APU BIOS Overclocking question (Ryzen 5600G)

[rfdzn]

I have been searching for an answer to this question I have but no luck:

"What metrics should I look at to determine that an overclock is safe for daily use?"

I'm currently running an APU only system on an Asrock B550M-HDV, Ryzen 5600G, and after a day of tuning I get these numbers to run with no crashes both on benchmarks and just me playing games (mostly Warframe and Monster Hunter World)

4.1Ghz all-cores @1.2V,
GFX clock 2300MHz @1.3V,
GFX Memory stock (1600Mhz)

PBO On

 

I'm on stock cooler and temperatures when gaming are at around max of 75C for CPU and 65-70 for GPU. prime95 max'd at 90C after 30min torture test w/ hyperthreading and avx on.

as a noob, I only make sure the voltages are within spec (below max of 1.5 as the BIOS said) and the temps are not crazy hot.
but is there something I am missing?

 

Thank you.

[Jurrunio]

1. VRM temperature. On a motherboard this typically refers to the power delivery of the CPU. This board has everything negative about this (no heatsink at all, weak power capacity) which means it is easy to make it run hot.

 

To make it worse, it's divided into two parts, the bigger part powers the CPU cores while the smaller part powers everything else. PCIe controller and memory controller are among them, but the power hog here is the iGPU which can easily draw more than half that of the CPU cores. On this board, you have 4 phases for the CPU and 2 for everything else. Bad news.

 

2. Voltage. SOC voltage is what powers the "everything else" I've mentioned including the iGPU. I dont recommend going beyond 1.2V for the sake of lifespan.

[BiotechBen]

48 minutes ago, rfdzn said:

I have been searching for an answer to this question I have but no luck:

"What metrics should I look at to determine that an overclock is safe for daily use?"

I'm currently running an APU only system on an Asrock B550M-HDV, Ryzen 5600G, and after a day of tuning I get these numbers to run with no crashes both on benchmarks and just me playing games (mostly Warframe and Monster Hunter World)

4.1Ghz all-cores @1.2V,
GFX clock 2300MHz @1.3V,
GFX Memory stock (1600Mhz)

PBO On

 

I'm on stock cooler and temperatures when gaming are at around max of 75C for CPU and 65-70 for GPU. prime95 max'd at 90C after 30min torture test w/ hyperthreading and avx on.

as a noob, I only make sure the voltages are within spec (below max of 1.5 as the BIOS said) and the temps are not crazy hot.
but is there something I am missing?

 

Thank you.

On stock cooling, those are pretty standard numbers, 70C is nothing to worry about when gaming, and 90C while power virus is pretty expected.

Personally, the PRO4 is a much better board choice. Plenty of peripheral cooling for components.

[rfdzn]

24 minutes ago, Jurrunio said:

1. VRM temperature. On a motherboard this typically refers to the power delivery of the CPU. This board has everything negative about this (no heatsink at all, weak power capacity) which means it is easy to make it run hot.

 

To make it worse, it's divided into two parts, the bigger part powers the CPU cores while the smaller part powers everything else. PCIe controller and memory controller are among them, but the power hog here is the iGPU which can easily draw more than half that of the CPU cores. On this board, you have 4 phases for the CPU and 2 for everything else. Bad news.

 

2. Voltage. SOC voltage is what powers the "everything else" I've mentioned including the iGPU. I dont recommend going beyond 1.2V for the sake of lifespan.

I didn't know this board only had 6 power phase. At the time I intend to run everything stock because I only need it for college so I believe I bought the best I could afford (which turns out to be pretty bare minimum).

Is VRM temp monitoring available on HWMonitor? I've been using it to track the metrics.

I left the SoC voltage to auto in BIOS and only change GFX Voltage. I assume the SoC voltage would just automatically pull what it needs?

Looking at how the power phases are divided, would it be fine to run the 4.1 all core @1.2V and probably dial back on the GFX to <1.2V?

I do want this system to last but at the same time I want to try my luck with a little bit more performance. I did notice an improvement on the two games I played the most, fortunately.

[rfdzn]

35 minutes ago, BiotechBen said:

On stock cooling, those are pretty standard numbers, 70C is nothing to worry about when gaming, and 90C while power virus is pretty expected.

Personally, the PRO4 is a much better board choice. Plenty of peripheral cooling for components.

Good to hear the temps are reasonable, it is on stock cooler right now but may upgrade in the future.

At the time I went for what I could best afford, because I didn't really plan on overclocking since I wasn't playing much games. Now I'm starting to play more and in hindsight the Pro4 would definitely be a better purchase.

[Jurrunio]

27 minutes ago, rfdzn said:

I didn't know this board only had 6 power phase. At the time I intend to run everything stock because I only need it for college so I believe I bought the best I could afford (which turns out to be pretty bare minimum).

Is VRM temp monitoring available on HWMonitor? I've been using it to track the metrics.

I left the SoC voltage to auto in BIOS and only change GFX Voltage. I assume the SoC voltage would just automatically pull what it needs?

Looking at how the power phases are divided, would it be fine to run the 4.1 all core @1.2V and probably dial back on the GFX to <1.2V?

I do want this system to last but at the same time I want to try my luck with a little bit more performance. I did notice an improvement on the two games I played the most, fortunately.

VRM temp monitoring is really unlikely to be there on such a cheap board. HWinfo64 in sensor mode shows more data than HWmonitor (I honestly dont understand the appeal of HWmonitor, it's like free version of HWinfo in terms of things it can't do but HWinfo is also free).

 

A BIOS error, GFX voltage and SOC voltage on Ryzen are tied. I assume the motherboard just runs the higher value of the two.

 

Stock cooler (as a downdraft) should keep the CPU core VRM at ok temps, just dail back the SOC voltage.

 

Do memory overclocking instead

[rfdzn]

48 minutes ago, Jurrunio said:

VRM temp monitoring is really unlikely to be there on such a cheap board. HWinfo64 in sensor mode shows more data than HWmonitor (I honestly dont understand the appeal of HWmonitor, it's like free version of HWinfo in terms of things it can't do but HWinfo is also free).

 

A BIOS error, GFX voltage and SOC voltage on Ryzen are tied. I assume the motherboard just runs the higher value of the two.

 

Stock cooler (as a downdraft) should keep the CPU core VRM at ok temps, just dail back the SOC voltage.

 

Do memory overclocking instead

I used HWMonitor just because it's kinda laid out fairly simply. I understand your point though.

I dialed GFX back to 2000MHz @1.05V and it's running well so far. For some reason I don't lose that much performance. only lost about 30 points and like 3 fps on Heaven bench. I think I'll keep this OC since the difference between 2000Mhz and 2300Mhz is quite negligible for what I use my PC for but still better than stock.

GFX Memory overclocking is rather daunting for me right now, there's so many things to tune and I don't  know where to start. I'll read up and watch more tutorials before doing anything with it.