Within the range of NVIDIA products based on its RTX 30 architecture or known as GeForce Ampere, there were two gaps to be filled due to the price distance between the RTX 3070 and the RTX 3080 on the one hand and from RTX 3080 to RTX 3090 The new RTX 3070 Ti is placed in the first hole and the RTX 3080 Ti in the second. Completing the medium and high ranges of the NVIDIA RTX 30.
NVIDIA GeForce RTX 3080 Ti
| Graphic card | NVIDIA RTX 3080 Ti |
|---|---|
| GPU | GA102-225 |
| Family | RTX 30 |
| Architecture | GeForce Ampere |
| SM / UC | 80 |
| ALU FP32 for SM / CU | 128 |
| Ray tracing unit | Yes, 1 per SM (80 in total) |
| Tensor units | Si, 4 per SM |
| TFLOPS FP32 | 34 |
| TFLOPS tensor (FP16) | 273 |
| VRAM type | GDDR6X |
| VRAM quantity | 12 GB |
| VRAM speed | 19 Gbit / s |
| VRAM bandwidth | 912 GB / s |
| PVPR | $ 1,199 |
| Basic speed | 1365 MHz |
| increase the speed | 1665 MHz |
The first and most powerful of the new NVIDIA graphics cards is the GeForce RTX 3080 Ti, which is the third model based on the NVIDIA GA102 GPU after the RTX 3080 and RTX 3090. In this case, it is based on the chip GA102-225
From VRAM, the graphics card has a 384-bit GDDR6X bus like the RTX 3090, but with 12 GB capacity, since Nvidia decided not to use x8 mode and therefore installed half of the memory chips as in its big sister. Another change is the somewhat slower memory usage, 19 Gbit / s instead of 19.5 Gbit / s. It therefore uses the same type of GDDR6X as the GeForce 3080. However, its consumption is 350 W, the same as that of the RTX 3090 and 30 W above the 320 W of the RTX 3080.
If we go to the GPU itself, we are faced with a Configuration 80 SM to one Base speed of 1365 MHz with a boost of 1665 MHz, which gives it a power of 34 TFLOPS and FP32, 67 RT-FLOPS y 274 Tensor TFLOPS. The first measurement comes from the computing capacity of the 128 ALUs in FP32 of each of the SMs, the second is a comparative measurement created by NVIDIA for its graphics cards and the third is the computing capacity of the Tensor Cores.
As for its price, it has a suggested retail price of $ 1,199which puts it almost twice the RTX 3080’s $ 699 MSRP.
NVIDIA RTX 3070 Ti
| Graphic card | NVIDIA RTX 3080 Ti |
|---|---|
| GPU | GA104-400 |
| Family | RTX 30 |
| Architecture | GeForce Ampere |
| SM / UC | 48 |
| ALU FP32 for SM / CU | 128 |
| Ray tracing unit | Yes, 1 per SM (48 in total) |
| Tensor units | Si, 4 per SM |
| TFLOPS FP32 | 22 |
| TFLOPS tensor (FP16) | 174 |
| VRAM type | GDDR6X |
| VRAM quantity | 8 GB |
| VRAM speed | 19 Gbit / s |
| VRAM bandwidth | 609 GB / s |
| PVPR | $ 1,199 |
| Basic speed | 1580 MHz |
| increase the speed | 1770 MHz |
The second model presented by NVIDIA at Computex 2021 is the RTX 3070Ti, which immediately becomes the most powerful mid-range graphics card in the RTX 30. Like the RTX 3060 Ti and RTX 3070 Ti, it is based on a variant of the GA104, specifically the GA104-400.
Compared to the RTX 3070, it has the full configuration of 48 SM, in addition to operating at a 1580 MHz base clock speed and 1770 MHz boost, which are higher than the RTX 3070. Biggest surprise? NVIDIA replaced the 256-bit GDDR6 memory interface with an interface of the same width GDDR6X therefore the capacity of VRAM is always 8 GB, but in exchange, it went from a memory of 14 Gbit / s to 19 Gbit / s, increasing its bandwidth from 448 Gb / s to 609 GB / s. However, this increase in bandwidth is accompanied by an increase in the power consumption of the card, which goes from 220 W to 290 W in all.
Regarding the technical specifications, we find a power of 22 TFLOPS in FP32, 42 RT-TFLOPS and 174 Tensor TFLOPS. The meaning of the three performance parameters being the same as we discussed in the RTX 3080 Ti and therefore the same as the rest of the NVIDIA RTX 30 line.
NVIDIA’s suggested retail price for the RTX 3070 Ti? $ 599.
Based on GeForce Ampère architecture
Like the rest of the RTX 30s that have hit the market in recent months, the RTX 3080 Ti and RTX 3070 Ti are both based on the GeForce Ampere architecture, which relies on three different pillars to achieve its massive power:
- Improved CUDA cores, now with a second set of FP32 ALUs switched with the entire ALUs, allows you to perform up to twice as many operations per SM and clock cycle as GeForce Turing, aka RTX 20s.
- New RT cores, the ray tracing ability has been improved to the point of doubling the performance in the face of beam-box and beam-triangle intersection.
- Third generation tensor hearts, which allows acceleration of DLSS algorithms and being able to use NVIDIA Broadcast at high speed and without affecting game performance.
These three elements make today the most powerful RTX 30 graphics architecture on the market to date, making the RTX 3080 Ti and RTX 3070 Ti graphics unprecedented power in their respective price ranges.
To all of this must also be added support for NVIDIA Reflex, which is designed to keep gaming latencies as low as possible, allowing refresh rates of up to 360Hz on G-SYNC monitors.
RTX 3070 Ti and RTX 3080 Ti join RTX 30 2nd
NVIDIA has completely renewed its RTX 30 in a second generation specially designed for the gaming market. The way to achieve this was to halve Ethereum’s mining capacity, a change that was also made by NVIDIA in the RTX 3080 Ti and RTX 3070 Ti.
Another change that NVIDIA has implemented is serial support for the ResizableBAR standard, this technology consists of giving the CPU access to VRAM. His utility ? It allows GPUs to read command list information much faster by not having to access system RAM to get it. This results in a reduction in frame time and therefore more FPS in games.
Table of Contents
