Changing subjects slightly, AMD's latest Ryzen announcements seem to put it firmly in the 'Not a disappointment' category. In eight days we'll have independently run benchmarks, so will be able to see Dolphin performance from Anandtech.
I have no trust on AMD marketing anymore. I'll only believe they improved something when I see said independently ran benchmarks...
My A10 6790k @ 5.0GHz (Pile driver core ) is as fast as Haswell @ 3.2GHz in Dolphin Benchmark
If 40% IPC improvement is true then 3.6GHz Ryzen will be as fast as my OCed A10 . Not really amazing but at least it will be much better in multi-tasking
And it should handle most Wii/GC games well
Maybe I should wait a few years for a even better AMD CPU so that I don't have to worry about the performance on the other demanding emulators like Cemu , RPCS3
They're claiming it's actually 52% now, not 40%, so (if the scaling is the same for Dolphin), that performance point is well in reach of parts at stock speed.
(02-23-2017, 03:05 AM)admin89 Wrote: [ -> ]My A10 6790k @ 5.0GHz (Pile driver core ) is as fast as Haswell @ 3.2GHz in Dolphin Benchmark
That's the point, after K10 (which at least were very competitive with their Intel counterparts), every new architecture AMD released after were garbage, your A10 already have a very high TDP and still have to run at a high overclock to match the performance of an equivalent Haswell CPU running at stock speeds with a lower TDP, for example, and I remember AMD saying otherwise on their Piledriver announcement, the same they also did on Bulldozer and Steamroller announcements. With that said, I'm giving them *zero* credibility on their Ryzen claims (no matter how good or promising they look) until I see proper benchmarks...
Still , 8 cores 16 threads is something you can not miss . My Core i7 is low end CPU now , damn it
I hope Intel will soon realize Dual Core is dead . Quad Core should be mainstream and 6 or 8 cores should be more affordable ...
Its not that simple. You don't just stick more cores in and it works. core interconnect gets more complex. takes space. cache connect pathing. I/O pathing, all gets much more complex and expensive. Bulldozer, Vishera, etc got away with it because their chips weren't remotely as complex as Intel's.
A lot of people ask "just get rid of the iGPU and put in more cores"
that's not the way hardware design works you freaking lunatic sdghjkghjsdftyibvghj
That said, I am very interested how AMD is supposedly managing to meet haswell perf levels with 8 physical cores on a process that small, if they are. And if they are, how much of a loss they're selling at.
What we're seeing so far isn't entirely unbelievable, assuming the "leaked" benchmarks are accurate. They are going to be using a 14nm process for ryzen so for the first time in 10 years Intel won't have a manufacturing process edge boosting their relative efficiency. So it really just comes down to design then. Ryzen has no IGP (which affords them more die space for the cores) and uses 8 smaller/simpler cores at lower clock rates rather than 4 bigger/complex cores at high clock rates. This yields better multithreaded performance but less singlethreaded performance. And that's what we see in the benchmarks. Singlethreaded performance is still a good 30% higher on Intels platform (and likely higher for dolphin) but multithreaded performance is 50% higher on Ryzen. Meaning a 6 core Ryzen will likely be comparable to a 4 core kaby lake and a 8 core ryzen will be comparable to a 6 core broadwell-E in multithreaded tests. Quite frankly Rzyen still seems overpriced, they are only price competitive with Intel for multithreaded workloads. But AMD is known for overpricing CPUs at launch.
Also to add to that synthetic benchmarks do not always equate to real world performance. Bulldozer actually did well on a lot of synthetic tests.
Edit: I take back what I said about AMD having caught up to Intel in fab tech. Apparently they are still using FinFET for their 14nm process. Whereas Intel is using Trigate. 14nm FinFET is equivalent to 22nm Trigate in performance. Intel is working on GAA FET for 10nm/7nm while AMD is working on Trigate for 10nm. So AMD is still a generation behind Intel in fab tech.