An introduction into graphics cards technologies : 56K unfriendly

[Agent]

The purpose of this post is not to compare between graphics Cards Company’s, or to try and persuade you that one is better than others, but to show what different technology’s are in use in the graphics card sector, and on your cards today.
Im going to try and keep this post at mid-level understanding, im not aiming to go really in depth here, but to brush on each one enough so you have an understanding of each technology. Some of the things I have written are over simplified, so keep that in mind when replying about mistakes . For those of you that already know this stuff, then ill apologize now for boring you

The role of drivers.

A driver is, in simple terms, instructions that the computer uses to talk to hardware devices within it. Pretty much everything in your PC will have a driver for it somewhere along the line, although windows has made much of this transparent to the user.
Think of it like this. If I gave you a page of writing in a foreign language you didn’t know, it wouldn’t make any sense to you. If I first translated this into a language you did understand, and then gave it to you, reading it wouldn’t be a problem.
That’s basically what a driver does. Games / 3D apps give the instructions out on what it wants the graphics card to do (Usualy done via either Direct X, or OpenGL), the drivers take this information and put it into a form that the graphics card will understand. Hence why you don’t have different versions of games for differing graphics cards.
In graphics cards today, not only can the drivers do the above, but also have the ability to directly program the graphics card itself. This is very useful for developers when they want to take a intensive algorithm away from the CPU and onto the graphics card.

It’s all about IQ, sunny Jim.

And no, that’s not the intelligence rating of your card, but the Image Quality it produces. While Frames per second may well be important, image quality is often underestimated by people when buying.
A fairly big mistake that people make when seeing a card is come out with comments like “The 6800’s image quality is pants”, or “Ati’s image quality is rubbish”. Image quality can vary from manufacturer to manufacture, and to a very small extent, from card to card. When using an analogue signal; it will leave the GPU, then pass through several “filters” to turn the digital signal (a 1 or a 0) into a sine wave for the monitor to understand. The varying quality of these filters causes the image quality to differ. If one manufacturer uses a better filtering technique, then it will produce a better image. This affects both the sharpness of the image, and the quality vibrancy / depth.
Once the signal has left the GPU, it has no control over what the final quality of your screen will be. If you want to test this out for yourself, go and rip a few of these filters off your card and check out the result
This mainly affects analogue signals. Digital signals still have to go through a form of filtering, but its different to the one an analogue signal will go though. The beauty of DVI is that the signal will remain in its digital form all the way to the monitor. Unfortunately, DVI does not have error correction, so it has the risk of picking up noise on its way to the monitor from other electronic devices. This problem becomes apparent if your in the professional imaging industry, where it has been proved time and time again that low quality cables + distance = problems.

The filters aren’t something that can be changed by drivers. They are physical devices attached to the PCB. With that in mind, lets see what we can do at driver level.

[Agent]

Holy blur, batman!


Sticking along the lines of image quality, we come to out first friend. Anisotropic filtering.
Anisotropic filtering is a method of making an image blur less as an objects distance to the vanishing point decreases. In simple terms, it helps distant textures look sharper.
This method of filtering is very intensive, and can cause a nasty hit to the FPS a card can produce. This is, of course, dependant on the game, size of textures used, how many textures are present, and the sampling rate that you require. A higher sampling rate will make the image look better, but cause a bigger performance loss at the same time.

To demonstrate this, I took some screen shots from UT 2004 using different sampling rates. All these shots are from the same spot, and are pretty much pixel accurate.
All of UT’s settings were set to maximum (holy sh*t settings), and the HUD, weapons and crosshair was hidden to keep the view uncluttered. Screenshots were taken at 1024 * 768 and cropped to the area shown.

I will make all full, uncompressed screenshots available shortly; I just need to sort some web space out. Until then, you can see the original here At the moment, ive cut and pasted parts of the image onto this smaller map, so its easier to see what the effect does. The image below is a PNG (hence lossless) which will try and show AF in action.



The easiest place to see this in action is on the walls towards the centre of the image. You can judge the effects for yourself
Keep in mind that this was done on a Nvidia card, and results on a Ati, or other graphics card will differ. It may be better, it may be worse, but we’ll leave that for another thread.


That line is about as straight as Graham Norton !

Pixels have one major problem when being rendered: They are, of rectangular nature. When objects in a scene touch render over / under another one, the edges of them often form a “line” around the edge of the object. This causes a stairway effect to happen in the scene, that is even more noticeable when the scene is moving.
The technology used to remove this is known as Anti-aliasing. The method employed by different card manufactures is different, but the end result is similar to one another.
Like AF, AA causes a performance hit that is dependant on your graphics card, drivers and GPU.

To demonstrate this, the same scene in UT2004 was used, with varying levels of AA applied. It is important to note here that AA should only really effect the “stair way” effect that its designed to eliminate, but due to the “blurring” technique used in some modes, it can effect texture quality slightly. How much is totally dependant on the mode used, the driver revision you have installed, and what GPU your card has.
The exception in these screenshots is the 8xS mode, This not only tries to remove the jaggies, but uses a form of AF to try and make the textures better.
Again I just want to stress the point that these images are only meant to illustrate the technology and not to draw comparisons between cards, driver revisions, or GPUs.





The easiest way to see the effect, is to look at the plank of wood, almost central to the screen. It should become apparent from that.

The major problem I had showing the 2 effects above was resolution. First off, the file size of the images is already huge (sorry 56Kers, but its very hard to show AF well with compression). Secondly, using a higher resolution would have caused less jaggies in the scene, and the entire point of the AA one, was to show the technology in action. I have had a very small amount of time to put this together to, so given the time I will try to make them clearer as soon as i can.

There is another methods for making an image better, I may cover these in the future, but the main 2 ones are above

[Agent]

I cant do it captain, I just aint got the power.

Todays graphics cards drink more power than every before. A “powerful” PSU isn’t really what they need, its more of a PSU that can supply stable flow to the card. It’s the reason that shuttle PSUs that are only rated up to a few hundred watts can power a system loaded with the latest gear, and a top of the range graphics card. Without wanting to shoot off on a tangent, a quality PSU at a lower rating is generally much better than a higher rated, poor PSU. If advice is needed on PSUs, post questions in the hardware forum.



AGP vs. PCI express

This seems to be a hot topic at the moment, so some things to keep in mind.
Only the latest of the latest mobos support PCI express. The socket A platform does not support PCI express, so if you really want to use it, your going to have to get a new motherboard, along with a new processor to go with it. For this reason, its probably best to hold off until you can upgrade them all at the same time.
At the moment, PCI express doesn’t really offer any huge gains over AGP, other than it can supply the graphics card with more power. However, it goes without saying that given the time, PCI express will take over and shine through. Even if you cant future proof yourself against it, its good to know what your up against.
Having said that, the huge benefit that PCI express has over AGP is…….

SLI

So what is SLI ? Many of you may hear those little 3 letters and have memories of 3dfx days. It is the same concept, but different technologies. The name “SLI” is intelligent marketing on behalf of nvidia (3dfx’s SLI was “scan line interleave”, while nvidias is “Scalable Link Interface” : they also operate different hardware wise).
Ati has also published details of its version of SLI, which current looks like it will be under the name “Multi Rendering”.
The idea is to link 2 graphics cards, to try and double the work done. In practice, twice the work will not be achieved, but it can come pretty close looking from results dotted around the web.

I may well update this post to add more technologies, examples, and information, just looking for some feedback on what’s currently been done.

[Stewart]

All good mate, all good.

[DR]

Good post

[myth]

great post! Well, its confusing.... You are talking about AF but jump to AA with pics Then you talk about AA and jump to pics of AF....

[Apex]

Cheers for that

[johnnr892]

excellent post , but i think he has got the samples the right way round

[ajbrun]

I can't tell the difference between the 1st set of pics . I can see the differences in the 2nd set though .

[exAndrzej]

I always think that these things are easier to get if you have ever used Photoshop

Say you have a reasonable sized image (640x480) - but it is quite 'jaggy' in places - then by making the picture physically bigger (say 1280x960) using change image - then choosing 'blur' - and then shrinking it again you can make the image seem smoother

If you have a picture that you want to use on a WEB site - and it contains text that the visitor needs to be able to read - then you may choose to 'sharpen' the image and then save it as a 'GIF' to preserve the detail. This technique is especially useful if the text is (a) not the main focus of the picture and (b) at a slight angle

Given that AA technology 'blurs' and AF 'sharpens' - it is cool that they both work together in harmony (or that they work together at all!)

[Iain]

That'll be DM-Rankin if I'm not mistaken then

[myth]

Ok, I got a question for ya... What does 2Q do?

[mike_w]

I'm sure the pictures are the wrong way round - it talks about AF first, then shows AA pictures, and then talks about AA with AF pictures shown.

Still a good guide - never understood before what AF did.

[myth]

Originally Posted by mike_w

I'm sure the pictures are the wrong way round - it talks about AF first, then shows AA pictures, and then talks about AA with AF pictures shown.

Still a good guide - never understood before what AF did.

Finnaly some one agress with me! And its about time too!

[johnnr892]

oh, i thoight that u meant the labels were the wrong way round, I misunderstood Myth, you are right