NVIDIA GeForce RTX 4060 Ti Founders Edition Review
The RTX 3060 Ti set the bar very high
The initial 8GB version of the RTX 4060 Ti is upon us, replacing the brilliant RTX 3060 Ti (its only flaw was availability). And while the internet has collectively decided that this VRAM capacity is unacceptable in 2023 – due mainly to the increasingly lazy unoptimized ports PC gamers have to deal with – it is telling that NVIDIA isn’t promoting this as a 2560×1440 solution.
Which of course brings up the next point of contention: the $399 price tag. I suspect that complaints about pricing will take a back seat to those concerning VRAM with this launch, as everyone knows that things cost more these days and VRAM is the hottest GPU topic of all time. Anyway, there will be more of a proper mainstream 1080p option with the RTX 4060 (non-Ti) at $299 USD, but not until July.
Regardless of the VRAM argument, complaints about pricing, more VRAM arguments, further complaints about VRAM, and various other VRAM-related topics, we will look at performance in a few titles on an all-new “mainstream” GPU test platform in this review. Our goal is to see where this card sits relative to the previous-gen RTX 30 Series, and look at a few Radeon cards as well. But first, a look at specs for available RTX 40 GPUs (up to the RTX 4080):
RTX 4060 Ti | RTX 4070 | RTX 4070 Ti | RTX 4080 | |
---|---|---|---|---|
GPU | AD106 | AD104 | AD104 | AD103 |
Architecture | Ada Lovelace | Ada Lovelace | Ada Lovelace | Ada Lovelace |
SMs | 34 | 46 | 60 | 76 |
CUDA Cores | 4352 | 5888 | 7680 | 9728 |
Tensor Cores | 136 (4th Gen) | 184 (4th Gen) | 240 (4th Gen) | 304 (4th Gen) |
RT Cores | 34 (3rd Gen) | 46 (3rd Gen) | 60 (3rd Gen) | 76 (3rd Gen) |
Base Clock | 2310 MHz | 1920 MHz | 2310 MHz | 2205 MHz |
Boost Clock | 2535 MHz | 2475 MHz | 2610 MHz | 2505 MHz |
Texture Units | 136 | 184 | 240 | 304 |
ROPs | 48 | 64 | 80 | 112 |
Memory | 8GB GDDR6 | 12GB GDDR6X | 12GB GDDR6X | 16GB GDDR6X |
Memory Data Rate | 18 Gbps | 21 Gbps | 21 Gbps | 21 Gbps |
Memory Interface | 128-bit | 192-bit | 192-bit | 256-bit |
L2 Cache Size | 32MB | 36MB | 48MB | 64MB |
Memory Bandwidth | 288 GB/s | 504 GB/s | 504 GB/s | 716 GB/s |
Transistor Count | 22.9B | 35.8B | 35.8B | 45.9B |
Die Size | 190 mm^2 | 295 mm^2 | 295 mm^2 | 379 mm^2 |
Process Tech | TSMC 4N NV Custom | TSMC 4N NV Custom | TSMC 4N NV Custom | TSMC 4N NV Custom |
TGP | 160W | 200W | 285W | 320W |
Launch Price | $399 | $599 | $799 | $1199 |
Some Performance Testing
While the following benchmark results are by no means comprehensive, we can at least offer a glimpse of what one might expect from the GeForce RTX 4060 Ti. I think you will notice the performance trend immediately, at least in “raw” performance. But first, a word (or fifty) on the test setup used for this review.
Every benchmark result in this review was the result of fresh testing over the past four days; literally from May 19 through May 22. As I had an Intel Z790 board on hand with DDR4 support, I had the idea to feature the Intel Core i5-13600K processor in a “mainstream” build, trying to keep things under $1000 rather than going with the $2500+ systems I normally test GPUs with.
Test Platform | ||||||||
---|---|---|---|---|---|---|---|---|
Processor | Intel Core i5-13600K (Stock, Power Limits Enforced) | |||||||
Motherboard | MSI MPG Z790 EDGE WIFI DDR4 BIOS v1.60, Resizable BAR Enabled |
|||||||
Memory | 16GB (8GBx2) PNY XLR8 REV DDR4-3600 CL18 | |||||||
Storage | Samsung 980 1TB NVMe SSD | |||||||
Power Supply | be quiet! Dark Power Pro 13 850W | |||||||
CPU Cooler | be quiet! Pure Loop 2 FX 280mm AiO | |||||||
Operating System | Windows 11 Pro, 22H2 | |||||||
Drivers | GeForce Game Ready Driver 531.79 – 531.93 (RTX 4060 Ti Press Driver) Adrenalin 23.4.3 |
I almost succeeded, but this particular mix of components ended up being closer to $1080 (if we pretend I remembered to switch to the Pure Power 12 PSU). With a less expensive motherboard, an air cooler for the CPU, and some other edits, one could build a similar platform (with 32GB of RAM) for much less.
Do we really need to see any more results to know that this RTX 4060 Ti is going to be close to the performance of the RTX 3070, making it a modest upgrade over the RTX 3060 Ti? Yes we do need to see more, actually. NVIDIA is leaning heavily on DLSS 3 with Frame Generation for the kind of performance boost we saw going from the 20 Series to the 30 Series, and it is certainly possible to get a lot more out of the games that support this tech.
Cyberpunk was the better example of the potential of DLSS 3 + FG, doubling the performance of the 3060 Ti at 1080 / RT Ultra. We didn’t see the same boost from F1 22 with frame generation enabled, but it made an obvious performance difference.
And now I will touch on what I consider to the most impressive aspect of this new GPU: power consumption.
Total power draw is down significantly with this generation’s 60 Ti card, as we can clearly see from the squiggly lines above. This RTX 4060 Ti is a very efficient card, obviously.
Final Thoughts
Looking back only a few years, I think a card like the RTX 4060 Ti would meet expectations for a xx60 Ti card – which is to say that it effectively matches the performance of the previous-gen xx70 card, and adds current-gen features. But we live in the post-RTX 30 Series era now.
While many actual gamers were left empty-handed during the dark times (f*** Ethereum, anyway), the RTX 30 Series was a BIG upgrade over the RTX 20 Series, and list pricing was very good for the performance level.
My favorite card last generation was the RTX 3060 Ti, and for its elusive MSRP of $399 it was the card I would have bought with my own money. Think about this: it was faster than the $699 (and up) RTX 2080, cruising past heavyweights such as GeForce GTX 1080 Ti and Radeon RX 5700 XT. And this begs the question, was the RTX 3060 Ti too good? It certainly set expectations for the next generation of GeForce cards very high.
Seeing only modest raw performance gains over the previous generation xx60 Ti card here isn’t very exciting, but there are architectural improvements with the RTX 4060 Ti that stretch the lead to more impressive levels. I didn’t cover things like content creation, where this generation offers a better experience.
This card wants you to use DLSS 3 + FG, and if you get it, use this. Regardless of what you’ve watched (or possibly even read) about DLSS 3 and Frame Generation, the tech does greatly increase the framerates and perceived smoothness of games, and in games that support the DLSS 3 + FG combination the RTX 4060 Ti crosses into enthusiast 2560×1440 territory – at least based on the FPS numbers I was seeing.
Now, about that VRAM thing. 8GB is certainly a useful amount, but there have been multiple (and heavily-documented) examples of recent titles that want as much as they can get. I would love it if this card had 16GB, and while I could pontificate about public companies maintaining margins on products amidst rising component costs, the fact is that gamers don’t care about how well company X is doing. They all just want cheap GPUs with lots of VRAM, as far as I can tell.
The fact that a 16GB version of the RTX 4060 Ti will be made available is definitely a good move, but it isn’t coming until July. I would have loved to see it launch alongside this card, but the additional $100 for the 16GB RTX 4060 Ti does push it into a different market segment. We will have to wait and see if AMD answers with something compelling, and creates some pricing pressure. I think we’d all love to see a price break on components for this increasingly expensive hobby.
The confusing thing is going to be the 4070 (12GB) vs the 4060Ti 16GB when it launches in July. How much better can the 4070 be performance wise to compel someone to purchase one over the 4060Ti 16GB? Is the bus width going to be a selling point (192 for 4070 vs 128 for 4060Ti)? I must assume that, if not true on launch, the 4070 will end up selling for less than the 4060Ti. I personally don’t plan on buying a 4000 series card (I’m sticking with a 3060Ti until the day I decide to go to 4k), but if I was going to, any marginal performance difference between the two wouldn’t persuade me to choose the 4070 over the 4060Ti 16GB given the extra 4GB of VRAM and the state of AAA game development (which is a bit overblown, but still a problem going forward as I expect things to get worse for PC ports).
Really, NVIDIA should have developed the 4070Ti to have 16 GB of VRAM, and let the 4060 just be what they have here. It would be linear and make sense.
I reckon we need the 5000 series before the stack makes any sense. But then, what do I know?