Hey there, does someone here know if the itx range of AM4 boards will receive an uefi update shortly, similar to the atx ones last week? I am on the fence about returning my board because of the issue with Raven Rigde since february.

I really want them to fix those known issues - like the boot issue after the power has been turned off.

When My PC has been unplugged I  have to reset the board with the jumper to be able to make it post and thereafter reenter all the settings to the BIOS.

That is so annoying when you - as me - have to move the PC around. That is exactly why I decided on the ITX form factor.

I was however glad to see the update for the ATX boards - That means that they haven't abandoned these boards completely for the newer ones.

SYS64738 wrote: As I wrote in my posts in this topic the settings reset is my biggest issue along with the missing agesa update.

The beta BIOS available for the AB350 ITX seems to have fixed the settings lost after power loss, IMO it works better than the current stable, I mean with the stable 4.60 every time I pulled the power cord next time at boot the RAM required the "training" cycle and I was like - "WTF? How AMD could call this sh!t ultrafast boot?". Beware tho that in this forum I've read it breaks something else (USB3 ports maybe?).

I check nearly every working day if there's a new BIOS available (I don't suppose they would release during the weekends), but with new chipsets (X470, B450) and new MB models I predict a looong frustrating wait.

Actually my new PC (coincidentally also with a Raven Ridge 2400G) is collecting dust on my desk for a bug on Linux during boot that I suspect is related with BIOS/AGESA (MBs from different brand or models don't expose the bug).

Sigh...  ;__;

wino wrote: Hello boys & girls! The gods have smiled upon us today! Let the journey begin!

Oh my, an update. Yay!  \^__^/

Let's see if I can install it without issues...

Edit: update completed. First thing I've noticed is I can't find the menu entry to configure the "RAM training attempts" that was available on the 4.51D beta version. Not a big deal.

Ok, I did also test the new UEFI. The new AMD Microcode makes a big difference so now you can actually game on the 2400g with stock settings. I did not test the ASRock OC for the IGP on the Main Page yet but it is nice that they included this setting.

My second major issue unfortunately was not fixed. Despite metioning it in the change log the RAM frequency gets still reset to the JEDEC frequency (1018 MHzx2 in my case) instead of keeping the XMP profile. Also the UMA Mode of the IGP and all settings after that (e.g. framebuffer size) get reset. Saving a profile is also no option as the UMA Mode (under NBIO settings) is not saved to the profile.

Steps to reproduce (with my Ryzen 5 2400g configuration, see signature, RAM is on QVL):

- turn system on and set default settings in UEFI. Save an reboot. Result: Default RAM fequency (no XMP), Framebuffer size of 1 GB (also default).

- Enter UEFI again, set XMP Profile and framebuffer size to 2 GB, save User Profile, save settings reboot. Result: XMP profile and according frequency is applied, Framebuffer size is 2 GB.

- Shut down, cut wall power from PSU, wait a minute, plug power into the wall again an turn on the system. Result: Fans spin up for half a minute or so, then the system shuts down. This repeats three times. Then, the system posts normally, but both the XMP profile and the framebuffer size are reset to default values. Entering the UEFI and loading the user profile reapplies XMP settings but not IGP settings.

I am really disappointed and will try to return the board. It is really a shame as the hardware is exactly what I want from an AM4 ITX board

EDIT: I played around with the RAM frequencies using the XMP profiles and it seems to me that profiles/frequencies above 3066 MHz do not stick and lead the resetting the IGP settings. The RAM does run perfectly stable with 3200 MHz but does seem to produce the behaviour on startup after cutting power. Is there any option to turn certain checks off? I think I read somewhere on this forum that there is an option to disable memory checks.

Do you mean the "memory training" that happens at boot? As I wrote above, there was an option on the (beta) FW 4.51D but I wasn't able to find it on the latest 4.70.

Yes, thats what I meant. Sorry I missed that you mentioned it. It is strange that one can stress test and use the system with XMP-3200 but not "cold boot" it after power loss. The XMP settings in the UEFI also preselect the 3200 profile.

dspcmn wrote: SYS64738 wrote: Is there any option to turn certain checks off? I think I read somewhere on this forum that there is an option to disable memory checks. Do you mean the "memory training" that happens at boot? As I wrote above, there was an option on the (beta) FW 4.51D but I wasn't able to find it on the latest 4.70.

-------------
https://valid.x86.fr/wgjzle" rel="nofollow">

WinWiz wrote: This new bios definitely improved the "fast ram cold boot" issue. Before this bios my 2x8 GB 3000mhz Corsair ram would reset to 1066 MHz if ram frequency was above 2933mhz. Now I have no issues at 3200Mhz

Nice. I am unfortunately just really confused with my setup. Using Typhoon Burner I could read from the SPD that the G.Skill Trident Z Kit I have is using single rank Samsung b-die chips which are supposed to be the most compatible option for Ryzen.  Ryzen Calculator also gives me timings on the save option which are way sharper (14-14-14-30 as V1 or 16-16-16-32 as V2) than the ones of the XMP-profile (16-18-18-38) and do not work also.

did you change all the fields given in ryzen calculator ? eg all the rqz and clldo etc

they all must be entered

Yes I did. The only thing which I was not entirely sure about were the tRFC(2/4) values. RC gave me floating point values here and the asrock settings only allow for integers. So I tried rounding up and down and also the values from the tRFC calculator which were entirely different.

Edit: And I even tried settings the values using ASRock OC (decimal) and AMD CBS (hex)

Before this latest bios I also had this issue so I do understand your frustration, but above 2600Mhz the performance increase is minimal, so I don't believe you would really fell the difference between 2933 and 3200Mhz ram.

Have you tried to increase the soc voltage on or two steps above the default?

If you really want 3200Mhz you could try command rate=2 -or enable gear down mode...

You could also disable ram training but then I'm guessing your system would hang/freeze on cold boot instead of booting with bios default speeds, might be worth a try...  

I know its strange how fast ram can seem stable in prime, memtest and games but fail cold boot, but I have experinced this on overclocked athlons many years ago, so its not unprecedented.

Any advice on how to disable memory training in this UEFI version?

gade wrote: 2. I take it that because I do not see the Memory Training option anymore that means it is off by default now?

Not from my experience (AB350 ITX).

Every time there's a power loss some UEFI parameters are lost, one of these is the XMP2 profile set for the RAM, and the "Memory Training" is executed next time the PC is started...

I find it bizarre that the official description of the new FW explicitly mentions

2.Keep the XMP setting after system resuming from G3 state.

and G3 is AC loss FWIK.

So did they failed to fix it or are we doing it wrong?

Unfortunately they failed to fix it. But why they wrote it is ok? This is pathetic... 

It only happens if I raise the clocks on my RAM above 3266Mhz. So something I your setups is not "boot stable".

I suggest you try a mild OC (like stock +5%) any verify that cold boot keeps this mild OC. If +5% is ok then try 10%-15%-20%... Until you find the spot where cold boot becomes and issue.

WinWiz are you using CSM mode in UEFI?

(edit: Actually Pstate OC is working,but CPU-Z only show P0 frequency)

What are your XMP setings?

I don't know if I should open another thread and honestly don't know if the results are different with previous FWs, anyway I've found out last week that chassis fan headers behaves differently (AB350 ITX FW 4.70).

I have a PWM fan that's rated for 1850 RPM (+/- 10% according to manufacturer): when connected on CHA_FAN1 (the one that can also power a water pump) the internal sensor reports only ~1400 RPM if set to full speed on the BIOS.

I didn't cared as in idle usually 500~600 RPM are enough to keep the system temperature sensor to 44 C (25~26 C room temperature) and the fan is inaudible at that speed, the problem was that I noticed after resuming from S3 (suspend to RAM) the fan didn't spin up, not even when the temperature sensor reached ~50 C (using a customized profile).

I spent a few days trying to find a reasonable compromise with fan speed and noise and trying different options without success, until I tried the CHA_FAN2 header. Surprise: when set to full speed the sensor reads ~1800 RPM and no more stopped fan after resuming from S3.

Seems that there is just a new uefi version 4.71 out for the AB350 ITX- I have no idea what it fixes but it is online on the German vendor site I think. There is a link to it in the post 

I am a little reluctant to just fire it up since I prefer to have others write about possible issues first ha ha- But ok- We will see what it fixes.

Good job ASRock (and J Z i think)  But please, take care of this second annoying problem...

The result is 4.71 is 0.4% faster than 4.70 

Greath job Asrosk!!

wino wrote: 4.71 UEFI finally has fixed one of the two most annoying problems - XMP reset on power loss  It was supposed to be fixed on 4.70, but was not.

Flashed the 4.71 beta and I can confirm the above on a AB350 ITX model. It also retain the iGPU configurations on power loss when you use an APU.

Wow. I use this MB from one year. So much waiting and complaining, and it was enough to open a ticket ... A ticket to paradise ;)

Do you have a link for 4.71A?

there is no Dram Boot Voltage option (and probably never was), but there is MEM VDDIO and VTT MEM.

Ok, after some testing this seems to be the most enjoyable experience so far. Now that the basics work I looked deeper into the options and noticed some bugs:

1. When the AsRock OC mode is used, you can not disable Core Performance boost since it switches back to Auto. This forces you to use AMD CBS mode if you want the Raven Rigde CPU portion run much cooler since your focus is on the iGPU. I think this may be an old issue but I am pretty sure somewhere on my update journey I had Core Performance Boost disabled with ASRock OC enabled.

2. The settings on the main page for iGPU clock speed and voltages seem to do nothing. I have to use the AMD CBS options hidden in the advanced tab to try different speeds.

Can anyone else confirm these issues or has found more bugs worth reporting?

I can not confirm any of your issues.

The GPU clock and voltage on main page of advanced settings works just fine, IF you leave the CBS GPU freq/voltage on auto.

SOC voltage sets the GPU voltage at idle/low load, and change to GPU value when loaded. -So for my 2400G I have SOC voltage at 1.05 and GPU voltage at 1.275. This enables the Vega 11 GFX to run at 1600MHz without generating a lot of heat when idle 

Good to know, to make it sure you mean 4.71 beta is the one that fixed cold boot problems?

On 3.00 I have "stable" memory (run stressapptest 3 hours, run HCI memtest up to 1000% on all instances and both no errors), but booting is somewhat tricky sometimes, even resets. But after the second training it always boots and I have never got my memory settings reset because of too many training failures.

It's annoying but I can live with it. But if later bioses solve this I guess I will try them in the future. Only thing is stress testing everything again is kind of a bother, as different AGESA can make or break stability of your current memory settings.

Do anyone know if after 3.6 bridge bios, you can "bridge back" to pre 3.6 if you follow the reverse bios flashing order?

I have constantly been tweaking my system since I got it so now I know how my systems twerks.

What RAM, CPU and cooler do you have and what is the goal of your Overclock, gaming performance? General (not 3D) performance? Or are you looking for a balanced OC?

Personally I sacreficed a little CPU performance to gain 3D perfomance so my 2400G is running like this:

CPU: 3800Mhz

Vcore: 1.375V

SMT: Enabled

DRAM: 3200Mhz

GFX clock: 1550Mhz 

(played some games today and had some driver crashes so lowered this down from 1600)

GFX Voltage: 1.275V

(This is a bit high but my GFX needs this for stability at 1550Mhz, and I figure if it kills my CPU in the long run I have an excuse to upgrade next year. And my Noctua NL-L9a-AM4 can handle the heat)

CPU SOC voltage: 1.1V

CPU Vcore offset: 0

VDDCR SOC offset: 0

DRAM Voltage: 1.25V

(My corsair CMK16GX4M2B3000C15 with Henix dies are rated for 3000Mhz-15-17-17@1.35V, but I can't run them above 3200Mhz-16-17-17 no matter how much voltage i feed them. And @1.25V I can clock my CPU to 3800 -1.35V DRAM voltage limmits my CPU OC to about 3.700MHz) 

2.5V Voltage: 2.5V

+1.8 Voltage: 1.8V

1.05V_Prom Voltage: 1.08 (auto)

Thats it.

Did you change any settings, like RAM frequency, going from bios 4.6 to 4.7 ?

I think it's only supposed to boost to 3.9 for short bursts, that is how it behaves when the GFX isn't overclocked. The permanent boost to 3.9GHz with 100% load causes very high temperatures and makes my system unstable.

But I can not remember if earlier BIOS versions behaved like this..?

With 4.71 on my board trying to set the GPU overclock to default will not work, a UEFI reset is needed, or fiddling in AMD/CBS like with older UEFI's

Yes that is the way its working.

As far as i know Asrocks OC menu is actually a more user-friendly interface for some of AMDs frequently used settings. It gathers the various settings in one page and lets the user enter values in decimal instead of hex vid's.

I don't know why Asrocks GFX settings will only increase AMD's values and not reset them to auto, in my book this is only a cosmetic bug.