Computer speed

Off topic discussion zone.

Moderators: winston, another_commander, Cody

Post Reply
User avatar
Alex
---- E L I T E ----
---- E L I T E ----
Posts: 767
Joined: Mon Oct 06, 2008 10:49 pm
Location: Oz. The land of some gold but mostly rust

Computer speed

Post by Alex » Tue Jan 12, 2016 7:47 pm

I've been working with computers for 30 odd years and have never got a proper answer about computer speed.

1 CPU clocking speed
2 RAM and/or ROM clocking speed
3 Motherboard clocking speed

They are not the same
Anyone care to explain?
As far as I've found, the mother board is the slowest and the result will never be faster than the slowest part?
Internet is worse, only as fast as the slowest router you are patched through

Is this the true?
LOOK OUT!!!
OOPS..
"Press Space" Commander
If you do not see "Press Space" more often than you want.. Your not trying!

NigelJK
Deadly
Deadly
Posts: 201
Joined: Tue Feb 16, 2010 5:07 pm
Location: Stockport, England

Re: Computer speed

Post by NigelJK » Wed Jan 13, 2016 11:20 am

I'll try. I've got a bias toward Acorns innovations during the 80's and 90's as I know their stuff well (was a BBC micro service engineer for a while).

The original Von Neumann model required just one crystal to clock the whole 'model'. It was soon found that for various reasons different parts of the machine could be 'clocked' at different rates. The BBC micro for instance used some clever circuitry to split the (IIRC) 4mhz crystal into three separate cycles (1mhz, 2mhz and 4mhz) with 4 for the main bus 2 for the cpu and 1 for the memory (which had it's own special circuitry that allowed for 2 banks of 16K, when only 16K could be addressed, and swapped using the 2mhz cycle). The bus (or mobo as some would have it) should be the quickest as everything else hangs off it. As far as I know Acorn stuck to this idea through out as until the advent of pre-emptive multitasking it's the cheapest way. From here on in, the US have their way and we get CPU's with their own clocks which allows for 'over clocking'. What this means in practice is that multi-operation instructions can be carried out within one cycle of the bus hence the 4x or 10x multiplier. This is hugely inefficient, but as even IBM struggled to get RISC processors to work, that's the only way the mainstream microcomputers can use 80's technology today. Memory has come a long way and is usually clocked at (or more correctly 'up to') the same speed as the bus, packaging is the only decider there and quality of the substrate used.

As an aside to this the RISC processors (particularly those designed by ARM) need no clock of their own in theory as they are one instruction = one cycle processors.

User avatar
Norby
---- E L I T E ----
---- E L I T E ----
Posts: 2543
Joined: Mon May 20, 2013 9:53 pm
Location: Budapest, Hungary (Mainly Agricultural Democracy, TL10)
Contact:

Re: Computer speed

Post by Norby » Wed Jan 13, 2016 12:18 pm

There are two main components of speed: computing power and data transfer. First mean intensive mathematical operations on small data at once like file compression, the second is the opposite like texture fill rate of videocards.

In mixed cases the slowest part will be the bottleneck, but many solutions aim to work around these:
-in computing the architecture an paralelism within the CPU to do more in a single clock cycle,
-in data transfer various caches aim that only the first access of a data be slow but others are much faster.

Faster cache storages are more costly so must use multi-level caches for any reasonably fast but payable compromise. So neither the slower memory on the motherboard will always slow down your programs due to the processor in most of the time read from the internal caches only, nor you will buy a CPU with 8GB internal cache for incredible cost.

In practice you should look the results of performance tests, I like cpubenchmark.net for computing and tomshardware.com for videocards.

For raw data I look ark.intel.com and videocards in wikipedia.

User avatar
spud42
---- E L I T E ----
---- E L I T E ----
Posts: 1208
Joined: Wed Mar 26, 2014 10:11 am
Location: Brisbane,Australia

Re: Computer speed

Post by spud42 » Sat Jan 16, 2016 3:10 pm

Alex wrote:I've been working with computers for 30 odd years and have never got a proper answer about computer speed.

1 CPU clocking speed
2 RAM and/or ROM clocking speed
3 Motherboard clocking speed

They are not the same
Anyone care to explain?
As far as I've found, the mother board is the slowest and the result will never be faster than the slowest part?
Internet is worse, only as fast as the slowest router you are patched through

Is this the true?
This is going to get messy.....

As Nigel pointed out in the beginning CPU did it all and had 1 clock speed. back in those days it was less than 10 MHz .
Then CPU communicated to the rest of the system over "buses" , which is a fancy way of saying alot of circuit traces dedicated to one purpose. so you have the address bus, the data bus, and a few others that i will lump together as the control bus. the data bus connects to the ram, but so does the address bus. the address bus also connects to various other chips on the board . now the control buss also turns these chips on and off. this was fine when the speeds were low as all could be run at the same frequency.

why we have different frequencies today is because the CPU CORE can operate at much faster speeds than ram, and ram at faster speeds than the peripheral cards. the peripherals have developed different standards over the years from ISA to PCI to PCIE but these run at 100MHz base clock . your processor has internally what is called a clock multiplier so to get your 3GHz it multiplies by 30. this is also why "K"variants of CPU's can overclock as the multiplier is not fixed.
now to complicate things the ram runs off whats called the FSB (front side bus) this has got faster and faster over the years as ram tech got better going through DDR,DDR2,DDR3 and now DDR4. speeds from 400MHz up to 3000 MHz on overclocked ram.
then we had the HDD connections with basic IDE (which some heathens insist on calling PATA),SATA.SATA2 SATA3 even now using Solid State Drives on the PCIE buss as M.2
the major bottle neck was always the speed of the read and write to storage, mechanical disk drives.
there isnt any ROM as such since the old days either back in the day the "operating system " was on the rom chips but as that change to being on the HDD you needed only a very small rom called the BIOS, basic input output system. basically a bootstrap to tell the cpu where everything is and how to talk to them to get the OS booted off the HDD then its not used. so the rom /bios doesnt affect the speed of a computer at all after boot.
this is a really quick of the top of my head explanation that glosses over several things. because things get really complicated inside the "processor" we now have multi core multi thread processors that do many things all at once as well as something called IPC instructions per cycle( clock cycle) .
anything i missed? probably . and others may explain better...

and as for the internet question yes you may have the fastest internet connection in the world BUT if you happen to be going through or connecting to a server that limits the bandwidth then that is as fast as it will get. Routers usually are not the problem its the servers that provide the website or the download that is the limit. the routers or the total number of hops you have to make to somewhere affects your PING time. not important for data transfer but a killer for online gaming where a large ping will mean you are shot before you see it coming...
Arthur: OK. Leave this to me. I'm British. I know how to queue.
OR i could go with
Arthur Dent: I always said there was something fundamentally wrong with the universe.
or simply
42

Post Reply