AMD provides a first glance at Zen Summit Ridge performance

[badass]

Don't automatically Assume that Final chips will be clocked higher than engineering samples:

http://wccftech.com/amd-fx8150-engin...mple-detailed/

Some FX-8150 ES CPU's had identical clocks to the finished product.

However, here's hoping for another 1 GHz

[watercooled]

AMD have said clock speed of shipping CPUs will be higher than the ES: http://www.anandtech.com/show/10585/...than-broadwell

We were told explicitly that these are not the final clock speeds, but it at the very least it puts the lower bound on the highest end processor.

Of course it could be 3.1GHz and still be true, but it's at least something. Even if they can get it to 3.5GHz all-core clocks that's pretty good going for an 8 core. Don't forget the 6900k it was compared to is only 3.2GHz base clock.

[Bagnaj97]

plus (based on ~ 80% of RX 460 performance) a GTX 750 Ti.

~750Ti in terms of core speed, but if it's still using system ram then a GDDR5 750Ti will still be well ahead in terms of memory bandwidth. Unless AMD see fit to add a stack of HBM(2), which would be awesome.

[watercooled]

At least they'd be on DDR4, and hopefully dual-channel.

[scaryjim]

~750Ti in terms of core speed, but if it's still using system ram then a GDDR5 750Ti will still be well ahead in terms of memory bandwidth. ....

At least they'd be on DDR4, and hopefully dual-channel.

I think a minimum of dual channel is a given, tbh. And whilst the peak theoretical physical bandwidth will be a lot lower than a GTX 750 Ti, the improvements in DCC over the last couple of generations will make up some of that (afaik the 750 Ti doesn't support DCC at all) - Polaris reckons to gain up to around 40% effective bandwidth where DCC kicks in. If that can be carried through to the Zen APUs (or even improved on), dual channel DDR4-3200 could give it up to 85% of the effective bandwidth of GTX 750 Ti....

[CAT-THE-FIFTH]

Is the quad core APU a 4 or 8 thread chip?

Dresdenboy made some estimates for the core size:

http://dresdenboy.blogspot.co.uk/

[watercooled]

I think a minimum of dual channel is a given, tbh. And whilst the peak theoretical physical bandwidth will be a lot lower than a GTX 750 Ti, the improvements in DCC over the last couple of generations will make up some of that (afaik the 750 Ti doesn't support DCC at all) - Polaris reckons to gain up to around 40% effective bandwidth where DCC kicks in. If that can be carried through to the Zen APUs (or even improved on), dual channel DDR4-3200 could give it up to 85% of the effective bandwidth of GTX 750 Ti....

What I meant by that was manufacturers hopefully using dual-channel as opposed to them crippling it with just one RAM stick like happens with a lot of current shipping systems.

Dresdenboy made some estimates for the core size:

http://dresdenboy.blogspot.co.uk/

That seems a bit small IMO - inclusive of L2, Broadwell is around 7mm2 and Skylake around 8mm2 (based on my own pixel-counting of die shots). Zen has twice the L2 cache and uses a less-dense lithography. Skylake may dedicate more area to AVX but I wouldn't think the core size would be so drastically different. I'd guess more like 8mm just based on that, maybe more.