Prebuilt [SFFn] ASRock's DeskMini A300 - Finally!
- Thread starter confusis
- Start date
@D3NPA are you able to set SOC VID? It may cause the issue? Just to try out? Before setting XMP profile after a fail recovery try to push SOC VID a bit and then F10 and in next boot try to set XMP and see if it boots
Just a suggestion.
Since it's a small tolerance between 2933 and 3000
Maybe it's really a bad ram batch, since I have the same, just 3200 XMP@CL18. They're tested to run with the speed said.
It may not work, but it would add XP to the memory/soc issue subject. Since it could also be a degraded IMC (integrated memory controller).
May I ask which CPU and RAM you used to run / are running?
Just a suggestion.
Since it's a small tolerance between 2933 and 3000Maybe it's really a bad ram batch, since I have the same, just 3200 XMP@CL18. They're tested to run with the speed said.
It may not work, but it would add XP to the memory/soc issue subject. Since it could also be a degraded IMC (integrated memory controller).
May I ask which CPU and RAM you used to run / are running?
Last edited:
already tried 1.15V
no luck
3200G
3.60O BIOS
don't get it wrong
cl16 works fine
cl18 not
I just got my A300 build together for the first time and it's running great, I was just wondering has anyone got solutions for adding more reliable storage to it other than using another M.2 SSD? I was planning add at least 2x 4TB drives running on unRAID but it seems all the 2.5" drives out there of that size are not only too big for the Deskmini but are also SMR. Is there any way to use the SATA ports to run to something like an eSata multi drive enclosure that's powered externally? I don't want to run a full fledged NAS off the Deskmini but would like to use unRAID for it's VM capabilities and for some file stashing.
which bios do you use ?@D3NPA are you able to set SOC VID? It may cause the issue? Just to try out? Before setting XMP profile after a fail recovery try to push SOC VID a bit and then F10 and in next boot try to set XMP and see if it boots
Just a suggestion.
Maybe it's really a bad ram batch, since I have the same, just 3200 XMP@CL18. They're tested to run with the speed said.
It may not work, but it would add XP to the memory/soc issue subject. Since it could also be a degraded IMC (integrated memory controller).
May I ask which CPU and RAM you used to run / are running?
If you prepare your desmini a300 case at the side where the hdd is you can place 2,5 Seagate 4 TB HDD. But the under site must be like a flat screen. . And you have to make the part of the case wher install the hdd flat too ..
Note . The part of the VESA holder you have to make this a little shorter.
Note : With One hdd you can use 3 screws for vesa holder. With two hdds 2 screws or was it one scew? to the Vesa holder. Please use any second glue or plastic glue or something to install the vesa bracket in vesa holder. I think this is safe enough .
what a shame on asrock that they don't change it on x300
Last edited:
Thank you very much, first of all. This is pretty weird. Cl16 works at 3000MHz but CL18 does not? What's that?!
Which RAM have you had before this g.skill
? Just wondering. I would try out the 3.60R (works best for me) ... But if Cl16 works, it's fine then, or what do you think @alles_alles now I know what you meant. Thanks for the link. The same behaviour you do experience, when the SoC voltage rail too weak is.
This is observable. I mean also using Ryzen 2400G. Must be a special AsRock built-in instability feature.
Ok, when A300 runs on power limit, it can not supply SoC voltage -> therefore the IMC degrades in terms of it's speed -> instability using higher speed -> lowering RAM CLK... what a nicework around. bravo -.- ... if the x300 has the same physical capabilities, I will be shocked.
There is tolerance in production, of course, and there is silicone-difference, but the mainboard is still facing the same problem as with 2000G Series running at high RAM CLK. SoC line Vdrops. If they remove SOC VID from BIOS options, I'm lost. It will run on 2933MHz, thought. Just remember, AsRock pushed this product with 2000G Series availability, which was spec'd at 2933MHz max. There may be just no room for silicone-tolerance from the mainboard side of things.
I bet this is not chip lottery on the 4750G - as it's probably manufacturing tolerance, which the mainboard should fiigure out / measure / be able to adjust slightly and recover from running at full speed. But what do I know.
Last edited:
It gives 4750g chips on the chip lottery they work with 1,0 SOC voltage stable. But this is the less of the 4750g that it works with. The most works with 1,1 Voltage , but it set to under 1,1 volt @ autoThank you very much, first of all. This is pretty weird. Cl16 works at 3000MHz but CL18 does not? What's that?!
Which RAM have you had before this g.skill
@alles_alles now I know what you meant. Thanks for the link. The same behaviour you do experience, when the SoC voltage rail too weak is.
This is observable. I mean also using Ryzen 2400G. Must be a special AsRock built-in instability feature.
Ok, when A300 runs on power limit, it can not supply SoC voltage -> therefore the IMC degrades in terms of it's speed -> instability using higher speed -> lowering RAM CLK... what a nicework around. bravo -.- ... if the x300 has the same physical capabilities, I will be shocked.
There is tolerance in production, of course, and there is silicone-difference, but the mainboard is still facing the same problem as with 2000G Series running at high RAM CLK. SoC line Vdrops. If they remove SOC VID from BIOS options, I'm lost. It will run on 2933MHz, thought. Just remember, AsRock pushed this product with 2000G Series availability, which was spec'd at 2933MHz max. There may be just no room for silicone-tolerance from the mainboard side of things.
I bet this is not chip lottery on the 4750G - as it's probably manufacturing tolerance, which the mainboard should fiigure out / measure / be able to adjust slightly and recover from running at full speed. But what do I know.
i have tested the 2400g with x300 Bios on a300 with high ramsettings . So the result is :
For a ramclock of 3600. I need a SOC Voltage more then 1,26 Voltage for Vega 11 Graphics also don't work stable because that makes no sence.
For a ramclock of 3533. Need SoC Voltage of more then 1,26 Voltage for Vega 11 Graphics. With thist it don't work stable
For a ramclock of 3466. Need SoC Voltage of more then 1,237 Voltage for Vega 11 Graphics. This Works extrem Stable . More then 1,24 makes no sence
For a ramclock of 3466. Need SoC Voltage of more then 1,237 Voltage for Vega 11 Graphics. This Works extrem Stable .
For a ramclock of 3400. Need SoC Voltage of more then 1,225 SOC Voltage for Vega 11 Voltage .Works extrem Stable.
But 3600 needs 1,225 Voltage If it works with the cpu part only..
On 4750 you need 0,1 less i thin at voltage . if you work with vega 8 graphics . On cpu side every ram configuration works stable on this cpu. You can set it up to 4133. No problem . But not on gpu side if you have to les SOC voltage
do you understand it now?
I think I understand you most of the time, thank you.
It's considered NOT SAFE if you go over 1.2V on SOC VID. It (may) degrades your memory controller. (Due to higher voltage the transistors get it, and are stressed more till they start fading out) Memory Controller is part of the SoC design, so the SoC voltage defines also the power for the memory controller.
I see what you are saying, but it all results into/hints at a bad SoC voltage rail / adjustment on the mainboard
Earlier it was simple: you had a memory controller in the northbridge on the mainboard.
Then Athlon 64 came along introducing built-in memory controller. Intel followed that path, also by using AMD x86-64 instruction set.
That's why it's sometimes called x86/i386 and amd64.
Now we have here System on a Chip, which means, the memory controller is not anymore responsible for CPU-RAM data only. It's responsible for the whole SoC design. And if there are additional components accessing the memory, it will go through the IMC. Hence Vega-issue is very sensetive to Voltage -- the CPU not I get it ^^
I think they had to deal with concurrency on the IMC which device has the priority to access data first, etc. This is very sensetive topic and requires logic glued into IMC by itself. IMC being part of SoC is kinda own logic inside a system. But mainboard should be able to tune it's voltage rails to accord the requirements (this is the case for up to 2933MHz on A300).
If the IMC and SoC becomes more and more insesetive to Vdrop on the other side, it would be more a development by AMD than AsRock Or the Vdrop caused by the silicone is less profound would also be AMD's achievement - maybe this is the "lottery" but I think most CPU's will run just fine with other mainboards since they have better power delivery. .. ah yea, that's what you say: there are 4750G's running with SoC on 1.0V but there most of them required SoC to be 1.1V - yea. Kommt hin
PS: Mine is running at SOC VID = 1.13750 at 3200MHz,CL18 just to keep away the Vdrop in GPU-apps. And still, it jumps between [1.033V and 1.131V in HWINFO64]
I think the chip which is responsible for the voltage-level-regulation is not the best one, i dont know, maybe @Danlopez1222 could take a look under VRM's
It's considered NOT SAFE if you go over 1.2V on SOC VID. It (may) degrades your memory controller. (Due to higher voltage the transistors get it, and are stressed more till they start fading out) Memory Controller is part of the SoC design, so the SoC voltage defines also the power for the memory controller.
I see what you are saying, but it all results into/hints at a bad SoC voltage rail / adjustment on the mainboard
Earlier it was simple: you had a memory controller in the northbridge on the mainboard.
Then Athlon 64 came along introducing built-in memory controller. Intel followed that path, also by using AMD x86-64 instruction set.
That's why it's sometimes called x86/i386 and amd64.
Now we have here System on a Chip, which means, the memory controller is not anymore responsible for CPU-RAM data only. It's responsible for the whole SoC design. And if there are additional components accessing the memory, it will go through the IMC. Hence Vega-issue is very sensetive to Voltage -- the CPU not
I get it ^^I think they had to deal with concurrency on the IMC which device has the priority to access data first, etc. This is very sensetive topic and requires logic glued into IMC by itself. IMC being part of SoC is kinda own logic inside a system. But mainboard should be able to tune it's voltage rails to accord the requirements (this is the case for up to 2933MHz on A300).
If the IMC and SoC becomes more and more insesetive to Vdrop on the other side, it would be more a development by AMD than AsRock
Or the Vdrop caused by the silicone is less profound would also be AMD's achievement - maybe this is the "lottery" but I think most CPU's will run just fine with other mainboards since they have better power delivery. .. ah yea, that's what you say: there are 4750G's running with SoC on 1.0V but there most of them required SoC to be 1.1V - yea. Kommt hinPS: Mine is running at SOC VID = 1.13750 at 3200MHz,CL18 just to keep away the Vdrop in GPU-apps. And still, it jumps between [1.033V and 1.131V in HWINFO64]
I think the chip which is responsible for the voltage-level-regulation is not the best one, i dont know, maybe @Danlopez1222 could take a look under VRM's
Last edited:
i don't think that the chips dies. I think they are very stable . and 0,1 voltage isn't very much at this pointI think I understand you most of the time, thank you.
It's considered NOT SAFE if you go over 1.2V on SOC VID. It (may) degrades your memory controller. (Due to higher voltage the transistors get it, and are stressed more till they start fading out) Memory Controller is part of the SoC design, so the SoC voltage defines also the power for the memory controller.
I see what you are saying, but it all results into/hints at a bad SoC voltage rail / adjustment on the mainboard
Earlier it was simple: you had a memory controller in the northbridge on the mainboard.
Then Athlon 64 came along introducing built-in memory controller. Intel followed that path, also by using AMD x86-64 instruction set.
That's why it's sometimes called x86/i386 and amd64.
Now we have here System on a Chip, which means, the memory controller is not anymore responsible for CPU-RAM data only. It's responsible for the whole SoC design. And if there are additional components accessing the memory, it will go through the IMC. Hence Vega-issue is very sensetive to Voltage -- the CPU not
I think they had to deal with concurrency on the IMC which device has the priority to access data first, etc. This is very sensetive topic and requires logic glued into IMC by itself. IMC being part of SoC is kinda own logic inside a system. But mainboard should be able to tune it's voltage rails to accord the requirements (this is the case for up to 2933MHz on A300).
If the IMC and SoC becomes more and more insesetive to Vdrop on the other side, it would be more a development by AMD than AsRock
PS: Mine is running at SOC VID = 1.13750 at 3200MHz,CL18 just to keep away the Vdrop in GPU-apps. And still, it jumps between [1.033V and 1.131V in HWINFO64]
I think the chip which is responsible for the voltage-level-regulation is not the best one, i dont know, maybe @Danlopez1222 could take a look under VRM's
Sorry, I welcome the fact when people try to use their knowledge to do things and stuff, but some things have to be read up.
I suggest you reading:
- How to Kill Your CPU with “Safe” Voltages | Raven Ridge SOC Voltage Guidelines.
- AMD CPUs AMD Ryzen (2000-Series) Overclocking Guide
Also very interestng video. Check beginning till ~3:00mins. (For Raven Ridge)
I'm not arguing with you, I'm just pointing out, that there was made an observation by couple of people and companies.
It's wide known in the community, go over 1.2V to o/c and test, but not on 24/7 system.
In the end, I am not trying to advice you to do something or not, for me it does not matter. I just wanted to shed a bit of light on this SoC VID topic.
Don't forget, when you speak about USB running at 5V - 0.1V is not much, that's right.
When the relative voltage is as small as 1.0V than 0.1V is like 10% more. That's rather "much" to me
PS.: Our AsRock mainboard has no LLC-Settings, which are actually used to prevent Vdrop.
Edit x: From Nexus Article Conclusions:
I suggest you reading:
- How to Kill Your CPU with “Safe” Voltages | Raven Ridge SOC Voltage Guidelines.
- AMD CPUs AMD Ryzen (2000-Series) Overclocking Guide
Also very interestng video. Check beginning till ~3:00mins. (For Raven Ridge)
I'm not arguing with you, I'm just pointing out, that there was made an observation by couple of people and companies.
It's wide known in the community, go over 1.2V to o/c and test, but not on 24/7 system.
In the end, I am not trying to advice you to do something or not, for me it does not matter. I just wanted to shed a bit of light on this SoC VID topic.
Don't forget, when you speak about USB running at 5V - 0.1V is not much, that's right.
When the relative voltage is as small as 1.0V than 0.1V is like 10% more. That's rather "much" to me
PS.: Our AsRock mainboard has no LLC-Settings, which are actually used to prevent Vdrop.
Edit x: From Nexus Article Conclusions:
Last edited:
Ok You are rigth but at SOC Voltage. ?? The Ram Voltage does run with 1,35 Voltage, because it is an 1,35 Module. And Rawiolli goes there Often @1,4 volt or so . And VEGA is an Powerful chip , that needs power . Before you kill the SOC / Ramcontroller and VEGA you will kill your Motherboard mosfetsSorry, I welcome the fact when people try to use their knowledge to do things and stuff, but some things have to be read up.
I suggest you reading:
- How to Kill Your CPU with “Safe” Voltages | Raven Ridge SOC Voltage Guidelines.
- AMD CPUs AMD Ryzen (2000-Series) Overclocking Guide
Also very interestng video. Check beginning till ~3:00mins. (For Raven Ridge)
I'm not arguing with you, I'm just pointing out, that there was made an observation by couple of people and companies.
It's wide known in the community, go over 1.2V to o/c and test, but not on 24/7 system.
In the end, I am not trying to advice you to do something or not, for me it does not matter. I just wanted to shed a bit of light on this SoC VID topic.
Don't forget, when you speak about USB running at 5V - 0.1V is not much, that's right.
When the relative voltage is as small as 1.0V than 0.1V is like 10% more. That's rather "much" to me
PS.: Our AsRock mainboard has no LLC-Settings, which are actually used to prevent Vdrop.
Edit x: From Nexus Article Conclusions:
This i got there with my 2400g on a biostar b450gt3. and here i have make lots of more oc then on the sweet deskmini i have looked up the video . Its logical if you overclocke your VEGA 11 at MHZ you need more power at fps on soc.Sorry, I welcome the fact when people try to use their knowledge to do things and stuff, but some things have to be read up.
I suggest you reading:
- How to Kill Your CPU with “Safe” Voltages | Raven Ridge SOC Voltage Guidelines.
- AMD CPUs AMD Ryzen (2000-Series) Overclocking Guide
Also very interestng video. Check beginning till ~3:00mins. (For Raven Ridge)
I'm not arguing with you, I'm just pointing out, that there was made an observation by couple of people and companies.
It's wide known in the community, go over 1.2V to o/c and test, but not on 24/7 system.
In the end, I am not trying to advice you to do something or not, for me it does not matter. I just wanted to shed a bit of light on this SoC VID topic.
Don't forget, when you speak about USB running at 5V - 0.1V is not much, that's right.
When the relative voltage is as small as 1.0V than 0.1V is like 10% more. That's rather "much" to me
PS.: Our AsRock mainboard has no LLC-Settings, which are actually used to prevent Vdrop.
Edit x: From Nexus Article Conclusions:
But i have only tweak the ram OC and oc the soc a bit . Soo 1,237 voltage should be ok at deskmini and the 2000 apu. But i have to set this because in otherwise it doesn't work stable.
I think we have to look up what the system set to auto on soc on a 4750g and set this first before we oc the ram . I don't think you need 1,2 Voltage at Renoir. Maybe 1,15 or so . I get some chip to test . So make an report
Yes, I can imagine that one on A300Before you kill the SOC / Ramcontroller and VEGA you will kill your Motherboard mosfets
After reviewing the video by myself, I have to admit, it's rather safe, but we do not know which Voltage is truly on the PINs.
This is also my understanding, Renoir's 7nm process will also help in terms of voltage. This has to be tested out, I agree.
F4-2400C16D-8GRS at 2933 17(16+gear down)-18-18 probablyWhich RAM have you had before this g.skill
Yes I found the posts for the extra USB ports, I already have one installed actually, but not the audio cable. Did a search for 'audio' in the two A300 topics as well but found nothing. I could have sworn I had found a listing on aliexpress once for a cable that looked exactly the same but last time I searched there the results didn't give anything useful. My current speakers are connected with usb so no direct problem but perhaps for the future I would want to use it.
got that cheap ram
cl18 profile doesn't even boot, lol
even with 1.35V
interesting, i have same ram sticks (reinserted screenshot here for easy comparision), but slightly older chip version means lower jedec dimm label & minimum timing delay specs:
i run them with 4650g on 3.60S @ 3200 mhz @ 16-18-18-43 1.2 V without problems. never tried cl18 though
i could not find any information regarding the bios version he uses for this build. 3700x in deskmini, how did he manage to run it in a300?
He didn't. He tried it and it didn't work so he switched to a 2400g.
Similar threads
