How can I tell if a PC processor is any good
The processor is the most important part of a computer, but CPU names and numbers don’t mean much to most people. Regular commenter 75drayton wants help figuring them out
I had no idea the Core i3-6100 you mentioned last week is faster than some of the slower i5 chips. Is there any chance of you writing an article that focuses on processors? I appreciate that PCs are more than just processors, but I would find it useful. 75drayton
It’s worse than that. There have been cheap Intel Pentium chips that were faster than Core i7’s! Intel uses BMW-style branding, where the Core i3, i5 and i7 are marketed as good/better/best. This is usually a fair reflection of current performance per watt of power used, but it doesn’t tell you the raw performance.
You also have to look beyond the branding, because today’s Core chips are different from last year’s. Intel usually brings out a new generation of processors every 12-18 months, and the Core range is now in its seventh generation.
Each generation of Core chips has its own code name, such as Sandy Bridge, Haswell, and Skylake. The latest is Kaby Lake. Each generation brings enhanced features, and some mark a shift to a new manufacturing technology, measured in nanometres. Smaller is better. Core iX chips have gone from 32nm to 22nm to today’s 14nm. Shrinking the transistors enables Intel to put more of them on each chip, and thus add new features.
The generation is shown by the first number after the dash in each Core chip’s name. For example, a Core i7-3770 is a third generation chip, while the Core i7-7770 is the seventh generation version of the same CPU. Intel says the rest of the number – in this case, 770 – is its SKU (stock-keeping unit) designation. Higher numbers usually indicate better performance and other features, but I don’t have room for all the details you can find online.
All Intel’s processors now include a graphics co-processor. This is also given a generation number, but it’s one step behind. For example, seventh generation Core i7 chips have HD Graphics 620 or whatever, while sixth-gen chips have HD Graphics 520. The best graphics chips get Iris branding.
Speed vs power consumption
People who need performance, including gamers, often complain that processors aren’t getting faster every year, like they used to. They are getting a little faster, but mostly they’re getting smaller and consuming less power. The key point, mentioned above, is “performance per watt”.
Chips that use less power generate less heat, so they don’t need as much cooling. This enables manufacturers to make thinner laptops with better battery life, which is what most people want.
Progress is now indicated mainly by the fall in TDP (Thermal Design Power) ratings. The second-gen Core i5-2500 in my 2011-vintage desktop PC is still faster than most current Intel chips, but it runs at 95W. It’s slightly slower than a brand new Core i7-7600U, but that runs at only 15W.
In other words, you can now put the same processing power in an ultra-thin laptop that used to need a tower system with big cooling fans.
Of course, chips with a higher TDP will generally run faster, other things being equal. In Intel’s current line-up, the fast quad-core chips with HQ after their names run at 45W or 47W, while the U chips are 15W or 28W designs. The Y chips consume only 4.5W: see below.
Under the Core
Intel was able to out-design and out-manufacture rivals such as AMD, but faced a tougher challenge when ARM processors started to dominate the markets for phones, smartphones and tablets. Intel’s chips were big. power-hungry and expensive; ARM chips were small, phenomenally power-efficient, cheap, and fast enough.
Intel sold ARM chips for a while before deciding to create its own range of small, cheap and power-efficient Atom processors to compete for this booming market. Atoms, unlike ARM chips, could execute the x86 instructions needed to run Microsoft Windows software.
Atom chips failed to penetrate the smartphone and tablet markets in significant numbers, but they were successful in powering cheap, small-screen Windows “netbooks” like the Asus Eee PC and Samsung NC10.
These early Atom chips were rather slow. When designs became fast enough for general-purpose use, Intel started branding them with historic names – Pentium and Celeron – that had more prestige. These two lines dominate today’s entry-level PC market.
Unfortunately, there doesn’t seem to be any logic to Intel’s numbering system, beyond the fact that X7 Atoms are faster than X5 Atoms, which are better than X3 Atoms. Pentium-Atoms should also be faster than Celeron-Atoms, but you’d have to run benchmarks to be sure.
In 2014, Intel announced a new family of processors, which it called Core M. It said that Core chips were for people who needed power, and Atom chips were for people who needed battery life, but Core M would bridge the two.
The Core M wasn’t a success in the Windows laptop market, though Apple used it in the MacBook – the model best known for having a single USB-C port. That may be why the M range has survived. (Digital Trend’s headline put it brutally: Nobody wants Intel’s Core M processor, and Computex proves it.)
Unfortunately, with Kaby Lake, Intel changed its naming system by putting a Y in some chip numbers. The Core m5 became the Core i5-7Y54 and the Core m7 became the Core i7-7Y75. People who may think they are getting a full-power 15W Core chip will in fact be getting a slower 4.5W Core M. Caveat emptor.
When buying a new PC, you can check the CPU’s name to get its age and approximate performance level. The blurb should also tell you the clock speed and whether it’s a dual core or quad-core chip. More GHz and more cores are always better, though a dual-core chip can still be faster than a quad-core.
After that, you can look for benchmark comparisons, especially if the benchmarks measure the kinds of things you normally do – maths processing, video rendering, gaming or whatever. The problem is that there are dozens of benchmarks and it can be hard to find results for new PCs.
However, you can find benchmarks for most processors at websites such as PassMark, Geekbench, and AnandTech. You can also compare CPUs head-to-head at CPU Boss, CPU World and AnandTech.
The quickest option is still the one I mentioned last week: go to Notebookcheck’s Comparison of Mobile Processors (CPU Benchmarks) and, other things being equal, choose the chip that’s highest in the table. Anything in the top 200 is good. Anything that doesn’t make the top 500 isn’t particularly good, but might be OK for casual tasks such as email, watching YouTube and BBC iPlayer, and web browsing. A CPU that would be frustrating in a $500 desktop can be perfectly acceptable in a $150-$250 tablet or a 2-in-1 laptop.
Notebookcheck’s table also includes AMD processors, which are worth checking out. Some of AMD’s cheaper chips have been both slow and hot, but processors like the A6-7310 look competitive. AMD also has the 14nm Zen, a “clean sheet” redesign, on the way. I wouldn’t bet on AMD regaining its former glory, but it may be perking up.