Thanks to @confusis for the instructive review. I just ordered an A300 and I pondering about which 2.5" SATA hdds would be good to make it a small RAID 1 server. It would be asleep most of the time because it would only be for personal use so it could do without special NAS disks. The review suggested using SATA SSDs. That would be beautiful, but two Ironwolf 110 SSDs would be more expensive than the whole package. I am thinking about two 2TB Seagate Barracuda disks -- what would you suggest?
-
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
confusis
John Morrison. Founder and Team Leader of SFF.N
Original poster
SFF Network
SFF Workshop
SFFn Staff
I'd go with the spinning disks for a NAS, especially with low usage and if you don't need outright speed 
Consumer grade disk's are generally ok for a low use NAS (just remember a NAS is part of a 3-2-1 backup plan. 3 copies, 2 off the workstation and 1 offsite)
Consumer grade disk's are generally ok for a low use NAS (just remember a NAS is part of a 3-2-1 backup plan. 3 copies, 2 off the workstation and 1 offsite)
Quick update. 3200MHz isn't totally stable, so I went back to 3000MHz with the XMP profile at 1.2v. Stable so far, and with the lower heat and power consumption versus 1.35v, I might stick with that. . . for now >
Silicone Lottery is a thing, but my A300 has the ram stable at DDR4-3400. It's definitely possible. However that didn't happen without upgrading to 3.60 BIOS and moving from the 2400G to 3400G.
It ? makes a difference. Just not anything anything Earth shattering.
3400 is great! What memory kit are you using to get those kinds of speeds? I know my 3400G can take faster memory, because I've cranked it pretty hard with a b-die kit (and even had the record firestrike score for a couple weeks) but with this kit, it seems pretty stubborn. Clearly the silicone lottery gods didn't bless me with this memory kit lol. Maybe running the memory at 1.55v helped though. . .
Last edited:
Quick question, does anyone know where the Nuvoton SIO on the A300M-STX get's the "CPU" temperature reading from? My understanding would be that it would have to get it's reading from a thermal diode or resistor somewhere on the board and have looked at high res shots of the board but could not figure it out.
Reason I am asking is that the sensors readings in HWInfo64, SIV, HWMonitor etc will differ massively depending on whether only the CPU is loaded, or whether the GPU (+CPU) are loaded. It does not correlate to power usage or processor internal sensor readings in isolated testing. E.g.:
CPU-Z AVX load: 45W CPU power in Ryzen Master, CPU internal sensor 70°C, Nuvoton "CPU" reads <50°C
Furmark load: 27W SOC power in Ryzen Master, CPU/GPU internal sensor 50°C, Nuvoton "CPU" reads 70°C
So for the higher thermal load 20°C under the internal sensor, for the lesser thermal load 20°C over the internal sensor. This doesn't make sense to me if the reading came e.g. from a thermal sensor under the socket or similar.
I find this both confusing and annoying as this reading appears to be at least partially relevant for fan control curve. Anyone noticed the same on theirs or willing to try?
Reason I am asking is that the sensors readings in HWInfo64, SIV, HWMonitor etc will differ massively depending on whether only the CPU is loaded, or whether the GPU (+CPU) are loaded. It does not correlate to power usage or processor internal sensor readings in isolated testing. E.g.:
CPU-Z AVX load: 45W CPU power in Ryzen Master, CPU internal sensor 70°C, Nuvoton "CPU" reads <50°C
Furmark load: 27W SOC power in Ryzen Master, CPU/GPU internal sensor 50°C, Nuvoton "CPU" reads 70°C
So for the higher thermal load 20°C under the internal sensor, for the lesser thermal load 20°C over the internal sensor. This doesn't make sense to me if the reading came e.g. from a thermal sensor under the socket or similar.
I find this both confusing and annoying as this reading appears to be at least partially relevant for fan control curve. Anyone noticed the same on theirs or willing to try?
Seeing the coverage of the 4000 series mobile APUs, what do you all think the chances of seeing a 8 core 16 thread 4800G processor are? I'm also wondering if ASRock will offer support on the A300W. An AMD 4800G would be an instabuy if the A300 could support it.
Hmm, from the A300 / AM4 perspective I was a bit underwhelmed. Zen2 + more cores would be super-welcome but almost all improvement on the iGPU front appears to come from higher clocks and higher memory speed support (2400 MT/s -> 3200 MT/s). The A300 was able to run 3200+ MT/s on the RAM all the time, which is only now coming to mobile, so an equivalent increase would mean we would have to hit 4266MT/s on the A300. That is not going to happen especially since the system will never support DIMM voltage higer than 1.35v. Add to that the less than great SOC voltage rail and I do not see much of the iGPU improvements actually being relevant to an A300 based system.
I absolutely think an 8C/16T APU will come to AM4, but it's performance will be less balanced and more skewed towards CPU than GPU compared to 2000G/3000G series. So value may vary depending on your usage. Best use case may be with an eGPU, making up for lack of many-thread CPUs due to ASRock stance to not support non-APU CPUs.
I hear you on that, I doubt we'll see much improvement on the graphics side. I run my server headless so I'm more concerned about CPU performance. I would like to be able to double the number of VMs I run currently. 3700X performance in such a small package would be pretty amazing.
The new APUs support LPDDR4X 4266MT/s so the necessary increase for a desktop platform to match that would need to be quite a lot bigger. Above 5GT/s at least.
I think these APUs will be quite good nonetheless, even if I was hoping for a bigger GPU. They have IF clocks decoupled from memory speed entirely, so as long as the memory controller is as good as on desktop Zen2 they should go quite high as long as one is willing to pay for good memory and/or sacrifice some latency. The fact that they clock to 1750MHz in a mobile environment is also promising for overclocking.
As for A300 support... I kind of doubt it. Given fundamental changes like decoupling IF from memory speed they will need significant testing for validation, and probably a good deal more bios work than previous chips. Here's to hoping the A300 gets a follow-up to coincide with the launch of 4000-series desktop APUs!
AMD should have synced the naming and technology for their desktop and mobile parts. Really dislike that they drifted apart.
A 4800G will surely arrive as 8C/16T, and perhaps if we are lucky, and to further confuse things, with a Navi chip instead of Vega.
A 4800G will surely arrive as 8C/16T, and perhaps if we are lucky, and to further confuse things, with a Navi chip instead of Vega.
RDNA based APUs are Van Gogh, not Renoir. 5000 series and therefore likely not AM4 anymore. There is no trace of them yet in the Linux drivers afaik. First traces of Renoir appeared like a year ago.
I am actually fairly positive on that. Most of the changes are internal to the processor, from a MB maker perspective, they just get the new AGESA blobs and run with it. The processor still boots in the same way (PSP gets power, looks for magic string in BIOS ROM, loads PSP/SMU firmware, trains memory and hands over to x86 cores). That is all supplied by AMD, from there it is normal BIOS operations but all the initialization has already happened. Given that ASRock has actually been really good about Zen 2 support on even their bottom of the barrel A320 boards, I would be negatively surprised if they would not support a new AM4 APU on the A300.
If Renoir doesn't come to the A300 I don't need it. I want a low energy SFF PC which is more powerful but as silent as my current Core i3-7100/H110 combo with 51W TDP cooled by a Noctua usually at 800 rpm. The 7nm Renoirs could give me a few more cores and threads at the same TDP and that would be more than enough for me. All I want from the iGPU is proper media handling with HDMI 2.0 4K@60Hz HDR and HDCP 2.2. No gaming, just office, surfing and media playback.
The A300 could give m that but the current APUs are too energy hungry and thus need too much cooling.
I am drawn to mini-STX because it houses a desktop CPU but has only one cooling fan to care about. Thankfully their power bricks are external, thus without a need for another noisy cooling fan, and there is no graphics card to cool.
BTW, my incoming replacement A300 seems most likely to be lost/stolen at the nearby DHL distribution center. I asked them to investigate.
The A300 could give m that but the current APUs are too energy hungry and thus need too much cooling.
I am drawn to mini-STX because it houses a desktop CPU but has only one cooling fan to care about. Thankfully their power bricks are external, thus without a need for another noisy cooling fan, and there is no graphics card to cool.
BTW, my incoming replacement A300 seems most likely to be lost/stolen at the nearby DHL distribution center. I asked them to investigate.
ASRock unveils Jupiter A320 mini PC with AMD Ryzen support - Liliputing
ASRock unveils Jupiter A320 mini PC with AMD Ryzen support
liliputing.com
Please can I kindly ask for advice, I am about to give up on the A300. I was looking into the Nuvoton CPU temp sensor thing I have posted before (very high temp readings on GPU load) and tested different BIOS versions for that (My observations were originally made on P1.20). What I found was:
Starting from version P3.40 the fan curves are decoupled from the Nuvoton CPU sensor and only follow CPU internal sensor temp (before that I think fan controller took the higher of CPU Internal Sensor / Nuvoton CPU sensor). Because my CPU internal temps aren't high anyway as I typically turn off boost and even underclock P0 and GPU load alone doesn't really heat up the CPU, that means my fan spins lower with such a BIOS on the same curve. That means wherever the sensor, that the Nuvoton gets the reading from, apparently gets less air / cooling. The result is, after a couple minutes Furmark, 60°C on the CPU/GPU but 90°C on the Nuvoton sensor What is that sensor reading? There are other sensor that clearly show bogus, like ~115°C all the time ("Motherboard"), but this one is not stuck on a bogus value but clearly following GPU load. It is not that it is just offset by a fixed amount either.
I also found - and can't really make any sense of - that SOC Voltage setting appears to have no effect for me in any of BIOS P3.40, P3.50 or P3.60. No matter what I set on the OC tweaker page or in Advanced menu, SOC is stuck at 1V according to HWInfo64 or SIV. Meaning memory settings that are stable in P1.20 with a little SOC voltage tuning aren't stable in P3.X0. I do not remember this being the case when I was on a P3.X0 BIOS before...
So before I take the thing apart to take external temp readings, email ASRock support, swap parts or try to RMA the thing... Is anyone able to confirm my observations or share different ones?
a) On high GPU load e.g. any 3D benchmark or just 3D gaming, the "CPU" sensor that is listed in HWInfo64 under the "ASRock A300M-STX (Nuvoton NCT6793)" section will quickly show temperature values far exceeding the temperatures that are reported from the internal CPU / GPU sensor. If only the CPU is stressed, the sensor will show much lower readings, more in line with actual CPU temp. Picture to show you what I mean:
b) SOC voltage set in BIOS in P3.X0 versions is not reflected in HWInfo64 readings of the "SOC Voltage (SVI2 TFN)" sensor. That sensor will only show around 1.0V, even if SOC voltage was set to e.g. 1.1V or 1.15V in BIOS.
Starting from version P3.40 the fan curves are decoupled from the Nuvoton CPU sensor and only follow CPU internal sensor temp (before that I think fan controller took the higher of CPU Internal Sensor / Nuvoton CPU sensor). Because my CPU internal temps aren't high anyway as I typically turn off boost and even underclock P0 and GPU load alone doesn't really heat up the CPU, that means my fan spins lower with such a BIOS on the same curve. That means wherever the sensor, that the Nuvoton gets the reading from, apparently gets less air / cooling. The result is, after a couple minutes Furmark, 60°C on the CPU/GPU but 90°C on the Nuvoton sensor
What is that sensor reading? There are other sensor that clearly show bogus, like ~115°C all the time ("Motherboard"), but this one is not stuck on a bogus value but clearly following GPU load. It is not that it is just offset by a fixed amount either.I also found - and can't really make any sense of - that SOC Voltage setting appears to have no effect for me in any of BIOS P3.40, P3.50 or P3.60. No matter what I set on the OC tweaker page or in Advanced menu, SOC is stuck at 1V according to HWInfo64 or SIV. Meaning memory settings that are stable in P1.20 with a little SOC voltage tuning aren't stable in P3.X0. I do not remember this being the case when I was on a P3.X0 BIOS before...
So before I take the thing apart to take external temp readings, email ASRock support, swap parts or try to RMA the thing... Is anyone able to confirm my observations or share different ones?
a) On high GPU load e.g. any 3D benchmark or just 3D gaming, the "CPU" sensor that is listed in HWInfo64 under the "ASRock A300M-STX (Nuvoton NCT6793)" section will quickly show temperature values far exceeding the temperatures that are reported from the internal CPU / GPU sensor. If only the CPU is stressed, the sensor will show much lower readings, more in line with actual CPU temp. Picture to show you what I mean:
b) SOC voltage set in BIOS in P3.X0 versions is not reflected in HWInfo64 readings of the "SOC Voltage (SVI2 TFN)" sensor. That sensor will only show around 1.0V, even if SOC voltage was set to e.g. 1.1V or 1.15V in BIOS.
Similar threads
