AMD's new RDNA 3 graphics (opens in new tab) is less than two weeks away from retail availability, while Nvidia's RTX 40 series (opens in new tab) has been on the market for nearly two months. But if I were to invest my personal money in a new GPU in the near future, there is no doubt what I would buy. It's the GeForce RTX 3080 (opens in new tab). Yes, the old fashioned 10GB card.

Very recently, the RTX 3080 has finally become available at a price close to the original MSRP. As I write this, the cheapest is the Peladn Gaming RTX 3080 (opens in new tab) at $720, slightly above the original $699 that RTX 3080 was recommended for in late 2020.

Certainly, the 12GB version is tempting. However, the cheapest non-refurbished version is currently $999, a bit too expensive for the meager profit. It also undermines what I think is, absurdly, an excellent value proposition for a 10GB card. This is the world we live in; a $700 GPU is now a "value" option.

Anyway, the appeal of the RTX 3080 lay from the start in the fact that it is based on the same GA102 Ampere chip as the $1,500 MSRP RTX 3090. (Open in new tab) When buying a GPU, a slightly cut-down version of the top GPU is always the best long-term bet, and the RTX 3080 proves that beautifully. In fact, as I recently explained (opens in new tab), the new RTX 4080 (opens in new tab) is based on Nvidia's top new AD102 Lovelace chip, not the second-tier AD103 GPU as is found in the powerful RTX 4090, which is a very big disappointment.

In other words, the RTX 3080 offers more than 80% of the hardware and experience of the RTX 3090 in metrics such as shader count and frame rate, while the RTX 4080 is less than 60% of the 4090. What's more, the cheapest RTX 4080 is currently over $500 more expensive. This is a premium of roughly 75% at current prices. And the RTX 4080 is almost 75% less fast than the RTX 3080.

Furthermore, if I'm paying $1,250 for a graphics card, I want something special, not a compromise like the RTX 4080. I would sooner wait a while for the RTX 4090 to drop down to near its MSRP of $1,599.

As for AMD's upcoming Radeon RX 7900 XT and XTX (open in new tab) boards, they simply are not doing it for me. The Internet is glowing with predictions that AMD's new boards will tear Nvidia apart thanks to their sharp price-performance proposition. But even at MSRP, the cards are $899 and $999, respectively. And there is a good chance that in the months following its release, it will be priced far above these numbers. That is if you can afford them and they don't sell out in a matter of minutes.

Even by AMD's own claims, the ray tracing performance of the new RDNA 3 architecture (open in new tab) is substandard; AMD expects a 1.5x to 1.6x improvement over RDNA 2's painfully weak ray tracing performance. The whole ray tracing thing is a bit of a nightmare, to be sure. Does ray tracing really matter? Is there a clear difference between ray tracing enabled and disabled in a given game?

Still, I find it very difficult to compute the idea of paying $1000 for a GPU with obvious performance weaknesses. My guess is that the RX 7900 XT or XTX (open in new tab) will be little if any better than the RTX 3080 at playing games that use ray tracing, if indeed they are better. The RTX 3080, on the other hand, is still a beast in older raster games. So I'd be much happier spending $720 on an RTX 3080 than spending $1,000+ on a new AMD board.

At launch in 2020, the RTX 3080 looked like a killer proposition. Only to be ruined by a massive spike in GPU prices. But the fact that the RTX 3080 is still so attractive more than two years later, even though it has just reached MSRP, is a testament to how good a spec it was.

Really, only the fact that Nvidia has locked in frame generation technology from DLSS 3 (opens in new tab) to RTX 40 series boards makes me think less. But I can live with it. Frame insertion does not improve latency, so in my opinion it is not a literal game changer as the DLSS resolution scaling has proven to be.

What I am less concerned about is the prospect of the next RTX 4070 or RTX 4070 Ti. Given the exorbitant pricing of Nvidia's RTX 4080 and the disappointing specs of that GPU, it is unlikely that the RTX 4070 and RTX 4070 Ti will dramatically beat the RTX 3080 in value. It's hard to imagine that the RTX 4070 Ti will be a rebadged version of Nvidia's infamous "unlaunched" (open in new tab) card, the 4080 GB, and the RTX 4070 will be even lower-spec.

My hunch is that the plain, non-Ti RTX 4070 will perform about the same as the RTX 3080 in raster games, maybe a little faster in ray tracing, and frame generation will be advertised in all comparative Nvidia benchmarks. And it will be more expensive than the $720 RTX 3080, at least at launch early next year and for a few months thereafter. In a worst-case scenario, it would be a slightly better buy. But not enough to really regret pulling the trigger on the RTX 3080 today.

So that's where I am now: two years after the RTX 3080's launch, with its replacement widely available and AMD's new GPUs just days after launch, the RTX 3080 will get my money completely, willingly, and without reservation. Am I missing something? Am I missing something?