Positioning & Architecture
NVIDIA has been busy—this week we have their third graphics card launch this year—the GeForce RTX 5070 Ti. With Blackwell, NVIDIA is launching their product stack from the top: first the RTX 5090 flagship, then the RTX 5080 and now the 5070 Ti. Just like the RTX 5080, the RTX 5070 Ti is based on NVIDIA's GB203 silicon, which is the second-largest consumer Blackwell chip in their arsenal. While the RTX 5080 came with all cores enabled (10752), the RTX 5070 Ti has 8960 of them active, a 20% difference. Other unit counts have been adjusted accordingly, you get 96 ROPs, 280 TMUs and 70 RT cores vs 128/336/84 on the RTX 5080. The memory subsystem is identical, the RTX 5070 Ti uses the same 256-bit 16 GB GDDR7 memory configuration, just the clock frequency is a little bit lower: 28 Gbps vs 30 Gbps.

Today marks the end of the review embargo for NVIDIA cards priced at $750 MSRP. Starting tomorrow, we can publish reviews for custom models that sell for more than this price.

The Blackwell architecture introduces several architectural improvements under the hood, like giving all shaders the ability to run FP32 or INT32 instructions, on Ada only half the cores had that ability. The Tensor Cores are now accessible from the shaders through a new Microsoft DirectX API, and they now support FP4 and INT4 instructions which run at lower precision, but much faster with less memory usage. There's numerous additional architecture improvements, we talked about all of them on the first pages of this review.

From a fabrication perspective nothing has changed though—Blackwell is built on the same 5 nanometer "NVIDIA 4N" TSMC node as last generation's Ada. NVIDIA claims this is a "4 nanometer process," but during Ada it was confirmed that NVIDIA 4N is actually not TSMC N4 (note the order of N and 4), but 5 nanometer. At the end of the day the actual number doesn't matter much, what's important is that NVIDIA is using the same process node.

Performance
We upgraded our test system in preparation for this wave of GPU launches, which is now built on AMD technology with the outstanding Ryzen 7 9800X3D. We've updated to Windows 11 24H2, complete with the newest patches and updates, and have added a selection of new games. At 4K resolution, with pure rasterization, without ray tracing or DLSS, we measured a 28% performance uplift over the RTX 4070 Ti, which is pretty good for a gen-over-gen improvement. While it's not as big as the RTX 5090, which is 36% faster than the RTX 4090, it's definitely better than the 15% that we got on RTX 5080 a few weeks ago. Just like with RTX 5090, NVIDIA achieves their "twice the performance every second generation" rule: the RTX 5070 Ti is twice as fast as the RTX 3070 Ti. This means the card matches performance of the RTX 4080 and RTX 4080 Super, and it's also beating AMD's Radeon RX 7900 XTX flagship by a wafer-thin margin. Impressive—NVIDIA's 3rd card in the lineup beats AMD's #1. And this is with pure rasterization—once you turn on ray tracing, the gap gets much bigger.

For this launch, NVIDIA provided us with the MSI RTX 5070 Ti Ventus OC, which, as the name reveals, is a factory overclocked card. This means that it has a small performance advantage—all the other comparison cards in our tests are clocked at reference. So, if you plan on buying a baseline card, subtract a percent or two from our performance numbers. Once cards appear in the market I will buy a pure base clock card, for comparisons in future reviews. There is no Founders Edition for the RTX 5070 Ti.

While RTX 5070 Ti is a very decent card for gaming at 4K, it's not a fire-and-forget solution. There are several titles that run at less than 60 FPS when maxed out (without RT and upscaling). I'd say RTX 5080 is a better choice for demanding 4K gaming, but considering the price differences, I think lowering details slightly or using upscaling / frame generation is a very reasonable approach. For 1440p, the RTX 5070 Ti is awesome, here it can achieve excellent frame rates and will be able to drive high-refresh-rate displays very well.

Ray Tracing & Neural Rendering
NVIDIA is betting on ray tracing and Blackwell comes with several improvements here. When comparing the RTX 4070 Ti with the RTX 5070 Ti, we see a really nice +49% performance uplift. One of the reasons for that is that the 5070 Ti comes with 16 GB, whereas the RTX 4070 Ti had only 12 GB. Interestingly, against RTX 4070 Ti Super, which has 16 GB, too, the RT performance delta is just 15%. In the comparison against AMD Radeon, the 5070 Ti makes short shrift of the RX 7900 XTX, beating it by over 42%. AMD has confirmed that for RDNA 4 they have put in some extra love for the RT cores, so hopefully they can catch up a bit. With Blackwell, NVIDIA is introducing several new technologies. The most interesting one is Neural Rendering, which is exposed through a Microsoft DirectX API (Cooperative Vectors). This ensures that the feature is universally available for all GPU vendors to implement, so game developers should be highly motivated to pick it up.

VRAM
While the RTX 4070 Ti came with 12 GB VRAM, which ended up a bit limiting in some titles, the RTX 4070 Ti Super bumped that to 16 GB. The RTX 5070 Ti retains that 16 GB framebuffer size, and it makes a lot of sense in this segment. Increasing the memory size further would have required a wider memory bus and a GPU design with support for the extra bus width and more pins in the design, etc. All these changes would make the card more expensive, for relatively small gains. More memory does not automagically turn into additional performance. You will see scaling only in titles that go beyond 16 GB in this case—which is extremely rare, and usually a sign of developer fail. Video games are developed for consoles first these days, and their hardware sets the bounds for what we can expect from VRAM usage on PC, too.

DLSS 4 Upscaling & Frame Generation
NVIDIA made a big marketing push to tell everyone how awesome DLSS 4 is, and they are not wrong. First of all, DLSS 4 Multi-Frame-Generation. While DLSS 3 doubled the framerates by generating a single new frame, DLSS 4 can now triple or quadruple the frame count. In our testing this worked very well and delivered the expected FPS rates. When using FG, gaming latency does NOT scale linearly with FPS, but given a base FPS of like 40 or 50, DLSS x4 works great to achieve the smoothness of over 150 FPS, with similar latency than you started out with. Image quality is good, if you know what to look for you can see some halos around the player, but that's nothing you'd notice in actual gameplay.

Want lower latency? Then turn on DLSS 4 Upscaling, which lowers the render resolution and scales up the native frame. In the past there were a lot of debates whether DLSS upscaling image quality is good enough, some people even claimed "better than native"—I strongly disagree with that—I'm one of the people who are allergic to DLSS 3 upscaling, even at "quality." With Blackwell, NVIDIA is introducing a "Transformer" upscaling model for DLSS, which is a major improvement over the previous "CNN" model. I tested Transformer and I'm in love. The image quality is so good, "Quality" looks like native, sometimes better. There is no more flickering or low-res smeared out textures on the horizon. Thin wires are crystal clear, even at sub-4K resolution! You really have to see it for yourself to appreciate it, it's almost like magic. The best thing? DLSS Transformer is available not only on GeForce 50, but uses Tensor Cores on all GeForce RTX cards! While it comes with a roughly 10% performance hit compared to CNN, I would never go back to CNN. NVIDIA's driver with Transformer model support has been public for a few weeks now and people are slowly become aware that there's a per-game DLSS override the NVIDIA App now—with very good results, even on older GPUs.

Physical Design, Heat & Noise
Despite being a cost-optimized model that sells for NVIDIA MSRP, I kinda like the Ventus design theme—it reminds me of the RTX 2080 Ti Founders Edition. While its plastic cooler shroud is definitely not as fancy as the competing metal designs, it definitely gets the job done, and the color theme ensures it will fit into all builds. There is no RGB lighting or dual BIOS though.

In terms of cooling the Ventus 3X does a good job, temperatures reach just 68°C, which is super low, especially for a card with such limited cooling capacity. Unfortunately the noise levels are quite high. With 41 dBA, the card is pretty loud, especially considering the performance class. It seems that MSI has focused on low temperatures way too much and couldn't find the right balance between temperature and noise. Usually MSI gets this perfectly right. For example, had they allowed 75°C, the card would be much quieter, and still far from "running hot." It's also surprising that NVIDIA decided to sample a card with such fan settings to press, as stand-in to represent their newest GPU. Had anyone checked (or cared?) it would have been easy for them to convince MSI to come up with better fan settings.

Our apples-to-apples cooler comparison test reveals that the MSI cooling solution is pretty weak, considerable weaker than the Galax 1-Click OC that we tested today, too. Tomorrow's custom designs with their bigger, more premium, cooling solutions will certainly improve this even further.

PCI-Express 5.0
NVIDIA's GeForce Blackwell graphics cards are the first high-end consumer models to support PCI-Express 5.0. This increases the available PCIe bandwidth to the GPU, yielding a small performance benefit. Of course PCIe Gen 5 is backwards compatible with older versions, so you'll be able to run the RTX 5070 Ti even in an older computer.

Just like we've done over the years, we took a detailed look at PCI-Express scaling in a separate article. Testing includes x8 Gen 5, for instances when an SSD is eating some lanes. The popular x16 4.0 was tested, which is common on many older CPUs and entry-level motherboards. Finally, some additional combinations were run, down to PCIe x16 1.1. The results confirm that unless you are on an ancient machine, PCIe bandwidth won't be a problem at all.

Power Consumption
Just like other Blackwell cards, the idle, multi-monitor and media playback power consumption of RTX 5070 Ti is higher than expected. Compared to the RTX 4070 Ti, the increase is +50%, against the RTX 4070 Ti Super it's still 29%—not good. Since we saw similar increases on other new cards, I hope this is a bug that NVIDIA can fix. Especially if you run your PC for many hours a day without gaming, then this increase can add up, even more so when you multiply it by a few million, which is how many units NVIDIA plans to ship of these cards. In contrast, the gaming efficiency is excellent, outshining all previous generation graphics cards and only being surpassed by the RTX 5080. NVIDIA is using the same process for Blackwell as for Ada, so there are no process-related power improvements here. With around 300 W for the graphics card alone, PSU requirements are nothing to worry about.

While other RTX 5070 Ti cards have the ability to increase the power limit setting beyond the default of 300 W, the MSI Ventus doesn't have this feature. It's not a big deal though, because it's somewhat rare for the card to draw that much power with normal gaming workloads.

Overclocking
Overclocking worked extremely well on all RTX 5070 Ti cards, with most of them reaching more than +10% in real-life performance. This is quite unexpected, especially for NVIDIA, who are usually the masters at eking the last bits of performance out of their GPUs, even at stock. Going from a real measured clock frequency of ~2800 MHz to ~3200 MHz is massive—why didn't they clock these cards higher? Memory overclocking is equally impressive and topped out at NVIDIA's artificial cap of +375 MHz / +3000 MT/s—the chips could certainly take more.

Pricing & Alternatives
NVIDIA's MSRP for the RTX 5070 Ti Series is $750, which is very reasonable for the performance you're getting. Actually, this MSRP is $50 lower than the $800 price point that both the 4070 Ti and 4070 Ti Super launched at. There has been lots of controversy about fake MSRPs, and this has been going on for years now, so do expect higher prices in stores. The primary driver for this is supply and demand, if everybody wants a product, its supply won't be sufficient and prices will go up. For the RTX 5090 and RTX 5080 supply was very low, too, making the situation even worse. I've plotted various alternative price points in our price/performance charts, reaching up to $1100, which, according to some early postings might end up being a realistic price point. We'll know more tomorrow, when sales go up.

MSI's RTX 5070 Ti Ventus 3X is priced at the NVIDIA MSRP, which is nice (as long as it's true and there's supply). Since there is no Founders Edition this time, there really isn't a baseline to compare to. Today we also tested the Galax RTX 5070 Ti 1-Click OC, which is MSRP as well, but comes with a much better cooler and much better noise levels. Still, the Ventus is definitely not bad. It is able to deliver the full RTX 5070 Ti experience, just with a little bit higher noise levels out of the box. Considering that, I'm having serious doubts whether I would be willing to spend, +$200, +$300 or even more for any custom design—we've seen pricing like that on some RTX 5070 Ti cards! Maybe $50-70 for a better cooler that runs really quiet, but that's about it.

There really isn't any alternative to the 5070 Ti in this segment, and NVIDIA knows that, and they designed the card with that in mind. No reason to give you +50% of anything if there's no competing product. AMD's flagship, the Radeon RX 7900 XTX currently sells for $820, with less performance, especially in RT, higher power draw and no DLSS. The RTX 4080 and 4080 Super are priced at around $1000 these days—no reason to buy them unless they are heavily discounted and end up below 5070 Ti pricing. What else is there? RTX 4090? Super expensive because people buy them for AI. RTX 5080 and 5090? Sold out, scalped to several thousand dollars. Let's hope that supply of RTX 5070 Ti is better and gamers can actually get their hands on these new cards.

AMD is set to release the Radeon RX 9070 Series shortly, but it probably won't match the performance of the RTX 5070 Ti. Instead, it seems it will be more comparable to the RTX 5070, which is also expected to be released soon. While these new cards cannot rival the RTX 5070 Ti in terms of performance, they are likely to be priced more competitively due to increased competition in this market segment.

Due to the cooler's noise levels, the Ventus only gets "Recommended," the Galax in today's second review does better and gets Editor's Choice.