Every time a new product or technology shows up, it's like people forget that time doesn't stop. In other words, should we assume people don't change or change slowly? What's it based on?
My thought is that consumer usage of CPU cores has been limited to what is available to their budgets, and to net the most sales, developers won't go out of their way and waste their resources to tailor software to as many cores as you can throw at it, should you get more cores later. Rather, they will aim for a best fit group; this is why games don't improve with more cores, they weren't written open-threaded from the beginning. Hardware isn't at fault, software is; but Intel is placing responsibility on user habits. Except the end user wouldn't know the hardware upper limits of their software, devs would, that's why they give out minimums and recommended marks.
But like any business, developers don't spend millions and then 'hope' it sells or take a 'let's see what happens' approach; they all already know it will sell, growth is the unknown factor (i.e. will new customers continue to purchase over time before a competitor jumps in or responds?).
A business will afford those expensive market research reports and look for trends of what hardware typically gets sold and aim software at people most likely to by it. It does seem catch 22, and granted, a business has to start the ball rolling for hardware and software-- one could conclude with regards to AMD and Intel, this is where we are right now. Most major OEMs still don't carry too many Ryzen lines, in a way, it isn't worth the marketing response from Intel-- unless they already knows their customers (OEM) are looking into it and trying to deter by remind them what software end-users use. Although, playing Devil's Advocate for Intel, 'desktop' doesn't automatically mean enthusiast or gaming grade, most people don't do much and tend to do the same things on their notebook or tablet.
From 2006 to 2017, it was the quad-core w/HT that was the gold standard at the $300-ish price point, dual-cores w/HT still existed below $200 and 6-core started out at $1000, dropped to about half in that time, succeeded by 8-core models. What did most regular folks get? 4-threaded processors were dominant in sales, dual-core w/HT and quad-core w/o HT. That is what Intel's Real World is based on; it isn't a coincidence or a 'supposed to be' ideology. It is just a result of what folks afford; it takes a few years to develop new software from the ground up, so a new hardware trend has to last a while before we start seeing programs that use a different hardware set. But this trend of new hardware adoption takes time by itself, it isn't like enthusiasts keep the market afloat, major software players have to wait for everyone else to change. That isn't going to happen until the cheapest models begin getting more cores, and we're not there yet. Ryzen 3 established and Core i3 followed making quad-core w/HT entry level, but they need to go up for real change to occur across the board.
With Zen 1.0, AMD knew their IPC was behind Skylake and maybe frequency wasn't as high as it could be, so the only strategy left is increase core count per dollar-- until it suits them, they are a business after all. After a while, even they will come off as greedy waiting for Intel to respond, that's just capitalism. These companies don't make their products for the customers, they do it to keep their customers away from competitors; if a competitor is unable, there is no reason to continually introduce new stuff at the same price points (might as well raise price points). That's why 3900X became 1800X's successor without doubling the core count, and 3950X's debut was postponed; and why nVidia introduced RTX at all.
People do change, and they change slowly, but that is provided an overhead exists to grow into. Sure, there are more consumers than creators, but will it stay that way? Yes and no, plenty of humans on the planet and people get educated somehow, got to love the internet. Intel's argument isn't against creators, they just figured those willing to pay for their premium models are few, as if not affording keeps them from creating as much. But they aren't considering those that want to be creators but can't afford-- this is of benefit to AMD. Frankly, I'd put myself in this category and prefer Intel pull their head out of their rear.
I get that Intel is just trying to easiest way keep their business with these tactics, they credit AMD for closing the gap and giving them the excuse for more SKUs with better core per dollar than before; but outside that, I'm sure they prefer we all ignore AMD on the basis that we don't need more. Intel's response to lead by raising core counts on their own, ahead of AMD, is delayed; in the meantime, criticize someone that can do something you can't. It is a very human response at it's core, pun intended...
What's at the $300-ish mark now? 6- & 8-core, who's okay of another decade of that?
Subject change: Let's take a look at Intel's twelve years of quad-cores from Core 2 Quad to Core i7-7700K at the $300-ish mark. Both of these were from Anandtech, reviewing
Kaby Lake and
Sandy Bridge, and Cinebench was present when it favored Intel and I'm sure back then no one really used it:
I like that frequencies were included in the SB review, when normalized, i7-2600K (boosts to 3.8GHz) was a 20% faster IPC than i7-920 (boosts to 2.93GHz) and 36% ahead of Core 2 Quad Q6600. But while 7700K had a boost of 4.5GHz, it was just 26.8% IPC ahead of 2600K. But hey, this is R10 and Ryzen didn't exist.
We could look at the IPC of
Coffee Lake and
Zen 2, scroll to the bottom. No IPC gain from 7700K to 9900K is sad, but fans don't care, they are superficially focused on the frequency. Speaking of which, it's too bad Zen 2 is so temperature sensitive on overclocking that far, considering most enthusiasts limit themselves to AIO's; but I don't see frequency going up with later versions of Zen. I think AMD's focus in increasing core counts, leaving frequency increases to Intel, which will benefit their legacy of most people not needing as many cores.
The last time AMD posed a series threat, Intel came at them hard, it'll be a while before this happens again:
These were all dual-cores, debuted in 2006, so dual-core had already been established as premium; of course these from
Anandtech's Core 2 Duo launch review are the single-threaded scores. I actually owned a D-920 at the time, blew my mind that a lower frequency could have a faster processor, that's the power of IPC, nearly 100% per core (Core 2 Duo E6300 versus Pentium Extreme 965).
post edited by lehpron - 2019/09/08 09:56:56