Best video card under 100$

[flingz]

I am looking for a card that will be good for some gaming but I do not need the best, just the most reliable and best card for under 100 bux.
thanks!

[gwallen4]

I'd go with the GeForce4-MX440. It's 64 MB, fast, stable, and made by a variety of manufacturers. It's one of the most popular cards made today.

I've seen them on newegg.com for less than $75 with free shipping and fast delivery time.

Here's a comparison of the common video cards at Tom's Hardware

http://www.tomshardware.com/graphic/02q2/020418/vgacharts-03.html

[BigBlue66]

GeForce3 Ti 200 or Ti 500 64mb or 128mb.

The GF4 MX series is nothing more than a GF2 with a couple of extra goodies.

[gwallen4]

Big Blue:

The GF3-Ti500 can't be found anywhere and if it could it would be well over the $100 limit.

The GF3-Ti200 can be found occasionally on the internet but not in most stores. The lowest price on it is about $95. It's a slightly better game card at a higher price than the GF4-Mx440 which can be found in any store and is often on sale.

[saphalline]

It's a slightly better game card...than the GF4-Mx440...
I don't think slightly is the word to use here! A GF3 Ti 200 is at least twice as powerful as any GF2 or GF4 MX. Not only that, but the GF3's and GF4 Ti's are DX8 parts, while the others are DX7 parts. And if you want to get into the nitty-gritty details, the GF3's & 4 Ti's also have Z occlusion culling and built-in FSAA settings. And while the GF4 MX series does have LMAII, they only use a dual crossbar memory hub, not the quad version used by the GF3's and GF4 Ti's.

Graphics cards under $100 is the touchy low-end range where spending a bit extra gets you much more performance! You are right, however, when you say that GF3 Ti 500's can no longer be found and would be overpriced anyway. I hold with BigBlue66, tho, on the GF3 Ti 200 choice. Better performance in just about every game, and the performance gap compared to the GF4 MX 440 just increases as detail/resolution go up. Plus, they run very cool and are highly overclockable! :D

[gwallen4]

Saph,

I bow to your superior wisdom on the subject of graphic cards. I was only looking at frame rates for games. I'm a little bit of an expert on where to find things on the internet and how much they cost, but not too good with all the technical mumbo-jumbo.

Anyway, you and Blue have me convinced. Good work!

[saphalline]

I bow to your superior wisdom on the subject of graphic cards.
Heehee :p, it's not my wisdom here, I just spend too much time reading the rediculously long graphics articles on Anandtech (www.anandtech.com/)! :eek:

Grab a snack and read for yourself, you eventually get the hang of it...

DX8 part - Any GPU that is DirectX 8 compliant, the requirements being (in simple terms) the ability to run a certain amount of vertex shading calculations per clock cycle. Vertex shading is basically being able to bend, twist, and otherwise deform a single 3D mesh multiple times, creating the illusion of, say, grass flowing in the wind. A DX8 card can use vertex shading algorithms on one blade of grass for the entire field, while a DX7 card must render each blade of grass individually (if it could even do it at all).

Z Occlusion Culling - The process of NOT rendering the polygons that are hidden, such as the back side of a box in front of you. No rendering = no performance loss for something that won't even be seen. This accounts for much of the high performance increase of current graphics cards, and it's so simple it makes you wonder why no one did it long ago! :rolleyes: Well, probably the transistor count, but still!

LMA - NVidia's memory bandwidth optimization "code word", I don't remember ATI's code word, but it's basically the same thing. You see, graphics cards have a certain "bit width" for their on-board RAM, and they can access the RAM in chunks the size of their bit width. As this has increased, a lot of RAM bandwidth gets eaten when the GPU only needs a 64 bit chunk, but must take a 128 bit or 256 bit chunk because that's all the RAM can do. But lately, the RAM bit width has been broken into smaller chunks so that different chunks can be taken from different parts of RAM. 128 or 256 bits must still be taken per RAM cycle, but the GPU can now choose which ones are taken.

In other words, let's say you want to borrow 4 books from the library, but the library requires that you must take 4 books of the same type. If all 4 of your choices are different types, that means 4 trips to the library! :eek: But if you can choose your 4 books, that means one trip to the library. In that analogy, you are the GPU looking for certain chunks of RAM, the 4 books is your "bandwidth", and the library is the RAM. Get it?

[gwallen4]

Saph,

Thanks for the primer on graphics cards. Great stuff!

[Whyzman]

Yep, thanks again Saph...excellent piece! ;)

Either your total immersion into the inner workings of things is wearing off on you, or you have a natural gift to break things down into bite sized chunks.

I'd say one of these companies analyzing hardware would do well to put you on as a staff or guest writer! :)