CPU [20/08/2024]ASRock DeskMini X600 BIOS "4.03.MEM01" (with RAM voltage settings!!!) - AGESA ComboAM5Pi "1.2.0.0a"

[HydrAxx747]

I have found the error!
Cause of error: Driver.

I have now tested all the latest drivers.
From version > Adrenalin 24.6.1 (27.06.2024) the problems occur.

No idea what AMD screwed up from then on and whether they will ever fix it.
For all who have similar problems: Up to version Adrenalin 24.5.1 (15.05.2024) everything works fine!

The only thing that bothers me now is that it only boots sporadically with more than 5200 Mhz.
Is this due to the RAM (bad kit), the CPU (bad CPU) or the DeskMini (bad DeskMini?).

As I said, with 6400 Mhz and 1.2V the computer runs like a charm when it boots.
No errors in MemTest86 and no problems in benchmarks.

Can you explain the problem you encountered because I don't see at all what you're talking about? In addition, your account is completely new so I don't see where you would have explained it on this forum, so sorry but what you report seems very vague to me but I would like to understand.

 

[DeiDei]

Can you explain the problem you encountered because I don't see at all what you're talking about? In addition, your account is completely new so I don't see where you would have explained it on this forum, so sorry but what you report seems very vague to me but I would like to understand.

Oh, sorry for the confusion.
I had found this thread of yours as I was hoping your released BIOS for the DeskMini X600 would solve my problem.

As described, I could not play Transport Fever 2 with crashes and Cities: Skylines II at all.


At first I thought it was the RAM (hence your BIOS) because there were problems with the memory of the graphics card, which comes from the RAM in an APU) and thought I would get help here.

But now I know it was a problem with the driver (see previous answers).

My question that remains now and where you may have valuable information:

Why does the DeskMini X600 with the mentioned memory (Kingston FURY Impact SO-DIMM Kit 64GB, DDR5-5600, CL40-40-40, on-die ECC) - no matter which RAM OC settings (5200, 5600, 6400) - not always boot? Sometimes it takes 30-60 seconds until it displays a picture and boots, sometimes I have to switch the DeskMini on/off several times until it is ready to boot.

The DeskMini copes with all stress tests for hours without any problems. The memory runs stable at 1.2V and 6400. The boot problems also occur with the standard clock (of course I have loaded the profile for this!).

So I only have the problem with booting. I am using your published BIOS 4.03.MEM01).

That's why I wrote in this thread.
Could I explain it a little better now?

 

[HydrAxx747]

Oh, sorry for the confusion.
I had found this thread of yours as I was hoping your released BIOS for the DeskMini X600 would solve my problem.

As described, I could not play Transport Fever 2 with crashes and Cities: Skylines II at all.


At first I thought it was the RAM (hence your BIOS) because there were problems with the memory of the graphics card, which comes from the RAM in an APU) and thought I would get help here.

But now I know it was a problem with the driver (see previous answers).

My question that remains now and where you may have valuable information:

Why does the DeskMini X600 with the mentioned memory (Kingston FURY Impact SO-DIMM Kit 64GB, DDR5-5600, CL40-40-40, on-die ECC) - no matter which RAM OC settings (5200, 5600, 6400) - not always boot? Sometimes it takes 30-60 seconds until it displays a picture and boots, sometimes I have to switch the DeskMini on/off several times until it is ready to boot.

The DeskMini copes with all stress tests for hours without any problems. The memory runs stable at 1.2V and 6400. The boot problems also occur with the standard clock (of course I have loaded the profile for this!).

So I only have the problem with booting. I am using your published BIOS 4.03.MEM01).

That's why I wrote in this thread.
Could I explain it a little better now?

From what I know (I still have little experience with how DDR5 memory works sorry), the long POST times would be due to the "Memory Training" option in the BIOS which would be used to calibrate the ideal timing parameters between the CPU memory controller and the RAM sticks in order to guarantee good system stability but is mainly useful when using the XMP or EXPO parameters pre-integrated in the RAM sticks because in the latter many timings and secondary parameters are left in "Auto" at the discretion of the system and therefore the Memory Training comes to test (when the PC is turned on) various settings to find those which are optimal to guarantee the stability of the system but this has the disadvantage of greatly extending the POST times depending on the hardware equipping the PC, the solution to considerably reduce the POST time at home was to manually adjust the maximum of secondary timings and other secondary RAM parameters (which are usually on "Auto" by default) and to deactivate the "Memory Training" option in the settings RAM in the BIOS, this allowed me to reduce the POST time to about a little less than ten seconds and I have no stability problems.

But sincerely as I already said earlier I am still far from being an expert in DDR5 memory settings/overclocking (I really miss the DDR4 era where everything was clearer and simpler) so I want to say that I do not guarantee that the settings I used on my DDR5 SO-DIMM RAM will not create any problems for you, so in your place I would rather ask advice on this subject from seasoned DDR5 RAM tuners/overclockers, but on the other hand specify to them that you are using DDR5 RAM sticks in SO-DIMM format because otherwise they could have timigs settings (and other secondary settings) that are perhaps a little too aggressive for this RAM format.​

 

[DeiDei]

@HydrAxx747

Thank you very much! This expands my knowledge enormously!

I didn't know that the DeskMini checks itself and its timings for stability.
No wonder it takes a while to boot up. I can live with that.

However, I had to lower my Mhz a bit because I was getting graphics errors and occasional crashes.

I use for the kit:
Kingston FURY Impact SO-DIMM Kit 64GB, DDR5-5600, CL40-40-40, on-die ECC

The 5600 mhz profile that comes with the kit.
However, I have now set the Mhz to 6000 (instead of 6400 previously) and the voltage to 1.35V. Everything else is from the Kingston kit profile. So far this is the most stable OC my DeskMini offers.

6000 Mhz is pretty close to the sweet spot (or it's even the sweet spot) so I'm leaving it that way. That's enough for me.
With 5600 Mhz you give away too much APU/GPU performance in my opinion.

There is certainly still some room for improvement, but the performance is good enough for me.

If you have any tips, please let me know.

 

[HydrAxx747]

@HydrAxx747

Thank you very much! This expands my knowledge enormously!

I didn't know that the DeskMini checks itself and its timings for stability.
No wonder it takes a while to boot up. I can live with that.

However, I had to lower my Mhz a bit because I was getting graphics errors and occasional crashes.

I use for the kit:
Kingston FURY Impact SO-DIMM Kit 64GB, DDR5-5600, CL40-40-40, on-die ECC

The 5600 mhz profile that comes with the kit.
However, I have now set the Mhz to 6000 (instead of 6400 previously) and the voltage to 1.35V. Everything else is from the Kingston kit profile. So far this is the most stable OC my DeskMini offers.

6000 Mhz is pretty close to the sweet spot (or it's even the sweet spot) so I'm leaving it that way. That's enough for me.
With 5600 Mhz you give away too much APU/GPU performance in my opinion.

There is certainly still some room for improvement, but the performance is good enough for me.

If you have any tips, please let me know.

I have a contact who uses the same model of DDR5 SO-DIMM RAM kit as yours and who advised me these settings for 8000MHz CL38 (yes! yes! ) where he holds more than 2 hours of Y-Cruncher (preset VT3) and more than 50 complete cycles of TestMem5 (preset 1usmus) stable, these are his everyday usage settings on his DeskMini X600 (bios "4.03.MEM01") and APU Ryzen 5 8500G, I have not yet had time to test them but if you want to try, you will gain 30~40% additional performance on your iGPU due to the significant increase in RAM/VRAM bandwidth, however you must remember to set your UCLK ratio to "UCLK=MCLK/2" (and not "UCLK=MCLK") to be able support such a high frequency, here is the ZenTimings screenshot he provided me which contains all the details of his voltage, timing and internal RAM resistance settings, to get 8000MHz CL38 @1.40v stable:

Correction: this is a Kingston FURY Impact SO-DIMM 5600MHz CL40 kit but in 2x16GB or 32GB (and not 2x32GB or 64GB like you) but there is a way with a few adjustments that it will still work for you.​

 

[DeiDei]

8000 Mhz. Wow!

I'll give it a try in the next few days and report back.
Thank you for the feedback!

P.S. Doesn't the RAM get way too hot?

 

[DeiDei]

8000 Mhz. Wow!

I'll give it a try in the next few days and report back.
Thank you for the feedback!

P.S. Doesn't the RAM get way too hot?

Current:

 

[HydrAxx747]

8000 Mhz. Wow!

I'll give it a try in the next few days and report back.
Thank you for the feedback!

P.S. Doesn't the RAM get way too hot?

I don't know exactly sorry but these are its everyday usage parameters.
On the other hand, inexpensive accessories that allowed me to lose almost 15°C at full load, this is what I use on my Kingston FURY Impact 6400MHz CL38 32GB (2x16GB) kit.
These are heatsinks specially designed for SO-DIMM RAM sticks, they are made of a pure copper main plate with a thickness of 0.5mm as well as a complete graphene coating improving their native dissipation, and on one of their sides there is also an adhesive thermal pad allowing them to be put in place easily and without screws (remember to remove the original Kingston labels and stickers for better conductivity of the RAM chips through the heatsinks), they are sold individually so remember to put 4 units in your basket, one last thing you may see other forms of similar SO-DIMM compatible heatsinks on AliExpress, but I do not recommend them because the latter generally have a much narrower thermal adhesive strip which does not allow to cover and make contact with a double row of RAM chips and their dedicated VRM which equips our Kingston FURY Impact models of DDR5 SO-DIMM kit (32 or 64GB). The one I present to you is the only one that can do the job, the others will prevent you from covering all the RAM chips and also the VRM.​

aliexpress.com/item/1005006726577782.html

 

[DeiDei]

Thank you.

But unfortunately the timings mentioned by your colleague didn't work for me.

The computer then restarts itself and loads at 5200 Mhz.
But I can't find all the variables in the BIOS that are listed in ZenZimings. Perhaps that is also the problem.

 

[martel80]

So I was finally able to assemble my x600.
For some reason, my M2 drive is recognize at every other reboot so I was not able to install windows on it. It's a shame because I now have Windows installed on a old 2,5'' drive which defeat the purpose of paying so much money for a 2tb Blazing M2.
I'm actually wondering if it's not an option from the bios that would make it act like that ?
Anyways, I've installed the latest BIOS 4.03 and I'm still getting 65W
It also says my RAM is clocked at 3200khz. I don't understand whats going on but something doesnt seems right.

If anyone have any pointers, I'd really appreciate.

 

[martel80]

Ok, I reinstalled everything and ran Geekbench 6.

ASRock X600M-STX - Geekbench

Benchmark results for an ASRock X600M-STX with an AMD Ryzen 7 9700X processor.

browser.geekbench.com

Single core score higher then the highest in the chart. Not sure why ...

Multi core score about equal to a 7945HX or a 14600K.

Core temp now read 105W peak. That is great improvement.

I think I'm back in business here.

Now I need to understand what happened to the M2 drive and why it's still not showing up.

 

[HydrAxx747]

Ok, I reinstalled everything and ran Geekbench 6.

ASRock X600M-STX - Geekbench

Benchmark results for an ASRock X600M-STX with an AMD Ryzen 7 9700X processor.

browser.geekbench.com

Single core score higher then the highest in the chart. Not sure why ...

Multi core score about equal to a 7945HX or a 14600K.

Core temp now read 105W peak. That is great improvement.

I think I'm back in business here.

Now I need to understand what happened to the M2 drive and why it's still not showing up.

Greetings friend, I see that you are using a Ryzen 7 9700X and unfortunately even if the latest BIOS "4.03" and "4.03.MEM01" support this latest generation of Ryzen CPUs, they are still equipped only with the AGESA firmware "ComboAM5PI v1.2.0.0a" which is far from optimal for managing the Ryzen 9000 series, moreover these AGESA firmware versions (and earlier) are known to have problems supporting certain models of M.2 NVMe PCIe Gen4/Gen5 SSDs and also in the CPU power management there are some issues...

In my humble opinion you would have done better to get a Ryzen 7 8700G APU (which has performances quite close to the 9700X, are much better supported, especially have an "iGPU" graphics part at least 8 to 10 times more powerful depending on the scenario and have an integrated NPU processor that can handle AI calculations, something that the 9700X cannot do because it is devoid of an integrated NPU processor unfortunately), but in the meantime you will have to be patient with your Ryzen 7 9700X because you will have to wait for a BIOS with the AGESA firmware "ComboAM5PI v1.2.0.2" to come out (which should not be too long now) so that most of the problems you encounter are fixed.​

 

[martel80]

Greetings friend, I see that you are using a Ryzen 7 9700X and unfortunately even if the latest BIOS "4.03" and "4.03.MEM01" support this latest generation of Ryzen CPUs, they are still equipped only with the AGESA firmware "ComboAM5PI v1.2.0.0a" which is far from optimal for managing the Ryzen 9000 series, moreover these AGESA firmware versions (and earlier) are known to have problems supporting certain models of M.2 NVMe PCIe Gen4/Gen5 SSDs and also in the CPU power management there are some issues...

In my humble opinion you would have done better to get a Ryzen 7 8700G APU (which has performances quite close to the 9700X, are much better supported, especially have an "iGPU" graphics part at least 8 to 10 times more powerful depending on the scenario and have an integrated NPU processor that can handle AI calculations, something that the 9700X cannot do because it is devoid of an integrated NPU processor unfortunately), but in the meantime you will have to be patient with your Ryzen 7 9700X because you will have to wait for a BIOS with the AGESA firmware "ComboAM5PI v1.2.0.2" to come out (which should not be too long now) so that most of the problems you encounter are fixed.​

We (the audio engineering community) are definitely waiting for that BIOS update to come out. iLok (the security software responsible for most of our plugins management ) have said that it was going to be in BIOS production by early November as it is currently being tested. As of today, most audio plugin don't work on AMD Series 9000 (AMD 5 for that matter) and it is creating quite a steer in the community.

Do you think this BIOS update will also fix my Blazing M2 issue ?

As weird as it may sound, it was originally recognized at every other boot when I inserted it in the M2_2 slot underneath the Mobo. So it was impossible to install Win 11 on it because it would crash at about 23% for whatever reason. Now that I've placed it in the M2_1 slot, it doesnt even show up anywhere and thats supposed to be the one that support Gen 5 M2....

Anyways, for the time being, I've installed Win 11 and a few random stuff on a old 2'5'' SSD.

Yesterday, the pain was a lot more profound when pretty much everything was going wrong. But I now have a stable machine, thanks to @Gidilo help and support.

I reverted back to the JDEC profile for my RAM which is set at 4800mhz.

All the XMP profiles would make the Multi-core Cinebench stress test do a Black Screen in the first minute of the test. That's quite sad as I bought those at 6400 mhz expecting to get some sort of benefit from it. Even the 5600mhz XMP profile would make the Cinebench Multi-core stress test crash and it's suposed to be supported by the CPU natively.....

It's the first time I run into such a difficult and and finicky computer. Issues with M2, Issues with RAM, issues with Security software ....

I hope it'll soon transform in some sort of wine and get better with age haha.

 

[martel80]

@HydrAxx747

Following your recommendation:

 

[Wakefull]

I just dropped by to say thank you for the excellent BIOS. And to show a photo of my Frankenstein. 8700G +

(Kingston FURY Impact 32GB (2x16GB) 6000MT/s CL38 DDR5 SODIMM KF560S38IBK2-32) OC 7800 with cooper heatsink​

 

[HydrAxx747]

I just dropped by to say thank you for the excellent BIOS. And to show a photo of my Frankenstein. 8700G +

(Kingston FURY Impact 32GB (2x16GB) 6000MT/s CL38 DDR5 SODIMM KF560S38IBK2-32) OC 7800 with cooper heatsink​

Thanks for your feedback, what is the thickness you chose for your SO-DIMM RAM heat sink model? And what are your settings to get a stable 7800MHz (a ZenTimings v1.32 screenshot would be ideal)?

 

[Wakefull]

Thanks for your feedback, what is the thickness you chose for your SO-DIMM RAM heat sink model? And what are your settings to get a stable 7800MHz (a ZenTimings v1.32 screenshot would be ideal)?

Heatsink 3mm version (From ali. put on search bar 3256803235523728 ) my memory has a one-sided design.
Note. Be prepared to grind down the heatsink! Some components are higher than the memory chips, which prevents proper contact with the heatsink and increases the risk of a short circuit. I used a grinder and insulated the protruding components with Polyimide Tape. For your information, the memory temperature will be lower than reported by HWINFO since the "SPD hub" won't make direct contact with the heatsink, and the tape will also act as a barrier.

Geekbench

 

[Stoney]

you are using to much voltage for the timings. you could run below 1.2v with that timings to keep everything cooler.

Edit: btw consider also testing 1950fclk for that sweet 2:1:1 sync which benefits latency a lot. I think i tested 2:1:1 sync vs 2:1:highfclk in timespy and the graphics score was slightly better with sync.

 

[Wakefull]

you are using to much voltage for the timings. you could run below 1.2v with that timings to keep everything cooler.

Edit: btw consider also testing 1950fclk for that sweet 2:1:1 sync which benefits latency a lot. I think i tested 2:1:1 sync vs 2:1:highfclk in timespy and the graphics score was slightly better with sync.

This is an interim setup. Absolute values haven't been determined yet. I plan to delid the CPU. My heatsinks and overall cooling system allow me to run these voltages. )))

 

[Wakefull]

Synchronization by the 2:1:1 formula won't provide read speeds above ~63,000. Only with FCLK overclocking can better results be achieved. The 8X00G is a completely different processor. Everything that applies to other Ryzen CPUs is either irrelevant here or has minimal effect.

I’m not sure if 8000 will be stable everywhere, but the 2:1:1 result is worse even at a higher frequency (timings were not changed).