on one configuration.

I am just a user sharing my modded UEFI, I am not qualified and or employed in the field of computing.

Download link for mod UEFIs.

Use flashback method of flashing UEFI.

i) Change UEFI filename as C7H.CAP or C7HWIFI.CAP depending on your model.

ii) Copy to USB stick, it must be formatted to FAT32. File must not be in a directory.

iii) Load UEFI defaults prior to doing flashback and let mobo repost and shutdown.

iv) Make sure USB stick is in Flashback port (ie one which has marking BIOS)

v) Press and hold the Flashback button (blue box, above image) ~5secs. Blue LED on the button will flash slow for a few seconds and quicken as it starts update process. At the end of process it will extinguish, you are ready to boot up. If the blue LED stays constant the UEFI file was not found on USB stick, so check USB stick is correct format and filename is correct.

Thank you much it is working great! About 70'C with Cinebench R20 multi, drops down to 64' with -0.08750 offset. Thanks to all involved! This is with 0002

[adr82]

gupsterg wrote:
Thanks to The Stilt's info I have managed to mod SMU FW 46.34.00 into C7H & C7HWIFI UEFI 2703 & 0002.

Just wanted to chime in with another "thank you" for this, it's nice to have these custom combinations of BIOS+FW available as the older AGESA certainly makes a difference.

Just flashed your modded 0002 on a C7H+3700X and spent a while messing around, some browsing plus a CB20 single+multicore run. I actually seem to be getting even better boosting than I previously did on 2406/2501. 1 core @ 4.475GHz, 1 @ 4.45GHz, 1 @ 4.425GHz, 2 @ 4.40GHz, 3 @ 4.375GHz. When I used 2406/2501 I never saw any core go beyond 4.425GHz. CB20 multicore is up by ~100 points vs 2606 as well.

Definite thumbs up from me!

[gupsterg]

mikev190 wrote:
I ran 12 instances of HCI with hwInfo running. I didn't run long as I don't have much time on tonight. The allcore speeds were higher than normal. Generally all core speed is a lousy 3950Mhz. Should be better tomorrow so will repost then when I have sorted my cooling etc. This is at stock with no PBO + 3600Mhz ram @ 1.4v SoC @ 1.05v cLDO VDDG @ 0.95v

To me it would seem your CPU sample is on totally differing voltage/frequency curve than mine :eek: , due to it's unique silicon characteristics.

See my stock CPU OC RAM and PBO+150MHz OC RAM WMVs in this ZIP. Compare average CPU frequency and average CPU Core Voltage SVI2 TFN.

madamaou wrote:
I feel I need to clarify that it's not exactly that I have no AMD CBS options, but only a severely reduced set.
I've been wondering if that's just how the new bioses are or if it's the case that a gen 1 ryzen are locked out of most options for whatever reason.
All I can acces is in the pic.

If the options have been suppressed and not removed I maybe able to bring them back. I will check if I have time today.

adr82 wrote:
Just wanted to chime in with another "thank you" for this, it's nice to have these custom combinations of BIOS+FW available as the older AGESA certainly makes a difference.

Just flashed your modded 0002 on a C7H+3700X and spent a while messing around, some browsing plus a CB20 single+multicore run. I actually seem to be getting even better boosting than I previously did on 2406/2501. 1 core @ 4.475GHz, 1 @ 4.45GHz, 1 @ 4.425GHz, 2 @ 4.40GHz, 3 @ 4.375GHz. When I used 2406/2501 I never saw any core go beyond 4.425GHz. CB20 multicore is up by ~100 points vs 2606 as well.

Definite thumbs up from me!

No problem, thank you for feedback and glad to read it shows gains 🙂 .

Just as a share to anyone wondering. This ZIP has ~7hrs P95 4K 4096K 27000MB non AVX run, frequency/voltage seems as it would be on unmodified AGESA 1.0.0.2/1.0.0.3 UEFI. So the SMU FW mod is not making CPU unsafe in high load/current situation from what I have seen.

[mikev190]

gupsterg wrote:
To me it would seem your CPU sample is on totally differing voltage/frequency curve than mine :eek: , due to it's unique silicon characteristics.

See my stock CPU OC RAM and PBO+150MHz OC RAM WMVs in this ZIP. Compare average CPU frequency and average CPU Core Voltage SVI2 TFN.

Your voltage is much lower! I wonder if there is anything I can do to lower it. I mean manually I can run 4.1GHz @ 1.256v so it feels like it is putting way too much voltage in for the clock speeds. I've just sorted my cooler so I will do some more tinkering! I've noticed when the CPU is around 30c it will boost to 4.2GHz but with that voltage it is almost impossible to keep it that cool without water. Doing offset voltage lowers the clock speed

Hwinfo new beta has added a CCD temp I've just noticed


HWiNFO v6.11-3900 Beta available.

Changes:

Fixed reporting of ES/Production stage for some Zen2 CPUs.
Fixed measuring of BCLK for some Intel CPUs under Windows 7.
Added reporting of CPU High Temperature Clock Limit for AMD Zen2 family.
Added reporting of CPU Automatic Overclocking Offset for AMD Matisse family.
Added a new (more reliable) method for measuring BCLK on AMD Matisse systems.
Added monitoring of Memory Controller Clock (UCLK) for AMD Zen family.
Added preliminary support of Intel Elkhart Lake.
Fixed reporting of NVMe capacity when LBA data size > 512B.
Added reporting of CPU Thermal Trip Limit and HTC Temperature Limit for AMD Zen.
Updated reporting of CPU VDD and SoC SVI2 current/power on some MSI mainboards.
Added monitoring of per-CCD Tdie temperatures for AMD Zen2.

https://www.hwinfo.com/download.php