How useful the RTX 5070 Ti REALLY is?

There has been a lot of talk about the Blackwell generation of NVIDIA GeForce RTX graphics cards, both in terms of price and performance, and even power consumption. However, it always comes down… to the users, that pays the money. And they pay for benefits, not discussions. Let’s take the ASUS PRIME GeForce RTX 5070 Ti OC as an example. An abstract model, without reference to the brand, SKU or number of fans. And now – I will explain how this graphics card can be useful in reality. BOTH for gaming AND for business.

TABLE OF CONTENTS:

DLSS 4 and Multi Frame Generation

I’ll start with the things that are unique to the new generation of NVIDIA GeForce RTX. For gamers, this is primarily, of course, NVIDIA DLSS 4 with Multi Frame Generation. We are all familiar with frame multipliers powered by artificial intelligence, right?

There are many types of those, from several manufacturers, some are hardware-accelerated, some are software-accelerated, and there are even some that allow you to combine two graphics cards more efficiently than SLI/NVLink! But specifically, newest NVIDIA DLSS 4, comparing to previous versions, consumes 30% less video memory, works 40% faster, and is available for ALL previous RTX models, including the Turing generation, i.e. RTX 2060, etc. To be as brief as possible, DLSS 4, DLSS Super Resolution and DLAA now are using “transformer” model, the same type that powers AI models of Flux, ChatGPT and Gemini. It greatly boost performance, latency and efficiency.

The Multi Frame Generation technology is unique and exclusive to RTX 50 Series. For RTX from the 2000 to 4000 series, Frame Generator allows you to set a frame multiplier of 2X, but Multi Frame Generator available with Blackwell GPUs allows for 3X or even 4X! I personally tested this in Cyberpunk 2077 and got 500 FPS on high settings! Even on the RTX 5060 Ti. While the picture on screen was looking better than what I got on my laptop with RTX 3080.

And that’s not all. DLSS 4 with Multi Frame Generation already works in hundreds of games. And not only in AAA projects such as God of War Ragnarök, Alan Wake 2, Returnal, REMNANT II, but also in S.T.A.L.K.E.R. 2: Heart of Chornobyl, Frostpunk 2, Ready or Not, NVIDIA RTX Remix and even POSTAL 4: No Regerts.

Actually, NVIDIA RTX Remix interests me personally, because seeing old 3D games with ultra-high-quality lighting and minimal visual artefacts is a pleasure unique, and even exquisite.

Of course, Multi Frame Generation is a new technology. And it is not yet 100% stable – S.T.A.L.K.E.R. 2, for example, does not support it if you play through Xbox Game Pass. And Fallout New Vegas through NVIDIA RTX Remix still looks terrible, because objects are not optimised for different lighting.

Also – don’t forget, that even peak stable MFG will need at least 50-60 FPS base, because less frames will not only bring more visual artifacts, but less responsive controls. It’s like giving Topaz Labs a tiny picture to resize it into giant one. You won’t get 16 times the details. But! The future for RTX and DLSS looks as bright and inevitable, as are God rays in Fallout 76.

Useful link: Looking for best steam keys? XD Deals – Cheap Steam Keys

GDDR7 VRAM

You may have noticed that, according to NVIDIA itself, the DLSS 4 model consumes 30% less video memory. This means that there should be enough VRAM for effective use of frame multipliers. Fortunately, the RTX 5070 Ti is equipped with 16 GB of fast GDDR7 memory. This is twice as much as the minimum system requirements for the most demanding games today, so gamers should have enough headroom for 3-4 years for 4K, and even more – for Full HD.

In addition, the new video memory standard is much faster – exactly twice as fast as GDDR6X, which means it is twice as fast as the previous generation flagship. 48 gigabits per pin allows, with time and improvements in manufacturing technology, to achieve a real speed of 1,5 terabytes per second.

Just think about it – NVMe PCIe 5, which I reviewed not so long ago, boasts speeds of 15 GB/s. And GDDR7 will soon be 100 times faster! And that’s not all. Compared to GDDR6, GDDR7 has 70% less thermal resistance, which allows for better heat dissipation and makes the memory less fragile under load.

As an example, the specifications for the NVIDIA GeForce RTX 5090 Founders Edition state that the maximum temperature for GDDR7 chips is 105 °C. The chips heat up to 95 °C, which is within the ‘safe temperature’ range. And since NVIDIA deliberately does not make its graphics cards perfect in terms of cooling, so as not to take away market share from its partners – even on the flagship custom RTX 5090 VRAM temperatures will be perfect!

However, everything has its price, and GDDR7 is extremely expensive. Almost like HBM2, which was once considered the most expensive of its generation. At the same time, GDDR7 has another advantage that justifies its price, and for which HBM2 was released too early. Yes, we are talking about artificial intelligence.

Artificial Intelligence

We have already published an article about how various types of artificial intelligence can be useful. I personally use it every day. Not chatbots, but AI plugins in DaVinci Resolve, intelligent fill in Photoshop, and also creating subtitles in volunteer videos thanks to StoryToolkitAI. So, if you are expecting a Luddite stance in this paragraph, defending against Skynet with stone axes and loud cries, you are in the wrong place. And in this field of expertise, NVIDIA dominates. We will delve into this next time, but rest assured – RTX is not for gamers only.

The GeForce RTX Blackwell generation has only reinforced this position. Fast video memory allows for faster data exchange, reduces bottlenecks, and its capacity allows for larger and heavier models to be loaded. In fact, 16 GB is exactly what is needed to store the best StoryToolkitAI model in memory.

Regarding the ‘raw’ performance of the RTX 5070 Ti for artificial intelligence processing, there is almost too much to say about the improvements in the new generation of tensor cores, which is why we are preparing another article on the topic. However, I will only mention the following. The performance of the NVIDIA GeForce RTX 2070 Super is just over 73 TOPS, or ‘trillion operations per second.’ The next generation, the RTX 3070 Ti, has 174 TOPS. The RTX 4070 Ti shows a bigger jump, up to 640 TOPS.

The performance of the RTX 5070 Ti? As much as 1400 TOPS. That’s more than twice as much. Thanks to this leap, we can enjoy a multiple increase in FPS in games thanks to DLSS 4, faster video rendering in DaVinci Resolve, and greatly improved text and photo/video generation in the corresponding AI models.

Summary on RTX 5070 Ti

Of course, along with the release of RTX Blackwell, several other important technologies were released, including RTX Skin and RTX Mega Geometry. But these technologies are much more useful for previous generations of GPUs, where older generations of tensor cores, CUDA cores, and media engines cannot even compare to the efficiency and performance of Blackwell, so any help they can get is almost life-saving.

As for the ASUS PRIME GeForce RTX 5070 Ti OC, you can see for yourself how useful this card can be. We have already reviewed many Blackwell models from ASUS, and neither temperature nor power consumption were a problem for the new generation. Considering its price and balanced performance, the RTX 5070 Ti will remain an interesting and profitable choice for both gamers and professionals for a long time to come.

Read also:

How useful the RTX 5070 Ti REALLY is?

DLSS 4 and Multi Frame Generation

GDDR7 VRAM

Artificial Intelligence

Summary on RTX 5070 Ti

Where to buy ASUS PRIME GeForce RTX 5070 Ti OC

New comments