Introduction
Remark: This review was first published on January 4, 2023.
NVIDIA only recently released the GeForce RTX 4080 and GeForce RTX 4090, both of which offer impressive ray-tracing performance and DLSS3 and Frame Generation boost in frame rates. However, these powerful graphics cards come with hefty price tags: The RTX 4080 has an MSRP of $1199, while the RTX 4090, the green company’s flagship GPU, costs $1599. These prices, along with the difficulty of finding cards at MSRPs, have made some gamers hesitant.
As a result, many look forward to the new GeForce RTX 4070 Ti as a more affordable option. NVIDIA has confirmed that the 4070 Ti will have a MSRP of €799, but it won’t have a Founders Edition. Instead, pricing and availability of the 4070 Ti will depend on custom models from manufacturers such as ASUS, MSI, and Gigabyte.
The RTX 4070 Ti is essentially a rebranded version of the previously announced, but subsequently canceled RTX 4080 12GB card. Both are basically the same cards, the main difference being that the 4070 Ti is priced about 10% lower than the ‘canned’ 4080 12GB model. In terms of specs, the 4070 Ti is specified lower than the officially available 4080 16GB model, with 7680 CUDA cores (compared to 9728), a 192-bit memory bus (compared to 256-bit), and 504 GB/s. memory bandwidth (compared to 736 GB/s). It does, however, have a slightly higher boost clock rate of 2.6 GHz (compared to 2.5 GHz) and a much lower TGP requirement at 285 W compared to the 4080’s 320 W and the 4090’s whopping 450 W.
| GeForce Graphics Card |
RTX 4090 | RTX 4080 | RTX 4070Ti (ROG Strix) | RTX 3090 ti |
RTX 3090 |
|---|---|---|---|---|---|
| GPU | Ada Lovelace (AD102) | Ada Lovelace (AD103) | Ada Lovelace (AD104) | Ampere (GA102) |
Ampere (GA102-300) |
|
Method
|
4nm (TSMC)
|
4nm (TSMC)
|
4nm (TSMC)
|
8nm (Samsung)
|
|
| Transistors | 76 billion | 45 billion | 35.8 billion | 28 billion |
28 billion |
| Streaming Multiprocessors (SM) | 128 | 76 | 60 | 84 | 82 |
| CUDA cores | 16384 | 9728 | 7680 | 10752 | 10496 |
| Tensor nuclei | 512 (Gen 4) |
304 (Gen 4) |
240 (Gen 4) | 336 (Gen 3) |
328 (Gen 3) |
| RT cores | 128 (Gen 3) |
76 (Gen 3) |
60 (Gen 3) | 84 (gen 2) |
82 (gen 2) |
| GPU base/boost clocks (MHz) | 2230 / 2520 | 2205 / 2505 | 2310 / 2760 | 1670 / 1860 |
1395 / 1695 |
| Memory | 24GB GDDR6X | 16GB GDDR6X | 12GB GDDR6X | 24GB GDDR6X | 24GB GDDR6X |
| Memory bus width | 384-bit | 256-bit | 192-bit | 384-bit | 384-bit |
| Memory bandwidth | 1,018 GB/sec | 716.8 GB/s | 504.2 GB/s | 1,008 GB/sec | 936 GB/sec |
| Couple | PCIe 4.0 | PCIe 4.0 | |||
| TDP | 450W | 320W | 285W | 450W | 350W |
| Price (at launch) | $1,599 | $1,199 | €799 | $1,999 | $1,499 |
So, is NVIDIA’s RTX 4070 Ti worth the money? Based on benchmark results, it seems like a good option with some caveats. However, it’s worth considering the other two options in the RTX 40 series as well.
For this review I have received ASUS’ ROG Strix GeForce 4070 Ti, which is almost too similar to its Strix 4080 and 4090 siblings. In short, it’s a huge card, even if at 337mm long, 124mm high, and 56mm wide, it’s just slightly smaller than both the 4080 and 4090 cards. So, fair warning if you have a small enclosure.
Like the other RTX 40 series GPUs, the 4070 Ti uses a 16-pin power connector, with a 2x8pin adapter in the box. NVIDIA recommends a 700W power supply for the card to function properly, despite the TGP requirement of 285W. The card also comes with three DisplayPort 1.4a and two HDMI 2.1a ports, although only four ports can be used. can be connected and used simultaneously.
While I would have preferred the inclusion of DisplayPort 2.1, which is available on the AMD 7900 XT and XTX cards, it’s not a deal breaker at this price point. DisplayPort 2.1 offers a higher bandwidth for 4K/480Hz or 8K/165Hz (compared to DisplayPort 1.4a’s 4K/240Hz and 8K/60Hz), but currently most monitors are not capable of such high refresh rates, so it’s not a major consideration here time. Plus, there probably isn’t enough framebuffer to manage on this range of cards.
