• Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!

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

rubicoin

Cable-Tie Ninja
Jan 12, 2020
164
104
First gen Ryzen (2200G is Zen 1) is very picky on memory and the model you have probably has Hynix MFR chips which are notoriously unstable on first gen Ryzen above DDR4-2933 profile overclock. The Hynix CJR chips used on for instance the Crucial Ballistix Sport LT DDR4 3200 CL16 are much less difficult and can run faster.

follow-up to my g.skill ripjaws 3000 mhz issue: ran thyphoon burner to check my modules, and they seem to have hynix cjr chips. so one more reason to blame shitty memory controller in 2200g for below spec ram speeds.

 
Last edited:
  • Like
Reactions: Phuncz

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
ok, i've just realized that my a300 buid has got minor ram issues ? i've been using this config for 4 months now and i had a couple of resets during light gaming (fallout 3: nv) and also got BSOD once, but simply ignored that (thought it's related to windows10/driver/game bug). today i got curious, ran memtest86 and i got errors @ tests 7 & 9 consistently.

my config: 2200g + 2x8gb g.skill ripjaws ddr4 3000 (F4-3000C16D-16GRS) + bios 3.60. timings are set to default xmp 1.0 (3000 mhz @ 16-18-18-43) but reported as 18-18-18-43 in memtest which is weird. cpu-z reports it correctly as 16-18-18-43.

before getting these errors i started to increase ram speeds a bit, first tried 3066 with same timings (stable but w memtest errors), then 3166 gave me BOSD during login and 3200 did not even post. these memtest runs made me to check my ram at g.skill factory settings.

after these tests and errors i tried downclocking. first put it to auto (2400 mhz, 18-17-17-39) and then tried 2933 (16-18-18-43). all memtest errors vanished this time. never changed default voltage from 1.2 V factory spec.

memtest errors are no big problem for me at the moment as i want to upgrade my apu to 4400/4700g (or to 3400g if renoir wont't be supported), which could change things i guess.

my questions:
  • is there a chance that upping dram voltage to 1.35 V would help?
  • is there a really high chance that changing to a newer generation cpu would help? official qvl for raven bridge only has 2666 mhz g.skill ram sticks.
  • is it enough reason to rma my memory kit (being stable on 2933 mhz instead of 3000 with factory timings), or should i blame 2200g in the first, a300's cheaper mobo in the second, and the ram sticks themselves only for the third place?
  • if i'm upgrading to ryzen 3000/4000, would it be better to choose a new kit from the qvl list and choose 3200 mhz ram? i guess for renoir it would not hurt.
anyway, the reason for choosing this exact g.skill kit was price, as i got it super cheap (~60 EUR). my initial build was planned to be as inexpensive as possible with decent noctua cooling for later cpu upgrade. but i knew that at some point i'll have to invest more money on it, maybe it will be time for a new ram kit for a new apu.
Hello, I'm using RipJaws 3200, CL18. Will have to look up the part number. I have also instability issues when running an this frequency and APU came in.

After lot's of trial and errors I've managed to enable the SOC Overclock flag in BIOS.

It will automatically push the SOC Voltage to 1.1V and the critical Vdrop is gone. Yes, you do have a bit more heat (maybe couple degrees, this is not the core)

I think a better Mainboard can handle Vdrop better, so you don't need to increase SOC voltage. But it case of A300 it did for me the trick. You may try it out.

I've read, that voltage up to 1.2V on the SOC is considered safe. Well, you don't need to run that high.

How can I edit my message?

Forgot to mention: Ryzen 2400G, RipJaws 16GB (8x2)

Sorry, still don't know how to edit existing messages. Therefore sorry for typos, since autocorrection on phone and my lazyness.
<a href="https://ibb.co/VTn1Rw9"><img src="https://i.ibb.co/VTn1Rw9/a300-3200cl16.png" alt="a300-3200cl16" border="0"></a>

This is how I run them, I have to admit @CL16, Vddr=1.2V, Vsoc=1.125V

Although I have to add, I had recently increased the Vsoc from 1.1V since there were instabillity issues in GTA5. I've benchmarked it at 1.1V but maybe the warmer weather in summer and GTA as a workload have forced it to appear. Valley UNIGINE Benchmark was the way to go (on lower res, to make the GPU deliver more frames, so the SOC I/O will be stressed - hence you know there has to be the problem) and additionally CPU benchmark on 4 Threads (I've used FurMark 1.20

The test is important, you want to stress the SOC Memory I/O, so let the GPU deliver more frames, so the bus will be full with GPU data and let it concurr with CPU bandwith, since the 16GB are shared. The I/O will be stressed - you may observe a drop of Vsoc in HWInfo32 below 1.0V on stock voltage.

I hope I could explain it a bit more. :)
 
  • Like
Reactions: rubicoin

akisu

SFF Lingo Aficionado
Jan 15, 2019
111
65
I built a DeskMini last year. I bought the cheapest modules I could find and ended up with 2x G.Skill RipJaws 4GB @ 2133MHz CL15. They are also Hynix Modules. I did some OC testing. While 3200MHz won't even start 3000Mhz was working stable. So I optimized my timings for that and ended up with CL18. But I didn't spend to much time into that. It wasn't meant to be a gaming machine (only Athlon 200GE) but the improvements were satisfying. The machine is now running fine somewhere in the heat of Brazil. :)
 

westdam

Chassis Packer
Jun 22, 2020
13
0
hi, marco from italy . i've bought a deskmini a300 bios 3.6 with ryzen 3400g with stock heatsink and awaiting ballistix 3200mhz cl16 ( 2x8gb) , a 256gb savage hyperx and a 1tb lowcost ssd. right now just to test i'm using a 2gb ddr4 2400 crucial and a 120gb 840 evo ssd with linuxmint 20 with kernel 5.4, 5.6, 5.7.
i've also bought the addon usb 2.0 cable and wifi-kit (not yet arrived). unfortunately the usb2.0 kit its pretty unusable for me: its too big and the cable is too long and hit on the mobo while closing,anyway with some tricks i was able to manage the cable near the heatsink ( but every time i need to open i'll need to rearrange all). the real problem its the power.. even if i plug a small wifi mouse the mouse " lag" , if i plug a wifi card the os freeze, the same with any usb dongle ( 4, 8 , 32 gb). i can only use for wired keyboard and mouse. seems a power problem. a friends suggest to use a bigger powersupply ( like 19v 9a 180w) do you think will solve the problem instead the stock one ? ( 19v 6.32a 120w) also i've got a horrible brusing from the mobo. it's quite udible.. it's not the fan itself, is something else. like elettrical components from the mobo.

i'm quite happy with this configuration but i've heard of a lot of problem, reboot, hungs, bios needs to be restored often and so on.
i cannot figure out for the noise and i've return the usb2.0 cable back it's too annoying to manage every time i open the case and it's pretty unstable and unusable. i'll use a usb 2.0 powered hub.

someone can help me?
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
Hello, Westdam.

I've used to run linux on the A300. No problems so far regarding USB-Sticks. (Ubuntu LTS 18.04.3) I do not have the Wifi card. Does the freezing occur when inserting USB without WiFi-Module installed?

Noise: yes, I do experience it also. Its brushing simultan to the Power LED flashing (e.g. in standby) this is the Mobo/VRM components of the Power supply on the Mobo. Unfortunately, you can not change this behavior.

What I'm trying to say is - a new power supply probably won't solve your issue.

The way I see is to RMA the A300. Because of the freezing regarding USB.

Have you tried to Load BIOS Defaults? Maybe something gone somehow wrong.

I doubt it will also resolve the issue. Maybe the WiFi board has a failure, but I don't think so too.

It's probably the sh*tty power supply on the Mobo.
 

ssouthall6

Trash Compacter
Jun 18, 2020
50
18
So the PC boots fine with either of the two sticks in the working slot? And not at all with either stick in the other? And if you boot up with RAM in both slots, what happens then - is all the RAM seen and accessible by the system, and is the system stable? Did you run memtest with both sticks installed?

Booting up with occasional fan spin and "trying" to boot indicates that something fails in RAM training. If either stick works fine in one of the slots, that's a clear indication that the slot itself is bad. There's nothing you as an end user can realistically do about this.
I think I knew deep down it’s a bad ram slot

it boots fine with either
So the PC boots fine with either of the two sticks in the working slot? And not at all with either stick in the other? And if you boot up with RAM in both slots, what happens then - is all the RAM seen and accessible by the system, and is the system stable? Did you run memtest with both sticks installed?

Booting up with occasional fan spin and "trying" to boot indicates that something fails in RAM training. If either stick works fine in one of the slots, that's a clear indication that the slot itself is bad. There's nothing you as an end user can realistically do about this.

I think deep down I knew that would be the case. Memtest shows both sticks are completely fault free. The system boots with either stick in slot 1. In any configuration where slot 2 is used, either the blue light comes on and I get no video, or the fan starts then stops, it never gets far enough for me to 'see' the second stick in the UEFI.

It's a bad board for sure. A nice little system but I think I misjudged it's capabilities. Maybe with a 3400 and an RMA when the gap in price closes it'll live up to its potential
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
@ssouthall6 I've read somewhere that the mobile RAM modules (SO-DIMM) not always have the compatible timing programmed in their SPD and thoughts do not get recognized properly. (More related to mobile / laptop market)

But in your case, somehow, it makes no sense. Sorry, I may need a follow up. Does the system boot with one stick only installed in slot 2?
 

ssouthall6

Trash Compacter
Jun 18, 2020
50
18
@ssouthall6 I've read somewhere that the mobile RAM modules (SO-DIMM) not always have the compatible timing programmed in their SPD and thoughts do not get recognized properly. (More related to mobile / laptop market)

But in your case, somehow, it makes no sense. Sorry, I may need a follow up. Does the system boot with one stick only installed in slot 2?

No it doesn't, and today it wouldn't boot the display in the configuration that has been working fine for months. Eventually, after removing all usb connectors etc and trying individual connectors one by one, I got it to boot with hdmi to output, but no displayport. If I plug in dp the button lights up and everything moves over to that monitor, so the system I guess is recognising it, but it won't output.
 

westdam

Chassis Packer
Jun 22, 2020
13
0
Hello, Westdam.

I've used to run linux on the A300. No problems so far regarding USB-Sticks. (Ubuntu LTS 18.04.3) I do not have the Wifi card. Does the freezing occur when inserting USB without WiFi-Module installed?

Noise: yes, I do experience it also. Its brushing simultan to the Power LED flashing (e.g. in standby) this is the Mobo/VRM components of the Power supply on the Mobo. Unfortunately, you can not change this behavior.

What I'm trying to say is - a new power supply probably won't solve your issue.

The way I see is to RMA the A300. Because of the freezing regarding USB.

Have you tried to Load BIOS Defaults? Maybe something gone somehow wrong.

I doubt it will also resolve the issue. Maybe the WiFi board has a failure, but I don't think so too.

It's probably the sh*tty power supply on the Mobo.

the strange of the USB freezing occours only on the usb 2.0 side port with the add-on cable. with the built-in usb port no problem so far.
the wifi card is not insterted yet. the kit arrived yesterday so the problem was without the wifi card plugged in.

i've RMA the add-on usb 2.0 cable, it's also difficult to manage inside the deskmini. maybe is there different versions for deskmini? ( a300, 110 ecc )

give me a suggestion : the ram i've choosen ( ballistix 3200 cl 16 ) seems difficult to find, right now. amazon doesnt have a vaild delivery date and i dont want to wait too long. i've found a gskill 3000 mhz cl18 ram module. would be a noticeable performance hit or the loss is not so marked?

today i'll try the wifi card. if i've had problem too i'll rma the deskmini.

oh,, sooner i'll grab the noctua slim cooler. it's noticeable the difference with temperature and noise?

thanks for the help!

marco
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
@westdam
on usb 2.0 add-on cable: I do have it too. :) Maybe by next needed BIOS reset I'll make a pic; I put the cables, - yes managements of those is not easy - when I slide the mobo back in kinda "on the side of CPU FAN" so it just holds there. Hmm, ok, maybe the wiring inside the cable itself gone loose or something, can happen.

on clear CMOS / Load BIOS Defaults: I don't know if there is a substantial difference between those two, but both should load/reset to valid initial values

on ballistix 3200 cl 16: Sorry, I do not have any xp with them. If they have a proper XMP profile, there should be no problem.

on difference 3000 / 3200: there is very much on the Internet to read into, but what I understand is that Ryzen likes lower latency. "3000cl15 ~ 3200cl18"
Was the initial Ryzen not designed for max. 2933 MHz memory?
I think by going the 3000cl15 way, the chances are less to xp instability problems because of Vdrop and still have low timings.
Somewhere was also a discussion wheather Ryzen's timings are only even - CL14,16,18... As you can see, there is pretty much to this topic.
But again even if, "3000cl16 ~ 3200cl18" with 3200 you have a bit more b/w, but still, the chance is pretty low your ever to notice the difference.
On the other hand, you can go CL18 and probably do also 3200 if you are enthusiastic to xp it :)
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
on stock cooler / noctua: I did also the same way, first with boxed cooler by screwing off the top platic cover to be able to fit in.
After some time, a Noctua. What should I say. It's quiter - oh yea! Temp is about the same - but system quiter. It may depend on your use.
Of couse the system is not quite when under load. But than that's under load. :)
If you have a M.2 - put it on the front slot, not the rear one. Front slot is being cooled from the side (CPU FAN) (my max M.2 temp, front: 85°C controller, (40+°C NAND) / read: 100°C controller, (50+°C NAND))... I also dont remember NAND temp not really, but it was not all that bad. I just remember after running for 5 min in rear slot (I was having Vartualization/NVMe issues) and checking HWInfo temp I decided to switch it back :D M.2: Crucial P1 1TB
 

akisu

SFF Lingo Aficionado
Jan 15, 2019
111
65
on difference 3000 / 3200: there is very much on the Internet to read into, but what I understand is that Ryzen likes lower latency. "3000cl15 ~ 3200cl18"
Ryzen likes all kind of fast memory. And the benefits are way higher than CPU OC.

That might be interesting for you:
 

Valantar

Shrink Ray Wielder
Jan 20, 2018
2,201
2,225
Ryzen likes all kind of fast memory. And the benefits are way higher than CPU OC.

That might be interesting for you:
It does, but there are also limits to the maximum stable memory speeds, dependent on both the CPU/APU (memory controller), motherboard design, and the memory ICs. My setup refuses to be stable above DDR4-2933 no matter what, despite the memory being nominally rated for 3200 (and using timings from 1usmus' DRAM calculator doesn't help whatsoever). It seems to work, and memtest and the like gives no errors, but I get random shutdowns if I try to run the memory at 3200. Zen (1) and Zen+ are much the same in this regard, with Zen 2 being much improved. Of course, all currently available APUs are Zen or Zen+. So, to sum up, especially with pre-Zen 2: YMMV.
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
It does, but there are also limits to the maximum stable memory speeds, dependent on both the CPU/APU (memory controller), motherboard design, and the memory ICs. My setup refuses to be stable above DDR4-2933 no matter what, despite the memory being nominally rated for 3200 (and using timings from 1usmus' DRAM calculator doesn't help whatsoever). It seems to work, and memtest and the like gives no errors, but I get random shutdowns if I try to run the memory at 3200. Zen (1) and Zen+ are much the same in this regard, with Zen 2 being much improved. Of course, all currently available APUs are Zen or Zen+. So, to sum up, especially with pre-Zen 2: YMMV.
Exatly! Thank you for sharing your xp. Which voltage is your SoC? Stock=1.0V? This is what I was also guessing by saying "Was the initial Ryzen not designed for max. 2933 MHz memory?"
I've looked that up yesterday, it applys for AMD Ryzen 2400G but also for the 2200G. The integrated memory controller was initially specd for up to 2933MHz. If ASRock designed the A300 with that specs in mind and being not that costly, it makes sence, that there is not "plenty" room for higher freqs in terms of the mainboard. And the mobo cannot handle the higher power reqiured by the SoC Memory I/O --> Vdrop and instabillity.

As I mention, when running @3000+ the SOC Voltage can drop below 1.0V (HWInfo) where System becomes unstable. You may try to enable the Overclock Flag. It will raise the voltage by 0.1V to 1.1V. The Vdrop can still break it down to 1.034V (possible e.g. value) - it may stay stable.

But this is exactly the point with this ASRock product. No good/done BIOS. Mainboard running on Limits alone in the stock config. I don't know. If I see no love at least, a clean BIOS from ASRock there will be no X300. That's for sure :)
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
Regarding timings from 1usmus' DRAM calc.

Agressive timings, may also result into bigger Vdrop.
You may try to enable Gear Down Mode. (helps stability)

I could decide. Wheather I can run the G.Skill 3333@CL18 or 3200@CL16 - both needs almost the same Voltage addition.

My phylosophy was - I've paid for 3200, I would them also run @3200. After tinkering with this system, I found the Vdrop being the reason #1.

Achieve UNIGINE Valley to run without flashing random colored polygons in the picture. There you can see the Memory Controller struggeling to keep on.
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
I cannot emphasize enough.
But this is exactly the point with this ASRock product.
No good/done BIOS. Mainboard running on Limits alone in the stock config.
I don't know. If I see no love at least, a clean BIOS from ASRock there will be no X300.
That's for sure. As a consumer I can not tolerate and financially support such a behaviour by spending money for something "new", where the old "new" was not even done.
 

westdam

Chassis Packer
Jun 22, 2020
13
0
so, with a ryzen 3400g , this mobo could run and use a 3200mhz ram stable or it's safe and better stay a 2993? ( regardless the more power with hi-end ram for ryzen)
 

gustav

Cable-Tie Ninja
Jun 18, 2020
193
90
I have no xp using 3400G, but Internet states it's memory controller is also specd for 2933 MHz. Sorry, you will have to try it out or wait untill someone replys who has actually tried it out already :)
Try it out using 3200. If it's unstable, enable SOC Overclock Flag - if you don't mind about +0.1V. Otherwise use "fault-safe" 2933MHz.