According to photos published on the Japanese hardware site Hermitage Akihabara, AMD’s Kaveri has a fairly weak CPU in terms of performance relative to a mobile Core i5 processor from Intel in Cinebench CPU raytracing benchmark. Admittedly, this benchmark has favored Intel’s CPUs for quite some time, but Kaveri is supposed to be AMD’s latest and greatest and the preliminary performance figures don’t look too good. They also published some preliminary CPU-Z specs which appear to conflict with the clockspeed figures provided by Cinebench. This could be due to the fact that CPU-Z hasn’t been updated for Kaveri use quite yet, or that their information simply isn’t consistent. Obviously, we won’t know the answer to either of those questions until we get Kaveri in-hand and test it ourselves, which we suspect will be very soon as the frequency and intensity of leaks increases.


Here, we see that CPU-Z identifies the chip as an AMD A10-7850K APU using the socket FM2+ socket and the 28nm manufacturing process, which is in line with the expected Kaveri code named APU. What’s interesting is the 0.912v core voltage, although this appears to be a lowered Pstate because the CPU is only running at 850 MHz which is the lowest multiplier that this CPU appears to operate at. It also shows support for AVX, but no AVX2 support. Unfortunately for us, there is no clear indication of what the L3 cache will be, but is interesting to see that AMD is marketing the GPU as R7 graphics.

Unfortunate for us, GPU-Z appears not to have been updated whatsoever for the Kaveri APU’s GPU, so we basically get zero additional information here.

In Cinebench, we see that the CPU is running at 3.7 GHz and is being tested on the single core Cinebench test. In Cinebench R15 the CPU is recognized as a dual core processor with four threads, which varies from CPU-Z. While I do believe that this chip is running at 3.7 GHz, I suspect that the single core ‘boost’ performance is likely greater than 3.7 GHz on this chip. Once again, we clearly can’t confirm this until we test for ourselves, but the fact that it underperforms when compared to an Intel Core i5 3317U CPU is a pretty sad state of affairs. The 3317U is a low power mobile processor, not a desktop APU like the Kaveri, so there shouldn’t even be a comparison.

In the end, however, the most important thing will be overall user experience with Kaveri and whether or not AMD can deliver an APU that delivers good performance from both a CPU, GPU and compute standpoint. If they can deliver a better experience with HSA features, then it really doesn’t matter if the CPU is weaker than Intel’s if Intel’s GPUs can’t do what AMD’s do. Hopefully we’ll have the answers to these questions soon and will do our best to assist in clarifying any uncertainties caused by these leaks.