Difference between CPU and processor

Assguard, shut up now.

A GPU, if you weren't to lazy to Google, or actually perhaps knew anything about anything, instead of consistently embarrassing yourself, is a GRAPHICS processor. It does matter, because you posted that you THINK I was wrong, whereas in actual fact you're an imbecile who won't save himself from looking foolish by searching the web.

Now please, go learn to spell.

 

Log in or Sign up to hide all adverts.

And it is still a processor so where is he wrong?

 

Log in or Sign up to hide all adverts.

He said 'when computers first came out there was a different "maths" proccesor to the CPU'.

Which is bullshit. Co-processors are a later addition to computers, most prevalent on 386 machines. While a 386 might have been Assguards first exposure to computing, they were not the first computers to 'come out'.

Second, he asserted GPU stood for 'Game Processor', which is bullshit.

Of course, if you had just re-read the thread, all this would have been apparent to you.

 

Log in or Sign up to hide all adverts.

You are the biggest joke that posts in this forum man.

 

That's your riposte? A cheap ad-hom?

No discussion of the fact that DEC made virtual memory a standard and must have, as market leaders? No discussion that DEC was the leader in Clustering technology? No discussion that the DEC Alpha chip was the first cheaply available 64bit chip, and the first to be available in a desktop computer(in 1992, pipping Intel by some margin!)? No mention of the PDP-11 RTOS still being used today? No mention that Patriot Missiles used MicroVAX computers as their guidance systems, and that NORAD used DEC equipment?

Most of the above, before Sun Microsystems existed as a company.

Don't get me wrong, I used to earn a living administrating Sun Solaris servers, I like the OS, but SPARC has not always been the industry leader. Maybe if you had more experience in IT you know that Sun products weren't always that robust, Solaris 2.3 took longer to patch after an install, than to install! That was 1993, btw.

Now, in industry, I see more Linux than Solaris deployed. The hardware is cheaper, the hardware is more familiar to most tecchies, and with Intel SMP options cheaply available, there is plenty of power to be harnessed, and it's no surprise that Sun themselves are offering more and more Intel based servers, over SPARC. Also, that SPARC processors only feature ONCE in the top 500 supercomputers list, so they aren't dominating the bottom end, nor the high end market, and are a bit flaccid in the midrange too.

Hardly 'market leaders', therefore.

But please, don't bother rebutting my rebuttal of your claim, let's have another 'witty' ad-hom!

 

all math in early computers was handled exclusively by software.
todays CPUs have limited hardware math capability.
the math coprocessor was introduced to primarily aid in floating point math.

i really didn't know how to answer the question without confusing him more.
there are indeed several "processors" inside todays computers.
the CPU itself is sometimes referred to as a "processor"

 

ahhh yes. the old FPU.

 

Don't mention the FPU! You'll only confuse Asguard!

 

phlogistician's cpu history is accurate, though "market leader" might be a confusing term, given the small market slice that technology targeted. maybe "technology leader"? The SPARC and ALPHA designs were the lust-targets of high-end workstation geeks for years.

A processor is something that contains logic gate circuits, and performs data manipulation. Modern computers have a whole bunch of them, including the CPU. Usual common usage of the term processor refers to the CPU, as it is the central processor, but technically, the north bridge, south bridge, network controller, memory controller, GPU, FPUs, DSPs and more are all processors in their own right.

The CPU commonly contains three sections, the CU (Control Unit), the generic ALU(Arithmetic and Logic Unit), and Cache. The Control Unit handles the management of the CPU's work such as fetching data from memory and keeping instructions organized; the ALU handles all sorts of data manipulation - depending on the kind of work the programmer wants it to do, it can be made to do lots of different things - and cache which acts as a quick-access storage facility for data that's going to be used a lot.

Non-CPU processors usually have a simpler design, with very simple CUs, a small amount of logical circuits very specific to a single task which are not programmable (and therefor not "generic"), and often no cache.

GPU (graphics processing units) started out pretty simple, designed to remove the burden of processing 3d graphics from the CPU, allowing it to handle sound, physics, and game logic. There were GPUs designed for 2d graphics and for content creation like movie rendering, but those were quickly merged into the generic GPUs we have today. As GPUs became more powerful, they also became more flexible, with programmable logic gates, more complex CUs, and huge amounts of cache.

In the past few years, GPUs have become nearly as complex as the CPUs they assist, and some people have started doing all sorts of data manipulation directly on the GPUs themselves. Array and stream processing, protein folding processing; anything where you do one operation over and over on lots of data; or SIMD (single instruction, multiple data). It is likely that in the next few years, GPUs will disappear altogether; replaced by another generic core in the CPU.

 

i doubt that will happen. first there is competition and choices to be had in the graphics chipp market and secondly the heat generated from a single chip will be rather high si i think that a cooling unit double the size would be necessary.

to add to that it wil be too proprietary for consumers and really why is it different than having the chip directly on the mainboard? doesnt seem like people like that too much.

 

We used to have a single processor, and then required an additional GPU to handle rendering graphics. Now we have quad core chips, .... one or two entire processors could do nothing but render graphics, and they would not require any extra cooling, quad cores are cooled normally.

Too proprietary? Generic processors being tasked via software?

Also, you aren't tied into using those processors for just graphics, are you, like you are with a Graphics Card, ... I have a quad core PC I use as a media centre, and I can record two TV shows at once with my dual channel DAB recorder, and play back another, all stutter free. Try and do that with a single CPU and a top end graphics card, the card sits there useless, but it cost you a lot. So going forward, it will be more advantageous to use quad core (and whatever follows) chips, and we will see graphics cards die off, unless we start getting into immersive 3D rendering. 2D stuff can be easily handled, so the future of graphics cards lies in their ability to deliver something a PC cannot do easily with it's native hardware. I think immersive 3D games are overdue, I want virtual reality fps, the Wii was step in the right direction for interactivity, I want more. Maybe that will keep the GPU market afloat.

 

Admittedly it has been years since i studied processors, never studied them beyond single core units. I does make sense, to me, to have the graphics rendering part seperate from the core CPU because for that matter you may as well say why not incorporate everything to a single chip, including memory?

The fact is that there is no benefit to it and now you are stuck with one single chip that can take a whole system down and is useless if\when it goes bad. The price becomes much higher to replace it and you cannot take the graphics card out and use it in another system.

A fast single processor can do that easily because the 2 movies being recorded, if they are beind downloaded and not played back simultaneously, and is dependant on the drive speed. why does the card sit there useless? :shrug:

 

You know, I never buy a system based on which bits I can recycle if/when it breaks! If I thought a system were that likely to break I'd be scavenging it for bits to re-use, I would buy something else.

Quad cores are no more likely to fail than single cores are, and do not cost the same as a single core+top end graphics card, so you get mroe processing power for your money.

and four do it even more easily. This is a synchronous operation, remember. If that fails due to a resource conflict, swapping or processor priority, you lose data, because we are recording a broadcast, and there is no retransmit. It must work flawlessly, and multiple processors allow this to happen with fewer interruptions.

BARP! They are not being 'downloaded' but recorded from a DAB broadcast, we can't just ask for dropped packets to be resent like we can when we download, we must capture each and every packet, flawlessly. If your signle processor starts maxing, you'll drop data, and we are talking about shifting a lot of data at the same time.

I use two different SATA disks, one for recording, and the other for storage and playback, to double up the throughput. Parallelisation, of disk and CPU is the best way to go for performance.

Apart from rendering fairly low quality images ('cos recorded TV isn't hires) the graphics card does very little in my scenario. It's a fairly powerful processor, with built in memory, but all it does is render graphics. In a single processor system, with a GPU, that processor can sit idle. I'd rather have a multi-processor system, with basic graphics, and have all the extra grunt to apply to all potential tasks. For instance, if you want to copy a DVD, quad cores rip through that, 'cos the application parrallelises wonderfully. A fast video card would be displaying the screen saver while a single processor system chewed on it for an hour or more.

 

System-on-a-chip setups are fairly common for embedded needs. The direct connections between system components result in very low latencies. the downside is that you have all of the parts tooled into a single factory line, which makes it more difficult to upgrade a single component.

Well, with a single-core CPU, if that core dies, then the CPU is useless. With a multi-core system, one or more of the cores can fail without rendering the entire CPU useless. AMD's triple-core CPUs are just their quad-core products with a single bad core turned off.

The issue in this example will be potential blocking on I/O or other process hiccups. By providing multiple CPUs/cores, you allow a single busy core to be bypassed, preventing a slow or hung process from stalling the entire system.

I'm going to go with a car example, even though they never quite work: Imagine you have two roads; one is a two-lane road (one lane each way) with a 65MPH speed limit, the other is a 4 lane (2 lanes each way) road with a speed limit of 45MPH. In perfect conditions, more cars will pass through the 65MPH single lane road in a minute, due to the higher speed.

However, if a car needs to turn left, it block one lane, which stops all traffic flowing in one direction completely. If this happens often enough, the average speed over the course of a day on that road will drop far below 65MPH. In that case, even though the top MPH on the 4-lane road is less, the availability for cars to pass means that you get more traffic through per minute on average, despite a few slowdowns here and there.

phlogistician mentions parallel hard drives, which serves the same purpose - by making available redundant functionality in the system, you provide a back-up method for completing work when the first cpu/hard drive/other is busy.

 

Article covering OpenCL - a general purpose programming API for easier GPU programming of non-graphics-centric applications. Namely supported by nVidia at the moment, with promised AMD/ATI support down the road.
http://arstechnica.com/news.ars/post/20081209-gpgpu-opens-up-with-opencl-1-0-spec-release.html

 

As an electronic engineer, who has designed processors, I can tell you that 'CPU' is the term for a general-purpose workhorse processor 'central' in a system such as a PC. 'Processor' refers to any hardware device that executes instructions. A personal computer will have CPU(s) for the user to run their programs. There will also be many, many other processors such as those on the soundcard, motherboard, harddrive, NIC and any other non-trivial subcomponent.

 

OK, i have to admit that i was dismissive of this thread question at first, but got me thinking;

Now I am not a hardware expert, but i do code for my bread and butter =).

CPU is a term that means it is central processing unit, so it is more of an abstract and logical term IMHO (although i know that CPU is used interchangeably with Processors by people these days), so CPU can comprise of million processors spread across accomplishing different tasks, working like organs for the whole body (which MAY or May not have a physical boundary)

Is that a correct assertion? I personally think so. (I don't know if I was able to explain this right to you guys ...)

Rick

 

In fact CPU is an incorrect term, it should be called (based on Current technology) Processing Unit Group or something ...

 

and take this :

http://www.eetimes.com/news/latest/...=SCQCNRSSNQBTKQSNDLSCKHA?articleID=197004697

Whats a CPU in this case?

Please Register or Log in to view the hidden image!

Rick

 

Asguard does not have a field of expertise. At best he has a small windowbox.