explain the basics

[keke]

In every VGA card review there's always shown the core & memory clocks of the cards. Usually also buswidth and how many pipelines it has. But who needs this stuff?

The core & memory clocks might be useful to know if you're thinking how much you'd need to raise the mhz to overclock a certain card to reach the level of same card that has the tag "ultra" or "pro" or something like that, right?

But what about pipelines & buswidth?

What do I do with the information whether a card is 128-bit or 256-bit?

Matrox Parhelia is 256-bit but does it beat 128-bit Radeon 9600 XT in games?
Atleast not because of its bandwidth.

I've got GF4 MX 440-8x SE that runs fine with core clock 300 MHz and memory clock 450 MHz. If I buy GF FX 5700 Ultra I know newer games will run smoother (better fps) with some special effects turned on, but what about older games like Counter-Strike that use 5 year old engines? Is it just the core & memory clocks that affect the performance in older games?

(And for those who think even GF2 MX's are enough for CS, no that's not true since I can see a difference in CS with standard core & memory clocks compared to overclocked values)

Here's a link to the latest VGA charts made by tomshardware: http://www6.tomshardware.com/graphic/20031229/index.html

 

[cleeve]

Video card performance is based on 3 things:

<b>ARCHITECTURE, SPEED, and DRIVERS</b>

<b>Architecture:</b>
This covers any hardware design decisions. Important stuff in this category that is commonly thrown around includes:
- # of Pipelines
Pipelines are generally thrown around as a baseline to compare complexity of GPUs. The more pipelines a GPU has, the mpre pixels it can manipulate per clock cycle.
- # of Texture Units
The more texture units a GPU has, the more layers of texturing it can apply per clock cycle
- Memory Interface (64, 128, 256 bit)
The higher the nandwidth of the memory interface, the faster the memory can be accessed and used. This increases performance a great deal, especially when using effects like antialiassing.
- Shaders
New GPUs have Vertex and Pixel shaders that process special effects. There is no real number to compare how powerful they are really, you just have to compare their results in games that use shaders heavily.

<b>Speed:</b>
Speed is how fast your video card's components are running:
- Graphics Processor Speed (AKA "Core" speed)
How fast your GPU can make calculations. This governs how quickly your card can process geometry, textures, and effects.
- Memory Speed
How fast your video card's memory can be written and read. This governs how quickly your card can work with the GPU.
Memory speed x

Speed isn't just for overclockers. Speed is just as important as architecture.

Let's say you have two videocards based on two different graphics chips.
If Card"A" is only half as efficient as Card "B"... but you run Card "A" at 3 times the speed you run Card "B"... then chances are, Card "A" will be a faster video card.

<b>Drivers:</b>
Everyone knows what drivers are, but if you're a newbie you should know that drivers can be the difference between a great video card and a mediocre video card. They're also very important for stability: the fastest video card in the world is no good if it crashes in every game, right?

<b>So what's the most important statistic?</b>
ALL of them. These are ALL important factors.

For example, the Radeon 9600XT VS the Parhelia:

Radeon 9600XT:
4 pipelines
1 texture unit per pipe
128 bit memory interface
500 Mhz core x 300 Mhz memory
9.6 GB/s bandwidth

Matrox Parhelia:
4 pipelines
4 texture stages per pipe
256 bit memory interface
220 Mhz core x 275 Mhz memory
18 GB/s bandwidth

Wow... looking at those specs, the Matrox should kick the 9600XT's butt...
BUT the Radeon 9600XT will beat the Matrox Parhelia everytime. Why?

First of all, notice the core clocks. The 9600XT is running more than twice as fast as the Matrox! So even if the Parhelia is more efficient at the same clock speeds, the 9600XT is running more than twice as fast and can get many more operations computed per second.

Note that the Parhelia has a huge bandwidth advantage over the 9600XT, due mostly to the Parhelia's 256 bit memory interface VS the 9600XT's 128 bit memory interface.
While this helps the Parhelia, the 9600XT still kicks it's butt. This is because of stuff like Drivers, shader power, and other issues.

So paper specs are far from everything. You basically have to pay attention to reviews to find out which video cards have which strengths and weaknesses, and use that information to get a better understanding of the performance you can expect from a given solution.

________________
<b>Radeon <font color=red>9500 PRO</b></font color=red> <i>(hardmodded 9500, o/c 322/322)</i>
<b>AthlonXP <font color=red>2600+</b></font color=red> <i>(o/c 2400+ w/143Mhz fsb)</i>
<b>3dMark03: <font color=red>4,055</b></font color=red>

 

[TheGreatGrapeApe]

Hey! Lay off the Parhelia man!


- You need a licence to buy a gun, but they'll sell anyone a stamp <i>(or internet account)</i> ! - <font color=green>RED </font color=green> <font color=red> GREEN</font color=red> GA to SK

 

[cleeve]

Ah, sorry dude. I forgot that was a sensitive issue for you.
I should have used the Volari as an example...

________________
<b>Radeon <font color=red>9500 PRO</b></font color=red> <i>(hardmodded 9500, o/c 322/322)</i>
<b>AthlonXP <font color=red>2600+</b></font color=red> <i>(o/c 2400+ w/143Mhz fsb)</i>
<b>3dMark03: <font color=red>4,055</b></font color=red>

 

[Nights_L]

Graphics Card Specialist=Cleeve

 

[Crashman]

Stuff like this helps you when you compare similar cards. For example, the difference between the 9500 Pro and 9700 is only that the 9500 Pro has a 128-bit memory path, but hte 9700 has a 256-bit memory path. And the 9500 Non-Pro has 4 pipes, while the 9500 Pro has 8.

This is especially usefull when new "sucker" products like the 9800SE come out with half the pipelines and half the memory bandwidth of the 9800!

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>

 

[phial]

But who needs this stuff?

thats like saying "who really needs to know the truth?"


of course everyone needs that stuff.


you may see a card that has a 4x2 architecture
and then one that has 8x1

they look similar... but really, theres a huge difference in performance and things that they can be used for.

-------


<A HREF="http://www.albinoblacksheep.com/flash/you.html" target="_new">please dont click here! </A>
dhlucke - "Phew...ok my wrists are hurting. I'm taking a break."

 

[keke]

thats like saying "who really needs to know the truth?"

I should've asked "Why do we need this?"

I mean, if 2 cards from the same manufacturer have same pipelines & bandwidth but different core & memory clocks, you can get the better brand of the cards if you manage to overclock the worse to the clock speeds of the better?

Is that the only thing where you need that information and is that stuff even enough to know that the cards are otherwise similar considering game speed?

I guess what I should ask is "Who reads that stuff?"
I think only those who think how much would it take to overclock a card to the "next level". But tell me if there's any other purpose for that stuff for normal consumer.

So paper specs are far from everything. You basically have to pay attention to reviews to find out which video cards have which strengths and weaknesses, and use that information to get a better understanding of the performance you can expect from a given solution.

As stated here, it's the benchmarks that count when it comes to game speed, so why is that technical information always covered in the tests? Is it for some other purpose than to compare the speed?

 

[cleeve]

Well, if you want a simple answer, it's that technical specs have a limited usefullness.
But that doesn't mean people aren't interested in them.

The intrecacies of the chip's architecture are also very important but there's no way to compare that numerically. So people focus on the specs that they can.

It's very simple, really.

For example, the GeforceFX 5200 ULTRA and 5700 look very similar on paper. But obviously, that isn't the case.

You just can't assign numbers to everything.

________________
<b>Radeon <font color=red>9500 PRO</b></font color=red> <i>(hardmodded 9500, o/c 322/322)</i>
<b>AthlonXP <font color=red>2600+</b></font color=red> <i>(o/c 2400+ w/143Mhz fsb)</i>
<b>3dMark03: <font color=red>4,055</b></font color=red>

 

[keke]

I might be a bit repetitive and dumb but I don't think anyone's yet given a good answer, who is this information REALLY FOR..

If you were writing a review, I bet you'd start thinking "to whom am I writing this stuff for" if you don't even know yourself why you would need it.

Of course pipelines & bandwidth might show that some cards can't just be overclocked to reach the same level as their big brothers. But if both cards have same pipelines & bandwidth there could also be something different than just core & memory clocks that aren't covered in the review.

Sure technical stuff is interesting but should it be covered in benchmarks? Why? You can still always read it from manufacturers web pages if you're interested, right?

 

[pauldh]

Not looking at the technical data is like reading a performance car review and not looking at the engine sizes, HP, torque, etc.

Edit: If Car and Driver didn't show specs in their reviews, but only showed 0-60 and 1/4 mile times, many readers wouldn't buy another of their magazines.

ABIT IS7, P4 2.6C, 512MB Corsair TwinX PC3200LL, Radeon 9500 Pro, Santa Cruz, Antec 1000AMG, TruePower 430watt<P ID="edit"><FONT SIZE=-1><EM>Edited by Pauldh on 01/13/04 04:27 PM.</EM></FONT></P>

 

[cleeve]

The information is REALLY FOR anyone who wants to see it. People like myself who would ask for it.

What's the alternative? Disallow video card manufacturers from quoting specs because it doesn't tell the whole story? That wouldn't make any sense.

And GOOD reviews will quote spoecs AS WLEL AS architectural differences and possible reasons behind performance disparity.

Most sites that quote pipelines and bandwidth and then follow up with only benches of the 3 most popular games aren't worth reading.

________________
<b>Radeon <font color=red>9500 PRO</b></font color=red> <i>(hardmodded 9500, o/c 322/322)</i>
<b>AthlonXP <font color=red>2600+</b></font color=red> <i>(o/c 2400+ w/143Mhz fsb)</i>
<b>3dMark03: <font color=red>4,055</b></font color=red>