Nvidia GeForce RTX 4080: Packs a Punch, but Costs Too Much
The RTX 4080 has finally arrived and as expected, it’s a seriously potent GPU. It’s officially the second most powerful GPU money can buy. That would normally be good news for gamers, but its $1,199 price puts it in an odd space within the market. The RTX 4090 offers a lot more performance, and you’re already in the ballpark as far as pricing goes with the 4080. This will leave a lot of gamers in a conundrum over which GPU to buy, assuming they can actually buy either one of them. Still, the RTX 4080 is undeniably powerful, and a big step up from cards like the RTX 3080. However, Nvidia has tacked $500 onto the price tag compared with its x80 predecessor, which is a big jump. Also, AMD’s RX 7900 GPUs are waiting in the wings at $999 and $899, which makes Nvidia’s pricing look even more questionable.
Although Nvidia endowed the RTX 3090 and 3080 with the same “big” GA102 die for Ampere, it took a different approach with Ada Lovelace. Only the RTX 4090 gets the big die, with the RTX 4080 sporting a much smaller AD103 chip. It’s roughly half the size – 379mm² versus 609mm² – of that in the RTX 4090. As you can imagine this leaves quite a performance gap between the two GPUs. Thus, it’s easy to see Nvidia eventually introducing an RTX 4080 Ti with a neutered AD102 die. Nvidia also clipped the RTX 4080’s wings as far as its memory bus goes. The RTX 3080 had a 320-bit bus, while the 4080 is limited to 256-bit. It compensates a bit by offering more L2 cache than the RTX 3080 though. It has 64MB compared with the 3080’s paltry 5MB.
The RTX 4080 is rocking 9,728 CUDA cores with 16GB of GDDR6X memory. Its base clock is 2.21GHz with a boost clock of 2.51GHz. The GPU is fabricated on TSMC’s 4nm process and has a target power consumption of 320W. That’s a bit easier to swallow than the RTX 4090’s 450W, and the same TDP as the original RTX 3080. It’s impressive that Nvidia kept the power target the same between generations, despite providing significant performance uplift. That’s expected with a node shrink, however. It’s the same size physically as the RTX 4090 Founders Edition. One small change is the adapter cable is a three-into-one design instead of a four-into-one for PCIe power cables. It’ll be interesting to see if this still results in cables melting.
For benchmarking, our colleagues at PCMag.com ran the card through its paces. Its test bench consisted of an Asus ROG Maximus Z690 Hero motherboard with a stock Core i9-12900K. It’s running 32GB of DDR5 memory clocked at 5,600MHz. A 1TB PCIe 4.0 SSD and 1,500W PSU from Corsair round out the system. The CPU is also liquid-cooled via a Corsair cooler.
The Benchmarks: Synthetic Tests
It’s been said there are usually few surprises in GPU launches. You look at the spec sheet and that’s how the cards line up, barring things like boosts from DLSS. In synthetic tests, that’s exactly what we see. The RTX 4090 and 4080 tower over the competition in every test. In Time Spy Extreme’s Graphic score, the 4080 is 60 percent faster than the RTX 3080 it’s replacing. The same delta exists between the high-end AMD GPUs as well. However, the RTX 4090 has a significant advantage over its sibling, to the tune of 40 percent more performance. Regardless, the 4080 has a commanding lead over every other GPU tested. In these tests, the 4080 is around 40 percent slower than a 4090 while being 40 to 50 percent faster than the RTX 3080.
AAA Gaming Benchmarks w/DLSS & FSR
Once again, there are no surprises here: the 4080 is the second-fastest GPU on the block. The same delta we saw in synthetic tests is present, with a 40 percent delta to the 4090, and in reverse to the 4080. However, Deep Learning Super Sampling injects some interesting performance results. Essentially, the RTX 4080 with DLSS can match an RTX 4090, at least in Guardians of the Galaxy. It also showed CPU bottlenecking at 1080p, which is not surprising.
Once again, not many surprises here in standard AAA gaming. At 1080p we see some CPU bottlenecking. Then at 1440p the RTX 4080 starts to gap the field. Then at 4K it really stretches its legs. Both the 4080 and 4090 are a clear step above every other GPU in the field right now. The 4080 is 50 percent faster than the RTX 3080 at higher resolutions. As always, there’s a 30 to 40 percent gap to the RTX 4090 as well. Still, the RTX 4080 goes from around 60fps with high-end Ampere to 100 fps and higher at 4K in every game except Total War: Three Kingdoms.
Power Consumption and Thermals
Most of our fears about Ada Lovelace GPUs resembling nuclear reactors have been assuaged at this point. Yes, they have big coolers, but that keeps them cool and quiet. Plus, the RTX 4080 has a 320W TDP, which is in line with previous high-end GPUs. Overall, it seems like Nvidia did some excellent engineering on its cooler. Even in the FurMark torture test, it peaked at 62C. In the actual game of Guardians of the Galaxy, it hit 63C. The total system power draw was 447W, which seems reasonable compared with the 4090’s 577W. It’s also a lot less than the RTX 3080s 535W, showing Ada Lovelace’s efficiency gains over Ampere. Overall, this is one category where Nvidia’s GPUs really shine.
In 2016 the GTX 1080 launched at $599. In 2018 the RTX 2080 launched at $699. In 2020 the RTX 3080 had a price tag of $699. We can all see something of a pattern here. However, cue the “record scratch” noise for the RTX 4080 launch at $1,199. That’s an outlandish price jump, and it has turned a lot of people off. What’s interesting is the price makes sense compared with the $1,599 RTX 4090. But if it had kept up the pricing scheme from before and sold it for $899 or so, it would be an instant classic. Clearly, Nvidia believes it can charge more for its GPUs than it used to. At least, it can do so for the high-end stuff. It’s probably right, as the RTX 4090 has been sold out since launch. We’re sure the same fate awaits the RTX 4080.
In a vacuum, Nvidia’s pricing makes sense in terms of its performance. However, back in the olden days, we’d see a new generation of GPUs offer more performance than their predecessor at the same price. That’s clearly not the case anymore. After the RTX 4090 launch, Jensen said those days were over since Moore’s Law is dead. Everything costs more now, he said (we’re paraphrasing). Still, it’ll take AMD GPUs for us to put Nvidia’s latest effort into proper context.
On the one hand, the RTX 4080 is a blisteringly fast GPU. It can easily handle both 4K and 1440p at high fps, and with DLSS 3 it’ll outpace Ampere cards even further, eventually. And it runs cool and quiet too. It’s just so expensive, especially when the RTX 4090 is right there. Nobody who buys it will be disappointed by its performance. But they also might be kicking themselves when the RTX 4080 Ti comes along, especially if Nvidia offers it at the same price.