News:

Willkommen im Notebookcheck.com Forum! Hier können sie über alle unsere Artikel und allgemein über Notebook relevante Dinge disuktieren. Viel Spass!

Main Menu

How Intel can get back on top

Started by Redaktion, July 30, 2020, 13:42:11

Previous topic - Next topic

Redaktion

Rocket Lake could be Intel's comeback, its victory run and its redemption. All Intel has to do is win the core count game in the mainstream space.

https://www.notebookcheck.net/How-Intel-can-get-back-on-top.483471.0.html

Pixie

Sorry in advance for how long this comment is. While this can certainly work if everything goes in Intel's favor, most sources are currently (and have been for some time, including on this site) indicating a 10-15% IPC increase out of Zen 3's design, with the potential for higher clock speeds as well thanks to the refined 7nm process. To dip a little deeper into the rumor mill, Zen 3 is also poised to significantly and specifically focus on improving Integer performance and latency, which have been the Achilles heel of the Zen architecture (when compared to Intel, and for the sake of gaming performance) since the beginning, and will be moving to an 8 core per CCX design.

Frankly, 14nm is at its limit here - they managed to squeeze a bit more frequency out of tenth gen, but the power required for these clocks is out of control, making even the Vishera chips blush when going for all core overclocks. Even if (and I stress IF) Rocket Lake can hit the same clocks, there's no way it's exceeding them significantly - and that's to say it can hit the same clocks at all - which we aren't even sure of; this is a backport of an architecture intended for a different node, don't forget. This is also assuming Rocket Lake's IPC gains can overcome the bump Zen 3 will deliver over Zen 2, which is already superior to Comet Lake in the IPC department.

And now we get to production costs: after all, there is no such thing as bad chips, just bad prices. Intel is at a MASSIVE disadvantage here, thanks to the fact that they are stuck on an older node and are still monolithic. Sure, they could up the core count further (with absolutely massive dies, mind you) but their yields would suffer heavily for it. Intel's supply shortages are already both widespread and well-known, to the point which they've had to spin up their trusty 22nm fabs once more for chipsets and the like. Not only is TSMC's plain-old 7nm generally understood to have excellent yields at this point, but they're now going to be using a further refined version of this for Zen 3.

You're suggesting Intel should just drop the price per core significantly by moving huge core counts down the product stack, but there are a few issues with that too. For one, the aforementioned yield problems - will these chips even be available or will it effectively be a paper launch?

Then there's the issue with pricing / value for the consumer - sure, they can probably drop the price to a loss to maintain market share, but there's no reason to believe AMD can't roll up their sleeves and start cranking out more cores for less money, too - don't forget that AMD was able to do exactly that, offering more cores for less money with Zen 1, BEFORE they had a 1 and a half node advantage, AND when they were also massively in debt. They could also sucker-punch Intel with an "AF" style launch of dirt-cheap Zen 2 processors after the Zen 3 launch like they did with the "1600AF" - remember that people bought those in DROVES since the price made it so attractive that the disadvantages of the comparatively outdated Zen+ design mattered very little to most people. Sure, Nvidia's 3000 series will surely raise the bar for bottlenecks here, but to someone in the market for a sub $150 CPU, that won't really apply. If they raise the bar on the lower-end stuff too much, people won't have much reason to buy higher than that and the upper range chips won't sell well at all - many already consider the i5-10600K to be "too good" to justify the 10700K or 10900K, with some even going as far as calling the i7 part "pointless."

Next there's the issue of sustainability - sure, this would just be a stop-gap solution, but the danger with selling these chips at a loss is the issue of self-competition. Their new stuff MUST offer better performance and value than their old stuff, especially now that they've got competition again, or it won't sell. What happens when Intel finally pushes out a 10nm desktop line? If they push too hard into loss territory, then the die shrink may only barely push them back into profitability, and that's only if yields are good enough. Unless they plan on operating at a loss until at least 7nm is out, then they can't give too much away without making either the value proposition or margins on the following generations of parts suffer.

Finally, and this is a big point, would Intel even do something like this? They have shown a seriously arrogant position, demanding higher prices for comparatively less than their much smaller competitor. They effectively lost the mid-range market and below to AMD during the 9th gen, while their whole HEDT platform was rendered effectively irrelevant by AMD's MAINSTREAM offering, and they seemed unwilling to adjust until the release of Comet Lake. The 10600K was a beautiful course-correction, but it made the aforementioned irrelevancy of the i7 an issue and it still offers comparatively worse value than a 3600, especially with current pricing. That's to say nothing of Z490 platform costs, which end up translating to $200 or more all-in on a comparable, complete system between the motherboard, CPU, and cooling angles. With this track record of slow adjustments and unwillingness to bow their pricing to external forces, I just can't see Intel even considering this plan unless their management performed a complete 180. It seems more like they are willing to shed market share until they can come back swinging with a mature 10 or 7nm product.

I want to see Intel deliver good products at good prices. With AMD back in the ring, if Intel can get their node issues sorted out, we stand to get some truly amazing competition in the coming years, just like the good old days of the 2000s. That's even ignoring the enormous potential of having a third graphics supplier in the industry for the first time in decades. However, things aren't looking too good for the next few years at this rate.

LL

Interesting, i am not knowledgable enough on the subject to critic, but would appreciate that the author w ill revisit this article here when the dice stops.

Stas

Hi, Pixie.

Thanks for such a detailed and interesting take. I wholeheartedly agree with most of what you said, and I would also like to see Intel competing on its 10-nm and 7-nm node.

However, I do believe that Intel can only win this one (barring any advances in manufacturing technology) if it starts delivering unparalleled value. This is why I said that there should not even be a Core i9 SKU in the 11th generation unless it provides value beyond gaming and can compete head-to-head with the Ryzen 9 series in production applications, because if Intel releases a 140-dollar "Core i5-10600K" with an up-to 20% boost in IPC performance, then there would be no point in getting a Core i9 even with a 3000-series RTX graphics card. Like many other PC enthusiasts/gamers, I am still rocking a quad-core Intel chip, but the advent of the next-generation consoles is forcing me to think about upgrading my quad-core chip, like so many other PC gamers. I think Intel can win really big in the 11th generation if it provides PC gamers with killer gaming chips at low prices and if it manages to avoid CPU storages.

Quick Reply

Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.

Name:
Email:
Verification:
Please leave this box empty:

Shortcuts: ALT+S post or ALT+P preview