CPU [24/03/2025]ASRock DeskMini X600 BIOS "4.06" & "4.08" & "4.10" - with new AMD AGESA ComboAM5Pi v."1.2.0.2a" & DDR5 SO-DIMM Voltage Settings available

[BigMax]

HWInfo

Thanks for an advise. So I did some research and will post results.

BTW, my x600 runs 4.03MEM01 BIOS.

First, to confirm here is where i change RAM profile:

Next, here are the settings and HWinfo logs RAM-6400

I have noticed here that VDD_SOC is set to 1.300 and marked red, changed it to Auto and made another HWInfo logs. From temps perspective there were no difference to previous - equal to VDD_SOC set to 1.300.

After that, i set the RAM profiles to 4800 and 5600 and captured HWInfo logs (4800 and 5600), both stayed same whith CPU ~10C cooler when idle compared to 6400. I have also made a screenshot of 5600 BIOS settings (4800 is identical):

Appreciate your help and ready to make another experiments

 

[Deleted member 18826]

Thanks for an advise. So I did some research and will post results.

BTW, my x600 runs 4.03MEM01 BIOS.

First, to confirm here is where i change RAM profile:

Next, here are the settings and HWinfo logs RAM-6400

I have noticed here that VDD_SOC is set to 1.300 and marked red, changed it to Auto and made another HWInfo logs. From temps perspective there were no difference to previous - equal to VDD_SOC set to 1.300.

After that, i set the RAM profiles to 4800 and 5600 and captured HWInfo logs (4800 and 5600), both stayed same whith CPU ~10C cooler when idle compared to 6400. I have also made a screenshot of 5600 BIOS settings (4800 is identical):

Appreciate your help and ready to make another experiments

First of all I wouldn't run vSOC at 1.3v. You can think of it as the Memory Controller's voltage and it needs to be high if you run high memory controller frequency. I have no idea why it is being set at 1.3v when running 6400 MT/s memory in 1:2 mode (this is, Memory Controller clock or UCLK = Half of Memory Clock or MEMCLK, which results in 1600 MHz in this case, very low memory controller frequency). You probably would need vSOC at 1.3v in 1:1 mode (because memory controller clock = memory clock = 3200 MHz, which is very high frequency for the memory controller and requires a lot of voltage).

vSOC has a direct impact on idle power consumption and hence temperatures, and so far it seems to be the reason of your ~10C difference, but I would check HWinfo sensor information first to be sure. HWInfo logs don't have much live information, and since modern CPUs are constantly changing all internal frequencies, voltages, etc, it is better to see the average of live idle data, like this (share all sensor information after around 30 seconds of idle to get a good average):

Also, since you don't need 1.3v SOC to run 6400 MT/s in 2:1 mode, ignore what the BIOS is doing and set it to 1.1v (I would be very surprised if it doesn't work, even 1.0v should probably work) and see if it improves the situation. High vSOC causes CPU degradation and 1.3v is very high for daily use that I wouldn't recommend unless strictly necessary.

Edit: I would set vSOC to 1.1v regardless if it improves the temperature situation.

 

[BigMax]

I suspected that 1.3v SOC is way too much, but trusted AsRock Anyway, i set it to 1.0v with RAM at 6400 and here is the screenshot after clean boot and being idle for ~10minutes:

After that i set the RAM to 5600 and SOC to 1.0v and repeated with clean boot and 10 minutes idle:

sadly but looks like if i lost some temp - then not more than 1C or so...

 

[Deleted member 18826]

I suspected that 1.3v SOC is way too much, but trusted AsRock Anyway, i set it to 1.0v with RAM at 6400 and here is the screenshot after clean boot and being idle for ~10minutes:

After that i set the RAM to 5600 and SOC to 1.0v and repeated with clean boot and 10 minutes idle:

sadly but looks like if i lost some temp - then not more than 1C or so...

I've been looking at your numbers for a while and I think I spotted some clues.

One of the things that stands out the most is the 2W power increase from the CPU package. 2W wouldn't tipically cause a 10C temperature difference unless all those 2W would go to a very small area of the CPU (like a single CPU core) which could result in substantial temperature increase from ambient due to power density (high power consumed by a very small piece of silicon) causing that specific area of the CPU to raise in temperature (34C is pretty much ambient and any substantial power density could cause that area of silicon to raise to 45C easily since CPU coolers are not very effective at cooling high power density areas). It would be weird because the CPU Core Power (which is supposed to be the sum of all power consumed by the CPU cores) only goes up by 0.3W and that doesn't sound too much to me, but the delta (difference) between iGPU (GFX) temperature and CPU cores temperature is bigger when running 6400 MT/s memory suggesting that the source of the heat comes from the CPU cores. You can also see that the effective CPU clock is higher, the CPU utilization is also higher and the delta between the "CPU core" temperature (highest temperature of all cores) and "Core temperatures" (which is the average temperature of all cores) is also bigger, suggesting that there is a single core that is hotter than the rest.

Could you expand the "Core powers" and "Core temperatures" to see if there is a single core consuming substantially more power and with substantially more temperature?

Edit: Also expand the "Core Effective Clocks" and "Core VIDs" to see how high each core is boosting to
Edit2: The only thing I could think of that could explain this is a problem with the CPU cooler's thermal paste and/or a problem with the seating of the CPU cooler, but that would be evident by extremely high temperatures under (any) load

 

[BigMax]

right, so here i have:
idle at 5600 vs 6400 expanted core poweres and core temps.

vs

stress tested with expande core powers and core temps 5600 vs 6400

vs

and "Core Effective Clocks" and "Core VIDs" under stress test 5600 vs 6400

vs

before taking idle screenshots it was left 10+ mins idling, stress test - waited untill temp gets stable +- 0,2C (around 5 minutes)

 

[Deleted member 18826]

right, so here i have:
idle at 5600 vs 6400 expanted core poweres and core temps.

vs

stress tested with expande core powers and core temps 5600 vs 6400

vs

and "Core Effective Clocks" and "Core VIDs" under stress test 5600 vs 6400

vs

before taking idle screenshots it was left 10+ mins idling, stress test - waited untill temp gets stable +- 0,2C (around 5 minutes)

This is very interesting.

HWInfo clearily reports CPU cores not being using a substantial amount of power, but half of your CPU cores are hotter than the other half. The iGPU temperature is also lower than the hot cores.

This is a die shot of the 8700G. IF (and this is a big if) the physical core layout is as the following, it could mean that the heat is coming from the "memory controller" side of the chip (and quite a lot of it).

The memory controller's power consumption should be included in SOC power consumption since vSOC feeds the memory controller (among other things). Also, SOC power consumption is lower when running 6400 MT/s which makes sense given that the memory controller is running at lower frequency (~600 MHz at 6400 MT/s vs ~900 MHz at 5600 MT/s on idle), so it shouldn't be the source of the heat.

The other circuitry in that area is the PHY (the physical interface of the memory controller) and that is powered by VDD_IO and VDDP which are set to 1.35v and 1.15v respectively at 6400 MT/s, compared to 1.2v and (probably) 0.8v respectively at 5600 MT/s.

I honestly have no idea how much the PHY is supposed to consume, but it sounds quite strange to me that it could consume so much power so that the bottom CPU cores gets so hot during idle. Also, I'd assume other users would've reported the same behaviour.

The 8700G CPU doesn't have a soldered IHS (integrated heat spreader, the nickel plated copper CPU cover) there is thermal paste instead between the IHS and the CPU die, so it could be a problem with the internal thermal paste of your specific CPU? I have no idea what to think at this point, everything is very strange.

Edit: All I can think of is trying the latest BIOS just in case.
Edit2: Whatever is generating the extra heat is also eating into the CPU power budget, as can be seen in the CPU not being able to provide enough power to the cores to boost as high as when running at 5600 MT/s.

 

[BigMax]

Your research looks like it deserves a Nobel prize thanks a lot for that and i appreciate your support very much. Great job.

i will play around with my cpu further, maybe i can find something as i understand it much better now, and of course i will try latest bios. Honestly, after today’s research i am quite comfortable to go and delid my cpu and change the thermal interface to see how it goes, but i’ll leave it for later.

thanks for help again, will post back if i find something interesting.

 

[Deleted member 18826]

Honestly, after today’s research i am quite comfortable to go and delid my cpu and change the thermal interface to see how it goes, but i’ll leave it for later.

In that case I strongly recommend you to watch this Der8auer video where he replaces the thermal paste between the IHS and the CPU die with liquid metal on a 8700G. I've even seen reports of people improving temps after replacing the stock thermal paste by a good one suggesting that the stock thermal paste is not ideal

 

[BigMax]

In that case I strongly recommend you to watch this Der8auer video where he replaces the thermal paste between the IHS and the CPU die with liquid metal on a 8700G. I've even seen reports of people improving temps after replacing the stock thermal paste by a good one suggesting that the stock thermal paste is not ideal

Thanks again, I’ve watched that already, moreover i’ve seen some reports on local forums that deliding 8700g is quite simple and straight forward, so no fear here.
I’ve also read all thermal paste tests from Igor’s lab and have some options to purchase good thermal paste locally i as am not big fan of liquid meral, so i am quite prepred

 

[Deleted member 18826]

Thanks again, I’ve watched that already, moreover i’ve seen some reports on local forums that deliding 8700g is quite simple and straight forward, so no fear here.
I’ve also read all thermal paste tests from Igor’s lab and have some options to purchase good thermal paste locally i as am not big fan of liquid meral, so i am quite prepred

If you end up doing it please report back your result. I'd be quite interested

 

[matti]

I also own Astock Deskmini x600, Kingston 6400MT/s 16GB x2 kit, 8600G, Thermalright AXP90-X47 FULL+Noctua fan. I have noticed that, those voltages from 6000 and 6400 XMP-profiles, will add about 10 celsius more to processor temps compared to 5600-profile.

This is not related to memory stuff, just writing about my experience: I noticed that using manual Curve Optimizer can lower processor temps. My best 2 cores are at -36, 3-4 cores are at -32 and two weakest cores -25. Weakest cores were unstable at -30. But doing this instead of using all cores, lowered my temps about 5-6 degrees. HWinfo tells you which cores are the best in Sensors->Core Clocks.

 

[BigMax]

If you end up doing it please report back your result. I'd be quite interested

Sure, cant promise it happens tomorrow, but will post here if i do that definetly.

 

[Deleted member 18826]

I also own Astock Deskmini x600, Kingston 6400MT/s 16GB x2 kit, 8600G, Thermalright AXP90-X47 FULL+Noctua fan. I have noticed that, those voltages from 6000 and 6400 XMP-profiles, will add about 10 celsius more to processor temps compared to 5600-profile.

This is not related to memory stuff, just writing about my experience: I noticed that using manual Curve Optimizer can lower processor temps. My best 2 cores are at -36, 3-4 cores are at -32 and two weakest cores -25. Weakest cores were unstable at -30. But doing this instead of using all cores, lowered my temps about 5-6 degrees. HWinfo tells you which cores are the best in Sensors->Core Clocks.

Those 10C more you experienced are on idle or load?

 

[matti]

Those 10C more you experienced are on idle or load?

This is with browsing internet and watching Twitch or Youtube. I have limited my CPU temperature to 65 celsius, so I don't know what it would be without limit. But I would say whole case feels cooler than it used to be. Temperature limit is set to 65 for a reason. My appartment gets really hot at summertime, I have to find settings which I will use. So I am just trying different ways to lower the heat.

My SSD's don't have heatsinks so, I added thermalpad to my SSD's controller chip. Without pad temps were about 70-80 degrees on controller. With thermalpad connecting to motherboard tray temps are about 40-50 on controller. Not that I use my drives that much. No thermalpads on NAND, they don't need cooling.

 

[BigMax]

I also own Astock Deskmini x600, Kingston 6400MT/s 16GB x2 kit, 8600G, Thermalright AXP90-X47 FULL+Noctua fan. I have noticed that, those voltages from 6000 and 6400 XMP-profiles, will add about 10 celsius more to processor temps compared to 5600-profile.

This is not related to memory stuff, just writing about my experience: I noticed that using manual Curve Optimizer can lower processor temps. My best 2 cores are at -36, 3-4 cores are at -32 and two weakest cores -25. Weakest cores were unstable at -30. But doing this instead of using all cores, lowered my temps about 5-6 degrees. HWinfo tells you which cores are the best in Sensors->Core Clocks.

I missed your post for some reason, while it has very useful info! It looks like i am not alone with this feature of different temps correlated to memory profile.

I have CO set to all cores -40 or 43 and it looks quite stable, but your approach is more comprehensive. I tried to go even lower but i loose stability so i may try to lower good cores as much as possible but leave weak as they are now... Btw, which line in the HWInfo do you reference to pick weak cores?

 

[matti]

I missed your post for some reason, while it has very useful info! It looks like i am not alone with this feature of different temps correlated to memory profile.

I have CO set to all cores -40 or 43 and it looks quite stable, but your approach is more comprehensive. I tried to go even lower but i loose stability so i may try to lower good cores as much as possible but leave weak as they are now... Btw, which line in the HWInfo do you reference to pick weak cores?

In sensors window, CPU [#0]: AMD Ryzen, Core Clocks is between the Core VIDs and Bus Clock. Every core has a Perf #number. You can have 2 or more cores with same Perf#number. Like I have Core 3 and core 5 with same Perf#1. My understanding is that system puts load on these cores more often. Then it will use Perf #2 and so on. I am not saying that Perf#2 core is bad or weaker->You just have to test which works for you.

I can set also -40 or even more on all cores, but those weaker cores will crash sometimes. Reason is that they would need more voltage for those clocks to be stable.

 

[BigMax]

thanks, understood. I will make some tests and post results here when i have time.

 

[BigMax]

In sensors window, CPU [#0]: AMD Ryzen, Core Clocks is between the Core VIDs and Bus Clock. Every core has a Perf #number. You can have 2 or more cores with same Perf#number. Like I have Core 3 and core 5 with same Perf#1. My understanding is that system puts load on these cores more often. Then it will use Perf #2 and so on. I am not saying that Perf#2 core is bad or weaker->You just have to test which works for you.

I can set also -40 or even more on all cores, but those weaker cores will crash sometimes. Reason is that they would need more voltage for those clocks to be stable.

Hi Gents, so i had some time to play around this theory and here is what i got.
I get the cores perfs from the HWInfo and it looked like this:

And in BIOS i set the per core curve optimizer according to this, where cores with Perf1 got -37 down to Perf7 that received -47 accordingly. Just to remind you that my previous setting was all cores -40, -45 crashed sometimes. Now i have it 2 days stable in any games, tests etc. Will continue lowering the curves for every core to find the limits.

What it gave me (CPU is limited to 65W, RAM 6400), is i actually lowered the fan curve by ~7-10% in Fan Control, so it idles at 44-45C with fan at 32% (barely audible), games run at 61-65C with fan at 50-50% so it's reasonable loud and stress tests run at 70-72C with fan at almost full speed or 74C with fan at 60-70%, my fan blows out. (room ambient is 23-24C)

What really impresses me much is the iGPU and overall performance of the 8700G... i play games that are 2-3 years old at maximum possible settings (like Doom Eternal - all to ultra nightmare and it never drops below 60FPS (i use v-sync to 60hz), TitanFall2 - same, for modern and new titles or heavy games like Cyberpunk, i would either lower the settings a bit or enable upscaling and again, it's comfortable to play - just checked out w2K Space Marine 2 and fantastic Expedition 33.

Happy for now

 

[NEVi]

What it gave me (CPU is limited to 65W, RAM 6400), is i actually lowered the fan curve by ~7-10% in Fan Control, so it idles at 44-45C with fan at 32% (barely audible), games run at 61-65C with fan at 50-50% so it's reasonable loud and stress tests run at 70-72C with fan at almost full speed or 74C with fan at 60-70%, my fan blows out. (room ambient is 23-24C)

hello, and if you want "dead silent" experience you can use 120mm x 15mm > Arctic P12 SLIM < I have him run on 20% until 50C, which means most of the time spinning at 500-550 rpm (blow in)

 

[BigMax]

hello, and if you want "dead silent" experience you can use 120mm x 15mm > Arctic P12 SLIM < I have him run on 20% until 50C, which means most of the time spinning at 500-550 rpm (blow in)

Hi! Thanks for and advice, i tried it already and from the first look i did not like the results honestly (see few posts up), but at that point of time my CPU was configured different way so i may need another try.