Know More About CPU Performance Factors - Part 1

Written by Shivam Kumar on Sun, Oct 13, 2013 11:00 AM

You all have processors under the hood of the PC. You’re using them to read this article. But how much do you really know about that part of your system?

Over the next three weeks, we will be taking you step-by-step through every aspect of your CPU; from factors that most affect performance to our views on how to choose from the multitude of processors available.

We all know how to go about choosing a mouse, or a keyboard, or a monitor. But choosing a CPU is a little more complex. So to explain, we will first sum up all the factors to consider. They are:

1. CPU Clock Speed

2. CPU Cache

3. Number of Cores

4. Instruction Sets

5. Heat Management

6. Processor Bandwidth

7. Front Side Bus (FSB) Speed

You may notice that we have missed a few points, including number of transistors, fabrication and architecture. These factors will be covered in the above seven factors, although if you have additional queries you can ask in the discussion area.

You might argue that generation also affects the processing power. Although this is true, it would be more specific to say with every generation improvements are made to the clock speed, heat management and other things which we listed above.

So let's start:

1. Clock Rate

“Clock Rate” or simply the processor speed (generally in GHz) is one aspect you can look at and judge in two different processors; but to what extent it that reliable? What we usually see is people going to retailers and trying to get a processor with higher clock rates. However, this is a little too simplistic.

In definition form, clock rate is determined by the frequency of an oscillator crystal, which is present in the chip. Now, no one wants to go that much deep into the world of processors, so here is the simple explanation.

Clock rate is the number of instructions (commands) processed by the CPU in one second. A simple netbook processor with 1 GHz clock rate can process 1 billion instructions per second. Although that’s a huge number, this actually has no direct relation with the performance of the CPU. For instance, a modern CPU with 1.5 GHz speed can beat an older processor with 3 GHz clock rate. So, how is that even possible?

It is so because there are many factors besides clock rate that determine a processor's performance. The instructions in a simple PC are carried out in three stages; input phase, processing phase and the output phase. The same is true for a processor, but the process becomes quite complex.

The clock rate is only effective in carrying these instructions throughout the processor, but there is a distance or a time lag between different [logical] parts of the processor. We can take the clock rate as your walking speed and the distance as a long road. Even if you walk extremely fast, if the road is lengthier, you will be slower.

Modern processors are trying to decrease this lag between the different stages of processing to deliver more performance at lower clock rates. With every architecture, new instruction sets are introduced which take less time for processing and execution.

Even after this explanation, it cannot be completely clear how to choose between two different processors. This is because day by day processing is getting complex. One cannot solely judge the CPU performance on the basis of just the clock rate and distance; it requires other factors to arrive to a clear decision. However, if two processors share all the same specifications, then going with the higher clock rate is beneficial.

2. CPU Cache

This is probably the one thing most of us take for granted when it comes to CPU performance. So the question arises, is cache useless?

Well no, it is not useless at all. Every part of your CPU is an asset that contributes to its performance. Cache plays a vital role in improving performance.

In general, Cache is the in-house memory of CPU; where it stores frequently used data and paths for quick access and to lessen interaction with the slower RAM. But this definition is not adequate if we are going to know its real value for the processor's performance.

The main function of cache is to reduce a CPU's dependency on RAM. It buffers data that is used "more than average", so user can perform those actions faster than other. But if you notice, whilst RAM has a much larger size (on average 4GB) but cache only ranges from 256KB to 16MB. So, how can such a small unit improve performance? It can because it does not need to store files or running applications. It needs only to store paths and file segments, which are far smaller in size.

First of all let’s understand how cache works. When a CPU has to work on something, it asks various different memory units to get the file up. This request first goes to the L1 Cache, then the L2 Cache, then – if there is one – the L3 Cache (these will be discussed later). Then, finally, it goes to the RAM. When it reaches the L1 Cache unit, if the call finds the correct file or path there, it returns to the CPU for further processing. This is called a "hit". But if, however, it does not find the appropriate data there it moves on. This is called a “miss”. The RAM, as well as the HDD, has no misses, given that there are no errors with them.

Several benchmarks indicate that cache has a greater part in performance improvement than CPU cores. Disabling the cache unit, for example, will result in greater performance degradation than disabling one CPU core. Moreover, cache is available in three types; Level1 Cache, Level2 Cache and Level3 Cache. All the three caches are almost the same when it comes to affecting performance.

L1 cache is the fastest cache unit as it is the closest cache unit to the CPU. But it has its own drawbacks. When the CPU calls for something, this is the cache unit which responds first. Here the chances of getting a “hit” are very low due to its very small size; misses are very frequent in this cache. This cache unit is mostly used to store instructions for various different actions.

L2 Cache is a mid-level cache unit. AMD puts a great deal of stress on this unit of cache, while Intel has already adapted L3 caches in their CPUs. This level of cache is comparatively slower than L1 cache, as it is not very close to the CPU on the die. In AMD processors, this cache type has much greater capacity (usually 4 MB) when compared to Intel chips. Here the chances of hits and misses are equal. In an AMD processor the chance of hits is much more due to their bigger size.

L3 is the most advanced and largest cache on the processor's die chip. L3 cache is the slowest CPU cache unit until now, yet it is still far faster than the RAM. This has a large size for storing a great deal of data for quicker access. Here the chance of hits is very high, but it does not mean that there are no misses.

With explanations out of the way, let's move on to benchmarks. If we take three CPUs which feature the same number of cores as well as having identical clock rates, we can do some raw explanations.

Although we cannot do any head to head comparisons, by individual benchmarks we can arrive at a conclusion. In gaming benches, most games will show an average increase of 5 to 10 FPS when a CPU with a 4MB cache is compared to one with 2MB. The amount performance will improve varies greatly for different games, though the larger CPU cache has a borderline advantage. Just like in the aforementioned gaming benches, in synthetic benchmarks higher cache has the lead over others. Real-life performance increases are more noticeable and important.

It is worth noting that, surprisingly, software used for purposes like compression, video playback, internet browsing and video editing do not show any improvement through higher cache. This is probably because their data is not stored in cache due to less its frequent use, and hence the higher cache has no benefit.

Finally, we get to the point that CPU cache has an important role to play in CPU performance in areas which we usually take for granted, such as system start up. Modern processors utilize CPU cache more effectively, and thus it’s quite important to get a processor with an adequate amount of cache. Benchmarks also show noticeable FPS increases, so it’s clearly better if you know what’s under the hood.

 

To learn a little more, head over to Wikipedia to learn everything related to cache not included in performance factors.

Next Sunday, we will be covering the next three points in our series on getting to know your CPU, with our final two points and some helpful tips coming the week after that.

Did you find this article helpful?

Is there anything we missed out or didn’t explain sufficiently?

Thoughts and queries in the comments, please!

 

Login or Register to join the debate

Rep
75
Offline
admin approved badge
23:13 Oct-19-2013

Great article! I hope you'll be discussing CPU architecture in the other 2 parts, something that I am very interested in! :)

0
Rep
157
Offline
09:53 Oct-17-2013

part 2 ?

0
Rep
157
Offline
admin approved badge
09:59 Oct-17-2013

The development has started but as its a long article covering three important points, it will take time.

0
Rep
2,003
Offline
admin approved badge
20:33 Oct-17-2013

Take your time :)

0
Rep
77
Offline
21:44 Oct-16-2013

Were is the like button here?

0
Rep
2,003
Offline
admin approved badge
07:33 Oct-17-2013

On the lower-right corner of the top image :)

0
Rep
273
Offline
admin approved badge
16:11 Oct-16-2013

the issue is, and the reason you don't find really big caches is that the larger the cache, the longer it takes to respond.
there is also L4 cache found in some chips but it is extremely rare in consumer parts.
in older computers (Pentium II and earlier), cache was found on the motherboard or on a separate die rather than on the CPU die itself.

0
Rep
59
Offline
18:00 Oct-16-2013

oh thanks for clearing that up. i was wondering why not just one large cache instead of 3 of different size.

0
Rep
250
Offline
admin approved badge
07:50 Oct-17-2013

As TjPj has already said, a bigger memory is often much more difficult to maintain and access. A cache of upto L3 can be accessed in under 25 cpu cycles(5 instruction cycles) in the average case

0
Rep
4
Offline
08:02 Oct-14-2013

Awesome article !!!

0
Rep
109
Offline
admin approved badge
07:26 Oct-14-2013

An awesome article Shivam! I didn't even knew that cache was also important in performance. I only thought that clock rate matters the most. Waiting for next Sunday....BTW is cache and CPU registers are the same thing?

0
Rep
157
Offline
admin approved badge
07:29 Oct-14-2013

No,, CPU registers are somehow different from this. There are many types of CPU registers and they all do different things. Their work is way different, head over to wikipedia to know more as I have to do some more important work. They are not much important in performance, though as I always say, every part of a CPU is important.

0
Rep
109
Offline
admin approved badge
07:35 Oct-14-2013

Thanks.

0
Rep
250
Offline
admin approved badge
09:46 Oct-14-2013

CPU registers store the premature calculation result and make them available to the whole bus for reading writing and modifying. They are by far the fastest but are mostly of the size 32x 8bit. so only 32 kB.

0
Rep
250
Offline
admin approved badge
14:29 Oct-15-2013

*32 byte

0
Rep
460
Offline
admin approved badge
05:57 Oct-14-2013

Excellently written Shiv Kumar... You are only 14 but honestly, after seeing some of the materials that some of my classmates handed in previous semester, you outdo them in how well put together this is... You can teach those lazy a$$ 23 year olds a thing or two about how to put together a well worded piece...

0
Rep
157
Offline
admin approved badge
07:22 Oct-14-2013

GD has helped me lot in preparing this type of article. I would like to thanks GD and everyone should appreciate GD for being such a nice site rather than praising me. BTW My name is Shivam Kumar

0
Rep
55
Offline
02:11 Oct-14-2013

excellent explanation! eager for the upcoming part. but i thought you might able to just be a bit more detailed and give some example of cpu and compare it.

0
Rep
94
Offline
00:11 Oct-14-2013

Awesome! if you could make some graphs-images with comparisons it would be great too!

0
Rep
157
Offline
admin approved badge
07:20 Oct-14-2013

I will try this thing next time. It requires plenty of time, which I barely get these days.

0
Rep
2,003
Offline
admin approved badge
13:13 Oct-15-2013

Take your time to ensure the articles are as good as we're used to reading from you ;)

0
Rep
69
Offline
admin approved badge
23:42 Oct-13-2013

Very useful keep it going :)

0
Rep
3
Offline
23:15 Oct-13-2013

i3 2120 ?

0
Rep
3
Offline
17:16 Oct-13-2013

"Even if you walk extremely fast, if the road is lengthier, you will be slower." You probably have a good point there, but distance doesn't determine speed. If 100m takes 10 seconds to run, 200m wont take 30 sec (think in processor power).
or is it actually like that? :)

0
Rep
19
Offline
18:12 Oct-13-2013

he meas if the road is relatively lengthier than the shorter road and you are walking not fast enough to compensate for the longer distance then you have more lagg

0
Rep
157
Offline
admin approved badge
07:19 Oct-14-2013

Thanks for the explanation TheGHOSTyA.
@Mmaaaad What I mean by saying this is take the clock rate as your walking speed and the distance as the process lag. Let's compare you and me (just example). Your running speed is 20KM/h and mine is 30KM/h. You have to cover 40KM and I was required to cover 80KM. So, in this you will reach first despite of the slower speed. This is what I meant here. Hope all clear now, ;)

0
Rep
3
Offline
12:15 Oct-14-2013

All clear, over and out.

0
Rep
319
Offline
admin badge
17:08 Oct-13-2013

very nice. i study computer engineering and your explanation of the cache levels and usefullnes was an exact match to what i have been taught in university.

0
Rep
1
Offline
16:38 Oct-13-2013

So advanced.... Head is exploding

0
Rep
19
Offline
18:13 Oct-13-2013

i am 15 years old and i can keep up with it and i am not a nerd

0
Rep
-5
Offline
18:14 Oct-13-2013

im 7 old year and I yes to cpus

0
Rep
2,003
Offline
admin approved badge
13:12 Oct-15-2013

Epic :D

0
Rep
59
Offline
18:33 Oct-13-2013

same here man :D,im 15 and failed in 4 subjects, but i probably have more knowledge than my the best student in my class.

0
Rep
8
Offline
20:39 Oct-13-2013

A: there's a slight chance he was sarcastic (it's hard to tell in text. I think there should be a sacrastic front!"
and b: the one thing that is exploding for sure is your freaking egos!
p.s. "I yes to CPUs"? bro, do you even grammar?

0
Rep
319
Offline
admin badge
00:16 Oct-14-2013

lol i totally agree with aviviv

0
Rep
48
Offline
admin approved badge
16:31 Oct-13-2013

Will i get any better preformance on my GPU if i OC it from 900MHz to 1000 or 1050MHz?
And do i need to increase my fan speed to not burn it?

0
Rep
157
Offline
admin approved badge
16:35 Oct-13-2013

Yes, a slight difference in performance can be observed. About burning it depends. If the fan of your GFX card is sufficient in that condition then its fine otherwise you need to improve cooling. BTW This is not the place to ask this thing, you could have shouted this out or asked on the concerned card's page.

0
Rep
48
Offline
admin approved badge
16:40 Oct-13-2013

There should be a special place in GD for tehnical questions, idk where to ask. And thx for your answer

0
Rep
425
Offline
16:41 Oct-13-2013

on the HD7850 page thats where you should ask that question.

0
Rep
48
Offline
admin approved badge
16:46 Oct-13-2013

On GD HD7850 page, or official hd7850 page?

0
Rep
250
Offline
admin approved badge
16:51 Oct-13-2013

on the GD one, and there will be no real difference in performance with the overclock you are planning

0
Rep
19
Offline
18:14 Oct-13-2013
    it will be around 2-5 fps more in the best case                                             
0
Rep
48
Offline
admin approved badge
18:28 Oct-13-2013

No reason for doing it then :D

0
Rep
12
Offline
17:08 Oct-13-2013

We got the same gpu and I oc my just do manual fan speed to 100 and the gpu will stay cool

0
Rep
12
Offline
17:11 Oct-13-2013

I don't feel much performance increased by I of just because in some heavy games I do 100 fan speed so it wont overheat and I just also over clock it cuz its already at 100

0
Rep
425
Offline
20:22 Oct-13-2013

actually going 1050 will give you a 5-6 fps consistently..i have done extensie OCing with my HD7850 and i know it..

0
Rep
26
Offline
16:04 Oct-13-2013

Very nice article : very informative ! cheers!

0
Rep
-5
Offline
15:59 Oct-13-2013

TL; DR: cpus are thingies that process all the stuff done in your pc

0
Rep
250
Offline
admin approved badge
16:02 Oct-13-2013

Such infinite wisdom, why dont you just keep that to yourself

0
Rep
-5
Offline
16:07 Oct-13-2013

I want to share my infinite knowledge to everybody

0
Rep
425
Offline
16:08 Oct-13-2013

hey arent you the guy with the User ID of Interrogater i remember he had the same rig as yours.

0
Rep
48
Offline
admin approved badge
16:13 Oct-13-2013

This isnt Interrogator, this is his laptop!

0
Rep
-5
Offline
16:14 Oct-13-2013

Interrogater
Are you dyslexic?
This my other account btw


0
Rep
19
Offline
18:16 Oct-13-2013
you seem to be an analfabetic<br />Quote: this (is) my other account btw                                                    
0
Rep
-5
Offline
18:22 Oct-13-2013

sorry, was I talking to you? I dont think so, go stick your dick somewhere else.
Also I dont think you understand the meaning of analfabetism, If I was analfabetic, I couldnt even read your **** ass comment or type any of this stuff, so please grab your comment and stick it on your ass.

0
Rep
46
Offline
15:55 Oct-13-2013

Very informative article! Thank you for this!

0
Rep
425
Offline
15:16 Oct-13-2013

Already knew all the stuff but a well written article..:D

0
Rep
48
Offline
admin approved badge
14:38 Oct-13-2013

That was really useful man! Good job! Keep up the good work :D

0

Can They Run... |

Core i7-10750H 6-Core 2.60GHz GeForce RTX 2060 Mobile 16GB
| 60FPS, Medium, 720p
Core i5-4440 3.1GHz Radeon HD 6670 v2 Gigabyte OC 1GB Edition 16GB
| 60FPS, High, 720p
Core i3-1005G1 2-Core 1.20GHz UHD Graphics 630 4GB
| 30FPS, Low, 720p
Core i5-10400F 6-Core 2.90GHz Radeon RX 560 4GB 16GB
Ryzen 5 2600X 6-Core 3.6GHz Radeon RX 5600 XT Gigabyte Gaming OC 6GB 16GB
100% Yes [1 votes]
| 60FPS, High, 1080p
Core i7-10700 8-Core 2.90GHz GeForce RTX 2060 6GB 16GB
| 30FPS, Medium, 1080p
Ryzen 7 4800H 8-Core 2.9GHz GeForce GTX 1650 Ti Mobile 16GB
| 60FPS, High, 1080p
Core i5-9300H 4-Core 2.4GHz GeForce GTX 1650 8GB
| 60FPS, Ultra, 1080p
Ryzen 5 3600 6-Core 3.6GHz GeForce RTX 3060 16GB
50% Yes [2 votes]
| 60FPS, Low, 1080p
Core i5-3470 3.2GHz Radeon RX 470 Sapphire Nitro+ 8GB 16GB
100% Yes [1 votes]
| 60FPS, High, 1080p
Core i7-10870H 8-Core 2.20GHz GeForce RTX 2060 Asus ROG STRIX Gaming 6GB 16GB
100% Yes [2 votes]
| 60FPS, Medium, 1080p
Ryzen 5 3400G 4-Core 3.7GHz GeForce RTX 2060 6GB 16GB
| 60FPS, Ultra, 1080p
Ryzen 7 3750H 4-Core 2.3 GHz GeForce RTX 2060 Mobile 16GB
100% Yes [1 votes]
| 60FPS, Ultra, 1080p
Core i7-6800K 6-Core 3.4GHz GeForce GTX 1080 Asus ROG Strix Gaming OC 8GB Edition 32GB
100% Yes [2 votes]
| 60FPS, High, 1080p
Core i7-10700 8-Core 2.90GHz GeForce GTX 1050 Gigabyte D5 2GB 8GB
0% No [1 votes]
Ryzen 5 3600 6-Core 3.6GHz GeForce RTX 2070 Gigabyte Windforce 8GB 16GB
100% Yes [4 votes]
Ryzen 7 5800H 8-Core 3.2GHz GeForce RTX 3060 Mobile 16GB
100% Yes [2 votes]
| 60FPS, Ultra, 1440p
Ryzen 9 3900X 12-Core 3.8GHz GeForce RTX 3070 EVGA FTW3 Ultra Gaming 8GB 32GB
80% Yes [5 votes]
| 60FPS, Low, 720p
APU A8-7410 Quad-Core Radeon R5 7410 8GB
100% Yes [4 votes]
| 60FPS, Ultra, 1080p
Core i5-11400 6-Core 2.7GHz GeForce RTX 3050 Ti Mobile 8GB
0% No [1 votes]