4 ms·
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 (
by BirAdam 4y 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.
- goosedragons 4y agoWith 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.
- BirAdam 4y agoThat's a really fair point. I support you could do something like: i82020GEM or i82020NNN so you have intel 8 core, 2020, GPU, ECC, multiplier or without... then we're back to really bad clusters of numbers and letters, but at least in that scenario it has meaning.
- selectodude 4y agoBut anybody who can parse that without looking it up obviously cares enough to look at the current product numbering scheme and hop on over to ark.intel.com all the same.
- Sakos 4y agoMy biggest issue with that is I couldn't find it documented anywhere and it feels like they come up with new suffixes every year. I was on board with F and K, but now there's a bunch of others that I just have no clue what they mean.
- LegitShady 4y agohttps://www.intel.com/content/www/us/en/support/articles/000058567/processors/intel-core-processors.html https://www.intel.com/content/www/us/en/support/articles/000...
- Tsiklon 4y agoThe 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.
- matja 4y ago> The First Digit bears no connection to anything One pattern is: If first digit is 3,4,5 then it has one AVX512 FMA unit, if it is 6,7,8 then it has two. ...except the 5122 (2), 5219Y (2), 5222 (2) and 6122 (1)
- Tsiklon 4y agogood shout - I'm not so up to speed with what AVX512 can accelerate, but is it such a distinctive feature that it warrants a full delineating model number?
- adgjlsfhk1 4y agoan extra avx512 fma unit means it can do twice as many z=a*b+c instructions per clock. this only really matters for matrix multiplication (for anything else you are memory bandwidth limited), but 2x matmul performance is a really big deal for hpc
- acdha 4y 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.