1660 vs 1070
It is believed that when Nvidia released its 16
series, it saved the GTX name from complete extinction. The RTX product line is imminent, and it seems like the end of GTX may not be far behind, but in many ways the 16 series only provides more competition in the market for the original GTX GPU.
By comparing its specs, build, and performance, we'll find out if the 1660 is worth the investment or if the 1070 should be put away first.
6 GB GDDR5
memory bus width
GDDR5 8 GB
43 GPU shared architecture bus
4 far cry. The GeForce 1070 is built using the GP104200A1 variant of Nvidia's 2016 Pascal microarchitecture, while the newer 16 series cards have the Turing TU116300A1 microarchitecture variant. The
Turing architecture is much more advanced than Pascal, but the main reason people are so picky about it is the dedicated RT and Tensor cores, while the 1660 is not. These capabilities are reserved for Nvidia's RTX series of GPUs, but they are not the only advantage of the Turing system. It is said that the performance of each CUDA core of the Turing kernel has increased by 50%. So, let's talk about the following.
Pascal bones from 1070 contain 1920 CUDA cores; not bad, especially considering that the 1660 only brought 1,408 units, but the next thing to consider is improvement. If you want to trust iNvidia, you will see that the core count is more like 2816. The rest of the
architecture specification follows a similar pattern. 1070 encapsulates more physical components. The number of transistors reaches 7.2 billion, more than 6.6 billion out of the 1660. It has 512 shader units, 32 texture units, and 16 render units more than the 1660 card. However, in terms of clock speed, 1660 uses 1530MHz as a base and improves 1758MHz, returning to the top of the list. The
1070 is a fan unit rated at 150 watts and can run very hot. If you push it to an unlocked frame rate, it's not uncommon for it to peak above 80 ° C and stabilize somewhere in the 70s, but with a 94 ° C capacity, it should be fine. If you are a GPU developer, you can try with an open case, but if you are looking for a quiet GPU, the liquid should keep the temperature in the area of 50 ° C and keep the fan quiet, even if it is below 100 % load.
1660 is more energy efficient, with a power output of only 120 watts and a heat capacity of 95 ° C. It is also an outdoor design. They tend to run at lower temperatures, but they do heat up your case, so you may have to deal with the case process to prepare for this GPU. With a standard fan, you can expect very similar thermal performance to the 1070. It will exceed 80°C under 100% load and is stable at 70°C for most gaming applications. The
chip size is 314mm2. Nvidia’s 1070 graphics card technically has the largest chip, because Turing technology allows a higher chip density than its predecessor Pascal, but how do these GPUs expand together? The
measures 4.37"(H) x 5.7" (L) x 2 slots (W). The 1660 is almost half the size of the GTX 1070, making it more suitable for larger situations, whether small or busy. The 1070 measures 4.376"(H) x 10.5" (L) x 2 slots (W), so it takes up a lot of space.
If we just look at the sheer numbers, it appears that the GTX 1070 has a significant advantage over 16-series cards, but you can't ignore Turing's advanced technology.
In the 1080p field, the GTX 1070 eliminated the newer 16 series cards, ensuring PUBG's FPS increased significantly by 38%, with an average of 109fps exceeding the 79fps average of 1660. Versus Known GPU-killing games like Overwatch showed no fear, maintaining an average level of over 100 fps, while the 1660 averaged 83 fps. In general, for 1080p games, you can expect the 1070 to perform 9 times better than the 1660.
switched to the second most popular gaming resolution, in the process of many popular games, GTX 1660 managed to reduce the gap to 8% . This isn't a violent comeback, but we've seen those Turing cores start to accelerate. The
pushed 4K up a notch, and the 1660 narrowed the gap again, bringing the difference between the two GPUs to 7%. This is fairly trivial, but still a good effort.
We mentioned in the architecture section that while the 1660 has those shiny new Turing CUDA cores, it doesn't actually separate the RTX GPU from the pack. Of course, we are talking about Tensor and RT cores, the latter allowing ray tracing in games that support it.
You can experience ray tracing using one of the Nvidia drivers developed in 2019. The driver allows you to enable ray tracing for any GTX card, but for those Pascal kernels this is hard work. The 1660 may work fine, but the frames per second from both GPUs will drop significantly.
Nvidia does not treat the GTX 1660 as GDDR5X memory, so although it was released in 2019, it actually shares the same GDDR5 memory configuration as the 2016 GTX 1070.
In fact, the GTX 1070 has a bus specification interface. even more impressive 256-bit and 256.3 GBps bandwidth. The 1660 has a 193-bit bus interface and 192.1GBps bandwidth, but thanks to the impressive Turing architecture, it matches the overall 8GBps memory clock speed of the 1070.
6 GB GDDR5
memory bus width
RT or Tensor core, it is hard to see someone choose it on 1070.
If you are already using GTX 1660 and you think the 1070 is a possible upgrade, you may better make some bigger jumps. When you change the GPU, you really want to distinguish the difference between the game and the game, but consider All things considered, 1070 can only provide you with a performance improvement of 7%.