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

xredlinexx1

Average Stuffer
Mar 18, 2019
86
23
I'm planning to build out a a300 3400g build and I'm new to both AMD and iGPUs. Is anyone getting stable ram speed higher than 3200 Mhz? I think I read someone getting 3400 Mhz. What performance gains can be seen from a 200 Mhz difference? Will the a300 eventually support much higher speeds, and what would the performance gains be seen from that (says 3200 mhz vs 4000 mhz)? Sorry for so many questions...i'm trying to wrap my head around ram speed and its relationship to ryzen and integrated graphics.
 

ConsolidatedResults

Average Stuffer
May 4, 2019
66
72
I'm planning to build out a a300 3400g build and I'm new to both AMD and iGPUs. Is anyone getting stable ram speed higher than 3200 Mhz? I think I read someone getting 3400 Mhz. What performance gains can be seen from a 200 Mhz difference? Will the a300 eventually support much higher speeds, and what would the performance gains be seen from that (says 3200 mhz vs 4000 mhz)? Sorry for so many questions...i'm trying to wrap my head around ram speed and its relationship to ryzen and integrated graphics.

iGPU performance will generally scale really nicely with RAM speed, how much will depend on the workload, i.e. the game or the benchmark or whatever.

E.g my results in Unigine Valley benchmark, Basic preset, running an A300/2400G/BIOS P1.20/GDM on so only even CL/Dual Rank:​
2666 20/19/19/19 (Auto Timing, 1.2V) - 43,9 FPS​
2933 22/21/21/21 (Auto Timing, 1.2V) - 46,3 FPS​
3200 22/22/22/22 (Auto Timing, 1.2V) - 48,7 FPS​
3333 24/23/23/23 (Auto Timing, 1.35V) - 49,1 FPS​
3333 16/19/13/16 (Manual Timing, 1.35V) - 52,5 FPS​
So for this particular benchmark in it's Basic preset, it looks like 10% more RAM speed are worth around 5% more FPS, and manual timings vs stock timings (JEDEC stock, not XMP stock), can also be worth around 5%. Also in this benchmark, Dual Rank is up 5% vs Single Rank and a 2400G is up 15% vs a 2200G. Funnily enough, also in this benchmark, BIOS P1.20 is up maybe 5% vs BIOS P3.X on a 2000 series APU, which isn't insignificant. Unfortunately 3000 Series APUs require BIOS 3.X.​
There are no massive jumps directly from one tier to the next, but it's always some % here and there, which in the end can add up or multiply quite a bit.

As for achievable stable speed, 3333 to 3466 generally appears doable. I don't know whether anyone got 3533 or even 3600 stable on an A300 so far and if so with which RAM / CPU / BIOS

Whether we are going to see much higher memory speeds in the future probably depends on a couple of things. The board generally being capable of such, wider range of faster SO-DIMM kits, ASROCK offering a BIOS that allows for RAM Voltage > 1.35V, and support for the Renoir APUs, which should come with the much better Zen2 memory controller.
 

ConsolidatedResults

Average Stuffer
May 4, 2019
66
72

Aaaand it's gone. Just the link though, it's still on the server: http://asrock.pc.cdn.bitgravity.com/BIOS/AM4/A300M-STX(3.60)ROM.zip

Well, whatever that was, link on the BIOS download page is back for me.

Checked the BIOS files, appears to not be AGESA 1.0.0.3 like 3.51 but 1.0.0.1 like 3.50. What we probably want for the A300 is actually AGESA 1.0.0.4 as that unifies support for all Ryzen lines (1000, 2000, 3000), see


Not sure I want to test installing it if the link has been pulled. But maybe it's the magic unlocked BIOS haha. No.

Apparently wasn't pulled

Anyone tested this yet?

Ah f it, I installed it. From a quick glance, menus look exactly the same as P3.50 to me and it is indeed AGESA 1.0.0.1. Performance regression vs BIOS P1.20, taking the same setup as in my previous post:

2666 20/19/19/19 (Auto Timing, 1.2V) - 43,9 FPS -> 41,7 FPS
3200 22/22/22/22 (Auto Timing, 1.2V) - 48,7 FPS -> 47,3 FPS

RAM also starts erroring out at the same speed.

So, not knowing what "system compatibility" has improved, I'll try reverting once more to P1.20
 
Last edited:
  • Like
Reactions: pappl

xredlinexx1

Average Stuffer
Mar 18, 2019
86
23
At the moment, is the a300 the only way to get a hold of an AMD STX motherboard? Unfortunately, looks like this is the case. Hopefully we see some other companies come out with some.
 

ShamedGod

Cable-Tie Ninja
Apr 21, 2019
147
77
BIOS 3.60 is finally out. Going to update today.

Edit: BIOS updated, FINAL DDR4-3400 settings loaded, running BOINC now. Will see if it crashes randomly...

Edit2: Over 24 hours of BOINC crunching, so far everything is stable with the DDR4-3400 Settings. No Random shut downs like before.

Edit3: Over 72 hours of BOINC crunching. No issues found

Edit4: Over 7 days with out a crash.

These DDR4-3400 ram settings where already proven stable and reliable with BIOS 3.50 but would randomly cause the machine to restart. With BIOS 3.60 those issues seem to be resolved.
 
Last edited:

ShamedGod

Cable-Tie Ninja
Apr 21, 2019
147
77
iGPU performance will generally scale really nicely with RAM speed, how much will depend on the workload, i.e. the game or the benchmark or whatever.

E.g my results in Unigine Valley benchmark, Basic preset, running an A300/2400G/BIOS P1.20/GDM on so only even CL/Dual Rank:​
2666 20/19/19/19 (Auto Timing, 1.2V) - 43,9 FPS​
2933 22/21/21/21 (Auto Timing, 1.2V) - 46,3 FPS​
3200 22/22/22/22 (Auto Timing, 1.2V) - 48,7 FPS​
3333 24/23/23/23 (Auto Timing, 1.35V) - 49,1 FPS​
3333 16/19/13/16 (Manual Timing, 1.35V) - 52,5 FPS​
So for this particular benchmark in it's Basic preset, it looks like 10% more RAM speed are worth around 5% more FPS, and manual timings vs stock timings (JEDEC stock, not XMP stock), can also be worth around 5%. Also in this benchmark, Dual Rank is up 5% vs Single Rank and a 2400G is up 15% vs a 2200G. Funnily enough, also in this benchmark, BIOS P1.20 is up maybe 5% vs BIOS P3.X on a 2000 series APU, which isn't insignificant. Unfortunately 3000 Series APUs require BIOS 3.X.​
There are no massive jumps directly from one tier to the next, but it's always some % here and there, which in the end can add up or multiply quite a bit.

As for achievable stable speed, 3333 to 3466 generally appears doable. I don't know whether anyone got 3533 or even 3600 stable on an A300 so far and if so with which RAM / CPU / BIOS

Whether we are going to see much higher memory speeds in the future probably depends on a couple of things. The board generally being capable of such, wider range of faster SO-DIMM kits, ASROCK offering a BIOS that allows for RAM Voltage > 1.35V, and support for the Renoir APUs, which should come with the much better Zen2 memory controller.

I have gotten 3466 to work, I have not gotten it stable.

3400 is 100% doable, but there was an issue with the BIOS... I am testing BIOS 3.60 to see if the "improved stability" meant RAM and Power :)

It does make a difference, enough to matter... *ShurG*
 

Quango

SFF Lingo Aficionado
Apr 6, 2019
102
34
At the moment, is the a300 the only way to get a hold of an AMD STX motherboard? Unfortunately, looks like this is the case. Hopefully we see some other companies come out with some.
STX seems unpopular with manufacturers. For Intel Kaby Lake, we had a choice of an Asrock and three Asus boards (H110S1 and S2, Q170S1), but now it is only Asrock we have for AMD and Intel as consumers.
 

Slowjim

Trash Compacter
Mar 12, 2019
53
25
Wanted to share an update here on my 3400g, not using the A300 anymore (now on Asrock B450 fatal1ty itx) but thought it may be useful to some folks here.


I have two issues I have been spending a bunch of time trying to figure out. Firstly I am seeing an over-voltage issue (peaking at 1.55v SVI2 TFN) which I have also experienced on two other boards as well - MSI B450I (which I returned) and asrock A300. So I don't think the issue is exclusive to any one mobo. Two things that have seemed to help the over-voltage at stock settings were disabling "SOC/Uncore OC mode" under AMD Overclocking menu, and using the 1usmus power plan with enabling global c-state control and low current idle in bios.

For the second issue, basically it seems any time I apply an iGPU oc it bypasses all boost and power saving features on the CPU and just sits at max speed/voltage. Again, I am seeing like 1.55v SVI2 TFN. Nothing I have tried so far (see https://sff.life/how-to-undervolt-ryzen-cpu/) is able to change this behavior. The only thing that works is disabling core performance boost, which I don't really want to do.

On the plus side, I have cheap 2x8gb micron e-die running stable at 3533c14 1.45v and 3600c14 1.5v.
 

LostEnergy

Caliper Novice
Sep 25, 2019
31
22
Here are some CPU cooler clearance tests with the A300: http://blog.livedoor.jp/wisteriear/archives/1073654956.html (Google Translate did a good job translating it for me). I'd be cool to fit a C7 Cu into the case; did anyone try modding it yet?

I have an A300 with a 3200G on the way. I'm going to use it as a headless server running 24/7, but it'll be close to idling most of the time (running Nextcloud and file sharing services). It's going to live close to my bed, so silence is an important factor. From what I read, it's possible to idle the whole machine under 10 W without much trouble, so I don't think heat and noise is going to be much of an issue - but the bigger the heatsink, the better. :)

For power saving settings I shamelessly recommend you read my initial post if you're going with Linux, which I assume.

Non-headless 7.2W are easy—and I've exactly the same setup (excluding memory) you're describing. 6W and less can be achieved by removing and downclocking hardware. ~3.5W if you're adventurous, but that'll be a setup primarily for idling, monitoring, cron-jobs.

[Starting with the 3.5W in mind.] Write to memory (disk cache), you're back up to ~4.2W; involve the NIC and it's ~4.5W; no DMA because you've to decrypt, ~5.4W. (Tip: power off cores it that's your use case (torrents, wsus, cache of sorts, backup destination). For example, the fourth core: printf 0 >/sys/devices/system/cpu/cpu3/online).

A bigger heatsink is not necessarily the better heatsink. Parameters are capacity (Cu is king), spread/conductivity (silver, heatpipes), dissipation (surface and airflow / geometry). The cooler [aka “Wraith Stealth”] which comes with the 3200G is good enough, imho, for what you're describing.

I've tested the Cryorig Cu and found that'll allow a machine to run longer spiky loads without having to spin up the fan [as soon as with the Wraith Stealth]—as expected, [the Cryorig Cu is] smoothing better [the heat curve]. (I didn't test what most PC magazines go for, the top-end: Saturating it with heat.) But it's not worthwhile [to buy it], given that [with the Wraith Stealth] the fan spins at its lowest RPM most of the time anyway, at least for me in server duty and office use, with the Wraith. (I've posted the fan-control settings here previously. Bump the lower temp from 55°C to 60°C, that's what I use now.)

I've toyed with a zero-rpm setup (page 25), but that too is not worthwhile. Once PBO [setting in BIOS] is out (as it triggers the overvoltage bug;, but people don't seem to read my posts; I don't imagine PBO/PBC works anyway), the power saving settings and fan control settings are in (posts above), you cannot hear the thing [using the Wraith Stealth] like always. And when playing games, especially with a headset on, I don't care much about it [spinning up].

(Running into fan spin-ups might be different with a 2400G or 3400G, which I didn't test. Whoever wanted mid-range or high performance should've saved the money [given that you cannot use the cooler it comes with and have to go for a Noctua in any case, bumping the price by +90%] and gone for a 3600–3800X imho. And whoever buys an Intel NUC these days either required by policy, loves the color, or is an illiterate bum who needs to subscribe to tech news in audio book format.)

The 12nm Zens [aka Zen+] are less picky when it comes to higher memory speed (3200 MT/s is 1600MHz btw. [alluding to the frequency barrier some forum users experience]), and the higher the frequency the more W=€ are spent [the more the sticks cost]. On a homeserver, downclock [memory] for marginal gains [in power savings]. (And don't get distracted by the laymen measuring goodput of data access patterns mistaking it for memory bandwidth, which is MT/s× 8 byte× #channels for DDR4 of course.) It's a revealing comedy like those “bicycle tyre rolling resistance” sites that strap wheels or vehicles to a motorized drum and claim “look ma, can do 60kph!!!”, and who drive the drum instead of using the wheel to drive the drum. Digressing. ;)

Edited 2020-04-02 for clarity.
 
Last edited:

NateDawg72

Master of Cramming
Aug 11, 2016
398
302
And whoever buys an Intel NUC these days either required by policy, loves the color, or is an illiterate bum who needs to subscribe to tech news in audio book format
I'm quite happy with my $135 used intel NUC, thank you very much :p
They have their place. What is it that you have against them?
 

LostEnergy

Caliper Novice
Sep 25, 2019
31
22
I'm quite happy with my $135 used intel NUC, thank you very much :p
They have their place. What is it that you have against them?

What on god's green earth made you think, stepping up presenting a crooked comparison were a good idea?
It's probably the thought of people like that which makes Linus from Tech Tips do unchristian things to himself to better his mood every morning.

I'm aware you might be as much jesting as I am.

----
Trying new vs used is the ‘crooked’ part. And then I encourage you to click through a price-comparison site, rockbottom, skinflint.co.uk, geizhals.de, whathaveyou; to compare the A300+cpu vs a NUC. The latter only wins for two meager Atom price-wise; feature-wise on two NICs, and TPM (some models) – which isn't about this –, else the A300 with its superior expand-ability, lower price point, less variants (firms don't have to hold stock of N different NUCs for replacements (i3, i5…; H, non-H), just the same A300 and different CPUs, which can be cTDP down if need be), and better efficiency. You didn't get an answer (no one is entitled to one, btw.) because even a cursory search revealed that, and you fishing for comedy has been more likely.

I welcome that you edited your authoritarian forum-sheriff–like answers down below.
 
Last edited:

NateDawg72

Master of Cramming
Aug 11, 2016
398
302
What on god's green earth made you think, stepping up presenting a crooked comparison were a good idea?
It's probably the thought of people like that which makes Linus from Tech Tips do unchristian things to himself to better his mood every morning.

I'm aware you might be as much jesting as I am.
Your rudeness was completely unnecessary and uncalled for.
 
Last edited:
  • Angry
Reactions: LostEnergy

fadsarmy

Caliper Novice
Oct 24, 2017
31
10
Hi A300 owners. Just need a little advice please. I'm trying to troubleshoot mic issues. When you plug in a microphone into the mic jack, do you get a pop-up saying "you just plugged a device into the mic jack"?

I'm having problems with my microphone and was wondering whether it's the mic jack or the microphone. When I plug in a microphone into the mic jack, I don't get a pop-up but I always got a pop-up in every pc I've previously had.
 
Last edited:

LostEnergy

Caliper Novice
Sep 25, 2019
31
22
Hi @fadsarmy: Plugging in a headset with microphone (four rings on the jack) when the machine has booted both get detected. Leaving it plugged in and powering on, the mic is not detected; it's audio-out only. o_O

Edit: bottom/left socket. The middle/top one seems to be audio-out only and doesn't detect the mic at all.
 
Last edited:

m4758406

Caliper Novice
Oct 13, 2019
31
20
#541
Hi fadsarmy.
My English is weak, i am from Austria.
Used Google translator help ...

I use Headset SteelSeries Siberia V2.

A300 connectors:
1.) Headphone/Headset Jack (AUDIO1)
2.) Microphone Input (AUDIO2)

Directly after pluged, my headset cables, both to "AUDIO1" and "AUDIO2" Windows 10 asked me if it is Headset (with micro) or earphone (without micro).

The funny thing was, you have to choose without microphone although you have one. If you choose with a microphone, it never worked for me. After restart it did not work anymore. So i told the stupid Windows 10 it is headphone only and until today my headset work with skype or telegram very good.

Windows 10 asked this question only one time, after first connection !
So you should delete audio drivers, reinstall them and tell the idiotic Windows 10 it is headphone only. (the popup window directly after first conneciton, there you have to click it, earphone only).
A miracle, then the microphone works permanently.
Now i can always keep it connected and even after Windows 10 reboot everything works.
I had ~8 hours work, to find out this stupid problem, some months ago.
 
Last edited: