>>
>> Partly because I want to finally get around to building my own, which I've always
>> found vaguely intimidated so haven't done, and partly because it annoys me that I don't
>> understand.
>>
I've built a good few over the years. It isn't as intimidating as it used to be. Compatibility is much less of a problem than it used to be, though some care is still needed on selecting components. (QVLs - Qualified Vendor Lists - provide decent compatibility reference)
For anyone who wants a technical challenge, it can be quite rewarding (always assuming that it works from the off ;-) )
>>
>> A good reference, but it doesn't really explain.
>>
>> e.g.
>>
>> Of the following two, apparently the i5 is better. Higher CPU mark and better rank.
>> I am ignoring price.
>>
>> Rank 1460 Intel Core i5-4200U @ 1.60GHz
>> Rang 1461 Intel Core i7-940XM @ 2.13GHz
>>
>> So the older gen, lower i number, slower clock speed is 'better'??
>>
I think you've got that the wrong way round. The i5 is some three years or so newer, and on a much revised architecture and reduced fab size. There are far too many factors to cover in considering performance, but in general, it moves on (even though the nomenclature, presumably for marketing reasons, retains an element of commonality).
>>
>> >> Speed of CPU
>> >> Generally unimportant,*
>>
>> OK, faster is hotter. Got that. But a CPU doesn't have to run at full
>> speed does it? Do would a faster CPU not run at a lower capacity? OR
>> do CPUs always try to run as fast as they can?
>>
>> And why isn't speed important? Bear in mind my CPU knowledge comes from 80s DEC
>> kit, and that wasn't very complicated by comparison.
>>
Faster is generally hotter (though heat depends a lot on architecture). A CPU will run at a level required to process the work that is thrown at it, and nowadays, it is common to quote a base speed and a further (turbo) speed to which it can rise for some time if the workload demands it (constant turbo might result in thermal throttling). the processing speed is certainly not constant.
>>
>> If a CPU has 4 cores, does that mean that each core gets 1/4 of
>> it's performance? Or is it that when the CPU is not running at full capacity
>> overall it can use some of that spare capacity by doing mroe than one thing
>> at once?
>>
>> >> Threads.
>> >> Unless your software is developed to use them properly (Windows isn't) unimportant,
>>
>> Are threads to cores as cores are to CPUs?
>>
Cores are physical cpus within the unit. AIUI, these are all rated at the quoted clock speed and the OS and hardware divis up workload between them.
Threads are logical cpus within each core (i.e. the OS sees a two core four thread cpu as four logical cpus, and again the work is divi'd up between them).
The action of utilising multiple cores and threads adds overhead. Ultimate performance depends on how the workload (and OS) lends itself to distributing workload both physically and logically, and multiple cores are considerably more efficient than multiple threads (workload allowing).
Multiple cores and threads can (in the correct circumstances) compensate for clock speed.
>> >> The AMD Ryzens are vastly superior in every respect as things stand.
>>
>> And seemingly reasonably priced. Is there any issue from going non-Intel? There used to be
>> potential issues a gazillion years ago.
>>
I wouldn't say that the Ryzens are superior in every respect, but they do have the advantage of providing more bangs for the buck, and the AMD motherboard chipsets support some nice things that Intel are just catching up with.
My last two builds have been AMD - I don't need gaming so a Ryzen APU (with considerably more competent integrated graphics than the Intel stuff) provides a good, solid hardworking base - graphics are as good as a spare older machine that was built with a good-quality discrete card (to support gaming) at the time.
Do your research as to suitability of AMD kit for your desired purposes, but I'd be surprised if it wasn't a good choice.
|
