NVIDIA’s RTX 4090 is a $1600 GPU that will absolutely smash any real-world workload you throw at it. It’s indisputably packed with clever tech and absurd levels of raw power. However, even if you have the money, you probably shouldn’t buy one.
The Problem With “Halo” Products
The “halo” effect is when positive perceptions of one aspect of something bleed into other aspects of it. It happens in all areas of life, but when we’re talking about brands and products it can have a real effect on how customers perceive a product.
We see this in the car industry, where manufacturers make a super-expensive model with ludicrous performance and features, which no one can afford. Then they sell entry-level cars that are part of the same product line in massive numbers, partly thanks to prestige linked to the apex model.
Halo products like this exist not because they will make the company a lot of money. They exist as a way to show off prowess. The company is essentially showing the world what the best they can do looks like, but it has little relevance to the products most people can afford or are willing to buy.
The problem here (and it may not be a problem for everyone) is that a product that exists mainly to be a showpiece and not first and foremost a good product, is unlikely to be a good deal. It will likely have many impracticalities and come with hidden costs.
If we stick with the automotive analogy a little longer, when you actually buy an apex sportscar you’ll find that it won’t go over speedbumps, uses a ton of fuel, uses a fraction of its potential on public roads, and have insane running costs.
Poor Value at the Top
The way semiconductor production currently works means that some of the chips in each silicon wafer will have flaws. There’s the ideal perfect version of the chip that offers all the performance possible for that specific design, but only a tiny percentage of the actual chips come out perfectly. Those with flaws have the affected subcomponents disabled and are sold as cheaper products in the line.
The RTX 4090, in particular, isn’t even the full flawless processor die, but the higher up the stack you go, the worse the “yields” are. This is one of the reasons top-end chips are so much more expensive, even though they might not pack that much extra performance over the next model down.
Yields tend to improve as time goes by, but on a dollar-per-frame basis, the value at the top is generally relatively poor. The GPU that sells for half the price often offers more than half the performance. This means that every product line has a price-performance sweet spot, but it’s nowhere near the top of the line.
Of course, absolute performance matters as well. Great value per frame doesn’t matter if the end result is unplayable. But in practice, modern GPUs are so powerful and games so scalable that subjectively you may be leaving little or nothing on the table by going with a sweet spot card.
The Hidden Costs of Owning a Monster
The RTX 4090 may be the most egregious example of the hidden costs that come with halo GPUs, but this issue isn’t new. A GPU is only one part of a system, and if you want to get the most out of that system, no other component should hold it back. This means that the CPU, RAM, chassis, power supply, storage, and cooling solution must be up to the task of feeding the beast and keeping it happy.
In practice, your other components also have to be at or near the top of their respective product lines, compounding the poor value across the entire system. When you could get 60 or 70 percent of the same performance at half the total cost, that’s a hard pill to swallow.
The final bit of sticker shock comes from the energy requirements to run a computer built around a component like the 4090. If you want to get the maximum performance from your investment (and why wouldn’t you?), it’s going to be expensive, especially when energy costs are rising dramatically.
The Future-Proofing Argument Doesn’t Make Sense
One common argument in favor of buying the best GPU in the product line is that it will have the most longevity. Setting aside that the type of customer that typically buys an apex card usually upgrades immediately when something better comes out, there are some issues with this logic.
The main problem is that the next generation’s mid-range cards will likely perform close to or on par with this year’s high-end part. Not only that, they are likely to include new features and technologies that the older card won’t. While people are still using GTX 1080 Ti cards in 2022, that older model doesn’t have new acceleration technologies such as DLSS or hardware ray tracing.
An RTX 3060 Ti offers a little more raw GPU power than a 1080 Ti, but once you factor in new features and technologies, it will easily outpace it in modern titles.
It’s Not Just the 4090
While we’ve singled out the RTX 4090 as the poster child of halo GPUs that offer poor value for money, it’s only one example of these products that are meant to be shown and not really to be sold. That’s the main reason “no one should buy it” because it’s not a “product” in the true sense of the word.
Of course, different people value different aspects of a GPU like the RTX 4090. If you want the highest-performance computer and blow your nose with $100 bills, it’s the right type of component to buy.
However, if you’re one of the majority who have to consider value for money at all when purchasing computer components, a card like the RTX 4090 should not factor into your purchasing decision at all since its function is to skew the perception of other cards that are meant for volume production and sale. Even the next GPU down, the RTX 4080, is a hard sell in terms of value for money. It’s the equivalent of NVIDIA’s 60- and 70- series cards that are real-deal GPUs almost everyone should be looking to.
