Prebuilt [SFFn] ASRock's DeskMini A300 - Finally!

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
Unfortunately, people don't really read suggested things and every two pages I read same issues over and over again. What has been discussed two pages ago is all of a sudden forgotten. But the issue remains. And not everyone seems to be affected

@limsandy as you were pushing your RAM towards 3200MHz, have you slightly adjusted SoC VID to like 1.1375V? To exclude bad memory layout on the mainboard?

Renoir has improved IMC, so again, it should be able to handle 3200+MHz with no issues. That's why I'm thinking it's the mainboard which is blocking you away from a success.

Your feedback would be appreciated.

I know, we all differ in our level of understanding. That's why a feedback is so valuable. So we can reach a common divisor, even if the troubleshooting is remote like this :)

These are things which can be ruled out, like for example missing S3 support, etc. but only if we work on an issue together. If not, it will stay a "black box" for the remote participant

Caution: 3.60S does not allow soc vid option to be set. 3.60R is the latest and newest AGESA Version which is able to do so. This is also the version I am using. Let's start somewhere :) I'm really willing to help ^^ and so to be able to collect experience here for everyone :)

Since people with Raven Ridge and Renoir experience the same system behavior, I suggest the common divisor to be the mainboard hence failing to supply stable voltage towards the SoC where the IMC resides in.

My post should be no offense nor any kind of fronting :)
 
Last edited:

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
Since bios 3.60L is any CPU Microcode and gpu Microcode in bios included . I see that of screens of one person who analysed the bios with special bios tools
Yea, I was able (+ next 3 posts) - and share the extract UEFI partition table, also I'm missing @Danlopez1222 who digged even more inside with logic analyzer.
It is not something I would rely on (backporting new microcode into the old UEFI releases)
If I get AMD and AGESA right, then the AGESA modules have to be updated (by AMD and afterwards customized by the board manufacturer) and implemented into a proper bios file. If there are internal dependencies (and there should be, as with each complex software), it may not be as trivial as copy and paste, unfortunately.
 
Last edited:

Hammerfest

Trash Compacter
Jul 15, 2019
47
40
Hello, one question. Is there any reason to buy an A300 now that the X300 is out? Thank you

X300 Pros: "Overclocking" support, "hopefully" wont be dropped like a bad habit like the A300 has?

Thats it, im not even going to include next gen APU support as they have so far fouled up with the X300 in that regard to even speak of the A300.
I had a chance to get an X300 for the same price as the A300 and passed on it.

I said earlier in this thread as long as the A300 got the proper AGESA updates and next gen APU support because they are basically the same product, I would buy the X300, hasn't happened, haven't bought it.

IMHO: Avoid BOTH at the moment because the latest "offical" BIOS release's from both are horribly outdated AGESA/Module wise and are buggy unstable disasters.

I run an A300W with the 3400G 32GB DDR4-3200 CL16 RAM 1xNVME and 2xHDD on the 3.60R (or is it S, i cant recall, whichever allows SoC voltage adjustment because the APU NEEDS IT FOR FCKS SAKE to be stable even at stock) firmware.
 

limsandy

Average Stuffer
Jul 3, 2020
71
31
Gustav,

Pardon my ignorance for not reading the previous few pages. After reading your post, I immediately flashed back to BIOS 3.60R and set SoC VID to 1.1000V. Set the memory to 3200 MHz and timing to 16-18-18-18 @ 1.35V and set everything else to Auto.... CMD 1T and gear down disabled. I ran memtest for 30 minutes while I had my lunch and when I came back still 0 error.

Then I set timing to 16-17-17-17 while keeping everything else the same. Memtest returned so many errors within 10 seconds.

So now I'm back to 16-18-18-18 and I am playing around with the other bottom settings. Bandwidth increased from 16.66GB/s to 17.15GB/s but I still have to test for stability.

Thanks Gustav.


Updates:

I have successfully set VRAM to 1GB instead of 2.

Whenever you run Memtest without errors but get this kind of error message while loading Windows, then it's probably your SoC VID is not high enough and needs some juice. I have now set my SoC VID to 1.125V


Edit: I am having a problem with VLC. Playing back a 1080p x265 video from my external hdd would stutter in my A300 but playing the same file in my main rig is fine. Display driver problems?
 
Last edited:

smolfactor

Chassis Packer
Apr 16, 2020
14
4
IMHO: Avoid BOTH at the moment because the latest "offical" BIOS release's from both are horribly outdated AGESA/Module wise and are buggy unstable disasters.
what would you suggest, then? I'd like to build a sff pc with two hdd in raid, an ssd, and an apu.
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
@limsandy Happy to hear about your success! This confirms even more the mainboard-related issues :) Greetings

One small note: When you set your memory timings @XMP the mainboard does set automatically +0.1V on the SoC VID (resulting in 1.1V by default@XMP) - but this is not enough, as we observed.

I did also some aggressive timings, but they were not stable at 16-17-17-17 and even soc vid at 1.1375V, so I returned to 16-18-18-18, 1T :) at 3200MHz.

I was able to boot at 3333MHz using 18-18-18-18 timings. But this was also not that stable and I did not like the Bus/Mem divisor - to many fractions out of sync.

I did basically never encounter BAD CONFIG DATA BSOD but sometimes system freezes randomly, something with THREAD_STUCK... BSOD thought :/
 
Last edited:

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
what would you suggest, then? I'd like to build a sff pc with two hdd in raid, an ssd, and an apu.
I cannot name you the specific components, anyway I don't also know what your budget is, but for orientation you could use following:

- Ryzen G-Series (later being able to upgrade to dedicated PCI-E gfx)
- asus/msi b550-based mainboard, MicroATX (with 1 to 3 PCI-E slots, IO shield with Displayport and HDMI, 2 better 4 DIMM slots)
- full size 16GB ddr4 (at least 3600MHz to be future proof)
- nvme drive like crucial p1 or wd sn750
- modular power supply, at least 600W
- case fitting your place

~700€ I think should be fine. And you can upgrade your graphics once they become available at MSR.
 
  • Like
Reactions: smolfactor

Valantar

Shrink Ray Wielder
Jan 20, 2018
2,201
2,225
- modular power supply, at least 600W
I have to ask: why on earth would someone wanting to build a tiny APU-based PC need a 600W PSU? That is downright crazy. mATX doesn't make sense either - remember, the person you're responding to wants an A300 alternative, not any desktop PC.

what would you suggest, then? I'd like to build a sff pc with two hdd in raid, an ssd, and an apu.
Not really the thread for it, but what's your use case beyond wanting room for storage? If it's a general-use PC without specific hardware needs but a desire for a powerful APU and small size, I'd find a small case you like (I went for the Lazer3D HT5 for my HTPC, and it's gorgeous and easy to build in, plus it fits everything you want) and go for a simple ITX+Renoir setup. The only real challenge would be powering it, but if you're fine with power bricks that's not an issue either, just get a HDPlex combo, ArchDaemon+suitable 12V picoPSU, or similar. You won't need a powerful PSU - my ITX 4650G pulls ~110W at the wall in gaming loads with a small iGPU OC and RAM at 3800MT/s.
 
  • Like
Reactions: smolfactor

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
I have to ask: why on earth would someone wanting to build a tiny APU-based PC need a 600W PSU? That is downright crazy. mATX doesn't make sense either - remember, the person you're responding to wants an A300 alternative, not any desktop PC.
Hahaha you're right. I was missing somewhere the use case and leaned to much out of the window :D I guess some time in the listing I applied my use case. 300W should be enough. But forget the dedicated GPU upgrade (PSU will be limiting)
Edit: right, there are PSUs in smaller format than ATX, like pico for example. I have to agree, they will be more than enough.
 
Last edited:

Valantar

Shrink Ray Wielder
Jan 20, 2018
2,201
2,225
Hahaha you're right. I was missing somewhere the use case and leaned to much out of the window :D I guess some time in the listing I applied my use case. 300W should be enough. But forget the dedicated GPU upgrade (PSU will be limiting)
Edit: right, there are PSUs in smaller format than ATX, like pico for example. I have to agree, they will be more than enough.
Lol, I had a whole post written up here before glancing up at the quote and spotting your edit. Serves me right for leaving tabs open for hours before replying :p But yeah, as I said, my ITX 4650G system peaks at 110W at the wall while gaming (I haven't measured any concurrent CPU+GPU torture loads, which would probably have hit 130 or so). That's with the GPU at 2100MHz and 3800MT/s RAM, a single SSD and a single 140mm fan. For another couple of 2.5" HDDs I would add 10-15W on top of that (20-25W if they were 3.5"). At the most, though lower with the iGPU at stock and/or slower RAM. So a PSU capable of delivering 150W continuously should do the job perfectly. I'm using a MeanWell RPS-200-12C + ArchDaemon for my HTPC, and it works like a charm in a case that would never, ever fit an SFX, FlexATX or TFX PSU. HDPlex also makes great power supply setups. Of course an external power brick rated for ~150W would also work great, and allow for the use of some really tiny cases.
 
  • Like
Reactions: gustav

limsandy

Average Stuffer
Jul 3, 2020
71
31
After 2 days of light use, I find that my system is not 100% stable. I have bumped my SoC VID to 1.15V and it still crashes from time to time. More testing is needed.

 
  • Like
Reactions: gustav

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
@limsandy Unfortunately, it can not be guaranteed that you will be able to reach a stable state of the system.
Looking forward into your findings.

@Valantar Yea, I can relate :) (F5 and browser-cache) I had to Google up your case "Lazer3D HT5" I find it appealing. What I find interesting, can you name the mainboard you're using in your HTPC?
Just curious. I can understand, you've basically designated the box to be a HTPC. I think I have still to decide what it should be in my case. I'm really happy with the performance for browsing the web, compiling tasks, VMs (since fixed NVMe issues with BIOS update/HyperV/Android Studio unable to use AMD-V extension properly), or for casual gaming.
A little bump in performance is always good, but the main point is the stability. (yes, I could go back to 2933 when it's only about that) Oh, and what RAM are you using? Non SO-DIMM? Since spending money on this rig's memory only to be used for this rig, as I do not have any use case for SO-DIMM DDR4 else is pretty much >.< I think it's understandable.

And then I received a new laptop (I mentioned earlier, Ryzen 3500U/16GB@3200/NVMe/WiFi 6) and it basically outruns the Ryzen 2400G :D amazing in both - sarcastic but also in astonished way

One more case, for under 70€:
Jonsbo V9, but those are 17,2L

Cheers

Here a bit #offtopic: if you ever wondered about the history of Windows's Task Manager by the creator himself, I may share (more nerdy than entertaining)
 
Last edited:
  • Like
Reactions: ratundtat

limsandy

Average Stuffer
Jul 3, 2020
71
31
So it turns out that the blue screens I keep getting is because of corrupted boot files. After getting fresh windows 10 installed, I haven't had any blue screens.

As someone said 3 pages ago, installing fresh Windows 10 fixed his problem. Well that's because if you didn't successfully load Windows, your boot files might be corrupted and yes the only sure solution is to re-install fresh Windows 10. And that's probably because your SoC VID is not high enough.

Long story short: Low SoC VID > blue screens during Windows boot > corrupted files

Edit:
...... which begs the question..... I can run Memtest all day to test the stability of the DDR4 memory. Is there a program I can run to test the stability of the SoC?
 
Last edited:
  • Like
Reactions: rubicoin

rubicoin

Cable-Tie Ninja
Jan 12, 2020
164
104
So it turns out that the blue screens I keep getting is because of corrupted boot files. After getting fresh windows 10 installed, I haven't had any blue screens.

As someone said 3 pages ago, installing fresh Windows 10 fixed his problem. Well that's because if you didn't successfully load Windows, your boot files might be corrupted and yes the only sure solution is to re-install fresh Windows 10. And that's probably because your SoC VID is not high enough.

that was my case i guess: after switching 2200g to 4650g windows didn't successfully load and boot files became corrupted, had to reinstall, no other workaround. good to know.
 
  • Like
Reactions: ratundtat

yuusou

SFF Lingo Aficionado
Mar 16, 2019
115
70
So basically, when you install the 4650G or 4750G, install 3.60R, and before first boot raise SoC VID? Any recommendations on voltage? I've got Crucial Ballistix 2x8GB 3200MHz kit (has 3200 XMP profile) and would like to keep it at 3200 with a 4750G. Currently running a 2400G.

EDIT: I'll be using an eGPU and Ubuntu, I don't know if that makes a difference.
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
@limsandy oh, that's very unfortunate.

I can run Memtest all day to test the stability of the DDR4 memory. Is there a program I can run to test the stability of the SoC?

Yea, memtest is only half the truth. As I mentioned earlier, even with stable memtest86+ (or whatever memtest you wish, but low-lewel) it stresses the DRAM channels not that much.

I recommend UNIGINE Valley benchmark at low resolution, so the APU tries to pull bandwidth through resulting in further a stress/stability test.
You may see artifacts shortly before your system may hang up. But then you know, the VID is not enough.
I remember MSI or ASUS stating (for Raven Ridge) that an SOC Voltage of 1.2V should not degrade your IMC. (in the long run)
Member here on the forum managed to push his Renoir up to ~1.22V on SOC rail. (I believe, use search please)

So, yea next step after properly booting into Windows is a 3D Stress test. Why UNIGINE Valley? Because only this was able to spot the voltage drops.
Furmark did run well, 3DMark did, but Valley did not. After I adjusted Valley, I tested GTA V - it was fine. But in GTA V Multuplayer I had to bump it a bit more resulting in now used 1.1375V :)

I'll be using an eGPU and Ubuntu, I don't know if that makes a difference.

Well, it's too early to say something, but I think that 2D-based Vega would increase system stability and 3D eGPU-use would not freak out the DRAM lanes too much. There may be an improvement but it's unknown to me :)

Again, if you use search function in this thread, you may spot some results on 4750G's voltage, but it seems to be a whole another story. Since the PSU is not even that powerfull to be able to satisfy the power demands.
 
  • Like
Reactions: Nunuji

alles_alles

SFF Lingo Aficionado
Aug 11, 2020
107
28
Have any one an idea . I would like to make a backup from my bios of my Deskmini X300 because i have the 1.46B on it. But how does it work ? have any one any idea?

Update : See in next Post how you can Save /Load your Bios on your HDD from your Motherboard.
 
Last edited: