• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Another upcoming 14nm++++++ Desperate CPU from Intel

The biggest thing a 10 core chip on 14nm would signal imo is that they aren't confident of getting desktop 10nm higher core/performance parts in 2019. It's already leaning more towards thinking end of 2019 will bring probably quad core mobile parts and maybe if Intel wants to still have dual cores, a dual core part.

It's worth noting though that a 10 core cpu with igpu would be even worse again for Intel with capacity issues in terms of further reducing their capacity, but if they finally brought back pure cpus to mainstream then 10 core cpu would be smaller than a 8 core + igpu die. It would even be better financially if they managed to reduce die size but stuck with similar pricing as to the 9900k. Probably 99% of people who buy the two two SKUs on desktop don't use the igpu so a 10 core cpu only chip would make a lot of sense for them but obviously for salvage reasons you'd probably end up with at a guess, 10, 8 and 6 core chips without iGPU

Still though there is the issue on time frame, based on when the 9900k came out, I can't see them launching a 10 core before well, end of H2 2019 at the earliest really, maybe even Q3 and for me that puts 10nm replacements well into Q2 or Q3 2020.
 
Intel roadmap until 2021 leaked:
https://twitter.com/witeken/status/1121052220072583174

D47FpkkWwAEmef-.jpg


10C desktop consumer CPUs until early 2021.
 
drunkenmaster said:
The biggest thing a 10 core chip on 14nm would signal imo is that they aren't confident of getting desktop 10nm higher core/performance parts in 2019. It's already leaning more towards thinking end of 2019 will bring probably quad core mobile parts and maybe if Intel wants to still have dual cores, a dual core part.

It's worth noting though that a 10 core cpu with igpu would be even worse again for Intel with capacity issues in terms of further reducing their capacity, but if they finally brought back pure cpus to mainstream then 10 core cpu would be smaller than a 8 core + igpu die. It would even be better financially if they managed to reduce die size but stuck with similar pricing as to the 9900k. Probably 99% of people who buy the two two SKUs on desktop don't use the igpu so a 10 core cpu only chip would make a lot of sense for them but obviously for salvage reasons you'd probably end up with at a guess, 10, 8 and 6 core chips without iGPU

Still though there is the issue on time frame, based on when the 9900k came out, I can't see them launching a 10 core before well, end of H2 2019 at the earliest really, maybe even Q3 and for me that puts 10nm replacements well into Q2 or Q3 2020.
Click to expand...
It's been common knowledge for a while that Intel's 10nm node won't perform as well as their 14nm node in its first iteration, at least. This basically means it won't clock as high, since Intel has seemingly abandoned IPC increases (presumably pending a brand new architecture at some point). This is part of the reason that I expect no surprises from Intel any time soon: even if 10nm happens in 2019, it won't be used for high end desktop parts. 10 core 14nm+++ chips are far more likely.

CAT-THE-FIFTH said:
Intel roadmap until 2021 leaked:
https://twitter.com/witeken/status/1121052220072583174

D47FpkkWwAEmef-.jpg


10C desktop consumer CPUs until early 2021.
Click to expand...
Ugh...so does that mean no 10nm chips until 2021 and no 10nm mainstream desktop parts until at least 2022??
 
I'm not sure what your point is. AMD are bringing 16 core this year, Intel might lose some clock speed but going to 10nm would bring the actual size of their dies down massively. This is important in every segment.

The actual highest clock speed chips are very low volume and pretty much focus on the niche highest end gaming/enthusiast segment. Any production/creation type stuff prefers cores over clock speeds in general and almost all OEMs want smaller, quieter boxes.

A 9400 with a 4.1Ghz turbo speed will sell 10 times as many chips as a 9900k with a 5Ghz turbo speed. 10nm would address 99% of their products with vastly superior products with performance/w, performance/$ metrics that blow 14nm chips out of the water entirely.

99% of the market would take a 95W 14-16 10nm chip with 4.5Ghz turbo over a 200W 10 core 14nm chip with a 5.1Ghz turbo. The entire server market would take 56 core chips at 200W over 28 core chips at 200W in a heartbeat and they are pretty much laughing at Intel's attempt to stick two chips together as 350W watercooled only downclocked what is it 48 core chips were they, 52 cores, can't recall.

The clock speed matters to an exceptionally small amount of the market and most OEM sales would prefer lower power higher core, better value chips that actually represent real upgrades as that pushes sales. Oh, $500 for an extra 2 cores... probably stick with what I have... $400 for double the cores... okay I'm in.

If 10nm was working Intel would launch desktop on it, absolutely zero question, vastly improved chips for 99% of the market and almost all their profit, slightly slower single thread in one pretty small segment, it's a complete no brainer.

However these recent lists are even casting a doubt that Intel can do mobile in 2020, let alone 2019.
I'm somewhat expecting another risk production run of whatever change they made to 10nm, launch another mobile model with higher power and slightly higher clocks, still be a joke and still only turn up in 2 random products no one really wants, shut down production and again try to fix the fundamental problems.
 
Double07 said:
Unless you are the internet where games "matter" and so CS >>>>> CC.
Click to expand...

Only true historically, as developers embrace more cores this is becoming less and less true...

Will be interesting to see if the price of them is still stupid given their position, death by arrogance.
 
Don't get me wrong, with the NUMA issues on the 2990WX and Windows scheduler being rubbish, the 28 core Skylake X might actually trump Threadripper in performance, but a 9900K with another 2 furnaces lashed onto it is supposed to counter 12 and 16 core Ryzen 3000? Did Intel not watch CES?
 
LePhuronn said:
Don't get me wrong, with the NUMA issues on the 2990WX and Windows scheduler being rubbish, the 28 core Skylake X might actually trump Threadripper in performance, but a 9900K with another 2 furnaces lashed onto it is supposed to counter 12 and 16 core Ryzen 3000? Did Intel not watch CES?
Click to expand...

Yeah NUMA is an issue on Windows, they scale perfectly well on Linux.

Anyway. I suppose its all they can do because the Ring Bus only supports up to 10 cores, beyond that they have go to their mesh thing like Skylake-X and clock for clock those things are significantly slower in games.

This is the problem Intel have vs AMD, AMD have the technology to scale beyond what Intel can do, beyond economies of scale that hasn't helped them in the mainstream but now with Ryzen 3000 AMD are pushing to where Intel can't go.
 
LePhuronn said:
Don't get me wrong, with the NUMA issues on the 2990WX and Windows scheduler being rubbish, the 28 core Skylake X might actually trump Threadripper in performance, but a 9900K with another 2 furnaces lashed onto it is supposed to counter 12 and 16 core Ryzen 3000? Did Intel not watch CES?
Click to expand...

It is one area AMD needs to work with as a lot of companies are still running older software systems that will never get rewritten (partly cost and partly because no-one understands them any more) where due to software architecture inadequacies to cope with modern demands they have to go crazy on hardware with 16+ core CPUs and lots of RAM spawning multiple server instances to cope with demands that a modern equivalent could probably cope on a single 8-16 core server! a lot of this software is seriously impacted by the scheduler inefficiencies as well.

Maybe eventually the software will get replaced but I dunno how soon - for instance it has taken us since 2006 to beat just one of around 10 of such implementations into a modern incarnation at work (some of the source apparently has date stamped comments from 1970s - though IIRC the bulk of it was written in 1986).
 
You must log in or register to reply here.
Back
Top Bottom