-
Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!
Prebuilt [SFFn] ASRock's DeskMini A300 - Finally!
- Thread starter confusis
- Start date
You were quite right my friend. Replacement arrived today and everything is running perfectly. Thanks for the assistance, also thanks @Valantar
64gb running nicely, idling a little high, although it is warm here in the UK right now for once, so may look at changing the wraith that came with the 3200g for a noctua LP cooler - haven't seen tests so will ponder that for now on my working a300. Thanks again guys
@ssouthall6 just curious about your RAM. 
May I ask, which freq and voltage? XMP&Stock?
@micky1067 if you follow the guide, suchende has shared, can you post some lspci output or tell where you encounter problems?
May I ask, which freq and voltage? XMP&Stock?@micky1067 if you follow the guide, suchende has shared, can you post some lspci output or tell where you encounter problems?
Last edited:
if you want increased cooling performance see my noctua chromax mod (14 mm stock nf-a9x14 pwm @ nh-l9a-am4 -> 25 mm nf-a9 pwn) posted in this thread back in january/february. because of the 2200g it made not much of a difference in temps (still waiting for a decent renoir apu), but helped me lowering fan speed even further.
for the ram modules i came up with a super cheap 0,5 mm alu heat spreading option (0.7$ / piece from aliexpress):
ok, not a real cooling solution, but may help a bit when upping voltages. or may look cool when painted full black or put the g.skill stickers on it from the ram chips or something.
Last edited:
Get the plain Crucial 3200 RAM. Nominally CL 22, dual rank modules for me run 3200 CL18 @ 1.2V or 3333 CL16 @ 1.35V on a 2400G
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.
One needs to verify SOC Overclock VID actually having an effect, whether the setting does something at all depends on the BIOS version and whether configured RAM speed exceeds SPD rating (at least for non-XMP modules). Check post #640 in this thread:
Prebuilt - [SFFn] ASRock's DeskMini A300 - Finally!
@ConsolidatedResults I haven't had any reason to check my temperature settings beyond the one time I ran stress testing in Fedora 32. Temps were fine under load at that time and I've had no reason to check the server temps now that I'm running Win 2012 R2 on my 2400G equipped A300W. There may be...
smallformfactor.net
I wasn't talking about the A300 specifically but pre-Zen 2 Ryzen generally - the limitations to memory clock speeds are universal, but obviously dependent on the characteristics of each implementation.One needs to verify SOC Overclock VID actually having an effect, whether the setting does something at all depends on the BIOS version and whether configured RAM speed exceeds SPD rating (at least for non-XMP modules). Check post #640 in this thread:
Prebuilt - [SFFn] ASRock's DeskMini A300 - Finally!
@ConsolidatedResults I haven't had any reason to check my temperature settings beyond the one time I ran stress testing in Fedora 32. Temps were fine under load at that time and I've had no reason to check the server temps now that I'm running Win 2012 R2 on my 2400G equipped A300W. There may be...
Completely understood, where you wrote that Memtest gives no errors but you still get random shutdowns that in my experience can relate to SOC Voltage being too low or dropping under load, as @gustav also observed, so i wanted to suggest to check that - when you try to run your RAM at 3200 and you set SOC Voltage to offset the voltage drop to help it becoming stable - the SOC voltage setting actually does stick. Because it could be that you set it to e.g. supply 1.15V but it actually only supplies 1V. That is all assuming that "my setup" refers to an A300 based platform. That was all
Edit: changed "1.5V" to "1.15V"
Last edited:
That might be worth a try, but since I'm running a 1600X any voltage changes for the CPU would set the CPU into "OC mode", disabling boost and generally making it act like a mid-2000s CPU, which I'm not keen on. It's a good thing they've changed how this works in later revisions. For now it's actually been stable for as long as I can remember, so I'm sticking with current settings until I can upgrade again later this year or early next year, and relegate the 1600X to NAS duties (at DDR4-2400 or 2666 for stability, obviously).Completely understood, where you wrote that Memtest gives no errors but you still get random shutdowns that in my experience can relate to SOC Voltage being too low or dropping under load, as @gustav also observed, so i wanted to suggest to check that - when you try to run your RAM at 3200 and you set SOC Voltage to offset the voltage drop to help it becoming stable - the SOC voltage setting actually does stick. Because it could be that you set it to e.g. supply 1.5V but it actually only supplies 1V. That is all assuming that "my setup" refers to an A300 based platform. That was all
@ConsolidatedResults hello
well, my conclusion ist that the mobo was not *in the first place* designed to run RAM at 3200.I'm on 3.60 BIOS.
SOC Overclock Flag is just "a click", to set to 1.1V. you can also manually set the VID (will have to look up) value.
We, well, at least I do speak about modules with XMP profile for their 300/3200 speeds. The same values can be ofc entered manually, if A300 boots. I do not blame at this point. It's just how it goes.
FYI: The XMP profile @2933+ is also outside JEDEC specs - the institution which is designing DDR RAM
it's just manifested as XMP, but is tested outside of specs. So yeahEdit: do not supply more than 1.2V to your SOC, as more can permanently degrade performance
also I encourage everyone to try UNIGINE Valley out and HWInfo32 and observe the SOC voltage drop and just post your min value. And just provide shortly your config. We can do a survey and figure it out.
Like this:
3.60bios/3200@CL18@1.2Vram/2400G/1.1Vsoc/0.984Vatdrop/gdr=off/cr=1t
Gear Down Mode = Off (stressing SOC!)
Command Rate = 1T
My stable:
3.60/3200@CL16@1.2V/2400G/VID=41/1.034V/off/1t
TODO: Search up VID offset table for A300, which describes voltages. Higher VID = lower Voltage. Values 45 to 41 were tried by me. Considered safe, since under 1.2V. (VID=41 is around 1.125-1.13V, HWInfo32. it may still differ from real Voltage which comes on the PIN of the SOC vs reported by sensor!)
It should at least give us a picture. UNIGINE is not the payload/test to run. You can run anything. Just observe
we just need like 20 samples. Draw a graph and find a sweet spot, where we can say anything further. If you do instability, you may increase the voltage and make one more sample, so we could even compare Just an ideaMaybe I do get it wrong. But if you set 1.5V it should supply 1.5V on idle. But when there is power required for computation, the chip draws it from the supply (12V,5V,3.3V,1.8V,... Power rails on your Mainboard). The supply has to keep up with the demand. There comes in such a phenomenon as 'voltage drop'
The supply cannot satisfy the demand.
Last edited:
@ConsolidatedResults hello
I'm on 3.60 BIOS.
SOC Overclock Flag is just "a click", to set to 1.1V. you can also manually set the VID (will have to look up) value.
We, well, at least I do speak about modules with XMP profile for their 300/3200 speeds. The same values can be ofc entered manually, if A300 boots. I do not blame at this point. It's just how it goes.
FYI: The XMP profile @2933+ is also outside JEDEC specs - the institution which is designing DDR RAM
Edit: do not supply more than 1.2V to your SOC, as more can permanently degrade performance
also I encourage everyone to try UNIGINE Valley out and HWInfo32 and observe the SOC voltage drop and just post your min value. And just provide shortly your config. We can do a survey and figure it out.
Like this:
3.60bios/3200@CL18@1.2Vram/2400G/1.1Vsoc/0.984Vatdrop/gdr=off/cr=1t
Gear Down Mode = Off (stressing SOC!)
Command Rate = 1T
My stable:
3.60/3200@CL16@1.2V/2400G/VID=41/1.034V/off/1t
TODO: Search up VID offset table for A300, which describes voltages. Higher VID = lower Voltage. Values 45 to 41 were tried by me. Considered safe, since under 1.2V. (VID=41 is around 1.125-1.13V, HWInfo32. it may still differ from real Voltage which comes on the PIN of the SOC vs reported by sensor!)
It should at least give us a picture. UNIGINE is not the payload/test to run. You can run anything. Just observe
Maybe I do get it wrong. But if you set 1.5V it should supply 1.5V on idle. But when there is power required for computation, the chip draws it from the supply (12V,5V,3.3V,1.8V,... Power rails on your Mainboard). The supply has to keep up with the demand. There comes in such a phenomenon as 'voltage drop'
Starting from the back, the vdroop issue on the SOC rail is known for the A300 and has been confirmed several times in this thread, so your observations are of course correct.
What further happens with the A300 is that the board may set SOC Voltage to 1.1V automatically or ignore manually set vSOC. The information on how the board behaves for different BIOS versions is in post 640 and can easily be confirmed with a monitoring tool like hwinfo. 1.5V was a typo, meant 1.15V, don't try to run SOC at 1.5V
None of the Gen 1/1+ Ryzen G officially support more than DDR2933 anyway, so board support may be secondary. I had Single Rank x16 RAM stable at 3466 with a 2200G on A300, why its not stable beyond that could be anything really. What is definitely bottom of the barrel is the DRAM power solution on the board.
What you are seeing in terms of timings appears to me to be at the upper limits of what you can expect at only 1.2V DIMM ans such low SOC voltage.
The full VID table as such:
5F = 0.95625V
5E = 0.96250V
5D = 0.96875V
5C = 0.97500V
5B = 0.98125V
5A = 0.98750V
59 = 0.99375V
58 = 1.00000V
57 = 1.00625V
56 = 1.01250V
55 = 1.01875V
54 = 1.02500V
53 = 1.03125V
52 = 1.03750V
51 = 1.04375V
50 = 1.05000V
4F = 1.05625V
4E = 1.06250V
4D = 1.06875V
4C = 1.07500V
4B = 1.08125V
4A = 1.08750V
49 = 1.09375V
48 = 1.10000V
47 = 1.10625V
46 = 1.11250V
45 = 1.11875V
44 = 1.12500V
43 = 1.13125V
42 = 1.13750V
41 = 1.14375V
40 = 1.15000V
3F = 1.15625V
3E = 1.16250V
3D = 1.16875V
3C = 1.17500V
3B = 1.18125V
3A = 1.18750V
39 = 1.19375V
38 = 1.20000V
37 = 1.20625V
36 = 1.21250V
35 = 1.21875V
34 = 1.22500V
33 = 1.23125V
32 = 1.23750V
31 = 1.24375V
30 = 1.25000V
2F = 1.25625V
2E = 1.26250V
2D = 1.26875V
2C = 1.27500V
2B = 1.28125V
2A = 1.28750V
29 = 1.29375V
28 = 1.30000V
@ssouthall6 just curious about your RAM.
@micky1067 if you follow the guide, suchende has shared, can you post some lspci output or tell where you encounter problems?
Hi gustav,
Result.. The VM does niot start. And the proxmox system hangup.
I have done this:
sudo lspci -nnk | grep "VGA\|Audio"
04:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Picasso [1002:15d8] (rev c8)
04:00.1 Audio device [0403]: Advanced Micro Devices, Inc. [AMD/ATI] Raven/Raven2/Fenghuang HDMI/DP Audio Controller [1002:15de]
Subsystem: Advanced Micro Devices, Inc. [AMD/ATI] Raven/Raven2/Fenghuang HDMI/DP Audio Controller [1002:15de]
04:00.6 Audio device [0403]: Advanced Micro Devices, Inc. [AMD] Family 17h (Models 10h-1fh) HD Audio Controller [1022:15e3]
Subsystem: ASRock Incorporation Family 17h (Models 10h-1fh) HD Audio Controller [1849:2233]
Graphic: 1002:15d8
Audio: 1002:15de
/etc/modules-load.d/vfio.conf
vfio
vfio_iommu_type1
vfio_pci
vfio_virqfd
/etc/modprobe.d/vfio.conf
options vfio-pci ids=1002:15d8,1002:15de
update-initramfs
sudo update-initramfs -u -k all
sudo nano /etc/default/grub
update-initramfs: Generating /boot/initrd.img-5.4.44-1-pve
Running hook script 'zz-pve-efiboot'..
Re-executing '/etc/kernel/postinst.d/zz-pve-efiboot' in new private mount namespace..
No /etc/kernel/pve-efiboot-uuids found, skipping ESP sync.
update-initramfs: Generating /boot/initrd.img-5.4.41-1-pve
Running hook script 'zz-pve-efiboot'..
Re-executing '/etc/kernel/postinst.d/zz-pve-efiboot' in new private mount namespace..
No /etc/kernel/pve-efiboot-uuids found, skipping ESP sync.
update-initramfs: Generating /boot/initrd.img-5.3.18-3-pve
Running hook script 'zz-pve-efiboot'..
Re-executing '/etc/kernel/postinst.d/zz-pve-efiboot' in new private mount namespace..
No /etc/kernel/pve-efiboot-uuids found, skipping ESP sync.
sudo nano /etc/default/grub
GRUB_CMDLINE_LINUX_DEFAULT="quiet amd_iommu=on"
sudo update-grub
Linux-Abbild gefunden: /boot/vmlinuz-5.4.44-1-pve
initrd-Abbild gefunden: /boot/initrd.img-5.4.44-1-pve
Linux-Abbild gefunden: /boot/vmlinuz-5.4.41-1-pve
initrd-Abbild gefunden: /boot/initrd.img-5.4.41-1-pve
Linux-Abbild gefunden: /boot/vmlinuz-5.3.18-3-pve
initrd-Abbild gefunden: /boot/initrd.img-5.3.18-3-pve
Adding boot menu entry for EFI firmware configuration
sudo reboot
In Proxmox VM Linux Mint
Hardware:
BIOS: OVMF
machine: q35
PCi Device: 0000:04:00.0 / All functions, primary Gpu, Rom Bar and PCIe is on.
Thank you for your responce @ConsolidatedResults! It brings me a bit up-to-date!
On BIOS 3.60, it gave me 1.1Vsoc with: RAM=XMP, SOC OVERCLOCK = Enabled. No further values changed.
In my case: VID=41 == 1.14375V ; HWInfo/Sensor reads "only" 1.125V. But it's 1.14..V what is deliviered.
My statement was in regard of the A300 mobo. Like if ASRock saw, Ryzen is officially up to 2933MHz - ok, we build up to 2933MHz - and that's the problem. The A300 has no head room for limits. Like it's on stock running on limits, hence the problems with RAM above "official".
I also agree on RAM power supply It's really not dynamic. And thank you very much for the VID Table!
Yea, I used to follow this thread and @ShamedGod's results on his memory OC. I may have missed it, thank you for letting me know
Ok, it was not clear to me, that the board has such a weird behaviour. There you see again, ASRock & their dev team. So awful. Hence I can not emphasize enough. Give us a stable, production BIOS. And not glued together and forgotten about.
On BIOS 3.60, it gave me 1.1Vsoc with: RAM=XMP, SOC OVERCLOCK = Enabled. No further values changed.
That's actually pretty neat. Do you remember your voltages? Just curious. I had in mind, that a VID=41 is not that high! But, it's still ok.
In my case: VID=41 == 1.14375V ; HWInfo/Sensor reads "only" 1.125V. But it's 1.14..V what is deliviered.
I disagree on that. As Ryzen was released, (I'm talking about Zen 1/ Zen1+,Raven Ridge), AMD had it officially running at 2933MHz, and after some time, they've updated their statement but the specs were never changed. As far as I know. AMD Page, Raven Ridge Specs - still 2933MHz; Anandtech, Ryzen 2400G, Review
My statement was in regard of the A300 mobo. Like if ASRock saw, Ryzen is officially up to 2933MHz - ok, we build up to 2933MHz - and that's the problem. The A300 has no head room for limits. Like it's on stock running on limits, hence the problems with RAM above "official".
I also agree on RAM power supply
It's really not dynamic. And thank you very much for the VID Table!
Last edited:
@micky1067
Ok, those are the steps in the guide.
So the problem is: The targeted VM does not start? (Which should have the GPU pass-through)
At which point does Proxmox hang up itself? After starting the targeted VM?
How do you measure that it hangs? Other VMs become unresponsive? no ssh into proxmox possible?
Can you execute lsmod on proxmox to see, wheather your vfio_... modules have been successfully loaded? As a first step of troubleshooting.
If they are loaded into kernel. I assume proxmox hangs directly after you start targetet VM?
Bash:
sudo lsmod | grep vfioIf the output of the command is empty, they have not been loaded. Else, could you please post the output.
Last edited:
Hello gustav,
I stoped now to manage gpu passthrough. My whole Proxmox system was broken. I must reinstall it. I think the problem is the integrated gpu.
I think it will be better to have a second graphic card inside. Is a pitty...
To you questions. All vfio modules running. And Proxmox hangup when I start a VM with the changes in hardware.
So i can not do something. Only I can go with ssh inside the host system to reboot the machine.
But one time ist was so extrem, that I need to reinstall the whole system.
I become a grub error. Fix minimal bash like line editing .. I can not rescue grub. Really much work.
My luck.. I had a backup of my lxc containers and VM...
Hello to all deskmini lovers, new to forum but reading you offen,
i'm writing from Greece and i would appreciate your help to my problem.
I have the a300 running fluently for 9 months with m.2 nvme adata xpg 6000, 3400g, hyper x 2x8gb 2933, stock cooler from 3200g,
i was playing games with no problems but suddenly when i load not a heavy game for example rise of the tomb raider the system crashes (restarts), memtest crashes, burnintest on 3d crashes....i've tried 3400g on another mobo with no problem, i tried on deskmini the 3200g and crucial 2666 2x8gb ram the same crashes, in all bios's in other ssd fresh win10 install, same crashes. I have also noticed an temp increase inside the case on idle too and the environment temp is the same, the heatsinks on cpu cooler and mobo burn allover, is that normal?
I've sent it to the vendor and the first answer was "the deskmini is not for games but only for office work" (made me furious) anyway after the check they said that it passes with no errors. I requested to make a more thorough test and explained that crashes on 3d and i 'm waiting.
Please can anyone tell me if there is a way to see if the deskmini has the problem or if it can be fixed any how?
And finally i'm about to buy a noctua fan as my last effort, what do you suggest the l9i or l9a?
Thank you in advance, and sorry for my english.
i'm writing from Greece and i would appreciate your help to my problem.
I have the a300 running fluently for 9 months with m.2 nvme adata xpg 6000, 3400g, hyper x 2x8gb 2933, stock cooler from 3200g,
i was playing games with no problems but suddenly when i load not a heavy game for example rise of the tomb raider the system crashes (restarts), memtest crashes, burnintest on 3d crashes....i've tried 3400g on another mobo with no problem, i tried on deskmini the 3200g and crucial 2666 2x8gb ram the same crashes, in all bios's in other ssd fresh win10 install, same crashes. I have also noticed an temp increase inside the case on idle too and the environment temp is the same, the heatsinks on cpu cooler and mobo burn allover, is that normal?
I've sent it to the vendor and the first answer was "the deskmini is not for games but only for office work" (made me furious) anyway after the check they said that it passes with no errors. I requested to make a more thorough test and explained that crashes on 3d and i 'm waiting.
Please can anyone tell me if there is a way to see if the deskmini has the problem or if it can be fixed any how?
And finally i'm about to buy a noctua fan as my last effort, what do you suggest the l9i or l9a?
Thank you in advance, and sorry for my english.
sorry i realized that l9a is for our setup with amd, did you find anything better?
Asrock just released a Beta BIOS for the A300, updates to AGESA Combo-AM4 1.0.0.4 Patch B.
I was waiting for a new BIOS to be released before the 07JUL20 XT CPU launch, as I was hoping that would also be when the new APU's would be officially announced (or at least be unveiled), however this BIOS is specifically targeting "gamers, overclocking lovers and PC enthusiasts" So I'm confused.
I'm going to D/L it and try it out but will be really sad if this in a BIOS "unlocking features" for the A300 as a Swan Song before ASRock EOL's the A300
EDIT: So far I have crashed the A300 several times trying to run benchmarks and stress tests in an hour. Have yet to get through 3DMARK Night Raid... Believe them when they say BETA.
EDIT2: Set the RAM to XMP Profile and it completed the benchmark.
EDIT3: Defaulted the BIOS and set the RAM to the XMP Profile... It gave me errors in MEMTEST and crashed the system watching YouTube videos. DDR4-3400 settings settings at 1.35V set to auto everything gave MEMTEST Errors. SO I went to DDR4-3333 Auto everything, No errors at over 1000% runtime. Tightened the Primarry timings to 18-20-20-20-40 and it still passed MEMTEST no problem. So I may try to tighten the timings more but it's really odd that the machine HATES the XMP profile of the RAM and will not run DDR4-34000 speed now with full auto timings. It may trained poorly but if I remember right DDR4-3400 was borderline so I'll stick to DDR4-3333.
Last edited:
I mean, their not wrong... I have had issues with my 3400G giving me a BSOD or freezing while Watching a YouTube video or streaming Netflix\Hulu and playing a "light" GPU intensive game like PAC-MAN 256. The Noctua is 100% the way to go. If it was available when I went to purchase mine I would have gotten the CHOMAX NH-L9a.
Amazon.com: Noctua NH-L9a-AM4 chromax.Black, Low-Profile CPU Cooler for AMD AM4 (Black): Computers & Accessories
Buy Noctua NH-L9a-AM4 chromax.Black, Low-Profile CPU Cooler for AMD AM4 (Black): CPU Cooling Fans - Amazon.com ✓ FREE DELIVERY possible on eligible purchases
amzn.to
Similar threads
