Besides the box art, I miss the days when 1) the graphics card didn't cost more than the rest of the components put together, 2) the graphics card got all of its damn power through the connector itself, and 3) MSRP meant something.
I'm not in the market for a 5090 or similar, but the other day I was looking at a lower-end model, an AMD 9060 or Nvidia 5060. What shocked me was the massive variation in prices for the same model (9060 XT 16 GB or 5060 Ti 16 GB).
The AMD could be had for anywhere from 400 to 600 euros, depending on the brand. What can explain that? Are there actual performance differences? I see models pretending to be "overclocked", but in practice they barely have a few extra MHz. I'm not sure if that's going to do anything noticeable.
Since I'm considering the AMD more and it's cheaper, I didn't take that close a look at the Nvidia prices.
Looks. I'm not joking. The market is aimed at people with a fish bowl PC case that care about having a cooler with a appealing design, a interesting PCB colour and the flashiest RGB.
Some may have a bit better cooling but the price for that is also likely marked up several times considering a full dual tower CPU cooler costs $35.
The manufacturer can use better fans that move more air and stay more silent. They can design a better vapor chamber, lay out the PCB in a way that VRMs and RAM gets more cooling. But still all that stuff should not account for more than $30-50 markup.
Hey, c'mon now - some of that is flooding the market so hard that it's ~8:1 nVidia:AMD on store shelves, letting nVidia be the default that consumers will pay for. That's without touching on marketing or the stock price (as under-informed consumers conflate it with performance, thinking "If it wasn't better, the stock would be closer to AMD").
>What shocked me was the massive variation in prices for the same model [AMD v. nVidea]
I am not a tech wizard, but I think the major (and noticeable) difference would be available tensor cores — that currently nVidea's tech is faster/better in the LLM/genAI world.
Obviously AMD jumped +30% last week from OpenAI investment — so that is changing with current model GPUs.
I just bought a RTX 5090 at MSRP. While expensive, it's also a radically more complicated product that plays a more important role in a modern computer than old GPUs did years ago.
Compared to my CPU (9950X3D), it's got a massive monolithic die measuring 750mm2 with over 4x the transistor count of the entire 9950X3d package. Beyond the graphics, it's got tensor and RT cores, dedicated engines for video decode/encode, and 32GB of DDR7 on the board.
Even basic integrated GPUs these days have far surpassed GPUs like the RTX 970, so you can get a very cheap GPU that gets power through the CPU socket, at MSRP.
Do yourself/me a favor, and give your 5090's power plug/socket a little jiggle test.
I'm a retired data center electrician, and my own GPU's has been "loose" at least more than once. Really make sure that sucker is jammed in there/latched.
> the graphics card didn't cost more than the rest of the components put together
In fairness, the graphics card has many times more processing power than the rest of the components. The CPU is just there to run some of the physics engine and stream textures from disk.
The existence of scalpers rather shows that the producer set the price of the product (in this case GPU) too low [!] for the number of instances of the product that are produced.
Because the price is too low, more people want to buy a graphics card than the number of graphics cards that can be produced, so even people who would love to pay more can't get one.
Scalpers solve this mismatch by balancing the market: now people who really want to get a graphics card (with a given specification) and are willing to pay more can get one.
So, if you have a hate for scalpers, complain that the graphics card producer did not increase its prices. :-)
