For someone who doesn't follow Intel chip generations closely, the chart of the generations of Core I3/I5/I7 chips was informative. Two Observations:
1. Why doesn't Intel actually name the chips to make them clearer?
2. The poor Mac Mini that Apple is selling in 2017 uses 2013 era chips and is way overpriced.
Edit: fixed typos
Because that would make it clearer which chip was better when an average non-techie person was in Best Buy. If it was clearer which chip was which, Acer and Lenovo and whatever low end computer maker is still around couldn't put a cheaper chip in without people noticing.
Picked a random Dell and see they are charging $66.30 to upgrade from a $117 i3 to a $192 i5. I'm not sure where the unique SKU idea comes from. They've always had some chips that are retail packaged and others that are only available in trays for OEMs.
Although the author points out that single core performance hasn't improved since Sandy Bridge (2011) and multi-core performance has been largely lackluster since then.
While I agree it's shameful for Apple to be selling the current Mac Mini, the performance from 2017 chips is going to be only ~20% better than what's in the Mac Mini now. [0]
It's not a completely fair comparison because the i7-7500U is 15W, but I can't find any Intel mobile CPUs released in 2015 or 2016 which were dual core and didn't have a 15W TDP.
[0] http://www.cpubenchmark.net/compare.php?cmp%5B%5D=2345&cmp%5...
Which honestly seems pretty fair. 4K wasn't mainstream in 2013, you needed to spend a lot of money to buy a 4K display or projector. [0]
I wouldn't say that most people purchasing a Mac Mini are planning to use a 4K display with it, considering its target market is people who want to buy a Mac (without an integrated display) but don't want to shell out for a Mac Pro.
The Mac Pro not supporting 4K is a problem, one that has been brought up by many other people as an example of how Apple is ignoring the Pro market segment.
[0] http://www.alphr.com/news/382774/asus-takes-pre-orders-on-fi...
4K display prices are down, way down, and well within the normal consumer range now - you can pick up a very nice 4K monitor for $300 or less. It's not unreasonable to expect a Mac being sold new in 2017 to be able to drive it...
I have a late 2013 MBPR (NVIDIA GT750M) with an external 4K monitor in portrait mode connected via DisplayPort, plus the laptop's display. Both are running at 60Hz. I don't see any sluggishness or mouse lag at all; everything is very smooth.
I mostly run Windows 10 on this machine, but it also works fine when I boot in to OSX.
A while back I tested a 2014 MBPR with the AMD GPU vs. my late 2013 model with NVIDIA, and the AMD was a bit faster.
Which means it still won't be as powerful as the Mini they sold five years ago:
https://browser.primatelabs.com/mac-benchmarks
At the beginning of last year, I expected to be replacing a Mini and a MacBook. I ended up buying a 2012 Mini and a 2012 MacBook Pro on eBay, bumping up the memory and swapping in SSDs. I never would have imagined that a few years ago.
Intel does name their chips. ex: Intel Core i7-6700K[0], Intel Core i7-5775C[1]
They also have families (Core i7), and generations (6th Generation i7 Processors), and microarchitectures (Skylake)[2]
As for making it clearer, I think the i3/i5/i7/Xeon branding is probably as clear as they can, given that they have a few dozen processors per market (Desktop/Server/Mobile/Embedded) per generation, each targeting a different thing (low TDP, high clock speed, more cores, hyperthreading, etc).
[0] - http://ark.intel.com/products/88195/Intel-Core-i7-6700K-Proc...
[1] - http://ark.intel.com/products/88040/Intel-Core-i7-5775C-Proc...
[2] - http://ark.intel.com/products/codename/37572/Skylake#@Deskto...
[0] - http://ark.intel.com/products/family/78583/Intel-Xeon-Proces...
[1] - http://ark.intel.com/compare/?ids=85766,85765,85764,85763,85...
E3/E5/E7 (market segment + socket), the first digit is the socket configuration (2/4/8), the next 3 digits are a model number and the vX at the end is the CPU generation.
The Xeon E3-1220v2 in my ProLiant ML10 is a lower-end, single socket Xeon in a fairly high performance segment.
The Xeon E5-2403v2 in my ThinkServer TD340 is a mid-range, dual socket Xeon in a fairly low performance segment.
The fact that the 3 digit identifier has no relation to overall performance is the biggest shame in the Xeon chips, considering Intel does this with the Core i3/i5/i7 products.
OTOH, Xeon's can vary so wildly it's hard to give a tangible performance or configuration metric to the model identifier. Things like overall CPU cache, core count, clock speed, expansion support (namely, number of PCIe lanes) all vary so much that you really have no choice but to look at ARK since there's no way to convey this in a more efficient manner.
EX = product line
vX = generation
The clock speed omission is a bummer, but Xeons have two different clock speeds on all processors (one turbo -- which only a certain percentage of the cores can hit simultaneously), so things are a bit more complicated.
[1] https://semiaccurate.com/assets/uploads/2011/04/xeonmodels.p...
The i3/i5/i7 is clear as market segmentation. But you never know which generation is the current one. Also there remains to be seen a clear differentiation between models "inside" i3/i5/i7 division
x is the brand family that most everyone is familiar with.
y is the generation. 6th gens with have a 6. 7th gens will have a 7.
zzz is roughly a capability indicator. The higher the number, the more powerful the processor (generally speaking). This is only comparable within the same generation. Comparisons will be off when comparing these identifiers to other generations, so it's suggested to us benchmarks.
A is a alphabetical modifier to indicate a feature like unlocked.
So an i7-6700k is an i7 family processor from the 6th generation (Sky Lake) at the 700 "level". The chip is also unlocked for overclocking. An i7-7700k is an i7 family processor from the 7th generation (Kaby Lake) at the 700 "level". This chip is also unlocked. Take all of the goodness from the previous generation's processor, and (more-or-less) give it the new optimization from the current generation.
And honestly, the SKUs "level" is a fair bit more random than you imply; you wouldn't know from the model number that i7-6567U was significantly faster than a i7-6660U, for instance. Or that a i7-6700k significantly beats a i7-6700, which significantly beats a i7-6700T.
It's only within the same TDP that the level is non-random, but you aren't going to know this without going on ark.intel.com.
If I'm at a store and I see a i7-6700 and a i7-7300 which one is better? a i7-6735 or a i7-6720A? It doesn't make any sense for the general public.
Intel used to have one segment, then two segments: the Celeron and Pentium. Then it went for the Core 2, and then other lines
Especially today, that clock speeds are pretty much immaterial, they could take all those models and turn it into only a couple.
This comparison[0] shows how a 4th gen processor that has a higher over-all benchmark. It's still non-intuitive knowing the breakdown.
[0]: http://cpuboss.com/cpus/Intel-Core-i7-6700K-vs-Intel-Core-i7...
Because it makes more money this way. "Cheap", "Expensive" and "Middle ground" is the optimal marketing strategy.
It feigns transparency but really it is just an opportunity to exploit a common bug in human-brain logic. Specifically, Decoy Pricing.
The i5 and i7 are both quad-core CPUs clocked at no more than 100MHz away from each other. But the i7 is $100 more to make the i5 look like a good deal. When consumers buy the i5 they "think" they got a good deal, when in reality, Intel priced the i7 to make them feel this way.
Read up on Decoy Pricing to learn about this effect: https://en.wikipedia.org/wiki/Pricing_strategies#Decoy_prici...
https://en.wikipedia.org/wiki/Decoy_effect
Also, don't complain about marketers taking advantage of human psychology. That's their freaking job. Instead, you should take advantage of it too.
Instead of people being confused and accidentally buying a more expensive PC than they need, they just stop buying PCs altogether because there is no clearly communicated benefits for buying a new one.
Intel's marketing is due for an update. But for these past seven years, the "Decoy Pricing" of i3 vs i5 vs i7 has upgraded tons of users to i5s without those users fully appreciating the benefits of i3 vs i5 vs i7.
Its rather difficult for me to explain the difference between all three chips, and my uncles / aunts often just settle on i5... or even i7 "just in case".
Trust me on this: the decoy pricing has worked this past half-decade. That the marketing has gotten stale over the course of seven YEARS is no surprise... but how many other systems have had the same freaking name for seven years at a time?
I'm not sure thats a fair comparison - clock rate isn't the only thing and the current high end i5/i7 chips (6600k and the 6700k) are 500Mhz apart, and the i7 (as always) has hyperthreading (in some applications, hyperthreading is actually very useful) while the i5's don't. And while the i7 has a 15% higher clockrate, it is ~20% faster, while being 40% more expensive. With the k-series specifically, the i7 having a higher base clock also means you can overlock it better.
FWIW, I don't find Intel's naming scheme unclear, but OTOH, I have been following CPU/GPU version numbers for over 10 years so I could be blind to the confusion.
The i5-5200U is a dual-core laptop chip under 20W TDP. Its very different from the i5-6600k that Nemothekid was talking about.
The Core i3-6320 is about $60-70 cheaper than the i5-6600k, but on applications that don't benefit (sufficiently) from multiple cores, it can perform nearly as well. Anecdotally, friends building PCs always compare i5 and i7 and never consider the i3, but for their real world uses (many games!) it would be just perfect (and the savings could be thrown at the video card where it'd make more of a difference.)
Would i be willing to pay 35$ extra per additional core if I knew I would want to replace the hardware within two years anyway? Questionable. But when hardware lasts half a decade? The premium per core per year becomes irresistably low. The i7-2600k will soon have it's sixth birthday...
However, whether "the core is going to be fully utilized" is another question. If you only ever need 2 cores, it is wasted money, plus, if your focus is on gaming, and your budget requires you to make choices based on right now, the extra $170 can be the difference between getting a Radeon 480 or a Geforce 1070.
(Since I hope my next build lasts another 8 years, I'll probably "pad" my build with an i5... I've fallen into the trap!)
Hopefully somebody who knows more about this than me can jump in and fill in the blanks here.
EDIT: I communicated my point poorly and the comments below mine explain this is much clearer terms
Really it is just a way to do differential pricing. Some people will pay $500, some only $100 while the part only cost $25 to make. The way to maximize profits it sell to both, and find a way to justify the price difference.
As a hypothetical example (I'm not 100% sure if this is how Intel does things), and i5 could just be an i7 except it couldn't handle the higher clock rate. So instead of throwing away a chip that's limited in clock... Intel will sell the chip as an i5.
Or perhaps in another chip, the hyperthreading portion got borked because of a piece of dust landed in the lithography. So instead of selling it as an i7, the hyperthreading portion is fused off and the chip is sold as an i5.
The IBM-Cell in the PS3 was a different approach to the same problem. The Cell was specified as 7-SPUs and 1-CPU, but in actuality... the manufacturing process made 8-SPUs and 1-CPU.
IBM assumed that one of the SPUs would fail during manufacturing, so they only promised to deliver 7 of them per PS3.
You can literally see the 8 SPUs: https://upload.wikimedia.org/wikipedia/commons/e/e0/Cell-Pro...
There's lots of ways to approach the problem. Manufacturing is an imperfect process: not everything is made identically. So the manufacturers have to figure out how to test for differences and... if necessary... sell the same "manufactured" chip in different bins.
Phenom.
https://en.wikipedia.org/wiki/AMD_Phenom
https://en.wikipedia.org/wiki/Phenom_II
Not all the hardware of Cell on PS3 was available if you run the official Linux port. Which was later blatantly removed.
Server: expensive, more expensive, you can't afford it
Desktop: nicely priced, expensive
Mobile: nicely priced, expensive, pretty expensive
Of course, if they made it that simple, then this would happen:
Cheap server: buy ARM
Cheap Desktop: buy an older PC
Cheap mobile: ARM
REALLY cheap mobile: buy last years' ARM
For #2, the performance isn't really much different for most people. Running office and surfing the net probably are more limited by typing speed and internet connectivity than the speed of your CPU. I think most people pay those prices because they want the simplicity (either real or imagined) that Apple has over the other options.
[0] https://forums.lenovo.com/t5/ThinkPad-T400-T500-and-newer-T/...
[1] http://pcpartpicker.com/products/memory/#t=9&Z=16384002&s=30...
It's on purpose to trick "normal" consumers into buying lower-end chips thinking they get higher-end performance. It actually tricks a lot of tech-savvy people, too.
Examples:
1) switching up the Core architecture in Celerons and Pentiums for the Atom architecture (but keeping retail/OEM prices the same)
2) making "dual-core" Core i7 chips for laptops (what's the point of all the different variants of Core i5 then ?!)
3) replacing H series in most laptops with lower-performance U series, and soon we're going to witness the switch to the even lower-end (more throttling ~1.5GHz base speed) Y-series
4) they've just started calling the Y-series-based Core M5 and Core M7, Core i5 and Core i7, also to trick consumers into thinking they're getting "Core i5/i7 performance"
I'm hoping AMD's Ryzen + Qualcomm's (fully) Windows 10 capable Qualcomm 835 chip will put some sense into Intel to stop ripping off customers like this, but I think they'll only do it kicking and screaming. In other words don't expect much better "value" from Intel chips at least until 2019-2020. They'll probably just wait for their 2020 streamlined/lower cost architecture before they drop prices too much, unless the competition is rapidly eating away at Intel's market share by then.
Higher frequency, larger caches, more instructions per core, more cores per die, higher RAM bandwidth, more FLOPS per Watt consumed?