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

[ShamedGod]

I can confirm the 3400G works in the A300 with zero issues on the 3.50 BIOS. It also has better thermals BUT it throttles down to 3.7 GHz after boosting for a bit. So the 3400G doesn't have much sustained performance boost over the 2400G. The only reason I upgraded was for the 10-12% graphic performance increase, which is real, but wouldn't have been necessary if the 2400G was overclock-able in the A300 :/

Hopefully the BIOS/Driver updates keep coming.

 

[GLSRacer]

Thanks for the info. Let's hope we get OC functionality on the A300. I've contemplated delidding my 2400G and running liquid metal but it's not really necessary until we can overclock the APU.

 

[Slowjim]

I can confirm, got mine in yesterday. Using the Wraith Stealth and Thermal Grizzly it was able to stay under 80c for the duration of the test, which is a couple degrees lower than my 2400g was. But it's not boosting to 4.2 - maybe a power limitation. One upside is, with my G.skill 3200 cl18, I was unable to OC past 3200 with my 2400g. I was only able to tweak timings down to 16 which helped a tiny bit. But with the 3400g I am able to easily get the RAM to 3400.

Fire Strike 3400g: https://www.3dmark.com/fs/19829883
Skydiver 3400g: https://www.3dmark.com/sd/5605656

Fire strike 2400g: https://www.3dmark.com/fs/19664278
Skydiver 2400g: https://www.3dmark.com/sd/5584829

 

[GLSRacer]

Yeah, I was wondering about thermal headroom in the A300 case. I suspected that a lot of the gains were due to the switch to a larger cooler for the 3400G, a cooler that won't fit in the A300. ETA Prime just did an initial comparison video with the 3400G. He got between 2 and 18 FPS improvement based on the game. He noted that the CPU often boosted to 3950 MHz during game play, this was using the larger OEM cooler. I suspect that real world performance would be similar if we could OC the 2400 to 4 GHz CPU and 1.45 GHz GPU. Buildzoid had his 2400G GPU up to 1.6 GHz fully stable. ETA Prime has a few more OC videos in the works for the 3400G with comparisons to an OC'd 2400G, he also mentioned a small form factor build using the 3400G.

 

[ShamedGod]

I can confirm, got mine in yesterday. Using the Wraith Stealth it was able to keep it under 80c for the duration of the test, which is a couple degrees lower than my 2400g was. But it's not boosting to 4.2 - maybe a power limitation. One upside is, with my G.skill 3200 cl18, I was unable to OC past 3200 with my 2400g. I was only able to tweak timings down to 16 which helped a tiny bit. But with the 3400g I am able to easily get the RAM to 3400.

Fire Strike 3400g: https://www.3dmark.com/fs/19829883
Skydiver 3400g: https://www.3dmark.com/sd/5605656

Fire strike 2400g: https://www.3dmark.com/fs/19664278
Skydiver 2400g: https://www.3dmark.com/sd/5584829

I have my ram set to the same specs as the 2400G @ 3333. With the 2400G I was able to get 3400+ but I notice decrease performance as the Ryzen memory calculator said the chip needed more voltage then 1.35 to run at that rate reliably. Have you found the performance superior?

Also, at no time did my 3400G hit the max boost of 4.2. I thought that was normal from the reports I'm hearing about the BIOS for Ryzen 3000 (even though this is a Zen+ chip) and how AMD is setting the boost spec for their chips to "theoretical max" and leaving little overclocking head room.

 

[Slowjim]

Interesting comparison here - 3dmark Sky diver w/3400g running 3200 ram at XMP (on the right) vs 3400@1.35v on the left. Both CL18 timings. Higher overall score with 3400 but the multithreaded tests all seem to suffer.

And here is a test of my 2400g OC'd in my previous board. A full 1000 points higher than my 3400g in A300.

I scored 13 586 in Sky Diver

AMD Ryzen 5 2400G, AMD Radeon Vega 6 Mobile x 1, 8192 MB, 64-bit Windows 10}

www.3dmark.com

 

[ShamedGod]

Yeah, I was wondering about thermal headroom in the A300 case. I suspected that a lot of the gains were due to the switch to a larger cooler for the 3400G, a cooler that won't fit in the A300. ETA Prime just did an initial comparison video with the 3400G. He got between 2 and 18 FPS improvement based on the game. He noted that the CPU often boosted to 3950 MHz during game play, this was using the larger OEM cooler. I suspect that real world performance would be similar if we could OC the 2400 to 4 GHz CPU and 1.45 GHz GPU. Buildzoid had his 2400G GPU up to 1.6 GHz fully stable. ETA Prime has a few more OC videos in the works for the 3400G with comparisons to an OC'd 2400G, he also mentioned a small form factor build using the 3400G.

I am using the Noctua and the chip will boost above 4 GHZ but it will quickly drop down. I do not believe the issue is with the cooler. A friend of mine built his own A300, but he is waiting to see if he wants to buy the 3400G or just use my 2400G, if he decides to get the 3400G I can compare the two machines to see if the results are the same. I'm not sure why it will not stay boosted, maybe I lost the silicone lottery, but Ryzen Master will report a Total Socket Power of Over 100% (98-108%) before throttling back to between 70-80 percent running cinebench. That leads me to believe it's a power delivery issue with the A300 :/

If the issue is power delivery I would expect ASRock would come out with a second Gen A300 for the 2020 APUs.

I am looking at this as a "Video Card" upgrade, not really a CPU upgrade though.

 

[ShamedGod]

Interesting comparison here - 3dmark Sky diver w/3400g running 3200 ram at XMP (on the right) vs 3400@1.35v on the left. Both CL18 timings. Higher overall score with 3400 but the multithreaded tests all seem to suffer.

And here is a test of my 2400g OC'd in my previous board. A full 1000 points higher than my 3400g in A300.

I scored 13 586 in Sky Diver

AMD Ryzen 5 2400G, AMD Radeon Vega 6 Mobile x 1, 8192 MB, 64-bit Windows 10}

www.3dmark.com

I got my G.Skill ram to run DDR4-3333 @ 1.35 volts with 16-18-18-18-38 timings, but I had to "train" the board to except the specs. Go with stock XMP, then move to 3200 with higher voltage & tighter timings, then 3333. I couldn't just go strait into 3333 with out issue and use CL16. Would be curious if your results would be better at 3333 with tighter timings as well...


I was trying to get 3200 CL14 and 3400 CL anything to perform better and be stable but this was what worked best in my setup. I have not readjusted for the 3400G though. If the system limitation is actually power delivery I don't think there is much performance left on the table.

 

[NateDawg72]

I do not believe the issue is with the cooler ... I'm not sure why it will not stay boosted, maybe I lost the silicone lottery, but Ryzen Master will report a Total Socket Power of Over 100% (98-108%) before throttling back to between 70-80 percent running cinebench. That leads me to believe it's a power delivery issue with the A300 :/

I believe you are correct. From what the guys who are trying to mod the bios of the A300 have said, it seems to have power delivery issues. Specifically I recall that it provides excessive voltage which inflates power draw and leads to power limit throttling. No one has been able to get undervolting working so far ?

 

[Quango]

Anyone tried a higher wattage power brick? Maybe the brick is not suited for OC?

 

[GLSRacer]

It would be nice to source a higher output power supply. Maybe something in the 175 to 200 watt range.

 

[NateDawg72]

I don't think a better power supply will help because right now only ram can be OC and the issue is motherboard BIOS / VRM

 

[Slowjim]

I have a dell 330w I can try... just need to do some checking and make sure it won't fry everything before i test it out.

 

[GLSRacer]

Nice, hopefully the voltages and polarity of the plug will be correct. I'll be interested to see how it works out.

 

[Slowjim]

I got my G.Skill ram to run DDR4-3333 @ 1.35 volts with 16-18-18-18-38 timings, but I had to "train" the board to except the specs. Go with stock XMP, then move to 3200 with higher voltage & tighter timings, then 3333. I couldn't just go strait into 3333 with out issue and use CL16. Would be curious if your results would be better at 3333 with tighter timings as well...


I was trying to get 3200 CL14 and 3400 CL anything to perform better and be stable but this was what worked best in my setup. I have not readjusted for the 3400G though. If the system limitation is actually power delivery I don't think there is much performance left on the table.

I tried this and got up to 12768 in Sky Diver, compared to 12561 at 3400 CL18. Nice improvement!

In regards to the Dell brick, it looks like it's 19.5v output and the Asrock one is only 19. Not sure I should take the chance. Barrel plug looks the same though

 

[slimjim]

I tried this and got up to 12768 in Sky Diver, compared to 12561 at 3400 CL18. Nice improvement!

In regards to the Dell brick, it looks like it's 19.5v output and the Asrock one is only 19. Not sure I should take the chance. Barrel plug looks the same though

Go for it!
Nice name btw

 

[ShamedGod]

I tried this and got up to 12768 in Sky Diver, compared to 12561 at 3400 CL18. Nice improvement!

In regards to the Dell brick, it looks like it's 19.5v output and the Asrock one is only 19. Not sure I should take the chance. Barrel plug looks the same though

Glad it worked

It's a coin flip on the power brick. As @NateDawg72 said: it's not JUST the power brick, it's also the BIOS and the VRM. I'm not Buildzoid so I can't be 100% sure but I would believe ASRock would design the motherboard specifically to handle the load of the included power brick and it's variances and no more.

Voltage regulator module - Wikipedia

en.wikipedia.org

After using third party and "other then intended/included" power bricks on other machines/devices, with mixed results, it's advisable to not mess with, or swap, the power brick unless your confident OR can afford a catastrophic failure.

 

[NateDawg72]

I actually think asrock designed a fairly robust VRM for this board and it is well cooled ? I wouldn't have a problem trying other power supplies with it.

I just think the ryzen CPUs are throttling themselves because of how the Bios is configured on this board. It really shouldn't be getting so much voltage and I think that's why it's throttling so much at the stock power limit.

 

[GLSRacer]

The excessive voltage could also be part of the reason the chip runs so warm at idle. I'm really hoping that Dan Lopez and others will be able to crack the BIOS issue. Over the years I've tried to make time to study firmware modification techniques but other more important tasks always come up.

 

[A300]

Hi guys,

I only concern about cpu performance. Is it worth to upgrade to 3400g?

Currently, my rig scored about 777 in cb r15, with 2666Mhz OC to 3000Mhz in BIOS 3.40.

To all of you that already upgraded to 3400g, can you please tell me your cb r15 score and memory speed when run it?

Thanks in advance.