How does it Work?

[frankthetank195]

This is pretty general thread, so I thought this would probably be the place to put it.

I just want to know, on a technical level, how all the components in a PC work. I know how to see which of two products is better, but I don't always know exactly why this is. So I'll put components I'd like to know about on a technical level and any specific question at the end of this post. I wanna know everything Anandtech knows. I learn quick, so don't be afraid to be very specific.

I'm doing this for two reasons. 1) I'm considering going for a career in computer engineering 2) I just would like to know.

CPU- Why is 22nm better than 32nm? What makes Intel for the most part better than AMD?
GPU- What is a Cuda core? What's the difference between Cuda and Stream Processor?
HDD- How does it read multiple platters? How do you increase capacity without increasing size?
SSD- What makes them faster?
What schools would be good for computer sciences?

All information is great, I wanna hear it all. Thanks guys.

 

[sperho]

I'll answer a few of the questions:

CPU- Why is 22nm better than 32nm? What makes Intel for the most part better than AMD?

22 nm lithography allows more transistors per unit area on the wafer, which enables more processing to occur on a similar-sized die than any larger lithography. Some of the space gains can be used for "System on a Chip" (SoC) applications. i.e. multiple ICs on the same die which are closer together allows for more rapid inter-IC communication. Also, in theory, smaller lithography allows for more efficient power usage.

HDD- How does it read multiple platters? How do you increase capacity without increasing size?

Multiple read heads operating in parallel enable this. Increased capacity can be achieved by making the physical bits smaller so the areal density is greater.

SSD- What makes them faster?

http://www.storagereview.com/ssd_vs_hdd in short: data can be read directly without a read head having to seek (which can take orders of magnitude more time than it does for an SSD to find the data) for the correct physical location. I'm sure others will chime in with more details.

 

[Blain]

All information is great, I wanna hear it all. Thanks guys.

Google is a good source for finding established information.

 

[piasabird]

22nm is like a process technology. Every time it gets thinner it theoretically could be using less power to do the same thing and therfore it should be able to run cooler and faster. Processor design is a very comples field of study. Intel also redesigns the chip designs to run more efficiently. You might read up on the Tick Tock design plan that MicrosoftIntel uses. When the Q6600 quad came out, all 4 cores basically ran at the same speed all the time. Then when Sandy Bridge came out, Intel introduced Turbo boost and techniques like load balancing between the cores allowing cores to speed up and slow down as needed saving energy and allowing for speeding up the processor as long as the thermal limits were not exceeded. So in essence Intel overclocked its own processors.

At the computex conference, Intel also announced they would do things in the future like Increase the size of the CPU Instruction Register size from 128 bits to 256 bits and send more instruction and somehow process them faster with (I Assume),more pipelines.

I have one computer with the 2500k processor and an Intel Z68 MB, but I purchased it because it had the better Intel HD3000 video (At the time). So I did not overclock. My wife liked it so we use that to watch Internet TV. She kind of stole it. So I am still using a core 2 duo 7200. It is still fast enough to do web surfing and stuff. I only play some really old games that will play on integrated video.

 

[cl-scott]

Then when Sandy Bridge came out, Intel introduced Turbo boost and techniques like load balancing between the cores allowing cores to speed up and slow down as needed saving energy and allowing for speeding up the processor as long as the thermal limits were not exceeded. So in essence Intel overclocked its own processors.

Just a minor note. Turbo Boost was introduced with the Core i series, of which the Sandy Bridge line is the second generation.

 

[dave_the_nerd]

Anandtech's review of the Radeon 2900 series did a pretty detailed explanation of Stream Processors.

Basically, Stream Processors / 5 = CUDA cores.

 

[mfenn]

Anandtech's review of the Radeon 2900 series did a pretty detailed explanation of Stream Processors.

Basically, Stream Processors / 5 = CUDA cores.

It should be noted that AMD's basic compute unit has radically changed with the 7000 series GCN architecture. To put it simply, it's a lot more like Nvidia's.

 

[mfenn]

This is pretty general thread, so I thought this would probably be the place to put it.

I just want to know, on a technical level, how all the components in a PC work. I know how to see which of two products is better, but I don't always know exactly why this is. So I'll put components I'd like to know about on a technical level and any specific question at the end of this post. I wanna know everything Anandtech knows. I learn quick, so don't be afraid to be very specific.

I'm doing this for two reasons. 1) I'm considering going for a career in computer engineering 2) I just would like to know.

CPU- Why is 22nm better than 32nm? What makes Intel for the most part better than AMD?
GPU- What is a Cuda core? What's the difference between Cuda and Stream Processor?
HDD- How does it read multiple platters? How do you increase capacity without increasing size?
SSD- What makes them faster?
What schools would be good for computer sciences?

All information is great, I wanna hear it all. Thanks guys.

I would recommend going through Anandtech's review archives and reading all the big announcement reviews (Intel and AMD CPUs, AMD and Nvidia GPUs, SSDs) from the past few years. They always go into depth about the architecture and why it matters.

 

[frankthetank195]

That was perfect, thanks guys. Never too much knowledge