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

[NateDawg72]

Asrock wrote very clearly that this is an A300 chipset, what problems do you have?

Ah but that's the thing, can you point me to the a300 chipset on the board?
Since the chipset appears to be just the ryzen chips themselves, there is interest from enthusiasts here to see if more features can be exposed. Especially since the current bios doesn't seem to allow much other than memory tweaking.

 

[Kmpkt]

Asrock wrote very clearly that this is an A300 chipset, what problems do you have?

Locked chipsets typically, in my experience anyhow, allow undervolting and simply lock out overclocking functions. As far as iGPU frequency goes, right back to the first Japanese reviews the option to do this was in BIOS which led me and others to believe this was also within the functionality of the A300 chipset and Mini-STX BIOS.

Considering that from a gaming perspective an undervolt of the CPU with an overclock of the iGPU would be absolutely ideal. This board has appeared to have those functions since coverage started and I simply feel disappointed these features aren't included. If you look at (as an example) Timespy numbers with GPU overclock you can see about 25% more performance. I know that wouldn't translate linearly to every game, but without a doubt a lot of performance is being left on the table.

 

[W4RR10R]

Asrock wrote very clearly that this is an A300 chipset, what problems do you have?

As I understand it the X300 and A300 are purely BIOS, the "chipset" on the AM4 platform is a super IO chip, it just expands the IO of the CPU.

 

[Kmpkt]

Find myself wondering if one of the older BIOSes might have these features enabled:

https://www.asrock.com/nettop/AMD/DeskMini A300 Series/#BIOS

 

[Debris84]

After playing around a bit, I've managed to get the RAM stable at 3333 MHz on the A300. as well as a stable undervolt (I believe) on the CPU at 1.1875 V and GPU running at 1700 MHz for a TimeSpy score of 1327.

What Bios are you on and what ram are you using?

 

[Kmpkt]

I'm using the most recent BIOS (3.40 I believe). For RAM I am using a pair of G.Skill's 3800 MHz BDie SODIMMS at 3333 MHz with 14-14-14-14-30 timings. I've tried 3400 and 3466 but they require more voltage to be stable. The A300 board has hard voltage settings of 1.20 and 1.35V so I don't think those speeds are an option at present.

 

[Windfall]

I'm using the most recent BIOS (3.40 I believe). For RAM I am using a pair of G.Skill's 3800 MHz BDie SODIMMS at 3333 MHz with 14-14-14-14-30 timings. I've tried 3400 and 3466 but they require more voltage to be stable. The A300 board has hard voltage settings of 1.20 and 1.35V so I don't think those speeds are an option at present.

Good data.

Would assume that perf loss using slower RAM is even more obvious on this board than ITX?

 

[NateDawg72]

Would assume that perf loss using slower RAM is even more obvious on this board than ITX?

Why would it be any different? Am I missing something? I'd expect results should be within margin of error given similar ram speeds & timings, unless you're getting at sodimms having looser timings if not overclocked?

 

[Windfall]

Why would it be any different? Am I missing something? I'd expect results should be within margin of error given similar ram speeds & timings, unless you're getting at sodimms having looser timings if not overclocked?

In @confusis's review, he mentioned a significant performance drop from ITX to STX.

 

[Danlopez1222]

Here's something interesting that I found in the CBS menu for version 3.20, which I'm using for better memory overclock.

0x15C4F             Variable 0x6 equals 0xF {12 86 06 00 0F 00}
0x15C55                 Not {17 02}
0x15C57             End {29 02}
0x15C59             Setting: OC Mode, Variable: 0x28 {05 91 23 00 24 00 09 70 00 50 28 00 14 10 00 05 00}
0x15C6A                 Option: Normal Operation, Value: 0x0 {09 07 25 00 00 00 00}
0x15C71                 Option: OC1, Value: 0x1 {09 07 26 00 00 00 01}
0x15C78                 Option: Customized, Value: 0x5 (default) {09 07 27 00 10 00 05}
0x15C7F             End of Options {29 02}

When I manually change the variable to 0x0 or 0x1 with RU.efi, the voltage for the CPU stays around 1.375ish. However, something with memory overclocking automatically triggers an event to set it to 0x5, and then CPU voltage skyrockets to 1.488, sometimes hitting 1.5v on idle (according to HWiNFO64). I can set it back to 0x1 after configuring my memory at 3200Mhz, and it reverts the CPU voltage anomalies.

I'd like to know ASRock's reasoning for this.

 

[Boil]

Man, sounds like I dodged a bullet (so far) by not getting one of these units yet...

I am seriously thinking of ditching my PC & going with nothing but a 12.9" iPad Pro, which is about a third of a liter, INCLUDING the display...! ;^p

 

[NateDawg72]

In @confusis's review, he mentioned a significant performance drop from ITX to STX.

Ah that's right, I remember that now. I don't believe the performance delta was expected though, it must be that something is off with how asrock's STX boards are configured (in firmware/bios, not user configuration). From what Danlopez1222 has been investigating it sounds like there might be issues with voltage and power limits.

 

[Kmpkt]

Since we can't OC/UV and iGPU is locked, I'm curious what people's power consumption is looking like on this unit. I've had a hard time getting it to pull over 100W from the wall with the HDPlex 160W AC-DC and am thinking of trying to run the system on the HDPlex 80W AC-DC? Crazy? Yes/No?

 

[Windfall]

Since we can't OC/UV and iGPU is locked, I'm curious what people's power consumption is looking like on this unit. I've had a hard time getting it to pull over 100W from the wall with the HDPlex 160W AC-DC and am thinking of trying to run the system on the HDPlex 80W AC-DC? Crazy? Yes/No?

Please try it.

For SCIENCE!

 

[Kmpkt]

So I did some stress testing on the HDPlex 80W AC-DC and it performed admirably. Prime95 ran for 1/2 hour without issue pulling around 90-95W from the wall constant. Timespy runs pulled between 75-80W at peak and there was no issues. From a cold boot the unit managed a Prime95/FurMark dual stress test for about 7 minutes before shutting down, most likely due to overtemp protection as the unit was in excess of 50C on it's casing at the time of reset. It is worth mentioning that during the dual stress test the unit was pulling about 115-120W from the wall which means at ~85% efficient it was outputting about 100W (20 over its rated wattage). I believe overcurrent protection on the 80W is set at 110W which means this is likely an ideal pairing for the Deskmini A300.

Anyone want to help me design a 3D printable brickless enclosure?

 

[W4RR10R]

So I did some stress testing on the HDPlex 80W AC-DC and it performed admirably. Prime95 ran for 1/2 hour without issue pulling around 90-95W from the wall constant. Timespy runs pulled between 75-80W at peak and there was no issues. From a cold boot the unit managed a Prime95/FurMark dual stress test for about 7 minutes before shutting down, most likely due to overtemp protection as the unit was in excess of 50C on it's casing at the time of reset. It is worth mentioning that during the dual stress test the unit was pulling about 115-120W from the wall which means at ~85% efficient it was outputting about 100W (20 over its rated wattage). I believe overcurrent protection on the 80W is set at 110W which means this is likely an ideal pairing for the Deskmini A300.

Anyone want to help me design a 3D printable brickless enclosure?

I'd love to design an enclosure for it.

 

[Windfall]

So I did some stress testing on the HDPlex 80W AC-DC and it performed admirably. Prime95 ran for 1/2 hour without issue pulling around 90-95W from the wall constant. Timespy runs pulled between 75-80W at peak and there was no issues. From a cold boot the unit managed a Prime95/FurMark dual stress test for about 7 minutes before shutting down, most likely due to overtemp protection as the unit was in excess of 50C on it's casing at the time of reset. It is worth mentioning that during the dual stress test the unit was pulling about 115-120W from the wall which means at ~85% efficient it was outputting about 100W (20 over its rated wattage). I believe overcurrent protection on the 80W is set at 110W which means this is likely an ideal pairing for the Deskmini A300.

Anyone want to help me design a 3D printable brickless enclosure?

Quite interested, even had a preliminary design for a case.

In my spoiler hit projects and then HTPC Micro.

 

[Primerib]

Does anyone have any information on what 3200mhz+ SO-DIMMS are B dies? I'm running the Kingston 2933 C17 set from the QVL and the timings when speed is raised to 3200 are to high for my taste and I know this effects the APU performance greatly. I would like to see something that can reach at least 16-17-17-18 or maybe even 15 at 3333 or 3400. I can only find B die list for desktop memory online.

 

[Danlopez1222]

Does anyone have any information on what 3200mhz+ SO-DIMMS are B dies? I'm running the Kingston 2933 C17 set from the QVL and the timings when speed is raised to 3200 are to high for my taste and I know this effects the APU performance greatly. I would like to see something that can reach at least 16-17-17-18 or maybe even 15 at 3333 or 3400. I can only find B die list for desktop memory online.

These are B-die: https://www.newegg.com/Product/Product.aspx?Item=N82E16820232725

Currently running them at 3266mhz 14-14-15-14-28 1.35v

 

[Kmpkt]

The G.Skill 3200 set linked above as well as the G.Skill 3800 MHz 32GB kit are the only B-Die SO-DIMMs I've been able to find.