Manufacturing costs of video cards?

armandh01

Weaksauce
Joined
Dec 31, 2009
Messages
123
Does anyone have any good articles of video card manufacturing costs, even if old? I am curious to how much profit they make per video card. It is proving difficult to google the answer.



Or if you have industry knowledge that could enlighten me, thanks.
 
I think you should rather not find that answer, it may depress you about how much we pay as consumers. Typically the market standard is to get an average of 10x profit. That means that if a card cost around $10 to manufacture, it would cost you around $100.. I use to work at a manufacturing facility and I would handle the order bills for special magnets. It cost us 8 cents to manufacture one rare earth magnet, It was sold on average for $35 or more.
 
There's a lot more that goes into making a video card than just putting it together, though. Imagine the costs of R&D. I'm sure that's where a majority of the cost of a video card lies, by a long shot.
 
I think you should rather not find that answer, it may depress you about how much we pay as consumers. Typically the market standard is to get an average of 10x profit. That means that if a card cost around $10 to manufacture, it would cost you around $100.. I use to work at a manufacturing facility and I would handle the order bills for special magnets. It cost us 8 cents to manufacture one rare earth magnet, It was sold on average for $35 or more.

This is incredibly short sighted.

While a company is out to make profit (It is a business after all); You cannot look at just the cost of a particular products BoM as being the *only* cost for this company.

Sure, it cost your employer 8 cents per magnet, but how much were they paying you? How much did it cost to lease the land and the building? One could go on and on.

When it comes to technology in particular the BoM does not tell anywhere near the full story when it comes to cost.
 
Q2/2009:
H5Q8X.jpg


Q3/2011:
vkRDt.jpg


Both from Mercury Research.
 
One thing you have to remember when evaluating this is that the profit is not their sale price minus their unit cost. There are two additional major factors:

1) Retail markup. The price you pay is not the price they get, retailers mark things up, usually around 50-100%. So if something sells for $10 to you and it costs the company $2 to make, they are likely getting only in the realm of $3 in revenues per unit.

2) Fixed costs, mainly R&D. It costs hundreds of millions of dollars (or more) to develop a new GPU. That money has to be recouped. So while a card might have a certain amount of revenue per unit, you also have to spread the fixed costs across units sold to get an actual profit margin.

A more minor factor is external costs, like the cost of advertising, having facilities, tech support, and so on. Given the fabless nature (ie outsourced manufacturing) of the GPU makers this isn't as large as the other two, but it is still not trivial. This again has to be spread across units.

Remember that profit is money left over only after ALL expenses are covered. Revenue is the difference between what something costs and what you sell it for.
 
Today I would imagine R&D is the majority of the cost that needs to be amortized. Didn't nvidia spend around a billion dollars for Turing R&D?

I believe some people have worked out die cost is probably around $150 for a Titan RTX. The actual cards aren't going to cost that much to make, and compared to cards like GTX 580 the profit margin on a Titan RTX/2080 Ti on a per card basis for Nvidia are enormous. However their R&D costs have gone up significantly since the 40nm days, as have their appetite for profits.
 
Today I would imagine R&D is the majority of the cost that needs to be amortized. Didn't nvidia spend around a billion dollars for Turing R&D?

I believe some people have worked out die cost is probably around $150 for a Titan RTX. The actual cards aren't going to cost that much to make, and compared to cards like GTX 580 the profit margin on a Titan RTX/2080 Ti on a per card basis for Nvidia are enormous. However their R&D costs have gone up significantly since the 40nm days, as have their appetite for profits.

There's actually two parts to the R&D that gets rolled in to cards. The R&D nVidia does, but also the R&D TSMC does. That's one of the reasons a new process is expensive. There's a fair amount of cost making the chips period (because of the precision involved, modern wafers are expensive) but then there's also the cost of the fab. It is over ten billion dollars to develop and build a modern fab, and of course TSMC is going to earn all that back.

There's a lot of R&D dollars in high end tech shit. Not to say nVidia isn't overcharging for their newest stuff, but there's a lot of money up front to make it happen.
 
Back
Top