facebook rss twitter

Intel Cannonlake touted to offer >15pc performance improvement

by Mark Tyson on 10 February 2017, 10:01

Tags: Intel (NASDAQ:INTC)

Quick Link: HEXUS.net/qadd64

Add to My Vault: x

At Intel's annual investor day yesterday the company showed a slide claiming greater than 15 per cent performance improvements would be delivered by Cannonlake CPUs. The slide was entitled 'Advancing Moore's Law on 14nm', and you can see it reproduced directly below.

Above you can see that Intel reckons Kaby Lake Core i7 CPUs delivered an approximate 15 per cent performance uplift compared to Skylake Core i7 CPUs. When Intel releases its first 10nm CPUs, the 8th generation Core processors (AKA Cannonlake CPUs) we can expect an even better performance uplift at the top end.

Intel showed off a Cannonlake powered PC system back at CES 2017, last month. At that time CEO Krzanich announced that "for those who are wondering if Moore's Law is alive, is 10 nanometres going to be here, the answer is absolutely yes." Looking at timescales, the above slide shows Cannonlake's arrival time as being in H2 2017. However, the same slide says that Kaby Lake was delivered in H2 2016, and we know that was a mobile/laptop first release with desktop processors only arriving in Q1 2017. Then Intel will continue to refine its 10nm CPUs until its 7nm facilitiy comes on line in 2020.

It's over a year ago since we heard that Cannonlake processors would be the first for the Intel consumer that go beyond quad-core. With AMD Ryzen and its 8C/16T processors due to launch soon, Intel better get a move on.

Intel Xeon Processor E7-8894 v4 with 24C/48T

On the topic of Intel multi-core processors, yesterday the firm launched a new Xeon with more threads than you can shake a stick at. The pro-targeted Intel Xeon E7-8894 v4 CPU (list price US$8,898) has 24 hyper-threaded Broadwell cores with a base 2.4GHz and turbo 3.4GHz frequencies.

Intel has designed these new 165W Xeon chips for servers with up to eight sockets supporting up to 24 terabytes (TB) of memory (or up to 32 sockets via node controller) . Such a system should easily take in-memory analytics and real-time insight processing in its stride.

HEXUS Forums :: 32 Comments

Login with Forum Account

Don't have an account? Register today!
Posted by Corky34 - Fri 10 Feb 2017 11:16
IDK Skylake to Kaby Lake was a 15% improvement. ;)
Posted by Ozaron - Fri 10 Feb 2017 11:21
Corky34
IDK Skylake to Kaby Lake was a 15% improvement. ;)
*cough* 1% *cough*
Posted by Saracen - Fri 10 Feb 2017 11:28
While part of me thinks “great!”, there's a more subtle part that thinks “15%. So what?”

And I wonder if Intel shares a feature in common with most sharks …. keep swimming, or drown.

I mean, given that there will be a few exceptions to this, including perhaps gamjng and certainly a few true power-user applications, but for the vast majority I wonder …. just what will 15% increase in performance do for you?

Historically, IMHO, the performance race was about one of two things :-

1) Productivity increases in your workload, or
2) Enabling something new to be done that previously wasn't possible, or not at a pragmstic performance.

Or both.

An early example of 2) might be CAD, where you could get AutoCAD to run, even on 1980's era XT-class machines, but it was so slow as to be useless for all practical purposes. Another example was viable voice recognition and consequent dictation capabilities, where software waited for hardware to catch up.

My daily PC life, these days, consists of mainly humdrum stuff, like WP, spreadsheets, email, and so on. Then, I do use Photoshop but even that runs acceptably fast for my needs on a Q6600-era machine with 4GB DDR2 and HD, not SSD.

Oh, and as a writer, I use voice-recog extensively, and that machine copes perfectly well.

I suspect, for me at least, that a 15% system performance improvement would pretty much amount to my machine going from 90% of cycles in idle, waiting for activity from me, to 92% idle waiting for me, and a 15% CPU gain is highly unlikely to amount to anything like 15% system-wide.

I'm interested to know what others, if anyone's excited about this, thinks the real-world actual benefit to their daily PC usage will be? What will it let you do you can't do perfectly well already, or what prodictivity gains will be really noticeable?

Or is it just proof, however technically clever, of a shark's steely and grim drive to keep swimming?
Posted by kompukare - Fri 10 Feb 2017 11:30
Yes, it would seem that they are on about be the same +15% they claimed for Skylake vs Kaby Lake?
Which was more like this:

Ah, so Intel's marketing department has 15% = 7.5%. (All due to the increased base frequency and turbo).) Looks like their maths is exactly half of reality?
Anyway, I am sure there are some Kaby Lake mobile i7 which in some loads do perform 15% better than their Skylake equivalents most likely 15W models mainly due to not throttling as much.
But that is just used be expected from silicon stepping revisions (https://en.wikipedia.org/wiki/Stepping_level) and that is what Kaby Lake was: a new stepping of Skylake with some new fixed function video decode hardware. Certainly did not deserve a 1000+ numbers: the Haswell Refresh was a lot more honest, and years ago it might have gotten a designation like E2 from B2.
Posted by virtuo - Fri 10 Feb 2017 11:50
I do a lot of CPU (and RAM) intensive development work - particularly around encryption/compression and recently emulation. I like to keep my CPU as quick as possible without having to tear down my whole system or getting in to a constant minor upgrade cycle. That way I actually notice the performance bumps and it doesn't cost ridiculous amounts for what is ultimately a hobby.

So, as an example I bought in to my 4790K and keep it overclocked and watercooled. It is starting to fall slightly behind newer chips now, even though it tickles 5ghz most of the time, less in summer when ambient temps go up. So 2017 will probably be my next big upgrade, depending on what AMD actually do and how Intel react - especially with affordable extra cores/threads.

Being 3 gens behind, I think a 15% performance increase on my current setup is probably achievable. Which to me would be 10% quicker encryption, or 10% more FPS in a software rendered emulation. With overclocking headroom I'd be set for a while. So yes, this is exciting for me, even with the large pinch of salt it has to be taken with.

SEE NEWER »