Graphics Graphics card guide sticky wanted.

hey, i was trying to search the forum for a noob guide to graphic cards, but couldn’t really find anything to the point, all the threads were related to users comparing a few cards or other stuff.
Google showed results of just reviews.

If there are any links please let me know, but if u got time then....read on

What i wanted to know is what are:

- pipes???
- difference between 8x and 4x (i know twice a fast, but about the visually - difference, is it a big issue)
- how worthy is overclocking? Advantages and disadvantages
- the different values of RAM (is 128mb good enough, i know more is better, but how good is this)
- all the different titles like GDDR, DVI, e.t.c
- what’s value for money.

If examples are given, that would be nice.

I just wanted to know all the details, so i could then mix and match all the aspects that are required by me. Plus i hope the guide will help other users. For example some may want to spend more, some may want not need all the extra features.

Theirs allot on the market, i don’t just don’t want to pick a few and say "guys which one?"

I know the cards depend on the overall pc, so for this guide just assume an average machine say

2 Ghz
512 ddr ram
120GB
350 psu
Average price of around £50-£100

Thanks in advance
If i have made a mistake , please correct me.

 

I dont think you'll get a sticky..but here it goes. I'll awnser what i can.


There is usually no difference between 4x and 8x AGP usually.

There is sometimes no difference between 128meg and 256meg videocard.
You usually have to look at reviews to tell.

Overclocking matters on the card, once again..look at reviews. Some cards overclock more than others.

GDDR is the memory the card uses, however..DVI is a connection for LCD monitors, VGA is a connection for CRT monitors.

Value for the money, no idea what you mean here.

 

If I find an hour or two spare tomorrow, I will try and write something up - it's my first day off in August though, so I might not get the time.

 

Wrong.

A VGA connection is an Analog connection.
A DVI connection is a Digital connection.

Each can be found on either CRT/LCD monitors.

 

Ahh i didnt know that, thanks for the info . I've never seen a CRT with a dvi connection though..but i have seen LCD's with analog's..guess i should of thought that out more, kek.

 

No Probs!

 

Well, VGA is digtal--analouge cable--digital which causes quality to suffer, whereas DVI keeps the signal digital from card all the way to your screen giving you a better picture. A good VGA cable is better than a poor/cheap DVI though cause of EMI.

 

Cheers guys.

Bigz take your time, am in no rush. Got no money at mo, just wanted to build my knowledge up, and know how much money to save, and what to look out for.

I read about the new game "F.E.A.R", and really want to play it.
At the mo i only got a 32MB card. lol, i have had that for about 4 years now.

I still would like to know what pipes are?

I guess i will have to read as many reviews as i can.
Keep the answers coming thou

 

Pipes are, to my understanding, part of the "how much data can you process?" formula, in effect. Number of pipes * core clock speed (hz) = bits per second that can be processed. So the 7800GTX would be 24 pipes * 430,000,000Hz = 10,320,000,000bits/sec = ~10Gb/sec.

err, well something like that I think. I used to know it better, now I just know that it's roughly take the highest card from one generation, roughly double the performance, and you have the highest from the gen after that (ie 6800 Ultra x2 ~ 7800GTX).

Meh, rambling, but FYI, my 6800GT SLI rig runs the FEAR demo choppy at 1600x1200 with 0xAA 0xAF and most settings at medium. Still looks amazing and the AI is unrivalled, but my point is you need a hell of a lot of rendering power to run that game at even high-res-no-effects.

 

it is hard to determine a pipe these days because they can come in so many different forms.

If you think of it as how many pixels can be processed each clock cycle, that is probably the easiest way to think about it. If you just think about it as bigger is better when calculating (# of pixel pipelines) x (core clock speed), then you won't go too far wrong with pixel pipelines... for the moment. You could delve in to ALU's and efficiency, but that would just confuse things even further.

Then there's the rasterizer, which is where pixels are written to memory - some times there are less of these than there are pixel pipelines. I could go on for ages, but it's tough to really dumb it down.

 

Or simply put, the more/faster, the better.

 

When a mummy graphics card and a daddy game engine love each other verrrry much...

Also: A day off? HAHAHAHA... you lie.

 

Are pipes like a 'bus', like a channel?? the more pipes the more data can be sent at the same time? and this data is the pixel value. the clock speed is how fast the pixel value is calculated.

Right or wrong?

And when buying a graphics card, should i really be looking out for number of pipes, not that i have really seen much info given. (Thats why i have no idea what it is)

 

You can work on one pixel in every pipeline, every MHz. If that makes sense.

1MHz = 1 clock cycle

 

clock is a quantity of cycles = what is done . Increasing the MHz increases the amount of clock cycles done per second, which means more pixels are calculated. Increasing the amount of pipes increases the "theoretical bus width" (although it's not technically a bus, as such, as it's not a delivery of data but a calculation of data), as it allows more pixels to be worked on in the same clock cycle. Although, like biggles says depending on the commands issued to the pixel pipes depends on when the "end result" exits the pixel pipe "bus" and ends up in the rasteriser.. etc

I know timski will explain it in his info article but here's my take:


Sooo the process sorta goes like above - (game) engine instructions get delivered to vertex shaders (8) which setout the scene (iirc), into cull/clip/setup, then Shader Instruction Path gets executed only if the pixels are seen (Z-Cull), no point in drawing pixels that are behind something, that's why you need to work out the vertexes beforehand. Into the pixel shaders (takes most time to shade, colour, apply SM instructions etc), out into fragment crossbar into the 16.. other things i forgot, Biggles did educate me once but as usual i tuned out towards the end (edit: ROP?? edit2: Rasteriser Output P.. P... P-p-p-p-p..something), then it gets lumped together in memory (frame buffer - think "frames per second") then out to your VDU. Sorta like that, i dont care for exact details
.
It's all about doing it all in p// so you get more done at once. Each of the pipes are like mini-cores within a GPU because the current generation has stuff like mini-alus etc (denoted by the couple of little green and orange squares in the big boxes) which instead of only having an range of integers (numbers) to work with calculate an exact, seemingly infinate number instead, hence, they are programmable by code and can be told to do algebraic sums on whatever is needed (like a CPU but more specifically tailored to graphics instruction execution) instead of having a specific instruction set to work on (of sorts) like the old generations of cards used to do.
CPUs have a more generic function so they cant be tailored and efficiency specified like GPUs can.

This is also why newer games look better than older ones, because more instructions are applied to provide far better pixel, colour, shade, lighting calculations giving a more realistic image.

It's not a 100% correct definition but it's kinda cut down and will do you until timmeh produces something better.

 

Cheers guys, i have some sort of idea. Good enough for me. Thanks for caring

 

This is Bit-tech, we all care. (kinda)