iXBT Labs - Computer Hardware in Detail






Choosing a Processor

Part 1: Theory.

July 13, 2009

<< Previous page

     Next page >>

The false criteria of choosing a CPU

We have just analyzed the basic objective criteria for choosing a processor. And before we proceed to the practical part of the article (tips and examples), we should say a few words about another important issue: false selection criteria. Awareness of false criteria and their reasons for being false can be just as useful as understanding objective criteria.

Power consumption

It has become trendy to measure power consumption of processors. But consequences of this fashion for unenlightened minds of common users often turn out to be so fatal that I sometimes think that it does more harm than good. Everything looks very simple: here is a diagram, where one processor consumes X Watts, while the second one consumes Y Watts. If X > Y, then the second processor is better. Why better? Well, it consumes less! Why?! Nobody analyzes how X relates to Y as well as how these values relate to consumer electronics, for example -- that's the magic of numbers: if X=11, Y=18, and the lower number is better, then X is better than Y. So in the beginning of this chapter we'll make a provocative statement: power consumption of a processor is actually of little importance (especially in our parts and for our mentality).

It's difficult to explain to people, grown up in the Soviet Union, how to save on power consumption, as it had been very cheap in their childhood, and the stereotype is still working, although the situation has changed drastically. The later generation leave room for this possibility already. Here is the practical part -- power consumption of even the most voracious CPU is about 130 W -- that is approximately 1.5-2 average usual (non low-energy) light bulbs, used by most people living on the ex-USSR territory. It must be noted that the above-mentioned 130 W is the maximum power consumption of a top quad-core processor, when all its cores are fully loaded. When idle (computer is running, but no one is working with it), typical power consumption of even the most voracious monsters is within 20-30 W depending on activity of background processes. We don't have statistics on average power consumption, but if you don't mind expert judgment, we'd estimate the average power consumption of even the most powerful processor in an average home or office desktop PC about 30-60 W -- that is a single light bulb (not very bright at that).

8 x 40 W light bulbs = 320 W. There are no processors with such appetites.

Thus, we don't try to talk you out of saving power -- just treat it logically and consistently: it's not reasonable to choose a processor consuming 60 W instead of 130 W, if you don't save power in the other fields. It makes no sense to care about power consumption of a processor, if you forget to switch off the lights in your room at night: a light bulb that you always forget to switch off (in the lavatory, kitchen, on the desk) and a top quad-core processor will have similar costs. The same concerns office computers: if your office saves every penny, it makes sense to think about low-energy CPUs. But if your secretary is in the habit of reprinting a 40-page report because of a mistake on Page 13, and the other people download gigabytes of video from YouTube, you will hardly manage to cut your expenses, even if you install 1-W processors in all computers. Besides, it's not the most reasonable start for the retrenchment campaign.

So why low-energy processors are still so popular despite the above-said? in most cases it comes down not to electricity bills, but to noise, which often directly relates to power consumption. A powerful processor grows very hot, so it need intensive cooling. A powerful CPU cooler is more expensive (which increases the price of a system unit). Besides, it's much noisier. And a price tag of a powerful but low-noise cooler is several times as high. Besides, even low-noise powerful coolers still pump much more air through the system unit, and air means dust. Dust settles on all system elements (including the cooler heatsink). Sooner or later, it starts to interfere with effective cooling -- so a system unit with a powerful processor has to be disassembled and cleared out much more often. So a system unit with a high-power processor costs more, it is often noisier, disturbing the owner and other people, and it needs maintenance more often. It's all these issues that make low-power processors more attractive, not the desire to cut down electricity bills.


We'll start with a simple axiom: there are no manufacturers of 'good' and 'bad' processors, 'reliable' and 'unreliable' processors, 'compatible' and 'incompatible' processors. At least we haven't heard of any such x86(-64) CPU manufacturer. Let's enumerate the most significant myths and track down their origins.

Myth #1: 'incompatible' processors

There is only one definition of compatibility that applies to a processor: its capacity to execute programs. Indeed, in time immemorial, some MS DOS programs (compiled with a certain compiler, which is not relevant now...) failed to run on AMD K5 processors. It happened in 1995 -- much time has passed since that time, but rumors about incompatibilities are still alive. From hearsay, the first VIA C3 processors contained an error resulting in PC freezes in some programs. However, these processors were discontinued long ago. Besides, VIA processors are not very popular in ex-USSR countries. There is also such thing as 'false incompatibility': the lack of support for this or that instruction set -- MMX, 3DNow!, SSE, SSE2, SSE3, SSE4, etc. Why is it false? Because prior to using any extended instruction set, any correctly-written program (we theoretically assume that there might exist some mythic incorrectly-written programs, but we have never seen one) finds out whether it's supported by a given processor. If this processor does not support it, the program chooses another branch that does not use this instruction set. Thus, support for a given instruction set or the lack of it has nothing to do with compatibility. That's a fact.

Myth #2: 'hot' processors

We'll start with the main point: it's not a myth -- there really exist processors, which consume about 100 W under full load. Cooling such processors is a serious problem. Wrong cooling may result in CPU overheating, which may lead to freezes, switching off, or serious performance drops. The myth consists in the following -- its purveyors usually associate 'hot processors' only with one manufacturer of x86(-64) CPUs. If they had bad luck with AMD Thunderbird (one of Athlon cores), it will be AMD. If a user 'suffered' from the Pentium 4 Prescott, it will certainly be Intel. And so forth. We can give you plenty of examples that occurred earlier or later. in fact, there were a lot of 'hot' processors in the history of both biggest manufacturers of x86(-64) CPUs, so it makes absolutely no sense to dispute which company had more of them, just as it's pointless to say that all processors from one manufacturer are 'hot', while all processors from the other manufacturer are 'cold'.

Myth #3: 'unreliable' processors

It's probably the most 'non-mythic' of all myths, as it's based on real facts: AMD Athlon processors with Thunderbird core (Socket A) had a fragile open die (unlike Pentium III Coppermine, which also contained an open die, but it was much more resistant to damage). So it was possible to damage it during installation of a cooler (notorious 'chipped processors'). Such processors could break down completely, or they could become 'inadequate' (constant freezes, sudden reboots, etc). Thus, speaking of AMD Athlon "Thunderbird" as unreliable processors, users meant that such processors required more delicate treatment, as they were prone to damage in those situations where other processors did fine. However, processors with the Thunderbird core had fallen into oblivion long ago. And the myth about unreliability of AMD processors exists only in some good-for-nothing heads, even though x86 processors with open dice are not manufactured anymore.

We are through with the myths, so let's return to the main topic: to consider processors from only one manufacturer means to limit your choice out of false criteria. Of course, there were periods in the history of both companies, when products of one manufacturer were superior to competing products. However, any company living through such times can still offer several successful models, which can compete with products from the 'favorite' manufacturer in price/performance (or even in performance alone). You never know, perhaps such models are available in the price segment you need... That's why under no circumstances should you rule out all products of a manufacturer from your options. No one specializes in solely 'good' or 'bad' processors.

Advertising and advertisers

It's possible to extract useful information even from advertising brochures. You should understand several simple truths:

  1. Advertisements rarely contain outright lies, as it's liable to prosecution, and competitors are up to every move. So the main subterfuge of any advertisement is to tell the truth, just not all of it.
  2. Promotional materials will highlight advantages of the product, of course. But you won't find the list of drawbacks there. Does it mean there are no any cons? I don't think so. Companies just 'forget' to mention them.
  3. Besides, companies rarely dare to compare their products directly with competing models. So keep in mind that listed advantages of a product are hardly exclusive -- competing products will certainly board of a similar list.
  4. When a new product differs from the old solely in the cap color, marketing guys resort to their favorite trick -- they use phrase like 'even better.' Better than what? Better how? The lack of specifics is the first sign that there is nothing specific.

CPU advertisement is a specific genre. However, it obeys all the rules. The same concerns talking to sales managers in stores, if you want to buy a CPU there. The simplest test that any educated sales manager should pass is to describe a competing product. Ask about an Intel processor and then ask what AMD can offer in the same segment (you can swap company names). If one of descriptions is purely positive, while the second is emotionally negative, the test is failed. A good consultant should describe pros and cons of any product in a clear, precise, educated manner, without personal emotions.

Write a comment below. No registration needed!

<< Previous page

Next page >>

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.