Single thread perf of AMD A10-7890k and other APUs...

[a7751218]

Hi,

Firstly, thanks for your charts and performance stats!

Can you please double-check the single-threaded stats for the A10-7890k APU (currently: 1720, based on "171 samples").
Verses for example the next model down, the A10-7870k, which is currently: 1571, based on "506 samples"...

I'm finding it difficult to believe that a 200mhz (5%) increase of the A10-7890k (base and turbo clock) verses the A10-7870k would result in 10% increase in single-threaded performance, unless the architecture was different?

Similarly, another 200mhz lower in base and 100mhz lower in turbo of the A10-7870k and we have the A10-7850k, which also scores around 1572 (same as A10-7870k)!?

I would have though a slight increase in clock speed of the same architecture would result in very minor gains (as shown in various benchmarks of the Kaveri architecture), and not a ~150 point increase like with the A10-7890k?

Can you please also double-check the single-threaded stats for the A6-6400k in which my bog-standard setup scored above 1500 points in single-threaded bench using PT8, while your table is showing it as averaging ~1400 points?


Just to breakdown the single threaded scores as per your chart...
A10-7890k = 1720
A10-7870k = 1571
A10-7850k = 1572
A6-6400k = 1403 | mine = over 1500


As you can see, I find the scores for 7890k a little difficult to believe.

Thanks for your help!

[David (PassMark)]

These results are averages from the real world. So in that sense they are accurate.
So the interesting question is why doesn't the real world align with the specifications.

There are many possible reasons for the differences. For example the A10-7890K was shipped much later than the other CPU models. It also shipped with a much better cooling system. (bigger fan and heat sink). As a result overclocking would also be better and thermal throttling less. Because it was released years later, people likely also matched it up with slightly faster RAM. Could also be that for the newer 7890K CPU, that more people were using 64bit operating systems (which give better benchmark results).

For the 7870K and 7850K being unexpectedly close together in performance. It seems the small 100Mhz difference in clock speed isn't enough to overcome other factors. Maybe one of the major manufacturers (HP, Dell, etc..) released a system that used one of these chips in a sub-par configuration (for example with single channel RAM). Enough to move the average by a couple of percentage points. You would need to do some deep research to identify the real cause, and even then it might be hard when looking at such small differences.

PerformanceTest V8 is an old release. You should be using V9 now.