My experience with C7H UEFI 2406 & R5 3600

[gupsterg]

Last update: 10/07/19 16:00 UK BST

I have a C7H given to me by ASUS at launch. It has had 3x 2700X used on it and lost count how many UEFI flashes I have done testing things. It is also the early board which has bugged chipset voltage readout in SW monitoring, ~+50mV from actual. So in UEFI screenie/HWINFO you will see ~1.1V when actually it is 1.05V.

From having read others posts I thought I would have a mare with swap over to R5 3600, to my delight my experience was to the contrary of what I had read .

As the current 2700X had been in the rig since Jan 19, I was expecting it to come out of the socket with the water block , but no it did not pull the CPU . I had ran the system up prior to warm her up .

Any how quick painless remove and swap out to R5 3600. I did not even reflash UEFI, did clear CMOS from rear IO and BOOM! first time she booted! .

The Hero everyone wants is the C7H!

This album link has some screenies from UEFI 2406, continuing now with other bits of testing, etc

OS was freshly installed yesterday when on 2700X, I just had a screen doing some updates once booted on R5 3600. Seems AOK so far and will update benches, etc here ASAP .

HW in use:-

- R5 3600 Batch Code: BF 1922 SUT
- C7HWIFI
- F4-3200C14D-GVK
- Sapphire HD 5850 Toxic 1GB
- Intel 660P 1TB
- Cooler Master V650

Cooling used:-

- EK Supremacy EVO (Acetal+Copper)
- Magicool 360 G2 Slim radiator
- 3x Arctic Cooling P12 PWM
- 1x Arctic Cooling F12 PWM
- EK DDC 3.2 PWM
- Coolant is distilled water with ~20% mix of Mayhems XT-1 clear concentrate
- Arctic Silver 5 spread by hand using plastic strip from a plastic card (OFC )

All fans are using a 4 in 1 cable, connected to CHA_FAN2 and powered by molex. The water pump is powered by molex and gains PWM from H_AMP header. All fans/pump run a UEFI based profile, based on water temperature, T_Sensor1 is used.

2406_R5_Base_setting.zip

Here HWINFO shows some idle stats, etc, room ambient temp is ~24C.

Windows 10 Pro x64 1903 with AMD Chipset driver 1.07.07, power plan is Ryzen one which has UEFI CPP2 link.

So far benches in this album are at base settings stated in ZIP above, room ambient ~24C.

Done an image of system to SATA SSD over USB adapter, now to scratch the OC itch of Matisse! .

Overall so far experienced smooth POSTs on mobo (done a few) and usage in OS for tests/uploading images/posting here, etc all nice. PWM on headers is behaving, so not having any issues with the SIO (ITE IT8655E), so fans/pump reacting as they should. OS has only had whatever updates it wanted via web, only driver so far installed is AMD Chipset from AMD site. Aspects of Wifi, NVMe, etc all working as they should.

Setting up 3533MHz using The Stilt 3466MHz timings has not been an issue so far. I have done some warm POSTs (ie restart from OS) and then some POSTs from shutdown, but with board having standby power (ie PSU wall plug still had power).

As Elmor had on OCN highlighted an issue on current UEFI I set up voltages first, saved settings, set timings and saved settings. When FCLK (ie fabric clk) was left [Auto], even with 3533MHz MEMCLK it did not maintain 1:1 ratio until I manually set it. FCLK not being 1:1 can have severe affect on RAM performance depending upon how far out it is from 1:1. This 3533MHz without 1:1 FCLK, link, then here is with 1:1, link. Below ZIP has settings used txt.

2406_R5_Base_3533S_setting.zip

Just like on 1xxx/2xxx gear down mode must be disabled for odd CAS not to be rounded to nearest even. Just like previous gen again tRDWR on [Auto] does train to differing values (link), I left it on [Auto] as The Stilt's preset do this.

I set SOC manually to match what I saw at stock via with DMM, so I could see what RAM MHz gains I could get at stock SOC, link. CLDO_VDDG is explained here by The Stilt link. I should have perhaps used 0.950, but set it ~40mV less than SOC, CLDO_VDDG does accept 1mV steps, at complete UEFI defaults RM shows 0.950V, link.

Today I have ended on 3600MHz, FCLK & MEMCLK 1:1, using The Stilt's 3466MHz timings, 1T, Gear Down Mode: [Disabled], AIDA64 bench result link. Latency is ~7ns behind 2700X for same RAM MHz & timings, strangely.

Stability testing album link, 3533MHz passed 1100% even though screenie shows 750%, same settings failed with 3600MHz, even though I encountered no POST issues, etc. Bumping VDIMM to 1.355V seems to have solved it as gained 1200% PASS. Just as a comparative I needed 1.385V VDIMM on same RAM kit & mobo, but using 2700X. This IMO shows clearly RAM & MOBO was not what held back 1xxx/2xxx, but IMC quality. Room ambient was ~26C for the tests stated.

Settings for 3600MHz using The Stilt 3466MHz timings in below ZIP.

2406_R5_Base_3600S_setting.zip

This has now been tested for 2400% in RAM Test and then on same POST I did 1200%, link 1, link 2.

3666MHz I have gained pass in RAM Test ~1100% so far, this was by increasing SOC/CLDO_VDDG/VDIMM and CAD Bus needed to be 30 A A A. Link 1 , Link 2, RM is showing some things incorrectly, but a lot is correct and currently only util showing as much as it does on Matisse.

So using The Stilt 3466MHz timings, Gear Down Mode: Disabled, 1T, below is so far what has passed stability testing in RT reasonably well.

3533MHz SOC: 1.025V CLDO_VDDG: 0.985V VDIMM: 1.35V
3600MHz SOC: 1.025V CLDO_VDDG: 0.985V VDIMM: 1.355V
3666MHz SOC: 1.043V CLDO_VDDG: 1.003V VDIMM: 1.365V CAD Bus: 30 A A A

UEFI 2406 does had PMU Training menu (aka IMC), link, further info ref Elmor's post here.

*** more to come soon , so stay tuned peeps ***

[NightHawkSoup]

Looking forward to your OC adventure.
especially your memory OC and infinity Fabric Settings.
Are you planing for a 1 to 1 setting for the IF with the 3200 Cl14 Sticks?

Still looking through your posted images,and LEARNING alot, AND I am liking what I am seeing so far.

another ??.... did you have any issues with the 2700X with the 2406 BIOS installed??

I have not updated my board to BIOS 2406 yet, still rocking my 2600X until new 3000 chip gets here.

NightHawkSoup

[bigtop1967]

Thanks for the info gupsterg.

I will share as much as I can when my 3900x arrives

[gupsterg]

Looking forward to your OC adventure.
especially your memory OC and infinity Fabric Settings.
Are you planing for a 1 to 1 setting for the IF with the 3200 Cl14 Sticks?

Still looking through your posted images,and LEARNING alot, AND I am liking what I am seeing so far.

another ??.... did you have any issues with the 2700X with the 2406 BIOS installed??

I have not updated my board to BIOS 2406 yet, still rocking my 2600X until new 3000 chip gets here.

NightHawkSoup

Yes plan is 1:1 FCLK with MEMCLK, I have seen my C7H & this RAM do 3666MHz C15 1T a few times with 2700X in ~3000%+ Kahru RAM Test, so yeah I wanna see what it can do on 3xxx IMC . I did not buy CPU for need, but more so to tinker and clock the nuts off the RAM with my amateur skill . Newer UEFI should give access to PMU Training menu (aka IMC), this is one reason why I wanted 3xxx as 1xxx/2xxx don't have this menu, but Threadripper 1xxx/2xxx do. I saw when I transferred a CPU/RAM from ZE to ZEA that menu helped me snag RAM MHz that I was doing on ZE, without tweaking that the TR1950X was having an issue hitting the same RAM MHz on ZEA.

I have shared a bench above of AIDA64 with and without 1:1 FCLK & MEMCLK, defo not pretty .

Only issue I had with 2406 with 2700X was my mouse did not function in UEFI, this is also the case with R5 3600. AGESA ComboPi-AM4 1.0.0.1 onwards fixes inter core latency issue which was not present in PinnaclePi-AM4. I could see ~77ns on borked AGESA vs ~70ns on early or latest AGESA with fix.

The R5 3600 is even tighter than 2700X and this ~63ns was at stock RAM (as higher RAM affect IF on 2700X it improved ns), link.

Thanks for the info gupsterg.

I will share as much as I can when my 3900x arrives

NP and looking forward to your shares for sure man! .

[Duvar]

Nice i have the same Setup as you and i am looking forward to test your great results.
Thx and greetings from Germany.

[Rockford]

Gupsterg:

Max out the Ryzen 3000 IMC.......... full out

4800mhz?

[gupsterg]

Nice i have the same Setup as you and i am looking forward to test your great results.
Thx and greetings from Germany.

No problem, glad you have found it useful.

Gupsterg:

Max out the Ryzen 3000 IMC.......... full out

4800mhz?

I can only try . I gotta say for a £189 CPU I am impressed so far. Even more so I am impressed how well even on somewhat early/buggy UEFI supporting 3xxx, usage/OC experience has been good.

[KeithMyers]

Newer UEFI should give access to PMU Training menu (aka IMC), this is one reason why I wanted 3xxx as 1xxx/2xxx don't have this menu, but Threadripper 1xxx/2xxx do. I saw when I transferred a CPU/RAM from ZE to ZEA that menu helped me snag RAM MHz that I was doing on ZE, without tweaking that the TR1950X was having an issue hitting the same RAM MHz on ZEA.

gupsterg, can you answer a question about PMU training for TR? Your post is the first I've heard about PMU training on TR that allows better RAM speeds. I currently run 3400Mhz at CL14 with Fast timings, but I could never achieve stable 3466 CL14 speeds without errors. I've seen the PMU training in my menu but I just avoided it since I didn't know anything about it. I have the Asrock X399 Fatality Pro Gaming motherboard with the latest BIOS.

Do you have any quick and dirty simple tricks and tips for the PMU menu settings that might allow my TR 2920X achieve 3466 CL14 Fast timings without errors?

[ZEEED]

Well the 7nm higher latency on memory comes from IMC being in separate I/O die I was sure it will have negative impact. I'm on 60.4 atm with 3533cl14 guess that wont last hehe

[gupsterg]

gupsterg, can you answer a question about PMU training for TR? Your post is the first I've heard about PMU training on TR that allows better RAM speeds. I currently run 3400Mhz at CL14 with Fast timings, but I could never achieve stable 3466 CL14 speeds without errors. I've seen the PMU training in my menu but I just avoided it since I didn't know anything about it. I have the Asrock X399 Fatality Pro Gaming motherboard with the latest BIOS.

Do you have any quick and dirty simple tricks and tips for the PMU menu settings that might allow my TR 2920X achieve 3466 CL14 Fast timings without errors?

Elmor posted info while back in OCN ZE thread, link.

I have not spent as much time with tinker the X399 rig as I do with say AM4 board. You'll see in the OP of ZEA thread on OCN what I did, link, there may also be a post further in thread on when I dropped to C14 on 3400MHz.

Well the 7nm higher latency on memory comes from IMC being in separate I/O die I was sure it will have negative impact. I'm on 60.4 atm with 3533cl14 guess that wont last hehe

Inter core latency is better than 2700X on R5 3600. Also see the red line vs green of 2700X here.