Problem with an AMD 4650 Pro APU

Summerbreeze

New Member
Problem with an AMD 4650 Pro APU

I have some strange readings in HwInfo64 7.16 and would like to clarify whether these values are perhaps being read incorrectly.

I own a Lenovo T14 Gen1 notebook with the above APU.16GB (Samsung) memory are soldered, 16GB (Crusial) are additionally installed as SO-DIMM.
The clock rate of the RAM is always displayed to me as 1598.8 = 1600 MHz = 3200 DDR rate.

The graphics memory clock is also constantly displayed with 1200 MHz. Only under heavy load does the display reduce to 800 MHz.
Today I also saw 1333 MHz once.
According to HwInfo, the 400 MHz of the graphics memory recorded in the picture is only very rarely present for a fraction of a second. However, the AMD graphics driver says that the 400 MHz should be present quite often.
What is correct now?
I know that the driver display is not always reliable. That's why I'm asking here.

Also the number of PCIe lanes and their load status are apparently not displayed correctly?
Until now I was of the opinion that the APU only had 8+4 PCIe3 lanes? But 16+4 PCIe4 are displayed?
If the values displayed for the main memory were correct, it would mean unnecessary consumption of energy.
GPU-z has read out the following about the graphics:
Vendor Id: 1002
Device Id: 1636
Subvendor Id: 17AA
Subsystem Id: 5081
PCI Revision: D3
Memory: 2048 MB
GPU Clock: 1500 MHz

I have the suspicion that energy is being consumed unnecessarily, as the battery life is not as good as once hoped for and was advertised by Lenovo in the first few weeks.
At that time, you could read about "up to" 15 hours.
Of course, 15 hours with a 50Wh battery is very sporty.Personally, I would be satisfied if I could get a steady 10 hours without limiting myself too much. Currently it's no more than 8.5.

Can anyone confirm that the values are read correctly?
If so, I would contact Lenovo.

Screenshot 2022-01-17 111305.jpg
Screenshot 2022-01-17 111014.jpg
Screenshot 2022-01-17 111122.jpg
 
Last edited:
The GPU driver is responsible for reporting of actual GPU memory clock and this is sometimes not reliable as has been also described in Radeon Software release notes:

Known Issues​

  • Radeon performance metrics and logging features may intermittently report extremely high and incorrect memory clock values.
 
@Martin
Ok.
Then I can assume that the values from HwInfo are absolutely reliable.
Then my system memory is actually always running fixed at 1600 MHz (DDR 3200) and consumes more energy than is actually necessary.

Even if the memory controller display is constantly moving up and down, I can assume that the memory chips are always running at a fixed clock rate.
And my memory controller is not running at the clock rate that should actually be possible with a Renoir APU? Up to 1600 MHz with DDR 3200 memory.

Normally IF, UCLK & MemCLK should run in sync for best performance?

Have I interpreted this correctly?

The highest measured UCLK clock at the highest possible load was 1200 MHz.


Screenshot 2022-01-18 114225.jpg

Then I can assume that Lenovo has configured the bios or its settings incorrectly and that I should now contact the Lenovo support?

I would like to thank you in advance for your efforts.
 
Well, the situation with APUs is a bit different. Since the iGPU shares the same system memory, the GPU and main memory clock should be same.
However AMD said in the past that for APUs one should trust the iGPU memory clock as the main memory clock reporting doesn't reflect some low-power states. However, we have no proof for this and AMD never provided further details. Moreover, we've also seen many times that the GPU memory clock wasn't reported correctly. So you might better ask AMD, though I'm not sure if they will provide a clear statement.
 
Well, the situation with APUs is a bit different. Since the iGPU shares the same system memory, the GPU and main memory clock should be same.
However AMD said in the past that for APUs one should trust the iGPU memory clock as the main memory clock reporting doesn't reflect some low-power states. However, we have no proof for this and AMD never provided further details. Moreover, we've also seen many times that the GPU memory clock wasn't reported correctly. So you might better ask AMD, though I'm not sure if they will provide a clear statement.
Thank you very much.
That at least gets me a little further.
So I will first see if I can get some information from AMD on Reddit.
I won't bother the normal support with this.
 
Back
Top