Graphics Card Comparison Guide

JaniceZ writes "These days, there are so many graphics card models that it has become quite impossible to keep up with the different configurations. Therefore, we decided to compile this guide to provide an easy reference for those who are interested in comparing the specifications of the various desktop GPUs in the market as well as those already obsolescent or obsolete."

Comments

speed of development

[tute666]

they'll be obsolete in 5 minutes anyways....

Re:speed of development

[NanoGator]

"wtf is a texel"

It's a 'textured pixel'. Yeah, I know, that's not too descriptive. In 2D, when you draw a point on the screen, it's known as a pixel. In 3D, when you're filling a polygon with a texture map, every one of the pixels of that texture map is considered a 'texel'. That texel may be drawn of several pixels, but it's still one unit of that polygon that's worth measuring.

Article content is medicore at best

[linux_warp]

This article does not provide benchmarks, just things like "transistor count" and the number of pixel pipelines. Check out http://graphics.tomshardware.com/graphic/20050705/ index.html has the same information and benchmark charts.

Re:Article content is medicore at best

[Coneasfast]

i like digital-daily, they have some good benchmarks:

PCI-E 2005
ATI 2003

for example

Re:Article content is medicore at best

[Seumas]

Fortunately, I've converted to the Mac, so I don't have to worry about stupid things like having a choice in graphics cards.

(Sadly, I'm not trolling or being a smartass.)

Re:Article content is medicore at best

[DJ-Dodger]

The Tech-Report has a similar chart, but theirs is sortable and each card is linked to a review of the card if they did one.

Re:Article content is medicore at best

[freidog]

If the CPU wars of the last 5-7 years have taught us nothing; it's that you really can't judge a product on stats alone. At some point you have to see it perform. If we went strictly by the Rojak tables, the 5900 Ultra is a clearly superior card to the Radeon 9800 Pro.


Radeon 9800 Pro 128-bits 380 MHz 3040 MTexels/s 256-bits DDR 340MHz 21.76GB/s
GeForce FX 5900 Ultra 450 MHz 3600 MTexels/s 256-bits DDR 425 MHz 27.20 GB/s
The 9800 Pro has 1 additional vertex shader pipe, but the raw pixel pushing of the 5900U should be a good 15-20% faster than the Radeon.

Clearly that is not the case in the real world

A modern graphics card has so many complex and intricate features and tradeoffs for performance and power and production, looking at a handful of stats isn't even a good comparison when we're dealing with GPUs of the same family, much less a wide ranging comparison.

If you want to know how something performs, there is no substitute for benchmarks.

We really wouldnt need this type of thing

[Crashmarik]

If the manufacturers didn't go out of their way to completely confuse the issue to the point where there are no definitive answers to the question.

Short list

[Brian Stretch]

If you want Linux compatibility, you want nVidia. Yes, nVidia's drivers are closed-source, but they're at the same level as their Windows drivers, right down to the overclocking controls.

If you want a fanless, low-power GPU that can also do light gaming, get a GeForce 6200 with as much onboard RAM as you can find (ignore the TC "TurboCache" crap).

If you want a midrange, not-too-power-hungry card, get the 6600GT. This is my favorite card.

If you need a high-end GPU, get a 7800GT. If you have money to burn, get the GTX version. Check to make sure your power supply is up to snuff (Seasonic S12 series is my favorite, highest efficiency I've found), especially if you did something silly like buy an Intel P4. If you can afford one of these you can afford a proper AMD 64-bit processor to go with it.

There, everything you need to know. The 6200 was a pleasant surprise to me. I put one in my parents' Shuttle SFF box (Athlon 64 3000+), replacing a Ti4200, and the lower power consumption was enough for the main system fan to slow down to its minimum 1000RPM most of the time. It's still good enough to play UT2004 Demo at full detail at 1280x1024 res.

They missed FPS

[Mishra100]

I hear video cards being rated in FPS(frames per second) in certain video games all the time. If person A can get 100 more FPS out of Doom 3 using an ATI at the same cost as a Nvidia, usually they are going to go for that.

So my question is why didn't they include this in there? They have a lot of good data but I just wish that someone would run all the video tests on each card and check out the FPS data on certain popular games and produce them in a nice chart similar to this one.

Ouch my wallet!

[Wino]

That list looks like my damn credit card bill for the last 10 years! What would be really cool is if the guide had check boxes and when you hit submit it tells you how much money you wasted on all those shiney new gfx cards over the years.

Wait... on second thought that wouldnt be cool at all.

Re:It _is_ a shame

[Achra]

True.. I may come off as being whiney and uninformed as well, but I can remember when the 3dFX Voodoo2 was measurably better than the Voodoo. The Geforce2 was better than the Geforce, the Geforce3 better than the geforce2... I mean, you're going to have your stinker pieces of hardware, but both Nvidia and ATI's product line now is absolutely not linear. The Radeon 9600 is slower than the 9500... The Geforce 5200 is slower than a Geforce4 ti..
For all those that are going to ask me for hard benchmark data... Find it yourself. :) I'm too busy trying to sell my Radeon 9250.

Re:wow, what an utterly useless article

[thirty2bit]

It's a useful reference for people who have graphics cards that are a year or two or more old and need a comparison guide for a new purchase. I'm pleased to see it includes Direct-X levels for each card. That is the most often overlooked attribute in reviews. The only other similar ref I have seen is an issue of Maximum sellout^h^h^h^h^h^h PC that is over a year old, listing DX levels by chipset. Before HL2 and Doom3 came out, plenty of people were wondering what DX version their cards were, and if they would be usable.

There is no 'article text' because this appears to be a set of comparison charts, not a card discussion, and there is no explanation of 'what is what' because it is assumed that if you are comparing stats, you already know your subject. Finally, there is indeed an article navigation control at the bottom of the page. At least there is in Firefox, and also no Google ads for me, thanks to the same.

YAG3DGCC!

[Jah-Wren Ryel]

Oh Boy! Yet Another Generic 3D Graphics Card Comparison! Like we've never seen one of those before!

How about something that covers new ground? How about evaluating the features beyond simple stats and 3D performance in various games I'll never play?

I want to see a comparison that looks at these characterstics without regard to 3D FPS...

1) Noise level, idle and under load
2) Heat level and/or power consumption, idle and under load
3) DVI signal quality when pushed to maximum resolution & refresh rate - i.e. how long a cable I can hang off of it at what resolution
4) Video acceleration - mpeg2, mpeg2 for hi-def, WM9, WM9 for hi-def, h.264 and h.264 for at hi-def resolutions
5) Video de-interlacing support and quality - 3:2 telecine at what resolutions, how about 2:2 telecine, etc
6) Video scaling quality -- how many taps for vertical, how many taps horizontal, any fancy algorithms, test-pattern measured quality levels

Anybody and his brother can put up a speclist of 3D features or run a set of semi-standard 3D benchmarks and they already have. How about somebody with some real tools - oscope, multi-meter, pattern tests, etc do something new and useful to the REST of us for a change?

Re:Thank God

[dslbrian]

I will avoid the nVidia cards like the plague.
...
An equivalent ATI card *just works*. Period.

I've had the opposite experience. The worst problem I've had with an Nvidia card has been trouble configuring the DVI display settings on a Linux machine. Eventually thanks to user forums I figured out the magic settings (and they released a driver update with those settings a few days later).

On the other hand, I've been burned by ATI twice on graphics cards. In fact on one of the cards (a card supporting TV in) they never even made functional Windows drivers, much less Linux. Even called ATI tech support on that one and they put me on hold on their charge-per-minute support line - yeah great support there. Funny thing is years after the fact some 3rd party wrote a generic driver for the chipset under Linux which made it work. (So in total ATI "official" Windows drivers never worked, and generic 3rd party unofficial Linux drivers did work)

So now I don't even bother to look at ATI specs. They could make whatever uber-card they want that outperforms Nvidia ten times over, and I still won't ever touch the thing. Twice burned is enough for me.

Nvidia Linux support

[DrJimbo]

A couple years ago, I noticed a memory leak in OpenGL apps when using Nvidia drivers and an experimental Gentoo Linux kernel. I sent Nvidia an email about it around 10 pm Saturday evening.

I got a response about 20 minutes later which included a patch for the Linux kernel I was using. I recompiled my kernel with the patch and it fixed the leak.

It is too bad their drivers are closed source, but I have to say that their Linux support is outstanding and on a par with the best support I've experienced.