Originally Posted by Kuivamaa
What if they have been at A0 for months already and they are about to jump to the next revision? Random leak is random.
This is likely, AMD would finish up with their testing of A0 silicon, know most of the bigger bugs (at least) and have A1 silicon either in their hands or on their way to AMD at which point OEMs that need to design the related subsystems get A0 to fully test said subsystems (eg. Motherboards) which is usually where we get leaks. There's also that we have better ways of testing silicon pre-production (even for ES') every single release, meaning less ES' are needed because they can test them in software before they actually send designs off to the foundry to be made. Even a minor update like their Ridge CPUs (Which launched on A1 and A0 silicon) would have likely launched on A2 silicon even just a few years back.
My personal bet is Zen is either launched on A2 or B0 silicon and that AMD is likely waiting for those designs to come back now, with A1 chips in testing at their facilities and ASUS, Gigabyte, MSI, Shuttle, Foxconn, etc having A0/maybe some A1 silicon to help design motherboards/BIOS' around which is usually where leaks occur. (Mobo makers get it and employees there leak it as it's much harder to trace it to an individual than it is when AMD/Intel are the only ones with the ES silicon, that's also why we tend to only get leaks of new GPUs/their coolers once the OEMs that actually sell the cards get them)
Man, I really can't wait to see Anands article on Zen...It's going to be fascinating regardless of how Zen ends up performing.
Originally Posted by The Stilt
That's actually incorrect. Bulldozer 8-core A1 stepping samples (prototype) ran up to 3.6GHz (ZD262046W8K43) and B0 stepping samples up to 4.1GHz (ZD312051W8K44). Stepping B2F/G was the release stepping and it only increased the clocks by another 100MHz. First Vishera samples (ZD338251W8K54) operated up to 3.9GHz and the stepping didn't change between the sampling and release point.
If AMD plans to ship any Zeppelins in 2016, they must have the final silicon at hand by now. Infact the validation must be nearly completed.
iirc BDs A0 was around 3Ghz but we didn't ever see much of its A0 silicon with most leaks occuring from the A1 silicon. Feel free to correct me though, I'm not 100% if I'm mixing it up with other leaks from the time. Q4 2016 is October to December, if they mainly paper launch (ie. Ramp up production very shortly before launch and keep most of the chips made prior for OEMs and reviewers) in December we'd still get a Q4 "launch" even if most of us can't actually get chips until mid to late January. Not ideal, but it's what I can see likely happening.
Originally Posted by cssorkinman
What is out there that can make use of DX 12 besides Timespy?
Um, Hitman? RIse of the Tomb Raider? Can we just please start ignoring 3DMark? I remember back when 2006 was still the big comparison, it didn't really show the differences between different GPUs that well and nothing has changed since.
Civ VI apparently will have support and will be great to show CPU bottlenecks because Civ games are traditionally CPU bottlenecked, having Mantle support in BE was a massive boon to turn times and probably the best feature of the game. (If only BE wasn't half the game V is)
Originally Posted by Kuivamaa
Link? I see from time to time core scaling gaming tests popping by and they always leave much to be desired. Eg. once they find that from a certain CPU core count and up there are no more FPS produced , they declare it as the end of scaling without checking whether they are actually maxing out the GPU used at that core count.
There's also the fact that DX12/Vulkan make it much more up to the developer, so even if one DX12 game is limited to 6 threads that might not be true of others.