It’s time to examine the new Nvidia GeForce RTX 4080, Nvidia’s $1,200 GPU representing a significant 71% price increase over the RTX 3080. Of course, Nvidia would like that we ignore the RTX 3080 entirely and instead concentrate on the $1,200 3080 Ti.
While the industry has undoubtedly altered since then, it will be intriguing to see what the RTX 4080 offers. There are many topics to cover and benchmarks to look at, but before we do that, let’s talk about the specs, look at the Nvidia Founders Edition model, and then go on to the blue bar graphs. We also acquired partner models from PNY, Asus, and MSI in time for this review, which we will highlight at the conclusion.
Pros
- GPU with the second-fastest speed (for now)
- A top-notch ray tracing performance
- It contains all of the Ada Lovelace upgrades
Cons
- Over-priced
- Needs DLSS 3 for true gaming performance
- Persistent worries about the 16-pin connection
The Nvidia GeForce RTX 4080 is the follow-up to the RTX 4090, released last month, and is now one of the greatest graphics cards and is ranked first in our hierarchy of GPU benchmarks. Of course, the melting 16-pin connections have taken a little bit of the shine off. The good news is that since the RTX 4080 consumes less power, it should be less likely to channel enough power to melt the plastic connection. The bad news is that it inherits the RTX 3080 Ti launch price, which we also thought was too excessive, and is still too expensive for most gamers at $1,199. It also marks a significant generational price rise.
The GeForce RTX 4080 is built on the 378.6 mm2 AD103 silicon, which is 38% smaller than the AD102 silicon used by the RTX 4090 and has 40% fewer transistors. However, the 4080 still has 62% more transistors than the 3090 Ti flagship of the previous generation. Given those numbers, it shouldn’t be surprising that the 4080 has 41% fewer CUDA cores than the 4090, a 33% narrower memory bus, and 11% less L2 cache.
As for core clocks, there isn’t much of a difference between the RTX 4090 and 4080; boost clocks are 2505 MHz versus 2520 MHz, with the 4090 having a slight advantage. The RTX 4080 receives faster 22.4 Gbps GDDR6X memory, a 7% improvement over the series flagship, but the overall memory bandwidth has been reduced by 29% to 717 GB/s due to the smaller 256-bit wide memory bus.
How is the RTX 4080 constructed?
The AD103 chip, a second-tier Ada Lovelace GPU, is used in this RTX 4080 16GB card. The AD104, a third-tier GPU that should practically only be used in an RTX 4070-series card and not a modified RTX 4080, was chosen as the GPU for the 12GB card. Which, I suppose, is where we’ll see it reappearing next, with or without the “Ti” suffix.
The RTX 4080’s AD103 GPU, though, has all the architectural awesomeness that made the RTX 4090 such a potent new chip, regardless of how the AD104 stuff works out in the end. In addition to the frame-generating magic of DLSS 3, you also receive the fourth-generation Tensor Core and the third-generation RT Core. Naturally, this somewhat smaller piece of silicon.
| GeForce RTX 4080 | GeForce RTX 3080 Ti | |
|---|---|---|
| Lithography | TSMC 4N | Samsung 8N |
| CUDA cores | 9728 | 10,240 |
| SMs | 76 | 80 |
| RT Cores | 76 | 80 |
| Tensor Cores | 304 | 320 |
| ROPs | 112 | 112 |
| Boost clock | 2,505MHz | 1,665MHz |
| Memory | 16GB GDDR6X | 12GB GDDR6X |
| Memory speed | 22.4Gbps | 19Gbps |
| Memory bandwidth | 717GB/s | 912GB/s |
| L1 | L2 cache | 9,728KB | 65,536KB | 10,240KB | 6,144KB |
| Transistors | 45.9 billion | 28.3 billion |
| Die Size | 378.6mm² | 628.4mm² |
| TGP | 320W | 350W |
| Price | $1,200 | £1,269 | $1,200 | £1,200 |
Source: pcgamer
Nvidia has chosen to utilize the same MASSIVE cooler used for the RTX 4090 Founders Edition, despite the GPU in the RTX 3060 Ti being smaller than the Ampere chip at its core. I have a PNY RTX 4080 that is the size of a tiny moon; the add-in board partners (AIBs) have also adopted this strategy.
Additionally, it uses a PCB that is quite similar to the RTX 4090 Founder Edition, a modified PG136. To handle the 320W Total Graphics Power (TGP) that the RTX 4080 generates, the board must have the same cutoff and be highly power-phased.
It has a common 16-pin power connector as well, but the supplied adapter is a triple-headed one rather than a quad-headed one like the RTX 4090’s. Although it’s not good for a very expensive, powerful graphics card, the cause of the melting 16-pin adapters in the wild is still unknown, Nvidia is continuing to work on the issue. Also, AIBs are.
4K gaming performance RTX 4080
Credit: Future
- Credit: Future
Is an RTX 4080 a good investment?
When all other factors are taken into account, the fact that Nvidia can remove such a large portion of the GPU’s best features from its AD103 silicon while still performing at a level that is only a third lower than the much larger AD102 core indicates that it provides what you would want from the RTX 4080. Although the shroud is unreasonably large, it is quick, the GPU is surprisingly compact, and I am absolutely in love with DLSS with Frame Generation.
Nvidia has a history of determining card prices solely based on performance rather than the actual silicon used. The company has improved performance and efficiency going back to the Kepler architecture and the GTX 680(opens in new tab), and as a result, priced lower spec GPUs higher nominally.