What is more important — the number of cores or the clock speed of the processor

If you take a look at the cur­rent proces­sor mar­ket, it may seem that chip com­pa­nies have decid­ed to enter the Guin­ness Book of Records. AMD Unveils 32-Core Thread­rip­per 2290WX Proces­sor! “Intel released the Core i7-8086K proces­sor, which can over­clock up to 5.0 GHz”! New per­for­mance records are set every year, and as a result, it’s easy to get con­fused about which is bet­ter — the num­ber of cores or clock speed?

Let’s try to under­stand this issue.

A little analogy

Sup­pose you need to trans­port a load. Lots of car­go. And there are two options — to trans­port one very fast car or sev­er­al slow ones.

Using sev­er­al slow cars allows you to trans­port a lot of car­go, but not very quick­ly. Let’s say he trav­els from one city to anoth­er for three days. But he will come all at once, and that’s great.

One fast car can car­ry a lit­tle less car­go. But from one city to anoth­er, she will go all day. Then she, how­ev­er, will have to return, pick up the rest of the car­go, bring it again, and again, and again.

The ide­al solu­tion to this prob­lem is to use sev­er­al fast machines! Then a lot of car­go will be trans­port­ed, and it will take only one day. But we live, unfor­tu­nate­ly, not in an ide­al world.

It would seem, what do the proces­sors have to do with it? Yes, just a “heap of car­go” is a flow of cal­cu­la­tions that a chip needs to per­form. “A lot of slow machines” is a few low-per­for­mance cores. And “one fast” is one high-per­for­mance core, with an increased clock fre­quen­cy.

And the solu­tion to the ques­tion “high clock fre­quen­cy or many cores” depends pri­mar­i­ly on the planned scope of use.

How the processor works

The proces­sor is engaged in the fact that it per­forms the sim­plest arith­metic cal­cu­la­tions. The oper­at­ing sys­tem and soft­ware envi­ron­ments turn user com­mands into the very arith­metic tasks that the chip needs to solve. This is done through inter­preters and algo­rithms that indi­cate links between high and low lev­el pro­gram­ming lan­guages.

The proces­sor receives a lot of instruc­tions all the time. Each user action requires sev­er­al cal­cu­la­tions at once (or even tens of thou­sands). And these com­mands must be exe­cut­ed, oth­er­wise the com­put­er sim­ply will not func­tion.

Teams can be very dif­fer­ent. For exam­ple, to inter­pret the HTML code of a page (includ­ing this one) into a graph­ic image on the screen. Or decrypt a com­pressed ZIP archive. Or even turn some­thing cryp­to­graph­i­cal­ly encrypt­ed into a dis­play in the clear. In gen­er­al, the tasks are very dif­fer­ent and require dif­fer­ent resources to com­plete them.

So, for exam­ple, con­vert­ing HTML code will require a min­i­mum of resources and can be per­formed quite quick­ly even on a rel­a­tive­ly slow chip. ZIP archives already require sig­nif­i­cant­ly more com­put­er time. And cryp­to­graph­ic algo­rithms like SHA-256 cal­cu­la­tions com­plete­ly load the unfor­tu­nate chip like God knows what, and at the same time can­not be exe­cut­ed on mul­ti-thread­ed con­fig­u­ra­tions.

The cal­cu­la­tion of arti­fi­cial intel­li­gence — for exam­ple, in games — requires a lot of cal­cu­la­tions. But they are small, iso­lat­ed, and there­fore can be per­fect­ly exe­cut­ed in a mul­ti-thread­ed ver­sion — on sev­er­al cores at once. Thanks to this, high per­for­mance in games is achieved.

But archiv­ing and unzip­ping files is an exclu­sive­ly sin­gle-thread­ed oper­a­tion. So one pow­er­ful chip is required, oth­er­wise this process will take a lot of time. Espe­cial­ly if the archive con­tains a lot of small het­ero­ge­neous files.

And yet, which is better — the number of cores or the clock speed of the processor?

The prob­lem is that com­put­ers are used in dif­fer­ent ways. In the morn­ing you down­load firmware from the Inter­net to your phone, and there are two thou­sand files of 500 kilo­bytes in one archive. At lunchtime, you sit and look at cat gifs on the Inter­net. In the evening you play, fight­ing against numer­ous ene­mies in vir­tu­al real­i­ty.

That is, in the morn­ing the com­put­er needs a proces­sor with a high clock speed, at lunchtime — just at least some “chip”, and in the evening — with many cores. And “poke” “stones” depend­ing on the intend­ed use is clear­ly not the best idea.

That is why man­u­fac­tur­ers are try­ing to pro­duce mul­ti-core con­fig­u­ra­tions with high clock speeds. So, for exam­ple, the Intel Core i7-8086K (the top one in the Kabu Lake line at the time of writ­ing) is equipped with six pro­cess­ing cores with Hyper­Thread­ing tech­nol­o­gy and a base clock speed of 4.0 GHz. He can do every­thing! Expen­sive, though — $ 425 at the time of release.

For home use, choos­ing which is bet­ter — clock speed or num­ber of cores — is not worth it. The ide­al solu­tion would be to strike a bal­ance. For exam­ple, buy­ing some kind of quad-core chip with a base clock speed of 3.0 GHz. Its per­for­mance is enough for the vast major­i­ty of every­day tasks.