AMD CES 2014: Kaveri APU Is Imminent

Article Index:   

APU Performance

All the usual caveats apply about vendor-supplied benchmarks; we'll confirm these results when we have the chip in-house shortly.



PCMark 8 shows the A10-7850K beating past last year's Richland APU by 9%, even though the Richland core has a faster Turbo Mode of 4.4 GHz compared to 4 GHz for the new APU. AMD claims both cores are faster than the Intel Core i5-4670K, which is surprising given that Intel cores have often had an advantage over AMD chips in recent years.

The GPU performance from the new APU should be excellent, and AMD's middle benchmark implies that this is so. The new Kaveri chip is nearly twice as fast as Intel in 3DMark and 50% faster than AMD. Similarly, the OpenCL test shows Kaveri outperforming Richland by over 50%.

We'll have a full suite of tests up in a matter of days, so stay tuned for the full review. For an example of just how fast Kaveri can be, check this next slide. It shows a workload accelerated for HSA, and run through the standard CPU.



Clearly in the right markets, HSA can make a huge difference. AMD is already shipping the chip to system builders and computers should be available on preorders. Features like Mantle will also be supported on the new APU, so customers can look forward to an additional graphics performance increase thanks to the new API. Rumored price on the new chip is $173, so AMD is positioning this as a more expensive option than the older A10-6800K chip was.


As with previous cores, the K-class chips are fully unlocked and can be overclocked. Both CPU and GPU can be pushed independently. Kaveri's performance is stronger than Richland's overall, even though the chip runs at 3.7 GHz base / 4.0 GHz Turbo, compared to 4.1GHz / 4.4GHz for the Richland APU.

If the new Kaveri can actually hit the 6800K's clock speeds, the performance difference should be even higher. We'll have more details to share soon.
Tags:  AMD, CES, APU, Kaveri, huma, CES2014

Image gallery

Related content

Comments

blog comments powered by Disqus