What we are talking about is Zen's pricing. I am with Shintai on that one. My personal opinion is that Zen will not be the magical performer that some in these forums are predicting. Some of the predictions are quite.......optimistic to say the least.
However, even more absurd is the continuing contention that if AMD does pull off this 5960x performance in a 95 watt envelope that it will somehow be priced well below x99. Just look at their recent pricing history for Fury X, the 9590, and even the latest APUs, every new top end model of which is being released at close to locked i5 prices. Even Su herself has said AMD no longer wants to be the bargain alternative. AMD is currently the "bargain alternative" for a reason...lack of performance and high power usage. I am sure initially Zen will be in short supply, and if it is the magical performer that some predict, they will sell every one they can. Not even AMD marketing is inept enough to sell a top end performing part, supply constrained, at a relative discount.
The following *assume* first, then predict. Zen average IPC exclude AVX is roughly Haswell~Broadwell. IPC at only Sandy Bridge level will not applied. (Assumption)
1) Intel will keep the performance crown at the top. AMD is the price setter for the $150-$600 market (simply due to Intel being unwilling to kill their profit margin and play the #of core games except for the extreme $1000 segment.) Too much cores in regular desktop -> not good for Intel server SKUs.
2) Fury X priced at $650 is near the high end GPU segment but not 1% $1000 segment. It's not too expensive from historical point of view. Initial price is due to HBM capacity constrained. How 9590 FX is priced is not too relevant because there's really not much taker anyway. ()
It's their $1000 product that nobody is interested.
3) When Oil price was high, Global Foundry can forgive AMD more often and not enforce WSA all the time, and obviously enforcing it at that time would not be wise as they have a stake at AMD and know when to strike the WSA to benefit both.
4) Given WSA agreement and the backing of Samsung fab (assumption) if GF is too capacity constrained, then capacity isn't a problem but how high it can clock at what wattage and the average IPC.
5) 8core 5960x at 3GHz 22nm can easily be overclocked to 4.5 using high-end air or watercooling. Since Zen ES is reported to be running at 3GHz, let just assume they are 140W as well. They can label it 95W or 140W all they want, I doubt the majority will care other than the max oc level at acceptable power consumption on a normal workload (not AVX virus).
6) Zen without APU die space is a 1xx mm max 200mm2 die. Between a Polaris 11-10 die size. The Polaris 10 is expected to be mid range SKU, selling for $400 Max? The $400 include all the extra cost excluding the GPU, which AMD sell the GPU for peanuts vs the retail price at $400/CPU?
7) Since Zen is a priority over anything else, and Polaris 10 die size is most likely bigger than Zen, should AMD fab a Polaris 10 over a Zen CPU? -> capacity constrained? ()
8) Finally pricing, just 1xx mm2 chip, let's blow the WSA wafer limit.
$1000 Not sure what's being offer, Gold binned chip, maybe watercooler bundled.
$650 2nd tier, just chip
$350-400 - 3rd tier - *mainstream high end* i7 competitor. 8C/16T, cache cutdown, default speed/W irrelevant to me and many others, other than attainable max oc speed <160W. That's when most aircooling start having trouble.
$200-$300 - 4th tier - *mainstream oc tier* i5 competitor K competitor. 6C/12T, cache cut.
$100-$200 - this is harder to predict, most damaging segment. I will avoid. :sneaky: