Home » Technology » Nvidia GeForce RTX 3070 Ti Review – Introduction

Nvidia GeForce RTX 3070 Ti Review – Introduction

Simultaneously with the RTX 3080 Ti, Nvidia announced the RTX 3070 Ti. However, the company spread the benchmark and review embargoes over two Tuesdays with a week in between. The latest card in the Ampere family should be 50 percent faster than its predecessor, the 2070 Super, and with additional CUDA, Tensor and RT cores, the card should of course also leave the regular RTX 3070 behind.

Faster memory for the GA104 GPU

By far the most eye-catching specification of the RTX 3070 Ti is the type of memory. Like the much more luxurious GeForce cards in the RTX 30 series, the 3070 Ti has GDDR6X memory, which offers even higher speeds than normal GDDR6. Thanks to a memory speed of 19Gbit/s over a 256bit bus, the 3070 Ti has 608GB/s memory bandwidth, where the 3070 has to make do with 448GB/s. This means that the RTX 3070 Ti has over 35 percent more memory bandwidth than the normal RTX 3070, while the number of enabled cores on the RTX 3070 Ti is only slightly more than 4 percent higher. Because modern GPUs are almost by definition limited in their performance due to the total bandwidth of the memory, the RTX 3070 Ti already seems a lot more interesting on paper than the regular RTX 3070. We will find out in this article whether this is also the case in practice. review.

The RTX 3070 Ti has a fully enabled GA104 GPU, with all 6144 cores active. This is a small step forward from the normal RTX 3070, which comes out at 5888 cores. Yet the tgp with 290W is more than 30 percent higher. How is that possible? This difference is almost entirely due to the GDDR6X memory on the 3070 Ti, which runs at much higher speeds and consumes considerably more energy than the normal GDDR6 memory as used on the RTX 3070, among others.

video card RTX 3080 Ti RTX 3080 RTX 3070 Ti RTX 3070 RTX 3060 Ti
Gpu GA102-225 GA102-200 GA104-400 GA104-300 GA104-200
Architecture Ampere
Process 8nm
Diesize 628,4mm² 628,4mm² 392,5mm² 392,5mm² 392,5mm²
Transistors 28 billion 28 billion 17.4 billion 17.4 billion 17.4 billion
CUDA-colors 10.240 8704 6144 5888 4864
Tensor colors 320 272 192 184 152
RT-cores 80 68 48 46 38
Texture-units 320 272 192 184 152
Rops 112 96 96 96 80
Boostsnelheid 1665MHz 1710MHz 1770MHz 1730MHz 1665MHz
Vram 12GB GDDR6X 10GB GDDR6X 8GB GDDR6X 8GB GDDR6 8GB GDDR6
Memory Speed 19Gbit/s 19Gbit/s 19Gbit/s 14Gbit/s 14Gbit/s
Memory bus 384bit 320bit 256bit 256bit 256bit
Bandwidth 912GB/s 760GB/s 608GB/s 448GB/s 448GB/s
Tgp 350W 320W 290W 220W 200W
Release date 2 june 2021 17 september 2020 9 june 2021 October 29, 2020 2 december 2020
MSRP
(Founders Edition)
1199 euro 719 euro 619 euro 519 euro 419 euro

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.