Market behaviour and pricing strategies for consumer GPUs

30 series was affordable because the Samsung 8nm node was cheap.
Nvidia went with Samsung because they knew the competition could touch them on tsmc and because it was cheap.
40 series on TSMC5 was a huge increase in price per wafer, to the point where cost/xtor was likely unchanged at best (3x more density for 3x the price).

Think about it -- if $/xtor stayed about the same then a 4070Ti (100% of a 36B-xtor AD104 die) likely has a higher BOM cost than a 3080 (80% of a 28B-xtor GA102 die) did. The sticker prices track very closely. Other products along the stack were shuffled but on average the prices we see reflect the reality of the underlying component costs.

What some people on these forums are demanding is for NVIDIA and AMD to charitably swallow these costs in order to provide an illusion of continued perf/$ scaling when in fact Moore's Law is utterly dead.

Going forward we'll still see incremental scaling from higher clocks, and possibly dramatic scaling from algorithmic innovations (or alg/hw co-design). We may still see raw perf increases at the top end -- for a price.
Calling a 60% gross margin charitable is certainly something. Im not going to say anything else as the numbers speak for themselves. No argument you pose really holds weight when compared with the numbers.
 
Nvidia went with Samsung because they knew the competition could touch them on tsmc and because it was cheap.
You need to decide which process you're going to use long before you have any clue where your competition will end up. It was a risk which paid off, but it just aswell could have backfired.
 
Calling a 60% gross margin charitable is certainly something. Im not going to say anything else as the numbers speak for themselves. No argument you pose really holds weight when compared with the numbers.

60% is nothing, when you have to invest billions to make a proper tape out.
 
60% is nothing, when you have to invest billions to make a proper tape out.
Profit per unit is a rather vacuous metric. The more relevant figure is the company's net profits as that includes all costs - the argument for an excess of profits will reflect in their net revenues, although that's somewhat obfuscated by further investments. eg. a company could makes...shoes. They could operate on 5% profits a year for decades, growing slowly from a family company to a notable chain with world-wide branding. Then a change of management comes in and wants to make more money, so they decide to branch out into anime. They greatly increase the profit margins on their shoes and channel that into their anime business. The corporate financials perhaps show losses, the C-suite perhaps make more money than ever, the shoe workers maybe take a pay cut, and the purchasers of the shoes lament increased prices and lower construction costs and don't feel their shoe needs should be 'taxed' to pay for the corporate anime aspirations or the C-suites new yachts.

Just one possibility.

Given the basis of every argument on this matter is heavily stooped in one's own values, and the parameters very complicated, I don't understand the value in trying to have such debates. No-one will convince anyone to see things differently through any number of examples and statements, in contrast to proper engineering debates where the data should drive towards logical conclusions.
 
Last edited:
I’ve worked for a very large tech company for a long time. If you’re in a business unit with low margins it is awful. A few, or one, bad quarter and things can get tense. I’ve also been in business units that funded the rest of the company. One unit with high margins and high profits gives runway to strategic units that aren’t generating profits yet. It keeps a lot of people employed during hard times. That’s why I’m skeptical of anyone that tries to draw a line in the sand about how much profit is too much. They’re also typically people that will complain about company layoffs after a bad quarter. It gets more complicated when you’re talking about basic necessities like food shelter and medicine, but we’re talking about gaming and graphics cards. Nvidia is rolling in money now, but like Intel that can change fairly quickly. I don’t think anyone would have predicted Intel’s current situation five to ten years ago. I’ve also been around business units that missed the mark on a product generation and never recovered. There’s not really a too big to fail from profits alone, maybe there isn’t at all. It takes subsidies and monopoly manipulation to push in that direction.
 
What some people on these forums are demanding is for NVIDIA and AMD to charitably swallow these costs in order to provide an illusion of continued perf/$ scaling when in fact Moore's Law is utterly dead
Consumers have a right to demand this by only buying products when they do. It’s not my job as a customer to worry about wafer costs.

That said, most don’t share this conviction and will buy it anyways.
 
Consumers have a right to demand this by only buying products when they do. It’s not my job as a customer to worry about wafer costs.
Absolutely. I’m not going to buy a $2.5K GPU either. Doesn’t matter to me if others do. It’s not for me.

What I was pushing back against were these mouth-frothing “grrr… greedy corporations…” remarks. I mean yeah corporations are designed to maximize profit but in this specific case of consumer GPUs it’s not like they are dialing it up 10x. It’s business as usual.
 
What I think is more relevant to the thread disucssion though is that the appeal of a product (or service) for a consumer is affected by how they percieve said product and the company offering the product outside of the instrinc functionality of said product.

So things like whether or not it's "fairly" priced, how they view the corporations, hiring practice, social practices, and etc. do the affect the appeal of a product and by extension has been a part of marketing for some time now.

Although what's interesting is the above is also a two way relation. In that how a consumer feels about the functionality of a product also infuences how they may feel about the other factors such as whether or not a company is "good" or "evil."
 
I’ve worked for a very large tech company for a long time. If you’re in a business unit with low margins it is awful. A few, or one, bad quarter and things can get tense. I’ve also been in business units that funded the rest of the company. One unit with high margins and high profits gives runway to strategic units that aren’t generating profits yet. It keeps a lot of people employed during hard times. That’s why I’m skeptical of anyone that tries to draw a line in the sand about how much profit is too much. They’re also typically people that will complain about company layoffs after a bad quarter. It gets more complicated when you’re talking about basic necessities like food shelter and medicine, but we’re talking about gaming and graphics cards. Nvidia is rolling in money now, but like Intel that can change fairly quickly. I don’t think anyone would have predicted Intel’s current situation five to ten years ago. I’ve also been around business units that missed the mark on a product generation and never recovered. There’s not really a too big to fail from profits alone, maybe there isn’t at all. It takes subsidies and monopoly manipulation to push in that direction.

From direct knowledge of Intel inner workings, the big difference between nvidia and Intel is their focus on progress.

Intel was happy to sit back, lose talent out of boredom, didn’t keep up with salaries and were inflexible with remote working. By the time they had to get out of bed to compete, there wasn’t much talent on the bench left.

ARL is a giant f’up due to layoffs in key departments during development. Couple that low morale and arl represents the current state of Intel a whole: a mess.

Even without the AI boom, nvidia was keen to keep its foot on the gas gen after gen. And they clearly haven’t still.
 
I wonder how much of that comes from the top or is it fully ingrained in the company culture.
Has to come from the top. Left to itself a company tends to gradually turn into a short-term optimization machine. Employees start optimizing for their own personal protection/growth — which is not *wrong* as a co-optimization objective but toxic if it’s the only thing driving you. Groups become competitive tribes. Bean counters gain more power, and engineers and middle managers with the worst instincts end up getting hired and/or promoted. The degradation is natural. To prevent it you need a strong benevolent dictator at the top who imposes a vision that everyone rallies around, weeds out the crabgrass way before it can become a problem, and fends off attacks from vultures (active investors, Wall St) during tough times.

I hear Intel was an exciting place to work when Andy Grove (a founder) was running it. There was some rot in middle management of course but it was noise. Once he left the momentum continued for a while but the rot began to spread.

It’s truly sad — I heard from multiple people that Gelsinger brought back some of that old Andy Grove mojo back. He cleaned out a bunch of junk but the cancer had spread too far. I suspect expectations from the board were also unrealistic.
 
What I think is more relevant to the thread disucssion though is that the appeal of a product (or service) for a consumer is affected by how they percieve said product and the company offering the product outside of the instrinc functionality of said product.

So things like whether or not it's "fairly" priced, how they view the corporations, hiring practice, social practices, and etc. do the affect the appeal of a product and by extension has been a part of marketing for some time now.

Although what's interesting is the above is also a two way relation. In that how a consumer feels about the functionality of a product also infuences how they may feel about the other factors such as whether or not a company is "good" or "evil."
Interesting perspective. I think these ethical considerations may ever so slightly nudge a buyer in one direction or the other but are not a dominant factor.

I need to think more about your two-way doublethink hypothesis.
 
Back
Top