Zarathustra[H]
Extremely [H]
- Joined
- Oct 29, 2000
- Messages
- 39,027
Hey all,
I am new to APU's and did some googling to figure out if my problem was normal, but haven't found much.
My HTPC uses a cheap "hold me over until Kaveri" Richland A4-4000. It's a dual core with stock clock at 3Ghz, turbo up to 3.2Ghz.
Last night I decided to test if my cooling was adequate. I wanted to keep my build cheap, so I used only components I had kicking around. I opted against the boxed heatsink as it looked WAY too wimpy. Instead I used an old bigger copper heatsink that came in the box for my Thuban Phenom II x6 1090T. I figured if this boxed cooler is designed for a 125W TDP CPU, it ought to be able to handle this 65W TDP CPU.
My test:
Monitor status with HWMonitor (for temps) CPUz (for CPU clocks), GPUz (for GPU load) and task manager (for CPU load)
1.) Fire up Prime-95 and start a worker on each core.
2.) Fire up heaven benchmark to load the GPU and the CPU at the same time.
Result:
The CPU started out at 3.2Ghz as expected and crept up toward 60C. As soon as I started the heaven benchmark, I ran into trouble. CPU utilization dropped from 100% to ~72%, and CPU clocks dropped from 3.2Ghz to ~2.2Ghz.
Shit - I thought - my cooling must be insufficient (even though my HTPC is in a very cold room)
I was about to give up and go order a beefier cooler online, when I decided to perform one more test.
This time I ran just the heaven benchmark without any additional CPU loading to see what would happen.
Before the APU even got hot, it clocked the CPU portion down to 2.2Ghz again. So this can't be thermal throttling. Maybe it's TDP throttling?
So I guess my question is, do AMD APU's automatically turn down the frequency of the CPU portion whenever the GPU portion is heavily loaded? Is there no way to load them both fully at the same time?
Is this just how they are designed? Or is there something faulty in my setup.
Much obliged,
Matt
I am new to APU's and did some googling to figure out if my problem was normal, but haven't found much.
My HTPC uses a cheap "hold me over until Kaveri" Richland A4-4000. It's a dual core with stock clock at 3Ghz, turbo up to 3.2Ghz.
Last night I decided to test if my cooling was adequate. I wanted to keep my build cheap, so I used only components I had kicking around. I opted against the boxed heatsink as it looked WAY too wimpy. Instead I used an old bigger copper heatsink that came in the box for my Thuban Phenom II x6 1090T. I figured if this boxed cooler is designed for a 125W TDP CPU, it ought to be able to handle this 65W TDP CPU.
My test:
Monitor status with HWMonitor (for temps) CPUz (for CPU clocks), GPUz (for GPU load) and task manager (for CPU load)
1.) Fire up Prime-95 and start a worker on each core.
2.) Fire up heaven benchmark to load the GPU and the CPU at the same time.
Result:
The CPU started out at 3.2Ghz as expected and crept up toward 60C. As soon as I started the heaven benchmark, I ran into trouble. CPU utilization dropped from 100% to ~72%, and CPU clocks dropped from 3.2Ghz to ~2.2Ghz.
Shit - I thought - my cooling must be insufficient (even though my HTPC is in a very cold room)
I was about to give up and go order a beefier cooler online, when I decided to perform one more test.
This time I ran just the heaven benchmark without any additional CPU loading to see what would happen.
Before the APU even got hot, it clocked the CPU portion down to 2.2Ghz again. So this can't be thermal throttling. Maybe it's TDP throttling?
So I guess my question is, do AMD APU's automatically turn down the frequency of the CPU portion whenever the GPU portion is heavily loaded? Is there no way to load them both fully at the same time?
Is this just how they are designed? Or is there something faulty in my setup.
Much obliged,
Matt