Things are heating up in the graphics card space, as today sees NVIDIA dropping their latest offering in the 50 series lineup, the RTX 5070 Ti. NVIDIA claims this GPU will give users still rocking an RTX 3070 Ti a reason to upgrade through uplifts in performance and the latest features, like DLSS 4 with multi-frame generation. However, at $749 and with AMD’s 9000 series on the horizon, it’s already starting off on the back foot, so it’ll be interesting to see what kind of gains the new Blackwell-based GPU brings us over not only its Ampere predecessor but also what increase we see generation on generation when comparing it to its Ada Lovelace sibling, the RTX 4070 Ti. But before we get into that, here’s a quick word from this video’s sponsor.
Okay, so before we get things started, I want to add a small caveat to this review. Our RTX 5080 review was nearly an hour long (YouTube), and while we at eTeknix believe it was one of the most well-rounded reviews on the internet, it was long and there was a lot involved. As we don’t have a Founders Edition for the 5070 Ti, I want to condense things somewhat and focus purely on the GPU and not the MSI Ventus 3X cooler that I have here, as we’ll have a separate feature for that, drilling down into the cooler, the overclocking potential and all of that good stuff. I’ll say for now that I believe most media outlets are getting the same card, so that should make it easy for you to compare with other reviews and see how things line up.
Now, before we get into the benchmarks, let’s talk specs, as the 5070 Ti does pack some power due to NVIDIA opting for the same GB203 core, just a slightly different variant with the GB203-200-A1, compared to the GB203-400-A1 on the RTX 5080. In terms of the nitty-gritty, this means we have the same die size at 378 mm² and the same 45.6 billion transistors.
This means that the 5070 Ti comes in with 8,960 CUDA cores, which is a 17% increase from the 4070 Ti and a 46% uplift from the 3070 Ti. The same increases follow for the texture mapping units, streaming multiprocessors, tensor cores and RT cores, and we see a healthy uplift in L2 cache, now matching the RTX 5080 at 64 MB, compared to 48 MB on the 4070 Ti and a lowly 4 MB on the 3070 Ti.
We also have 16 GB of GDDR7 on a 256-bit bus, matching that of the RTX 5080, though with a slightly lower memory bandwidth of 896 GB/s. This also means that we now have 6 GB more than the 4070 Ti and double that of the 3070 Ti, so those looking for an upgrade from the 3070 Ti could see that as a huge selling point.
In terms of clock speeds, the 5070 Ti comes in with a base clock of 2,295 MHz, a boost clock of 2,452 MHz, and a memory speed of 1,750 MHz or 28 Gbps effective. However, with the 3X OC model we have here from MSI, we see a slight increase in the boost clock to 2,482 MHz, with the same memory speed of 1,750 MHz. All of this bundled up gives us a TDP of 300 W, and for this card, NVIDIA recommends a 700 W PSU as a minimum. Though, as always, depending on your other components and mainly your processor, I’d always recommend going slightly higher, around the 850 W mark, to give yourself some headroom and to future-proof your system.
So, on paper, there’s the steady flow of increases in specs that we’d expect from generation to generation, and hopefully, this all equates to a good, strong performance uplift as well.
