Confused about 256-Bit vs 128-Bit Memory Interfaces

[Player_13]

I've been learning more and more about a computer's innards as I go along in my interest concerning them. Building and upgrading computers, reading about PSUs, GPUs and CPUs, etc. But one thing that has confused me for a while is the difference between a 256-Bit and 128-bit memory interface, specifically when it comes to graphics cards.

I assume it has something to do with memory bandwidth (I see that term thrown around a lot) and I know there's a large performance gain, but it seems to always be accompanied by a huge price increase. Why is the difference between the interfaces price so large?

Anyone care to enlighten me?

[Ultima]

I think it's best to explain this in a simple way.

The wider the memory bus, the larger the GB/s throughput.

Same goes for system ram, look at before, AMD socket 754, has single channel ram (64bit wide), which comes in, in theory, @ 3200MB/s.
Then came socket 939 along with dual channel (128 bit), resulting in theory into 6400MB/s

The reason why the price difference between those things in videocards is not because of the extra throughput, but because the card usually needs extra layers of pcb to get this working just fine, and more layers of pcb results in higher price.
I mean, it's not like they pull a little switch somewhere on the card and say, now it's 128 bit and now it's 256bit.

On the outside the cards may look the same, but for instance a card that knows both, the ATI 9800. the 128bit and 256bit version may look the same, but on the inside, they are really different.

And that the throughput is higher, that will indeed help the price go up, because it's also a marketing thing, bigger is always better in the eyes of the unknowing consumer.

[SoloCamo]

Well actually, Ultima, when it comes to comparing the same/similar card, especially when it comes to memory interfaces, bigger is always better.

[Player_13]

Hmmm. I suppose that makes sense. I guess I always thought of an interface being something fairly simple between two components of the card. I thought it would be simple to switch between the two.

Another reason I asked is that I was in the market for a new video card and while looking I noticed a Geforce 8 series card that was fairly well priced. I checked the specs on wikipedia and noticed it actually had a Max Memory Bandwidth only about 4GB/s higher than my Radeon 9800 Pro!

I checked the specs of the two cards and noticed that the 8600M GT (card I was looking at) and my 9800 had one main difference, and that was the 128bit vs. 256bit memory interface.

I hope that my 8600M GT (in my new Dell Laptop) will still outperform my Radeon in a lot of cases (at least in DX10 apps). I love my Radeon, it performed super well, but it's getting on in the years.

[Ultima]

Quote:

Originally Posted by SoloCamo

Well actually, Ultima, when it comes to comparing the same/similar card, especially when it comes to memory interfaces, bigger is always better.

actually I never did say that they are the same, I also didn't say bigger is better, that I forgot, but....

in this case, bigger is indeed better, unlike the amount of ram on the card, that stops being better at a certain point, depends on the chip that is on the card ofcourse.
for example, a 256MB 9800XT is better then a 128MB version, but any bigger then that is plain overkill, the resolutions you'd need to play at to make use of the extra ram are so high that the gpu just ain't powerfull enough to handle it.