(09-14-2017, 04:43 AM)zxcvbad Wrote: CPU bound situations only (GB4) aren't really good indicator of actual thermals, especially GB4 tries to avoid throttling by pushing short scenarios with pausing between them. Also keeping in mind Dolphin won't use CPU only. CPU + GPU scenario/workload more realistically triggers actual thermal throttling. I'm quite opposite thinking that A10 was indicative about Apple's direction and A11 is an expected and logical transition from A10 with it's hexa-core design and with increase of all the cores frequencies. Smaller node allowed Apple to squeeze more space for the extra cores (likely keeping Monsoon cores at the same size as the Hurricane) + higher peak frequencies for all of them. A8/A9 were limited by the node size hence clocked low winning by the huge IPC. I'm not too optimistic about A11
A9 was on 16nm, same as A10. Node size was not an issue then, the jump from 20 to 16 was huge and allowed them to completely surpass existing performance metrics. The A10 was a one off because it made the changes one would expect from a node jump (higher clocks, more dense) without a node jump. The A11 brings a node jump to the table and also brings the Apple GPU with a supposedly far lower TDP (which would mean real world use fares much better). The A10 was a mistake, but also not the first one, and it hasn't been indicative of direction before. Apple is very clearly doing the multi core bit hard, but I sincerely doubt single core perf will ever be downgraded from a previous release. The A9 is the baseline, already near enough to run dolphin. A11 will not be worse and will likely be better. Also where the hell did this 2800 GB stat come from? Seems oddly specific considering dolphin has really never had a specific perf target like that before, was always real world results first.