I just wanted to share the results of my fairly small A8-6410 APU performance analysis. For those who don't know this is a Puma+ based quad core APU. This is the successor to Jaguar which are the 8 core cpu's in the xbone and ps4 as well as the kabini desktop quad core variant (and some other laptop apu's).. This APU is running at a boost of 2.4ghz and a base of 2.0ghz. The gpu core runs at 800mhz. I am testing between 1600mhz CAS11 and 1866mhz CAS10. The results were a bit better than expected on some, and fell flat on others. I will be updating this as I get more results but as there is very little out there on how these APU's actually perform I'd figure I'd do everyone a favor and show these are fairly decent, very affordable 15w tdp APU's.
The main testing I did can be seen in the below chart, it is for the in game benchmark of the source engine.
Edit: graph is blurry for whatever reason.. see below:
DDR3L 1600 CAS11
LOW SETTINGS - 640x480 NO AA 144.5
MAX SETTINGS - 1368x768 NO AA 75
MAX SETTINGS - 1368x768 2X MSAA 57.33
MAX SETTINGS - 1368x768 4X MSAA 55.33
MAX SETTINGS - 1368x768 8X MSAA 53.33
DDR3L 1866 CAS10
LOW SETTINGS - 640x480 NO AA 149.75
MAX SETTINGS - 1368x768 NO AA 85
MAX SETTINGS - 1368x768 2X MSAA 66
MAX SETTINGS - 1368x768 4X MSAA 64.33
MAX SETTINGS - 1368x768 8X MSAA 62
As you can see, the bandwidth does play a role almost netting at 10fps gains at best and 9fps on average when the stress is more on the GPU portion. This was more than I expected, but being that this architecture is unfortunately limited to single channel the extra memory bandwidth really helped here.
On a side note, it is amazing to me that the scaling going from from 2x > 4x > 8x MSAA was a 2fps drop each time. I reran all of the tests multiple times to confirm this but it averaged that each time. No AA is clearly still best with these APU's, though.
Unfortunately, when I was doing the more cpu bound portion I was disappointed in the minor 5fps gain here. Again I reran this multiple times to be sure but it averaged around 5-6fps gained each time. I was thinking this cpu to be more bandwidth bound than it is, but with such high frame rates already it made almost no proper gains in that kind of scenario. At least for the source engine.
Though it's not in a graph form yet, please also see the below
DDR3L 1600MHZ CAS11 6GB (4GB & 2GB)
Score Graphics Physics Combined1754 1786 1578 1844
DDR3L 1866MHZ 8GB CAS9 (8GB)
Score Graphics Physics Combined2035 2036 1977 2123
This was at the default Skydiver 1.0 settings set to 1366x768
I also did a Firestrike Physics test at the lowest possible settings and 640x480 and the scores were near identical so no gains there. Scores were both around 2950
Anyways, I will be adding more tests to this but I wanted to share it with you all, especially those considering an AMD laptop. I paid under $300 for mine new at bestbuy so I consider it a great value. The battery life is great as well. Lenovo G50-45 (see specs in sig).
Please let me know if you can think of other benchmarks that would be more useful! Thanks!