Overclock.net › Forums › General Hardware › General Processor Discussions › What is advancing in CPUs?
New Posts  All Forums:Forum Nav:

What is advancing in CPUs?

post #1 of 11
Thread Starter 
Hi, I am elemein, I am currently trying to give myself a crash course on trying to learn about computers and how they work and probably will be doing this for a year or two until I gain a very good understanding (I've gotten good at self-learning things after I learned more about cars than I'll ever need in my life last year...)

Anyway, I am curious about a few things:

Mainly, what gets better every year in PCs?

Now, I am going to keep this restricted to silicon chips as they are the primary chips I am trying to learn about.

Anyway; the reason I ask this question is mainly because every year I only see an increase of a few things:

1: Clock Speed & 2: Number of Cores

Now, clock speed, to my understanding, is the interval in which a CPU processes information and fires off electrical signals throughout the die in the CPU with certain inputs of information in order to get certain outputs; those outputs are given to other components which I will learn about later (I'm focusing on processors at the moment.) So basically, the faster the clock speed, the higher the amount of intervals done in a specific amount of time.

Now, faster clock speed obviously results in a faster computer; that makes sense.

Now, what I also see is the number of cores increasing. This is, to my understanding, multiple processor sets put into one die; so that a multitude of intervals can be completed in the same amount of time.

Now that's all fine and dandy, but why aren't the number of transistors increased? I understand that the industry has moved from the 45 nm transistor the the 32 nm transistor lately (probably even smaller by now), but why are the numbers never advertised? To my understanding (what this particular video teaches me: http://www.youtube.com/watch?v=VBDoT8o4q00&feature=related ), having twice the amount of transistors allows for twice the amount of instructions to be completed in each interval; no? Logically, if two chips are identical except one uses 64nm transistors and the other uses 32nm transistors (therefore the 32nm can fit twice the number of transistors on the chip), wouldnt the 32nm CPU be able to do twice the amount of work in the same time? I dont understand why this stuff isnt advertised as much as it should be... Usually only clock speed and cores are advertised. Am I wrong in my understanding?

Also, one quick question; this video teaches me that every character has a corresponding binary code (am I correct?), is the interpretation and production of these codes exactly what the CPU's purpose is to do?

Sorry for all the questions; I'm just trying to learn... smile.gif Thank you all who reply!
Little Monster
(9 items)
 
   
CPUMotherboardGraphicsRAM
i7-4710MQ Intel HM87 * Clevo nVidia GTX 860M 8 GB x 1 @ 1600 MHz 
Hard DriveOSMonitorKeyboard
500 GB 5400 RPM Hitachi Windows 8.1 13.3" IPS 1920x1080p Full Backlit Keyboard 
Mouse
Microsoft Mouse 4000 
CPUGraphicsGraphicsRAM
AMD A8-3520M X4 with Radeon HD Graphics AMD Radeon HD 7450M AMD Radeon HD 6620G Samsung 2+4 GB RAM 
Hard DriveOS
Toshiba MK6475GSX ATA Windows 7 Enterprise 
  hide details  
Reply
Little Monster
(9 items)
 
   
CPUMotherboardGraphicsRAM
i7-4710MQ Intel HM87 * Clevo nVidia GTX 860M 8 GB x 1 @ 1600 MHz 
Hard DriveOSMonitorKeyboard
500 GB 5400 RPM Hitachi Windows 8.1 13.3" IPS 1920x1080p Full Backlit Keyboard 
Mouse
Microsoft Mouse 4000 
CPUGraphicsGraphicsRAM
AMD A8-3520M X4 with Radeon HD Graphics AMD Radeon HD 7450M AMD Radeon HD 6620G Samsung 2+4 GB RAM 
Hard DriveOS
Toshiba MK6475GSX ATA Windows 7 Enterprise 
  hide details  
Reply
post #2 of 11
My understanding is probably not as great as some other of the great people on these forums, but considering you're starting out learning things I feel I can help.
First thing when it comes to processors you need to understand is that clock speeds don't necessarily make for a faster processor when compared to a chip with a different architecture. Let's give an example:
On your way to becoming a computer master you might have heard about the competition (or lack there of) between intel and AMD.
At this point AMD is struggling to compete with intel in the mid-high performance range as far as processor speed/power goes. This is because even though AMD's new chips have 8 cores and a higher clock speed, their architecture and manufacturing process is inferior to that of intel. AMD is sticking with a 32nm manufacturing process while intel has shifted to a 22nm process with their new ivy-bridge line of CPUs which were released this year. This "shrinking of the die" allows for temperature reductions as well as power usage reductions.
But enough of that, let's move back to clock speeds.
To reiterate, clock speeds should not be a deciding factor when comparing AMD and intel at all. HOWEVER if 2 of the same processors are compared, with one running "overclocked" for example, the overclocked processor will, obviously, run faster. Clock speeds are also used when comparing generational successors to a previous chip. For example, it wouldn't be uncommon to compare the "clock for clock" difference from the previous generation chip to the new one. ie: 8120 comapred to 8320 clock for clock.
As far as number of cores goes, having a higher number of cores can be beneficial to a degree. For example, if you're a gamer you typically won't need more than 4 cores. Most games utilize 1-2 cores (Although this is slowly changing). You don't typically see 1-core processors anymore because 2 is so much more efficient. Although when it comes to cores too many could prove useless. Most applications use 1-2 cores or even maybe 4 with newer games and applications. However an 8 core processor like the ones at the top end of AMD's FX linecan be utilized when it comes to heavily "threaded" (using a lot of cores) applications like those used for rendering of video and such.
I hope this at least helped you out a little bit. Knowledge is power! thumb.gif
     
CPUMotherboardGraphicsRAM
i5 2500k @ 4.0ghz Asrock Z68 Extreme 3 Gen 3 Diamond HD 6950 2gb @ 900MHz core 1400MHz mem Gskill Ripjaws 8gb 1600MHz 
Hard DriveOptical DriveCoolingOS
Hitatchi 500gb Liteon DVD 22x Arctic Cooling i30 Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
Acer 23" 1080p 5ms response Adesso MKB-135B Mechanical keyboard hec X-Power Pro 650 Antec Three Hundred 
MouseAudio
Engage Wired Optical Mouse Plantronics Gamecom 780 
CPUMotherboardGraphicsRAM
RP2803 1.79Mhz NES motherboard RP2C02 5.37MHz 2KB onboard RAM 
Optical DriveCaseMouse
Cartridge Reader NES case NES controller 
CPUMotherboardRAMHard Drive
A10 5800k stock Gigabyte A85X Crucial ballistix 8gb @ 1866MHz Western Digital 250gb 
Optical DriveCoolingOSPower
Gateway DVD drive Stock fan Windows 7 Home Premium 64-bit corsair Builder series 430W 
Case
Raidmax Tornado Blue 
  hide details  
Reply
     
CPUMotherboardGraphicsRAM
i5 2500k @ 4.0ghz Asrock Z68 Extreme 3 Gen 3 Diamond HD 6950 2gb @ 900MHz core 1400MHz mem Gskill Ripjaws 8gb 1600MHz 
Hard DriveOptical DriveCoolingOS
Hitatchi 500gb Liteon DVD 22x Arctic Cooling i30 Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
Acer 23" 1080p 5ms response Adesso MKB-135B Mechanical keyboard hec X-Power Pro 650 Antec Three Hundred 
MouseAudio
Engage Wired Optical Mouse Plantronics Gamecom 780 
CPUMotherboardGraphicsRAM
RP2803 1.79Mhz NES motherboard RP2C02 5.37MHz 2KB onboard RAM 
Optical DriveCaseMouse
Cartridge Reader NES case NES controller 
CPUMotherboardRAMHard Drive
A10 5800k stock Gigabyte A85X Crucial ballistix 8gb @ 1866MHz Western Digital 250gb 
Optical DriveCoolingOSPower
Gateway DVD drive Stock fan Windows 7 Home Premium 64-bit corsair Builder series 430W 
Case
Raidmax Tornado Blue 
  hide details  
Reply
post #3 of 11
Quote:
Originally Posted by elemein View Post

Hi, I am elemein, I am currently trying to give myself a crash course on trying to learn about computers and how they work and probably will be doing this for a year or two until I gain a very good understanding (I've gotten good at self-learning things after I learned more about cars than I'll ever need in my life last year...)
Anyway, I am curious about a few things:
Mainly, what gets better every year in PCs?
Now, I am going to keep this restricted to silicon chips as they are the primary chips I am trying to learn about.
Anyway; the reason I ask this question is mainly because every year I only see an increase of a few things:
1: Clock Speed & 2: Number of Cores
Now, clock speed, to my understanding, is the interval in which a CPU processes information and fires off electrical signals throughout the die in the CPU with certain inputs of information in order to get certain outputs; those outputs are given to other components which I will learn about later (I'm focusing on processors at the moment.) So basically, the faster the clock speed, the higher the amount of intervals done in a specific amount of time.
Now, faster clock speed obviously results in a faster computer; that makes sense.
Now, what I also see is the number of cores increasing. This is, to my understanding, multiple processor sets put into one die; so that a multitude of intervals can be completed in the same amount of time.
Now that's all fine and dandy, but why aren't the number of transistors increased? I understand that the industry has moved from the 45 nm transistor the the 32 nm transistor lately (probably even smaller by now), but why are the numbers never advertised? To my understanding (what this particular video teaches me: http://www.youtube.com/watch?v=VBDoT8o4q00&feature=related ), having twice the amount of transistors allows for twice the amount of instructions to be completed in each interval; no? Logically, if two chips are identical except one uses 64nm transistors and the other uses 32nm transistors (therefore the 32nm can fit twice the number of transistors on the chip), wouldnt the 32nm CPU be able to do twice the amount of work in the same time? I dont understand why this stuff isnt advertised as much as it should be... Usually only clock speed and cores are advertised. Am I wrong in my understanding?
Also, one quick question; this video teaches me that every character has a corresponding binary code (am I correct?), is the interpretation and production of these codes exactly what the CPU's purpose is to do?
Sorry for all the questions; I'm just trying to learn... smile.gif Thank you all who reply!


God bless wiki.....

CPU tech is not as simple as you put it.....Even cars, not all engines have the same power even if they have the same number of cylinders and same capacity....comparing cars just because they have the same sized engine is kinda childish and not at all logical, and you're doing it to CPUs......unholy dude. And no matter how 'fast' the CPU is, if the application(s) can't take advantage of the new tech/threads, the results will be slower than expected.

Read reviews on different CPUs like Guru3d or Tomshardware

2 Main rivals at the moment (not including ARM) which are Intel & AMD. They both provide different CPUs at different prices for whatever need you have. Just don't forget that CPUs are only a small part of the system......
TheSadFlute
(16 items)
 
DyingHeart
(17 items)
 
 
CPUMotherboardGraphicsRAM
i7-6700K TUF Sabertooth MSI 1080 Ti 16GB 2400MHz C15 
Hard DriveHard DriveHard DriveCooling
Corsair 480GB LE 3TB WD Green Sammy 500GB 850 EVO H100i 
OSMonitorKeyboardPower
Windows 7 Pro  BenQ XR3501 Logitech G710+ EVGA G2 750W 
CaseMouseAudioAudio
Cooler Master Stryker Logitech G602 Asus Strix Nakamichi 5.1 
CPUMotherboardGraphicsRAM
Phenom II X4 965BE  Sapphire 790X Zotac GTX650  4GB Corsair 1066 DDR2 (1GBx4) 
Hard DriveHard DriveCoolingCooling
G.Skill Phoenix Pro 120GB WD 2TB Green Stock 3x120mm UQ fans 
OSMonitorKeyboardPower
W7 Pro SP1 x64 LG 21' 1680x1050 Genius  Corsair CX600 
CaseMouseMouse PadAudio
Gigabyte  E-Blue Mazer XFX WarPad On-board 
Audio
Logitech G35 
  hide details  
Reply
TheSadFlute
(16 items)
 
DyingHeart
(17 items)
 
 
CPUMotherboardGraphicsRAM
i7-6700K TUF Sabertooth MSI 1080 Ti 16GB 2400MHz C15 
Hard DriveHard DriveHard DriveCooling
Corsair 480GB LE 3TB WD Green Sammy 500GB 850 EVO H100i 
OSMonitorKeyboardPower
Windows 7 Pro  BenQ XR3501 Logitech G710+ EVGA G2 750W 
CaseMouseAudioAudio
Cooler Master Stryker Logitech G602 Asus Strix Nakamichi 5.1 
CPUMotherboardGraphicsRAM
Phenom II X4 965BE  Sapphire 790X Zotac GTX650  4GB Corsair 1066 DDR2 (1GBx4) 
Hard DriveHard DriveCoolingCooling
G.Skill Phoenix Pro 120GB WD 2TB Green Stock 3x120mm UQ fans 
OSMonitorKeyboardPower
W7 Pro SP1 x64 LG 21' 1680x1050 Genius  Corsair CX600 
CaseMouseMouse PadAudio
Gigabyte  E-Blue Mazer XFX WarPad On-board 
Audio
Logitech G35 
  hide details  
Reply
post #4 of 11
Thread Starter 
Quote:
Originally Posted by eXXon View Post

God bless wiki.....
CPU tech is not as simple as you put it.....Even cars, not all engines have the same power even if they have the same number of cylinders and same capacity....comparing cars just because they have the same sized engine is kinda childish and not at all logical, and you're doing it to CPUs......unholy dude. And no matter how 'fast' the CPU is, if the application(s) can't take advantage of the new tech/threads, the results will be slower than expected.
Read reviews on different CPUs like Guru3d or Tomshardware
2 Main rivals at the moment (not including ARM) which are Intel & AMD. They both provide different CPUs at different prices for whatever need you have. Just don't forget that CPUs are only a small part of the system......

I see what you guys are meaning; and I like how you related it to cars. You're correct; capacity of an engine does dictate how much potential for power you can make; but airflow characteristics, thermal efficiency, volumetric efficiency, compression efficiency and many other factors determine what comes out of the crank...

I suppose the layout of the chip is also important; or what you guys would call "architecture"... I personally am not really sure what differences in architecture mean; just by my own guess, I would assume the difference would be in how binary codes are interpreted and used or perhaps how certain patterns are used in the die... I am not really sure seeing as I've only doused myself in this stuff for a little under a week...

I will definitely look at the wiki and reviews! Thanks everyone!
Little Monster
(9 items)
 
   
CPUMotherboardGraphicsRAM
i7-4710MQ Intel HM87 * Clevo nVidia GTX 860M 8 GB x 1 @ 1600 MHz 
Hard DriveOSMonitorKeyboard
500 GB 5400 RPM Hitachi Windows 8.1 13.3" IPS 1920x1080p Full Backlit Keyboard 
Mouse
Microsoft Mouse 4000 
CPUGraphicsGraphicsRAM
AMD A8-3520M X4 with Radeon HD Graphics AMD Radeon HD 7450M AMD Radeon HD 6620G Samsung 2+4 GB RAM 
Hard DriveOS
Toshiba MK6475GSX ATA Windows 7 Enterprise 
  hide details  
Reply
Little Monster
(9 items)
 
   
CPUMotherboardGraphicsRAM
i7-4710MQ Intel HM87 * Clevo nVidia GTX 860M 8 GB x 1 @ 1600 MHz 
Hard DriveOSMonitorKeyboard
500 GB 5400 RPM Hitachi Windows 8.1 13.3" IPS 1920x1080p Full Backlit Keyboard 
Mouse
Microsoft Mouse 4000 
CPUGraphicsGraphicsRAM
AMD A8-3520M X4 with Radeon HD Graphics AMD Radeon HD 7450M AMD Radeon HD 6620G Samsung 2+4 GB RAM 
Hard DriveOS
Toshiba MK6475GSX ATA Windows 7 Enterprise 
  hide details  
Reply
post #5 of 11
I'm going to put it as simply as I can.

Intel has a much larger R&D budget than AMD, and therefore they can consistently put out better chips.
Max & Sandy
(17 items)
 
Secondary
(15 items)
 
XP Machine
(9 items)
 
CPUMotherboardGraphicsRAM
i7 2700K @ 4.6GHz 1.4V Gigabyte GA-Z77X-UD3H MSI 980Ti Gaming Edition Corsair Vengeance 16GB 
Hard DriveCoolingCoolingCooling
Crucial M550 1TB XSPC Raystorm EK FC 980Ti TF5 XSPC D5 Bay Res 
CoolingCoolingCoolingCooling
Swiftech MCP655-B XSPC EX360 XSPC EX240 5x Sythe Gentle Typhoon AP-15 
CoolingOSMonitorPower
Primochill Primoflex Advanced LRT Windows 7 Ultimate Yamakasi Catleap Q270 Cooler Master V1000 
Case
NZXT Switch 810 
CPUMotherboardGraphicsGraphics
Phenom II X6 1090T ASUS Sabertooth 990FX HIS 6850 HIS 6850 
RAMHard DriveHard DriveHard Drive
G Skill Ares 32GB OCZ Vertex 4 256GB WD Black 1TB WD Green 1TB 
Hard DriveOptical DriveOptical DriveCooling
Seagate Barracuda 1TB ASUS Sony Optiarc Corsair H80 
OSPowerCase
Windows 7 Ultimate Corsair TX750 Cooler Master Storm Scout 
CPUMotherboardGraphicsRAM
Core 2 Duo @ 2.8GHz Some Intel PoS XFX GeForce 9600GSO 2x 2GB Kingston 800MHz 
Hard DriveOptical DriveOSPower
Seagate 750GB Plextor Windows XP SP3 Cooler Master SIlent Pro 600 
Case
Dell Inspiron 
  hide details  
Reply
Max & Sandy
(17 items)
 
Secondary
(15 items)
 
XP Machine
(9 items)
 
CPUMotherboardGraphicsRAM
i7 2700K @ 4.6GHz 1.4V Gigabyte GA-Z77X-UD3H MSI 980Ti Gaming Edition Corsair Vengeance 16GB 
Hard DriveCoolingCoolingCooling
Crucial M550 1TB XSPC Raystorm EK FC 980Ti TF5 XSPC D5 Bay Res 
CoolingCoolingCoolingCooling
Swiftech MCP655-B XSPC EX360 XSPC EX240 5x Sythe Gentle Typhoon AP-15 
CoolingOSMonitorPower
Primochill Primoflex Advanced LRT Windows 7 Ultimate Yamakasi Catleap Q270 Cooler Master V1000 
Case
NZXT Switch 810 
CPUMotherboardGraphicsGraphics
Phenom II X6 1090T ASUS Sabertooth 990FX HIS 6850 HIS 6850 
RAMHard DriveHard DriveHard Drive
G Skill Ares 32GB OCZ Vertex 4 256GB WD Black 1TB WD Green 1TB 
Hard DriveOptical DriveOptical DriveCooling
Seagate Barracuda 1TB ASUS Sony Optiarc Corsair H80 
OSPowerCase
Windows 7 Ultimate Corsair TX750 Cooler Master Storm Scout 
CPUMotherboardGraphicsRAM
Core 2 Duo @ 2.8GHz Some Intel PoS XFX GeForce 9600GSO 2x 2GB Kingston 800MHz 
Hard DriveOptical DriveOSPower
Seagate 750GB Plextor Windows XP SP3 Cooler Master SIlent Pro 600 
Case
Dell Inspiron 
  hide details  
Reply
post #6 of 11
Smaller transistors can make you put more transistors on the same chip.
Basically you can use those for bigger caches, than can improve performance a lot. Another option is to transform your extra transistor in new cores.

Basically that's it.

The architecture can be very complex, but you can understand it in a simple way:

Imagine you have two numbers and you want to get their sum. How do you that? Well, there can be many ways, depending on your resources. But basically that is pretty simple.
Now what about multiplying those two numbers? Many more options on how you can do that. Since this math is done by the CPU, there can be many ways of doing it in there, and this is your architecture.

Of course that is not everything, but basically is how you do things, how do you use your resources: in this case your millions or billions of transistors.

Another way of looking at the architecture is where you put things. For example, if you are going to put your cores close to their cache or close to each other? Are they going to share resources or have their own?

Will you have 32 bit registers or 64 bit ones? There are many things related to architecture.

As for the importance of developing better architecture:
Intel suffered in the past for creating new interfaces with new ways of doing things. The problem is you need to write software capable of taking advantage of that new way of doing it:
it's of no use to write an * operand to your CPU if all software written only uses the + operand (for example).
AMD had little advances in benchmarks with bulldozer because of this very same thing, which could have been easily predicted, as far as I know.


There is also the Moore's Law thing, that says computer processors double in complexity every two years. Some say that we are close to the limit of this law because the components are becoming way to small, so small it will be impossible to manipulate them. This opens the possibility for the development of new technologies for processors, that are not based on silicon.

So far this law has been correct. However, my opinion is that, in the future we will either slow down completely or have a great jump in this velocity.
I would bet my money on the later, specially if it's backwards compatible with the current technology, because that would mean easy money for whoever own some patents biggrin.gif
post #7 of 11
Warning: Spoiler! (Click to show)
Quote:
Originally Posted by EduFurtado View Post

Smaller transistors can make you put more transistors on the same chip.
Basically you can use those for bigger caches, than can improve performance a lot. Another option is to transform your extra transistor in new cores.
Basically that's it.
The architecture can be very complex, but you can understand it in a simple way:
Imagine you have two numbers and you want to get their sum. How do you that? Well, there can be many ways, depending on your resources. But basically that is pretty simple.
Now what about multiplying those two numbers? Many more options on how you can do that. Since this math is done by the CPU, there can be many ways of doing it in there, and this is your architecture.
Of course that is not everything, but basically is how you do things, how do you use your resources: in this case your millions or billions of transistors.
Another way of looking at the architecture is where you put things. For example, if you are going to put your cores close to their cache or close to each other? Are they going to share resources or have their own?
Will you have 32 bit registers or 64 bit ones? There are many things related to architecture.
As for the importance of developing better architecture:
Intel suffered in the past for creating new interfaces with new ways of doing things. The problem is you need to write software capable of taking advantage of that new way of doing it:
it's of no use to write an * operand to your CPU if all software written only uses the + operand (for example).
AMD had little advances in benchmarks with bulldozer because of this very same thing, which could have been easily predicted, as far as I know.
There is also the Moore's Law thing, that says computer processors double in complexity every two years. Some say that we are close to the limit of this law because the components are becoming way to small, so small it will be impossible to manipulate them. This opens the possibility for the development of new technologies for processors, that are not based on silicon.
So far this law has been correct. However, my opinion is that, in the future we will either slow down completely or have a great jump in this velocity.
I would bet my money on the later, specially if it's backwards compatible with the current technology, because that would mean easy money for whoever own some patents biggrin.gif

Nice way of explaining things thumb.gif

One more, if you're wondering why AMD is slower than Intel in cases where it shouldn't, they both have patents on their CPUs tech and as you guessed, Intel has waaay more than AMD...
TheSadFlute
(16 items)
 
DyingHeart
(17 items)
 
 
CPUMotherboardGraphicsRAM
i7-6700K TUF Sabertooth MSI 1080 Ti 16GB 2400MHz C15 
Hard DriveHard DriveHard DriveCooling
Corsair 480GB LE 3TB WD Green Sammy 500GB 850 EVO H100i 
OSMonitorKeyboardPower
Windows 7 Pro  BenQ XR3501 Logitech G710+ EVGA G2 750W 
CaseMouseAudioAudio
Cooler Master Stryker Logitech G602 Asus Strix Nakamichi 5.1 
CPUMotherboardGraphicsRAM
Phenom II X4 965BE  Sapphire 790X Zotac GTX650  4GB Corsair 1066 DDR2 (1GBx4) 
Hard DriveHard DriveCoolingCooling
G.Skill Phoenix Pro 120GB WD 2TB Green Stock 3x120mm UQ fans 
OSMonitorKeyboardPower
W7 Pro SP1 x64 LG 21' 1680x1050 Genius  Corsair CX600 
CaseMouseMouse PadAudio
Gigabyte  E-Blue Mazer XFX WarPad On-board 
Audio
Logitech G35 
  hide details  
Reply
TheSadFlute
(16 items)
 
DyingHeart
(17 items)
 
 
CPUMotherboardGraphicsRAM
i7-6700K TUF Sabertooth MSI 1080 Ti 16GB 2400MHz C15 
Hard DriveHard DriveHard DriveCooling
Corsair 480GB LE 3TB WD Green Sammy 500GB 850 EVO H100i 
OSMonitorKeyboardPower
Windows 7 Pro  BenQ XR3501 Logitech G710+ EVGA G2 750W 
CaseMouseAudioAudio
Cooler Master Stryker Logitech G602 Asus Strix Nakamichi 5.1 
CPUMotherboardGraphicsRAM
Phenom II X4 965BE  Sapphire 790X Zotac GTX650  4GB Corsair 1066 DDR2 (1GBx4) 
Hard DriveHard DriveCoolingCooling
G.Skill Phoenix Pro 120GB WD 2TB Green Stock 3x120mm UQ fans 
OSMonitorKeyboardPower
W7 Pro SP1 x64 LG 21' 1680x1050 Genius  Corsair CX600 
CaseMouseMouse PadAudio
Gigabyte  E-Blue Mazer XFX WarPad On-board 
Audio
Logitech G35 
  hide details  
Reply
post #8 of 11
Thread Starter 
Quote:
Originally Posted by EduFurtado View Post

Smaller transistors can make you put more transistors on the same chip.
Basically you can use those for bigger caches, than can improve performance a lot. Another option is to transform your extra transistor in new cores.
Basically that's it.
The architecture can be very complex, but you can understand it in a simple way:
Imagine you have two numbers and you want to get their sum. How do you that? Well, there can be many ways, depending on your resources. But basically that is pretty simple.
Now what about multiplying those two numbers? Many more options on how you can do that. Since this math is done by the CPU, there can be many ways of doing it in there, and this is your architecture.
Of course that is not everything, but basically is how you do things, how do you use your resources: in this case your millions or billions of transistors.
Another way of looking at the architecture is where you put things. For example, if you are going to put your cores close to their cache or close to each other? Are they going to share resources or have their own?
Will you have 32 bit registers or 64 bit ones? There are many things related to architecture.
As for the importance of developing better architecture:
Intel suffered in the past for creating new interfaces with new ways of doing things. The problem is you need to write software capable of taking advantage of that new way of doing it:
it's of no use to write an * operand to your CPU if all software written only uses the + operand (for example).
AMD had little advances in benchmarks with bulldozer because of this very same thing, which could have been easily predicted, as far as I know.
There is also the Moore's Law thing, that says computer processors double in complexity every two years. Some say that we are close to the limit of this law because the components are becoming way to small, so small it will be impossible to manipulate them. This opens the possibility for the development of new technologies for processors, that are not based on silicon.
So far this law has been correct. However, my opinion is that, in the future we will either slow down completely or have a great jump in this velocity.
I would bet my money on the later, specially if it's backwards compatible with the current technology, because that would mean easy money for whoever own some patents biggrin.gif

Mmmm this architecture thing seems to play a far larger role than first thought. I definitely need to look into it more.
Little Monster
(9 items)
 
   
CPUMotherboardGraphicsRAM
i7-4710MQ Intel HM87 * Clevo nVidia GTX 860M 8 GB x 1 @ 1600 MHz 
Hard DriveOSMonitorKeyboard
500 GB 5400 RPM Hitachi Windows 8.1 13.3" IPS 1920x1080p Full Backlit Keyboard 
Mouse
Microsoft Mouse 4000 
CPUGraphicsGraphicsRAM
AMD A8-3520M X4 with Radeon HD Graphics AMD Radeon HD 7450M AMD Radeon HD 6620G Samsung 2+4 GB RAM 
Hard DriveOS
Toshiba MK6475GSX ATA Windows 7 Enterprise 
  hide details  
Reply
Little Monster
(9 items)
 
   
CPUMotherboardGraphicsRAM
i7-4710MQ Intel HM87 * Clevo nVidia GTX 860M 8 GB x 1 @ 1600 MHz 
Hard DriveOSMonitorKeyboard
500 GB 5400 RPM Hitachi Windows 8.1 13.3" IPS 1920x1080p Full Backlit Keyboard 
Mouse
Microsoft Mouse 4000 
CPUGraphicsGraphicsRAM
AMD A8-3520M X4 with Radeon HD Graphics AMD Radeon HD 7450M AMD Radeon HD 6620G Samsung 2+4 GB RAM 
Hard DriveOS
Toshiba MK6475GSX ATA Windows 7 Enterprise 
  hide details  
Reply
post #9 of 11
Quote:
Originally Posted by elemein View Post

Mmmm this architecture thing seems to play a far larger role than first thought. I definitely need to look into it more.

Well architecture is actually only a tenth of the CPU system. Architecture is mainly improved do to improvements in transistor design. Better transistors with less leakage means that you can cram more onto a CPU and increase the performance. Then once you have packed as many transistors into a core as possible, THEN you worry about architecture and adding more cores and more threads and more and more and more.......etc.

As for learning on how a CPU works, in short. It is a BUNCH of transistors that can act as switches for electrical signals. I think this video not only shows how technology is moving forward, but also some of the very basic principals of how computers work.

http://www.youtube.com/watch?v=65XpeuLA4Rg&feature=relmfu
Yin&Yang
(19 items)
 
Soon
(15 items)
 
 
CPUMotherboardGraphicsRAM
i7 4820K Gigabyte X79 UP4 GTX 660 (waiting for new AMD cards) Samsung green  
Hard DriveHard DriveCoolingOS
Adata SP900  Seagate 2TB hard drive (the power of 1 edition)  Custom WC Loop Linux (various versions) 
MonitorKeyboardPowerCase
Samsung SyncMaster T240HD Coolermaster XT Green switches XFX 1000w Corsair 800D 
MouseMouse PadAudioAudio
Razer Naga Epic 2014 edition Razer Vespula Stax Lambda Nova Classic Custom DIY DAC 
AudioAudioOther
300b PP monoblocks Snell J3 NZXT Hue 
CPUCPUMotherboardGraphics
8 core xeon E6 @ 3.5ghz 8 core xeon E6 @ 3.5ghz EVGA X99 SR?  AMD 390x 
GraphicsRAMHard DriveCooling
AMD 390x 128 GB ECC DDR4  1TB PCI-e SSD Custom WC Loop 
OSMonitorKeyboardPower
Linux 16:10 Samsung PLS 4K monitor Corsair RGB keyboard with custom cherry MX supe... EVGA 1600 watt titanium  
CaseMouseAudio
Caselabs Magnum STH10 DIY my own mouse Snell J 3 
  hide details  
Reply
Yin&Yang
(19 items)
 
Soon
(15 items)
 
 
CPUMotherboardGraphicsRAM
i7 4820K Gigabyte X79 UP4 GTX 660 (waiting for new AMD cards) Samsung green  
Hard DriveHard DriveCoolingOS
Adata SP900  Seagate 2TB hard drive (the power of 1 edition)  Custom WC Loop Linux (various versions) 
MonitorKeyboardPowerCase
Samsung SyncMaster T240HD Coolermaster XT Green switches XFX 1000w Corsair 800D 
MouseMouse PadAudioAudio
Razer Naga Epic 2014 edition Razer Vespula Stax Lambda Nova Classic Custom DIY DAC 
AudioAudioOther
300b PP monoblocks Snell J3 NZXT Hue 
CPUCPUMotherboardGraphics
8 core xeon E6 @ 3.5ghz 8 core xeon E6 @ 3.5ghz EVGA X99 SR?  AMD 390x 
GraphicsRAMHard DriveCooling
AMD 390x 128 GB ECC DDR4  1TB PCI-e SSD Custom WC Loop 
OSMonitorKeyboardPower
Linux 16:10 Samsung PLS 4K monitor Corsair RGB keyboard with custom cherry MX supe... EVGA 1600 watt titanium  
CaseMouseAudio
Caselabs Magnum STH10 DIY my own mouse Snell J 3 
  hide details  
Reply
post #10 of 11
Quote:
Originally Posted by elemein View Post

Mmmm this architecture thing seems to play a far larger role than first thought. I definitely need to look into it more.

Something about architecture that I heard and I didn't verify if it was true is that when you don't know what to do with your extra transistors you turn it into cache, which will increase performance in every scenario. But it seems that the cache is also the part of the chip that generates heat the most!

This is what I wanted to check if was true or not.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: General Processor Discussions
Overclock.net › Forums › General Hardware › General Processor Discussions › What is advancing in CPUs?