Fresh from its first barrage of Blackwell GPUs, Nvidia is preparing to unleash yet another cavalcade of cards on to the scene with GeForce RTX 5070 Ti leading the charge. Given its lower price point, this could be the pixel pusher of choice for more gamers than the relatively expensive 80 and 90 class alternatives.
If you’re considering grabbing a GeForce RTX 5070 Ti come launch, whether by choice or necessity given availability of its peers, here’s everything you need to know about the graphics card. Nvidia’s revealed much about its offering but I’ve plugged any gaps with credible rumours and educated guesses. Expect the final word on it via the Club386 review in due course.
Release date
GeForce RTX 5070 Ti will hit store shelves on February 20 2025. Nvidia confirmed this release date via an update to its website, following accurate rumours that pinned the card’s launch to that date.
GeForce RTX 5070 Ti will be the third card in the RTX 50 Series to launch, following the release of RTX 5090 and 5080 before it. Subsequently, RTX 5070 will join its siblings on March 5 but there’s no official word on when RTX 5060 pixel pushers will join in on the fun.
Specs
GeForce RTX 5070 Ti uses the same G203 GPU as found in RTX 5080, albeit with 17% (14) fewer Streaming Multiprocessors (SMs). This leaves the graphics card with a smaller amount of CUDA, RT, and Tensor cores, relative to the more-expensive and more-performant 80 class model. However, it does retain much of its sibling’s memory configuration and requires less power.
| GeForce RTX | 5070 Ti |
|---|---|
| Launch date | Feb 2025 |
| Codename | GB203 |
| Architecture | Blackwell |
| Process | TSMC 4N (5nm) |
| Transistors | 45.6bn |
| Die size | 378mm2 |
| SMs | 70 of 84 |
| CUDA cores | 8,960 |
| Boost clock | 2,452MHz |
| Peak FP32 TFLOPS | 43.9 |
| RT cores | 70 (4th Gen) |
| RT TFLOPS | 133.2 |
| Tensor cores | 280 (5th Gen) |
| FP16 Acc TFLOPS | 351.5 |
| Peak FP4 TFLOPS | 1,406 |
| ROPs | 96 |
| Texture units | 280 |
| Memory size | 16GB |
| Memory type | GDDR7 |
| Memory bus | 256-bit |
| Memory clock | 28Gb/s |
| Bandwidth | 896GB/s |
| L2 cache | 49,152KB |
| PCIe interface | Gen 5 |
| Video engines | 2 x NVENC (9th Gen) 1 x NVDEC (6th Gen) |
| Power | 300W |
| MSRP | $749 |
Looking at GeForce RTX 5070 Ti specs in more detail, the graphics card will boast 8,960 CUDA cores, 70 RT cores, and 280 Tensor cores. Given that SMs house each of these components, it should come as no surprise that their reduction relative to RTX 5080 also translates to -17%.
Power consumption falls to 300W, another 17% cut. From a generational perspective, this leaves GeForce RTX 5070 Ti sipping an additional 15W relative to RTX 4070 Ti Super’s 285W TDP. The graphics card will use a 16-pin connectors like other RTX 50 Series models with board partners likely including adapters that will allow for two 8-pin PCIe cables to provide power.
Nvidia includes 28Gb/s memory modules with GeForce RTX 5070 Ti, down from the 30Gb/s found in RTX 5080. This impacts memory bandwidth albeit only slightly, dropping to 896GB/s. The pixel pusher thankfully still boasts a 256-bit bus and a 16GB GDDR7 buffer, bolstering its ability to handle high resolution gaming and heavy ray traced workloads.
Other notable reductions include a 25% cut to L2 cache, down to 49,152KB. Nvidia has also trimmed one of its NVDEC video engines from GeForce RTX 5070 Ti.
Price
Nvidia sets GeForce RTX 5070 Ti pricing at £729 / $749, marking a generational price cut of £40 / $50 relative to RTX 4070 Ti Super. The company will not produce a Founders Edition and so adherence to RRP / MSRP is entirely in the hands of board partners.
Compared to other GeForce RTX 50 Series cards, RTX 5070 Ti is $250 (-25%) cheaper than RTX 5080 and a whopping $1,250 (-63%) more affordable than RTX 5090. However, it is $200 (+36%) more expensive than RTX 5070.
Given the limited availability of GeForce RTX 5080, it’s entirely possible that similar stock shortages will affect RTX 5070 Ti owing to much of their shared silicon.
Performance
The generational uplift GeForce RTX 5070 Ti offers is up to 2.9x in games and 3.7x in AI workloads according to Nvidia. It’s important to note, however, that the former improvement is while using the full DLSS 4 suite (including Multi Frame Generation).
The only benchmark Nvidia has provided sans any upscaling is in Resident Evil 4, suggesting GeForce RTX 5070 Ti will be around 20% faster than RTX 4070 Ti in ray tracing in that title. The company claims that the card will offer similar gains in raster performance using DLSS Super Resolution in Horizon: Forbidden West.
Recently, GeForce RTX 5070 Ti appeared in the results matrix of Blender and Geekbench. The graphics card was 23% faster than its forebear and a mere 8% behind RTX 4080 Super in the former of the two benchmarks. It managed a more impressive performance in the latter, scoring 248,739.
