Why are desktop card so big compared to mobile?

I've been wondering for quite some time what's different between the PCBs. A Mobile card is quite simple, it has circuits linking the memory, the GPU and the port.
On the other hand, a desktop card (especially high end) is really big. There seems to be a bunch of circuits, big metal cylinders and all that stuff. What are they there for?

 

Desktop graphics cards are a lot more powerful, they need a lot more power, produce a lot more heat and typically have more features than mobile cards (multiple ports, connections, etc.)

 

The circuits are pretty much the same, it's just the desktop version can use larger, and thus cheaper, components (for example... those "large metal cylinders" are capacitors).
It costs more money to make the same circuit physically smaller, which is why prices on laptop cards are so steep. If you've got the space, might as well use it to keep the price low.

 

Ok thanks. That's the answer I was looking for.

 

So they can be better and cheaper.

Nuff' said.

 

Bigger stuff = Cheaper to manufacture and that means money saved.

 

Plus more surface area for cooling.

 

Current top of the line desktop single cards are roughly 4 times faster. Dual cards (they have 2 GPUs and twice as much memory on one PCB, basically single board SLI/Xfire) are between 6-8x faster.

A Desktop card outputs to up to 6 monitors. A laptop card will only do 2.

That's um.. why.

 

Plus that is why you see a lag between the desktop tech and laptop tech. It takes more time to scale the tech down too, and as others have said, cooling and cost.

 

Wouldn't it be the opposite? Less materials used = cheaper to sell?

I do get what people are saying by desktop cards being cheaper in terms of the manufacturing process, but wouldn't the amount of materials used in a desktop card negate this?

 

While less material, it still has to do the work of a much larger component. Thus miniaturization = expensive.

 

But the work it does still isn't equivalent. The desktop card is also a lot more powerful. You would think that it would also add to a desktop card being being more expensive.

And notebook cards aren't all that expensive. You can grab a 3650 or a 9600M GT off Ebay for under $100.

It's only stuff like the GTX 280M and the 4870 that are expensive, obviously because they are powerful.

 

Engineering to get even near the power of the desktop cards is where the money is at. As said before, the desktop equivalents are almost x4 faster. I understand where you are coming from, but the money isn't in the components themselves, but rather the thinking and engineering behind them.

 

The material costs are close to nothing. The engineering costs are what make the card cost money. You've got $5 in materials on a large desktop card and $2 in an MXM card. Engineering these materials to be 1/8th the size and have the same function is what drive up the price.

 

It's not as much more powerful (although that is a big factor), but newer cards cost more just due to supply and demand and like anything else, top end or new parts cost more to cover the cost of development and manufacturing, not to mention yields. Older tech has higher yield (less waste) chips than newer tech. There's tons of factors, but it all comes down to cost of miniturization.

 

Because they can be less efficient.

 

Bedsides engineering the a smaller card, the equipment that has the precision and accuracy (since it's a smaller area, you can only have lower tolerances) needed to manufacture it is even more $$$

 

because desktops have SPACE, they arent restricted at all! its as simple as that, just because it can be as small as a mobile doesnt mean it has to be, also they are all universally PCIEs, even though motherboard configs are different the slots are the same. but the placement can be different.

mobile motherboards differ per company per model, if notebooks were universally the same then it would be alot easier to swap out parts

 

Not in the slightest. Lower production numbers and miniaturisation (and the cost of doing so) raises the costs much more than the small amount of materials saved. The actual die is one of the most expensive components, along with the actual design/engineering.

 

Indeed, thats why ati hd4650 cards for laptops dont drop below 200 euros, while the dekstop versions costs like 60 euros.
SIgh...I'll just wait another year till I buy that ati

 

Like I said, miniaturization is an expensive endeavor.

 

As others have said, no.

 

Surprised no one has pointed out that die shrinks often bring the cost down.

 

Only because the cost per chip is lowered, as (usually, low yields not withstanding) they can produce more chips per wafer, which have a fixed cost. That difference is still not always passed onto us, and doesn't avoid the costs involved in moving to the new process, nor sometimes making some slight tweaks to the design, plus it's pretty much the only opposition to this general rule. Further more, the dies used in mobile and desktop cards are often very similar or the same, just with different clocks, and perhaps cut elements, so if the die is shrunk, then the desktop card potentially becomes cheaper too, resulting in the same deficit in price difference.

Aside from that however, its largely a case of smaller = more expensive.
The same can be seen in audio technology, not just computers

 

Desktop cards are more powerful than mobile cards.

For example, the GTX 280M(mobile card) is about 40% weaker than the GTX 280(desktop card). The GTX 280M(mobile card) is equivalent to the GTS 250(desktop card).

 

But the decrease in price has already been passed on to us. A 8800m gtx can cost up to 700$ from reseller. When the 9800m cards hit the seen they were almost half the price. Now you can get a gtx 280m in the 400$ range.

 

No disagreements on smaller dies being cheaper, but thats because of the wafer point again (fixed wafer cost, smaller die gives more dies per wafer, dependant on yield). Die shrinks are main exception to the rule, and even then it's not always clear cut as you have to factor in lower yields etc. For example, the 40nm currently used by TSMC for the 5 series is hitting a roughly 40% yield last I heard, other older processes have higher yields, meaning sometimes the larger die is actually cheaper as you're not paying for anywhere near the same number of failures. <actually this was apparently one of the reasons the 2900XT was so unambitious when it arrived, due to TSMC messing up they were forced to use a larger die size than intended, which forced up cost, heat and necessitated a reduction in core clock speeds>
Last I heard it was approximately $5000 per silicon wafer, so however many chips can be taken from one wafer equates to a large percentage of the final cost for that chip. The older cards were 55nm, however if the yield was over 70-80% in which case they theoretically would be cheaper than 40nm 40% yield, despite the larger size. If you're looking at lower yields, throw in the cost of miniaturising parts beyond the die and lower manufacturing numbers then you encounter a premium.

I didnt say the cost reduction is never passed to us, just that its not always seen straight away, nor in the same product (the 55nm GTX260 was more expensive than the 65nm version IIRC when it came out), as you're not just paying for the fab costs, you're paying dependant on other components and wafer yields.

Its all fun and games!

 

... Till someone gets wafer'd.

 

Die shrinks don't necessarily mean lower costs per chip. To take advantage of die shrinks, they gotta retool the Fab line. Rather they get better yields and thus sell more. So it is more like lower cost per wafer.

 

Aye, that was one of the things I tried to get at.

 

The core of each desktop and mobile card is actually the same. The only difference is the other parts on the PCB around the core.





It isn't exactly that the cards are harder to make, but that the market for add-in notebook cards is just not there as notebook manufacturers don't wan't people holding onto one notebook for a long time, when they could be paying for a new one every year or two.

The extra space is usually for the extra capacitors, resistors etc. that feed extra power onto the card so that it can run these higher clocks than their mobile varients, and create more surface space to spread heat.

 

Spreading the heat might be just about the only thing worth of having on larger desktop GPU's ... at the same time ... decent cooling in laptops will keep temps stable, and there shouldn't be any issues (otherwise they wouldn't be viable to begin with).

The real reason is what 'unknown555525' said:
'It isn't exactly that the cards are harder to make, but that the market for add-in notebook cards is just not there as notebook manufacturers don't wan't people holding onto one notebook for a long time, when they could be paying for a new one every year or two.'

Translation: the manfuacturers do it because they can and want to make a profit.