Cannon Lake: Intel’s Forgotten Generation

https://chipsandcheese.com/2022/11/15/cannon-lake-intels-forgotten-generation/

by rbanffy • 4 years ago

62 33 4 years ago

Sakos3• 4 years ago

Honestly, the past 5-ish years of Intel generations have blended together for me. I'm not even sure if they've already released a new one this year or what it might be called. Everything is some kind of Lake, they're all i3, i5, i7, i9, and they have some weird x000 naming scheme (then there are the utterly indecipherable xx00nn ones).

Am I the only one who prefers the simplicity of Zen 1, 2, 3, 4?

onli2• 4 years ago

On the desktop side Intel's naming is not so bad. We are at 13000 now, 12000 was before, 14000 follows. They call it 13th gen accordingly, which is incorrect but memorable.

Zen isn't all that better. Sure, Zen 1, 2, ... seems clearer, but can you tell me on the top of your head which processor has which architecture? Including nice outliers like the Ryzen 5 1600 AF? Now do the same for the mobile processors, the naming there is a total clusterfuck.

Also, maybe a bit more on topic: The current processor generation is actually nice and gives a real performance increase. Enough so that all three (i5-13600K, i7-13700K and i9-13900K) take the top three positions in my meta benchmark (for games). AMD follows right behind that (and needs less energy to do so), so it's really a good race now, just like the article concludes.

Sakos3• 4 years ago

but can you tell me on the top of your head which processor has which architecture?

Not sure how that's relevant. By name, I have no idea if Cannon Lake is before or after Skylake. I only know because I've heard of Skylake longer than Cannon Lake. I'd have no idea how to place any of the post-Skylake architectures amongst each other chronologically or performance-wise. I know that Zen 1 is older then Zen 3, because that's how the naming scheme works. The same way I know that the PS5 came after the PS4.

It's like the Xbox bullshit. The only reason why I know Xbox One is newer than 360 is because I'm a gamer who pays way too much attention to this shit.

jerrygenser• 4 years ago

Have you tried a power limited version of Intel to compare?

I have my 7700x limited to 65w with very little performance loss -- it's very easy with eco mode. I've seen lot of benchmarks with 7000 series chips on eco mode but not so much on Intel side.

These chips really should have been power limited with a little less multicore performance instead of being power drains.

adrian_b2• 4 years ago

To be fair, Intel also uses "13th generation Core" instead of Raptor Lake, "12th generation Core" instead of Alder Lake, "6th Generation Core" instead of Skylake, and so on, while AMD also has meaningless code names that are used for products before launch, e.g. Raphael for Zen 4, Vermeer for Zen 3, Matisse for Zen 2, Pinnacle Ridge for Zen 1+, Summit Ridge for Zen 1.

In both cases, the shorter and easier to remember names have won in the conscience of most of the public, i.e. it is easier to remember Zen 2 than Matisse, but it is easier to remember which was Kaby Lake instead of which was the 7th Generation Core.

Because both companies launch products typically every year, I would prefer something like Intel Core 2017 instead of Coffee Lake and AMD Zen 2022 instead of Zen 4.

sys_64738• 4 years ago

This is what happened when a company has marketing departments making such decisions unchecked. Sun had the same problem with their idiot marketing dept naming the Sun Niagra systems bizarre names like CoolThreads 1000 and systems with numbers like 5420. Larry came in and said it was stupid and to name them precisely what they were so they became T3-1, T4-2, etc. to indicate the chip generation and the number of chips on the system board.

mark-r• 4 years ago

It's a good thing they fixed that problem and became the powerhouse that they are today. </sarcasm>

BirAdam2• 4 years ago

The Intel "core" series is just stale branding. Really stale branding. That name started being used in 2006 for an iteration of the Intel P6 microarchitecture (first implemented in 1995 with the Pentium Pro). Intel's NetBurst (Pentium 4) wasn't great, so they brought something old back, made improvements, made it dual core, and we had the Intel Core Duo. They then did the usual licensing back and forth with AMD for AMD64, and they released the Core 2 (Penryn) just a little later in the same year. The Core iN series started with the Nehalem microarchitecture in late 2008. This was a major improvement over, but still an iteration on, P6. However modest these initial changes were, we are now so far removed from those earlier designs that carrying the "core" branding is just bad. Intel really needs refreshed branding. They should also do away with the largely irrelevant numbers and letters that they shove on the end of their CPU family names.nI would even be okay with just "Intel 8core 3.6gHz 2020" and so on, but Intel tends to prefer cringe apparently.

goosedragons3• 4 years ago

With the way Intel segments CPUs you need the stupid numbers. "Intel 8-core 3.6GHz 2020 no integrated GPU, no ECC with unlocked multiplier" is a bit of a mouthful.

Tsiklon2• 4 years ago

The last canonical example of Intel refreshing their naming convention was with their Xeon line up which is now utterly incomprehensible, at a glance you used to be able to understand what was going on;

EX-YZZZ vA

E3 - Rebranded Desktop Class parts with ECC support

E5 - High-Desktop & Regular Multi-Processor Server parts

E7 - High-End Multi-Processor Server

Y - Denotes Socket Support. ZZZ - Denotes a model number, indicating from low to high increases in features, cores and/or clock speed. vA - Denotes Generation.

Xeon E5-2980v2 - is a high-core count server processor for use in dual socket servers, belonging to the Ivy Bridge Generation of parts.

The new naming "convention" is as follows - from a glance can you identify any information from this;

Xeon Bronze 3204

Xeon Silver 4214R

Xeon Gold 6212U

Xeon Gold 6342

Xeon Platinum 8280

Xeon Platinum 8380

Xeon W-2295

Bronze, Silver, Gold, Platinum - are so unclear as to be meaningless - there are dual socket Silver processors more capable than some Gold processors. W prefixed parts are Workstation parts. The First Digit bears no connection to anything (except in the platinum series where it denotes 8 socket capability), perhaps PCI-E Lanes? The Second digit apparently denotes generation. The Suffix denotes that it's only capable of running in a single processor configuration - except when it doesn't.

Ohh, and the 'E3-**** v*' convention still stands for desktop rebranded parts.

It's a complete mess.

acdha2• 4 years ago

Intel's NetBurst (Pentium 4) wasn't great, so they brought something old back, made improvements, made it dual core, and we had the Intel Core Duo.

Core was based on the mobile-focused Pentium-M design from their Israeli team. That lineage traces back to the Pentium III but both the Pentium-M and Core designs had significant modifications so I wouldn’t describe it as bringing something old back — more like admitting that the NetBurst big gamble had failed and they switched to a design more like what everyone else was making. The P4 was kind of interesting as Intel’s second bust around the turn of the century based on a design which tried to hit higher performance assuming everyone had access to highly-sophisticated compilers and would use them carefully. As with Itanium, I do wonder whether anything would have been different if they’d open sourced icc – the profits weren’t significant and while that could have benefited AMD, I doubt that would have outweighed two major flops.

ur-whale• 4 years ago

Agree that Intel naming schemes are very confusing and very few people understand them or care, even.

However, AMD isn't any better ... is it a Vermeer or is it a Cezanne, a Chagall or a Milan?

matja2• 4 years ago

Ryzen 5xxxG(E) or Ryzen 5 5500 -> Cezanne

Ryzen 5xxx -> Vermeer

Threadripper 5xxx -> Chagall

EPYC xxx3 -> Milan

dannyw• 4 years ago

AMD will be selling laptop chips of various older generations under one name soon in order to deceive consumers who aren't paying attention: https://youtu.be/MipuN4RU4m8

A 7000 series laptop chip could be Zen 4, 3, or 2...

georgeburdell• 4 years ago

I worked at Intel’s main R&D site from Sandy bridge to Meteor Lake bring-up on a team looking at electrical data. As I recall, the first Cannonlake wafers produced data relevant to me around New Years Eve 2015. This was not the first 10nm wafer by a long shot —- the process had been in development for years at that point and with many test chips run. Still, the initial data was so unbelievably bad, due to the lowest metal layers (supposed to be 10nm’s secret sauce), that even just looking at the basic line resistor test structures it looked like the measurement machines had malfunctioned or had the wrong hardware configuration.

Worst bring-up in my time. I feel bad for the architecture team. I know Intel is getting dunked on my AMD right now but internally the worst of it was years ago. I retain confidence in the direction things are heading now

n7pdx• 4 years ago

LOL I remember working on Cannonlake A0. The dozen (not plural) parts we got passed sort but all bricked. Turns out the PLL was randomly losing lock due to variation alone. Also the analog designers refused to run timing checks on digital control logic (if we just match the layout it would work amirite?)

I left a year after that... tbh don't see things getting any better from talking to friends. We'll see.

Cannon Lake: Intel’s Forgotten Generation

Related Stories