Originally Posted by Alatar
Too bad that the jaguar anand is comparing mullins to there runs at 1000MHz (or just a tad over) when all cores are stressed. So even if the CPU performance had gone up around 35% the chip would only be performing on par with a 1300-1400MHz quad jaguar. Not even close to a 2ghz jaguar. Meaning that a quad A15 at 2.3ghz would easily surpass it.
However what they're showing there is PCMark8 which will not give you an accurate representation of the CPU only performance increase. Other things are also included.
That and I'd much rather trust a 3dmark physics test (which stresses all the cores very nicely at once) when it comes to CPU power than kraken bench will will do some random small tasks and is run in a freaking browser. Hell, just take a look at the ULV i5 in that bench. We both know that in reality it isn't three and a half times as fast as a 1.5ghz Jaguar...
tl;dr bad bench to compare with. Stress all cores and jaguar comes out ~39% faster than A15 clock for clock. When the tweaked (no performance changes, just better power consumption) jaguar cores in mullins run at ~1400MHz when all of them are loaded and the A15s n K1 run at 2.3ghz you're left with ARM cores that are clocked 60% higher.
PCMark8's Home tests are lightweight benchmarks. http://www.futuremark.com/benchmarks/pcmark8
Stuff like web browsing, text viewing and editing, some light gaming, and "photo editing" (filters, color grading, cropping, very basic things). Most of the tasks probably won't load more than 1 or 2 cores - the exceptions being the video chat and encode test and the gaming test (I tested it)- in which case the A6-1450 can clock up to 1.4Ghz. In fact, looking at the benchmark results of the A6-5200 running at 2Ghz, I think the A6-1450 probably clocked in at 1.4Ghz for most of the benchmark. The A6-5200 clocks about 43% higher (2.0 vs 1.4) and performed about 39% better (1861 vs 1343).
Assuming Puma+ doesn't have architectural improvements, unlikely, it will probably be able to Turbo up to 2Ghz. Oh, and I really have no idea what you're talking about. Anand posted the scores for a 2Ghz Jaguar right there and the 4.5W mullins part virtually matches it. The A6-5200 is a 25W part that clocks in at 2Ghz flat.
Oh, and even in a fairly intensive, multi-threaded test like Cinebench R11.5 (mutli-threaded test) the A6-1450 can maintain 1.1GHz presumably across all cores.
Among the CPU exclusive tasks only Prime95 brought it down to 1Ghz. Furmark+Prime95 obviously brought it down even lower due to thermal throttling.
In any case I think it's fair to say for most real world tasks, even those that are multi-threaded, it wouldn't be unrealistic for the mullins equivalent of the A6-1450 to maintain at least a small boost due to Turbo.Edited by geoxile - 1/13/14 at 1:39pm