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

gustav

SFF Lingo Aficionado
Jun 18, 2020
105
47
I mean the BIOS is cheating when you set oc values. It reports set but does not. That's why my advice to check.. maybe you are speaking of ram clock... Don't know right now
 

alles_alles

Average Stuffer
Aug 11, 2020
67
19
I mean the BIOS is cheating when you set oc values. It reports set but does not. That's why my advice to check.. maybe you are speaking of ram clock... Don't know right now
In Otherwise it can be an error in bios. Its like you install an older bios . bios set it to to pci e 3.0 x4 but in windows you have to fix it with driver . Because if you don't it set it too pcie3.0 x2 or so hmmmm maybe i need to install chipset driver of asrocks site? don't know hmm
 

gustav

SFF Lingo Aficionado
Jun 18, 2020
105
47
I do not think X300 will install other chipset and gpu drivers than A300. :) But this needs to be verified.
 

rubicoin

SFF Lingo Aficionado
Jan 12, 2020
104
64
just compared win10 download section on official A300/X300 support pages with a little chaotic results:



i guess A300 won't be updated ever again after being discontinued, and it's clear that "Date" column for X300 only means date of upload :)

anyways, regarding chipset divers: where to get them officially outside windows update? and please do not link A320 page on AMD website, as it clearly did not work out good for me...
 

gustav

SFF Lingo Aficionado
Jun 18, 2020
105
47
As I understand there is only one Radeon graphics driver you should get for your APU (also from AMD, APU dedicated download selection)
It will install Adrenaline package und that's all. That's how I run. Never had the need for an explicit chipset driver.

Since you have APU and there is APU dedicated Adrenaline driver package, I think this is regarded there, as APU can act with it's own SoC chipset-like.

Edit 3: could dig into that package and try to find drivers related to 'SMbus', 'PCI Root Complex' or similar. That would be reasonable point of that mentioned above.
 
  • Like
Reactions: rubicoin

alles_alles

Average Stuffer
Aug 11, 2020
67
19
just compared win10 download section on official A300/X300 support pages with a little chaotic results:



i guess A300 won't be updated ever again after being discontinued, and it's clear that "Date" column for X300 only means date of upload :)

anyways, regarding chipset divers: where to get them officially outside windows update? and please do not link A320 page on AMD website, as it clearly did not work out good for me...
hmm the a320 driver work s perfect on the last a300 bioses like 3.60 N and over. PS : if you can install it you have to uninstall the amd chipset driver before. You can't install it if there is an older version on system.
 

alles_alles

Average Stuffer
Aug 11, 2020
67
19
@alles_alles Yea, U need to get an updated version.
It is seems your AMiBCP is outdated and supports only 1 BIOS language

But than again, want is your desired outcome sind this tool?
- aquire some XP
- try to get something running? - I doubt that's possible. Maybe you have to enable more items for the SOC VID to show up. I kept mine with least changes. Therefore it did flash but no wanted outcome was visible.

Btw. I'm still having issues @XMP profile (3200, cl18) using 3.60S. although the vdrop is only 0.1V (which increases stability) unfortunately my sample wants a higher voltage to be applied. 0.1375 vs 0.1V the 3.60S gives us. ->> I'm back on R

This behavior may change when I upgrade to 46xx/47xx or it won't change - then I know I have a faulty power delivery on A300 mainboard.

[3+2] phase design was like 3 phased for the CPU, One phase for SoC and one for APU? Or 2 phases for SOC+APU. I don't remember.

A change in phase-configuration could achieve desired stability results. We'll have somehow to acquire high-res x300 mainboard pictures

Is JZ from JZ computer de also here in forum? I know he has one X300 with P1.40 currently testing.

Maybe we could ask him to provide some high res pics + the "activator chip" should have a different marking (at least one char should differ)

Greetings
where can i get a newer version ? XD
 

alles_alles

Average Stuffer
Aug 11, 2020
67
19
I find it weird that on JZ Electronic they warn that you can't go over 2666Mhz with the new 4750G:

"* This APU (Renoir) does not support the S3 mode (energy saving) in the DeskMini A300!
Info! The Ryzen 7 Pro 4750G can currently max. can be set with PC2666! "
I get an answer of JZ.

the problem is the quality of chips. Some chips are very good and it works, like golden sample. Enough chips are a little worse then the good chips and it doesn't work correctly . A Solution is at work. :) Problems have A520 Boards too and they work on it to fix it.

don't panic ;)
 
Last edited:

rubicoin

SFF Lingo Aficionado
Jan 12, 2020
104
64
i don't understand this. you've just recently linked guckstv video who achieved 4000 mhz ram speeds. no newer beta bios version is available, so why this caution now? JZ/AsRock suddenly got too many feedbacks about this issue? 4750g is not even available to purchase separately anymore.
 
  • Like
Reactions: alles_alles

gustav

SFF Lingo Aficionado
Jun 18, 2020
105
47
You never eat as hot as something is being cooked :) let's just wait a bit more. General Availability is my keyword

@alles_alles I got it from Internet. In post #1204 you can see the version number which did open the BIOS :)
 

alles_alles

Average Stuffer
Aug 11, 2020
67
19
i don't understand this. you've just recently linked guckstv video who achieved 4000 mhz ram speeds. no newer beta bios version is available, so why this caution now? JZ/AsRock suddenly got too many feedbacks about this issue? 4750g is not even available to purchase separately anymore.
because it gives samples that are very good and it gives samples they aren't so good as not so good. Guckstv have luck and get an real good sample. but it gives sample they need more voltage with more ram. They can fix the error in firmware. JZ have get feedback of his customer . Not all 4750g apus are affected
 
  • Like
Reactions: rubicoin

gustav

SFF Lingo Aficionado
Jun 18, 2020
105
47
There is always a tolerance, also producing semiconductors. That's why a 8 core CPU may be re-sold as 6- or 4-Core CPU if there are defects in the silicon disallowing the other cores to run properly. This CPU may be sold as 6-Core or 4-Cores and branded as such.

This way the CPU manufacturers still earn money, even from chips which were designed to be 8C/16T but ended up being 4C/8T or 6C/12T :)

Maybe you mix some things up? I dunno
 
  • Like
Reactions: rubicoin

alles_alles

Average Stuffer
Aug 11, 2020
67
19
There is always a tolerance, also producing semiconductors. That's why a 8 core CPU may be re-sold as 6- or 4-Core CPU if there are defects in the silicon disallowing the other cores to run properly. This CPU may be sold as 6-Core or 4-Cores and branded as such.

This way the CPU manufacturers still earn money, even from chips which were designed to be 8C/16T but ended up being 4C/8T or 6C/12T :)

Maybe you mix some things up? I dunno
No i don't. :) I have asked JZ and he answer that to me that its luck of the its silicon lottery and binning. You have to See that 4750g is not very aviable at free market. So I think amd have problems to produce this cpu at a good binning. And they work on a fix. JZ and AsRock. I think if they would be fixed on the a520 chipset and deskmini a300 and x300, they have to set a lower voltage at same or a bit lower frequenz as standard in firmware. So this calls undervolting by default. And this would works with chips that are not so good.
 
  • Like
Reactions: rubicoin

gustav

SFF Lingo Aficionado
Jun 18, 2020
105
47
I may get you also wrong, but is it not "Ausbeute" what you mean? So we try to find consensus :)

That the Ryzen 4750 is hard to get - what's I think you are saying in #1253 - may be the reason behind it. Maybe we two mean the same :)
 

alles_alles

Average Stuffer
Aug 11, 2020
67
19
I may get you also wrong, but is it not "Ausbeute" what you mean? So we try to find consensus :)

That the Ryzen 4750 is hard to get - what's I think you are saying in #1253 - may be the reason behind it. Maybe we two mean the same :)
" Ausbeute" , "Chipgüte" , "Silikonlotterie " is that not the same? :D He said "Chipgüte" : Some chips have a good "Chipgüte". Some not so good as good :D
 

Valantar

SFF Guru
Jan 20, 2018
1,314
1,097
Regardless of what you're saying here, Renoir chips not supporting DDR4 past 2666 sounds ... highly unlikely. Given that the entire range (mobile + Pro and non-Pro OEM desktop) is rated for 3200 MT/s officially, and U and H parts support LPDDR4x at 4266, any noticeable portion of silicon not even reaching the publicized base DDR4 spec would be a indicative of a significant manufacturing problem.
 
  • Like
Reactions: gustav and rubicoin

rubicoin

SFF Lingo Aficionado
Jan 12, 2020
104
64
Regardless of what you're saying here, Renoir chips not supporting DDR4 past 2666 sounds ... highly unlikely. Given that the entire range (mobile + Pro and non-Pro OEM desktop) is rated for 3200 MT/s officially, and U and H parts support LPDDR4x at 4266, any noticeable portion of silicon not even reaching the publicized base DDR4 spec would be a indicative of a significant manufacturing problem.
this is exactly my point. if 2000g series with shitty memory controller had native 2933 mhz and renoir is rated for 3200 mhz, limiting 4750g to 2666 has to be more reason than silicon lottery :p and why just 4750g? also it's hard to imagine that it could be fixed with motherboard fw if this was a manufacturing problem.
 
  • Like
Reactions: gustav and Valantar

alles_alles

Average Stuffer
Aug 11, 2020
67
19
this is exactly my point. if 2000g series with shitty memory controller had native 2933 mhz and renoir is rated for 3200 mhz, limiting 4750g to 2666 has to be more reason than silicon lottery :p and why just 4750g? also it's hard to imagine that it could be fixed with motherboard fw if this was a manufacturing problem.
no it hasn't more reasons. good 4750g cpus needs less voltage. bad 4750 needs more voltage to run stable at higher ram frequenz. So the A520 chipset and deskmini x300/a300 have an voltage limit. So it comes to less voltage and the cpu doesn't run stable. there is no more reason
 

Valantar

SFF Guru
Jan 20, 2018
1,314
1,097
no it hasn't more reasons. good 4750g cpus needs less voltage. bad 4750 needs more voltage to run stable at higher ram frequenz. So the A520 chipset and deskmini x300/a300 have an voltage limit. So it comes to less voltage and the cpu doesn't run stable. there is no more reason
The CPU cores and memory controller run on entirely separate voltage rails (vCore and vSOC respectively, with the DIMMs themselves having separate voltage regulation circuitry close to the slots), regulated independently, and controllable independently, so what you say here makes no sense. The cores needing less voltage to reach/sustain their clocks should have zero effect on the IMC's voltages.

Besides, the 3200 spec for Renoir is at 1.2V, the stadard JEDEC DDR4 voltage. 2666 or even 2133 doesn't use a lower voltage than this. Of course, many XMP/DOCP settings apply 1.35V or higher for high clocking memory, which is less of a guarantee, but any modern IMC should handle that just fine (1.35V over 1.2V is a 12.5% increase - if the Renoir IMC can't handle that, they have a very serious manufacturing or design issue on their hands).
 
  • Like
Reactions: gustav and rubicoin

alles_alles

Average Stuffer
Aug 11, 2020
67
19
The CPU cores and memory controller run on entirely separate voltage rails (vCore and vSOC respectively, with the DIMMs themselves having separate voltage regulation circuitry close to the slots), regulated independently, and controllable independently, so what you say here makes no sense. The cores needing less voltage to reach/sustain their clocks should have zero effect on the IMC's voltages.

Besides, the 3200 spec for Renoir is at 1.2V, the stadard JEDEC DDR4 voltage. 2666 or even 2133 doesn't use a lower voltage than this. Of course, many XMP/DOCP settings apply 1.35V or higher for high clocking memory, which is less of a guarantee, but any modern IMC should handle that just fine (1.35V over 1.2V is a 12.5% increase - if the Renoir IMC can't handle that, they have a very serious manufacturing or design issue on their hands).
Yes But it does't concerns the cpu part . It concerns the gpu part . I forgot to say that. So I know how instability VEGA can run if you set the wrong settings. And this error that jz write to i can redjust on my 3400g to if i overclock ram settings too high. ;) So this part is a bit in trouble . the igpu part of cpu.
 
  • Like
Reactions: rubicoin