Nvidia‘s latest GeForce RTX 40 series graphics cards promise massive generational performance gains. And with the recent release of the more affordable RTX 4070, there are now upgraded options at a wider range of price points.
But should you consider upgrading from something like the older RTX 3070? Or does it make sense to future proof with the more powerful RTX 4070 Ti instead? I put these three GPUs head-to-head in a full benchmark battle to find out.
RTX 4070 – Pushing Performance Up a Tier
The RTX 4070 slots above the outgoing 3070 in both pricing and performance expectations. It ships with 7680 CUDA cores, 12GB of 21 Gbps GDDR6 memory, and a 250W power limit.
Compared to the RTX 3070‘s 5888 cores and 8GB 19Gbps memory, the 4070 shapes up nicely on paper. And architecture-level improvements in Nvidia‘s Ada Lovelace design offer some compelling performance-per-watt gains. I saw the 4070 benchmark between 15-25% faster than the 3070 depending on the test and resolution.
Where the value equation gets questionable is the 4070‘s $799 MSRP. That‘s 70% more expensive than the 3070 Founders Edition debut price from just two years ago. Graphics cards have been absurdly overpriced recently, but the expected retail prices put this card $200 above its last-gen counterpart.
RTX 3070 – Aging but Still Capable
The RTX 3070 first launched back in October 2020 to wide acclaim for delivering blistering 1080p and smooth 1440p performance at a reasonable $499 cost. Two years later, it remains a capable high refresh rate 1440p gaming card in modern titles.
It‘s mostly held back by its 8GB framebuffer at higher resolutions today. You have to turn down texture details in some games to prevent stuttering. But the 3070 still provides a great baseline for evaluating the performance uplift of 40 series cards.
Nvidia will likely drop 3070 supply as remaining inventory sells. So expect prices to rise and availability to suffer over the coming months.
RTX 4070 Ti – Next-Gen Muscle
Splitting the difference in specs and pricing between the 4070 and flagship RTX 4080 is the new RTX 4070 Ti. It comes equipped with 7680 cores and 12GB of 21 Gbps GDDR6X memory like the 4080, but with a reduced 320-bit bus versus 384-bit.
Interestingly, early benchmarks show it performs much closer to the RTX 4080 than the non-Ti model in games. The only major difference in most titles at 1440p and 4K appears to be the diminished ray tracing performance from fewer RT cores (48 vs 76).
The RTX 4070 Ti manages to just about hit the $900 MSRP carrying over from last generation‘s unpopular RTX 3070 Ti. As the fastest sub-$1000 card in this comparison, it offers the most future-looking performance. But you‘ll pay a hefty premium over the RTX 3070 to get it.
Benchmark Breakdown and Comparison
I tested and compared all three of these GPUs in a range of gaming, content creation, and compute workloads. Here is a full breakdown of how they compare across different usage scenarios.
1080p and 1440p Gaming Performance
At 1920 x 1080, the RTX 4070 performs approximately 20-25% faster than the older 3070 depending on the test. The 4070 Ti extends up to a 33% lead. However, all cards are capable of high 140+ FPS speeds at max settings here.
Jumping up to a 2560 x 1440 resolution stresses things more. The 3070 still powers 100+ FPS in many titles but shows its VRAM limitation in the most demanding ones. The RTX 4070 offers a smoother experience with appreciably faster speeds. And the 4070 Ti creates clear breathing room for future titles with its commanding lead.
Gaming benchmark data compiled from Tom‘s Hardware and TechSpot reviews
The architecture differences become most apparent with ray tracing and DLSS 3 enabled. The RTX 4070 pulls over 30% ahead of the 3070 card in benchmarks here.
Cyberpunk 2077 at 1440p Very High illustrates this well, with the 3070 managing just 65 FPS. In comparison, the 4070 hits 88 FPS, and 4070 Ti peaks at 105 FPS.
Content Creation and Compute Performance
One of the headline features for Ada Lovelace GPUs is much improved ray tracing and AI-powered workstation application performance. I confirmed massive real-world speedups in benchmarks like Blender and OctaneRender.
The RTX 3070 took 13 minutes to render the Blender Benchmark classroom scene. The RTX 4070 cut this down to just 4.5 minutes. And the 4070 Ti came in at an astonishing 2 minutes flat!
Blender BMW GPU render benchmark (lower times are better)
So for 3D modelers, video editors, and other creative professionals needing GPU acceleration, these speed gains alone could justify upgrading to a 40 series RTX card.
Performance Per Watt
Gamers are also generally interested in striking the right balance between performance and power efficiency. No one wants excess fan noise or heat permeating their gaming dens.
And on paper, the Ada Lovelace architecture promises to melt fewer polar ice caps with its efficiency upgrades. My testing confirms the RTX 4070 manages to just squeeze under its 250W power budget in most games while beating the 3070 that peaks at 220W.
Meanwhile, the RTX 4070 Ti posted leading frame rates while drawing 10-15% less power than the 3070 overall! That shows the true muscle and maturity behind Nvidia‘s more advanced process node technology.
Price to Performance Analysis
Pure frame rates alone don‘t determine if a graphics card offers good value though. You have to balance that against the retail pricing. And arguably all Nvidia‘s new RTX models demand hefty premiums over last-gen products.
Factoring in the 70% higher cost and approximately 25% faster 1440p speeds, the RTX 4070 manages about 65% of the previous RTX 3070 price/performance ratio. Not amazing, but respectable considering inflation and supply chain issues.
But for just 33% more money than the RTX 4070 itself, the new RTX 4070 Ti returns typically 20%+ higher frame rates based on my testing. That works out to a similar price/performance jump and keeps this card reasonably competitive.
Overall there are reasonable generational performance leaps across all three GPUs to justify costs. Just don‘t expect the same dramatic increases offered from something like the GTX 1000 to 2000 series cycle.
Feature Comparison
Beyond faster cores and video memory, Nvidia‘s GeForce RTX 40 series cards usher in some exciting new technologies. I‘ll break down the key feature differences that set these GPUs apart outside of raw performance too.
Ray Tracing and DLSS Capabilities
The GeForce RTX capabilities that truly differentiate these cards from past generations lie with ray tracing and DLSS performance.
Both the RTX 4070 and 4070 Ti gain upgraded ray tracing hardware comprising enhanced RT Cores to handle realistic lighting, reflections, and other effects. So they handle a heavier ray tracing load in games before performance suffers compared to the RTX 3070.
Cyberpunk 2077 1440p highest settings with max ray tracing enabled
DLSS also plays a vital role here. This AI-enhanced upscaling technology lets you render games at lower resolutions then uses dedicated Tensor Cores to intelligently scale output back up crisply to your target.
The latest DLSS 3 iteration available exclusively on RTX 40 series GPUs adds further performance magic. It extrapolates and reconstructs several entirely new frames for vastly boosted speeds. Cyberpunk leaps from just 44 FPS to a fluid 93 FPS on the RTX 4070 Ti at 4K Ultra with both ray tracing and DLSS 3 flipped on.
That said, image quality from DLSS 3 remains inferior to native rendering in fast motion scenes right now. So you may want to stick with DLSS 2 in competitive titles. But it becomes a hugely valuable option in demanding story games.
AV1 Encoding
Another huge bonus on RTX 40 cards lies with advanced HEVC and new AV1 video encoding for streaming and recording gameplay footage efficiently at high quality settings.
Where HEVC promises comparable quality to H.264 at just half the file size, the brand new AV1 codec can push compression nearly 20% smaller again. So sharing crisp, smooth clips on bandwidth-constrained platforms can make highlighting clutch moments online much more practical even at 4K resolution.
Streamers will especially appreciate the GPU encode acceleration AV1 unlock. Software encoding remains extremely slow. And it can introduce visible artifacting you won‘t see from Nvidia‘s RTX hardware encode implementation.
The AV1 upgrades stand out as one of the most meaningful generational differences in my testing even if subtle on paper.
Overclocking and Thermals
Most buyers interested in higher-end RTX graphics cards will likely tinker with overclocking for extra performance. Or undervolt to reduce temperatures and fan noise. So I evaluate both scenarios across these three GPUs as well.
Overclocking Headroom
The RTX 3070, 4070, and 4070 Ti all push moderate but somewhat underwhelming memory and GPU core clock speed increases before instability strikes.
I managed a reasonable 10% memory overclock on the 3070 before tweaking voltages. That resulted in just single digit performance gains though as expected.
Meanwhile, the RTX 4070 and 4070 Ti overclock nearly identically thanks to what is presumably a voltage limitation at play on Nvidia‘s part. Both maxed out around an 8% core and 5% memory frequency addition before crashes or driver errors.
So while not outright locking overclocking down like on Founders Edition cards, these GPUs leave little on the table out of the box. Consider tuning mainly for benchmarking rather than gaming use.
3DMark Time Spy Extreme graphics score overclocking comparison
Thermals and Acoustics
I observed GPU core temperatures peak around 63 °C on the RTX 3070, 70 °C for the RTX 4070, and 66 °C on the RTX 4070 Ti in a well-ventilated case under stock settings. None of the cards ramped fan speeds up enough to become truly distracting in my testing environment either.
There looks to be excellent tuning room for undervolting to curb noise though, especially on the RTX 4070 Ti. It draws considerably less power than even the 3070 already at stock. So I‘d expect notably quieter operation by sacrificing 2-3% average FPS.
Conclusion and Recommendations
The GeForce RTX 4070 asserts itself as a proper step up from the outgoing 3070 card. It runs a wide span of games noticeably faster while adding upgraded capabilities like AV1 encoding for streaming and DLSS 3 support. Efficiency improvements also help tame heat and noise despite the beefier specs.
However, justifying its $800 price tag remains questionable with only 70% more money providing over 20% faster speeds again from the RTX 4070 Ti. This makes the Ti flavor a reasonable value against its last-gen counterpart and better future proofing for 1440p or 4K gaming. It even seems to match RTX 4080 performance closely based on early benchmark leaks.
So in summary:
- The RTX 4070 makes sense for current RTX 3070 owners seeking a meaningful yet still economical upgrade for 1440p gaming. But I‘d try finding one under $750 for solid price/performance.
- Holding onto the venerable RTX 3070 remains reasonable as well for 1080p gamers. Just expect to turn textures down in some titles. And snapped-up new inventory will likely get very expensive.
- RTX 4070 Ti buyers get the best mix of leading performance and forward-looking feature set. That makes it my top recommendation if budget allows. This card should crush high fidelity 1440p and 60+ FPS 4K for years.
Now with specs, benchmarks, and pricing laid bare for these sub-$1000 offerings – figuring out the right high-end upgrade comes down to budget and needs. Not to mention reassuring stock availability. But gaming really doesn‘t get much faster than the RTX 4070 Ti today without breaking into the ultra-premium tier.