
The launch of Nvidia’s first two RTX Ampere 30 series GPUs has not gone smoothly, to say the least. The RTX Ampere 3080 sold out within eight seconds of its launch, the RTX Ampere 3090 was advertised of being capable of processing games at an 8K resolution, though benchmarks showed it couldn’t even handle doing so at a mere 30 frames per second, and then it turns out that several of the aftermarket RTX Ampere 30 series GPUs appear to be experiencing instability and crashes during overclocking. The latter issue appears to be caused by AIB manufacturers using cheaper POSCAPs, without MLCC capacitor groups in between, which ultimately creates interference that is amplified during overclocking, resulting in the reported instability and crashes: An issue that could have been avoided if Nvidia had provided the proper specifications required for these GPUs to their manufacturing partners. A special mention must go to ASUS, for using an all MLCC layout and not opting out for any of the cheaper POSCAPs.
But all negativity aside, Nvidia does have the opportunity to gain back its favor with the release of the new RTX Ampere 3070. If some of the early benchmark reports are to be trusted, this GPU appears to have one of, if not the best price to value ratio among all three announced RTX Ampere 30 series GPUs. In fact, it shows such high framerate performance that that many are comparing it to the RTX 20 Series flagship: The RTX 2080 Ti. Some YouTube channels are showing similar performance between the two GPUs, while others are showing a significant advantage for the RTX Ampere 3070, but most of these appear to be fake. Whether or not they are true, the RTX Ampere 3070 does have the potential of becoming the first mid-range GPU that allows mid-budget builds to experience the luxury of gaming at 1440p or 4K resolutions, without sacrificing framerate. In order to see how well the RTX 3070 compares against the RTX 2080 Ti, let’s first compare the two GPUs in terms of specifications, and find out to what potential the RTX Ampere 3070 can perform in comparison to its RTX 20 Series counterpart that is 170% more expensive.
RTX 3070 VS RTX 2080 Ti – Specifications
In comparing these specifications, we will take into account their memory, their processing power, and finally their power consumption. Some separate inherent advantages of the new generation of RTX graphics cards that give the RTX 3070 a slight advantage against the RTX 2080 Ti are the new 8nm Samsung manufacturing method (compared to the 12nm FinFET of the RTX 2080 Ti) and the new Ampere architecture.;
Memory
When it comes to VRAM, unlike its RTX Ampere 3080 and 3090 counterparts, the RTX 3070 does not make use of GDDR6X, so it does have the same GDDR6 found in the RTX 2080 Ti – but much less of it. The RTX 3070 comes with 8 GB of RAM with a 14 Gbps memory speed, a 256-bit interface, and a 448 GBps bandwidth. The RTX 2080 Ti, on the other hand, has 11 GB of GDDR6 VRAM, at a 14 GBps speed, a 352-bit memory interface, and 616 GBps bandwidth. The RTX 2080 Ti does have a clear advantage over the RTX 3070 in this department, but when it comes strictly to gaming, the memory will not play that much of a role, especially at lower resolutions. VRAM is mainly used to store textures and 3D objects at a distance, and not having enough VRAM will mean that textures may appear late, or not load properly. However, if the memory cap isn’t hit, then no difference will be felt. 8 GB of VRAM, for most, if not all, gaming applications is more than enough to handle the task. The only exception here is that some triple-A titles are slowly reaching the 10 GB VRAM cap at the 4K resolution; and as the new generation of games release, the 4K VRAM requirements may also see an analogous increase.
Processing Speeds & Performance
When advertising the generational leap between the 20 and the 30 series of GPUs, Nvidia released a graph, where the x axis is price and the y axis is relative performance, and they plot a variety of their GPUs. In this graph, the RTX 3070 appears as slightly faster than the RTX 2080 Ti, and the render configuration does show a small advantage.
The RTX 2080 Ti has 4,352 CUDA cores – which are configured separately from its INT32 cores, which are also 4,352 in number. On the other hand, the RTX 3070 has 5,888 cores, but half of these are dedicated CUDA cores, while the other half can operate both as FP32 and INT32 cores. Therefore, the RTX 3070 may have a larger total CUDA core count, but if both INT32 and FP32 cores are needed in a gaming application or otherwise, then the RTX 2080 Ti has a larger total number of cores to work with. That being said, the cores of the RTX 3070 are of better quality, as they operate at higher clock speeds. In terms of texture map units (TMUs), the RTX 3070 trails slightly, with 184 TMUS versus the 272 TMUS of the 2080 Ti. It does have more render output units (ROPs) though: 96 versus 88.
In regards to other computational power, the RTX 2080 Ti has far more Tensor and Ray Tracing (RT) cores, but the quality of these cores is far lesser than the quality of the equivalent RTX 3070 cores. The RTX 2080 Ti has 544, 1st generation, Tensor cores, and 68, 1st generation, RT cores. The RTX 3070 has 184, 3rd generation, Tensor cores, and 45, 2nd generation, RT cores. If what Nvidia advertises for their new generation Tensor and Ray Tracing cores is true, then the Tensor cores offer 2.7 times better high-performance computing (FP64), while the RT Cores offer 1.7 times more RT-TFLOPS each. This means that even with the lesser number of cores, the RTX 3070 will have the equivalent of 78 RT cores and 497 Tensor cores – versus the 68 RT cores and 544 Tensor cores of the RTX 2080 Ti.
The clock speeds of the RTX 3070 range between 1,500 MHz (base) and 1,725 MHz (boost), whereas the RTX 2080 Ti has a base clock speed of 1,350 MHz, and a boost clock speed of 1,545. Both GPUs have the same memory clock speed of 14 GBps.
Power Consumption
Though the economic aspect is a much easier argument to be made in favor of the RTX 3070 – since it retails for less than half the price of the RTX 2080 Ti – there is still a chance that one can find these two GPUs selling for around the same price. Indeed, many users appear to be panic-selling the RTX 2080 Ti, for anywhere between $500 and $600, due to the release of the RTX Ampere 30 Series GPUs and the promise that they hold. In terms of thermal design power, the RTX 2080 Ti draws 250 W of power from two 8-pin connectors, while the RTX 3070 draws 220 W of power from a 12-pin connector – so the RTX 3070 does appear to have a better power-design to performance ratio, but only slightly.
Conclusion
Taking all specifications into consideration, we see that both these graphics cards have their advantages and disadvantages; and a clear winner cannot be declared; at least not yet. The RTX 2080 Ti has more and better VRAM, more total number of cores (though they are less versatile and slower in clock speeds), more Tensor cores, and better FP64 performance. On the other hand, the RTX 3070 is built on a new Samsung processing node, it has new generation Tensor and RT cores (the latter being more in number as well) and is more energy efficient. If we were to guess which would perform better on the launch day of the RTX 3070, we would bet that the RTX 3070 would fair slightly better (or about even) at 1080p and 1440p resolutions, and will trail slightly in 4K resolutions.
If you can find a used RTX 2080 Ti for $500 and are wondering which of the two GPUs to go for, then it must be said that the answer really does depend on how you prefer to game. If you want to game on a 4K monitor, the RTX 2080 Ti is a much safer bet, as it has enough VRAM to be able to play triple-A games for now, and most likely for several years to come, whereas the RTX 3070 may be capped at lesser resolutions (or in-game quality settings) later down the road.
If you prefer to game at 1080p, or 1440p, at framerates higher than 60 frames per second, then the RTX 3070 is the better choice. It may not show a significant advantage at launch, but as developers begin to take advantage of the new generation of RT and Tensor cores, it will surely pull ahead in the long run. Also, since any VRAM trouble at 4K can be fixed by lowering the game-quality settings, it may even just be the better GPU overall. If Nvidia doesn’t botch the release of the RTX 3070, it would be the easier choice of the two, but if you can’t find the 3070, the RTX 2080 Ti at $500 is an excellent price.