The Nvidia RTX 4070 Ti is a card that is going to make the inexperienced staff some huge cash. For one factor it is a GPU in a position to ship gaming efficiency both stage with or forward of the RTX 3090 (opens in new tab), and usually past the dearer new AMD RDNA 3 graphics playing cards, too. Importantly, for Nvidia’s cash individuals, it is also a slice of Ada GPU silicon that is solely somewhat larger than the chip contained in the RTX 3050 however is retailing for a heady $799.
These die space and pricing numbers add as much as a GPU that ought to be comparatively low cost for Nvidia to provide and can then provide excessive margins to please shareholders and Jen-Hsun alike.Â
So, excessive efficiency and excessive margins; the holy grail for any graphics card making firm. And while you throw within the twin advantages of DLSS 3 and its black magic Body Era tech, you additionally get the flexibility to throw hyperbolic efficiency advertising and marketing numbers at it and nonetheless be capable to again it up with precise benchmark knowledge.
It is also a card that represents one thing new from Nvidia—one thing akin to a willingness to confess when it is made a mistake. This card initially glided by the title of the RTX 4080 12GB till Nvidia unlaunched it (opens in new tab) with models already made and sitting in companions warehouses ready to ship. Group inexperienced lastly realised it was complicated having two very completely different GPUs beneath the identical title, although solely after it was already introduced and the trade had exaggerated a collective spit-take.
Now it is rebadged and the Nvidia RTX 4070 Ti, née RTX 4080 12GB, is lastly in our check rig with the very same specs as earlier than however with a $100 cheaper value level. There is no such thing as a Founders Version for the RTX 4070 Ti, and so we have an MSRP-level Gigabyte RTX 4070 Ti Gaming OC (opens in new tab) card for our testing of the most recent Nvidia GPU.
Nvidia RTX 4070 Ti verdict
Is the RTX 4070 Ti any good?
The unlaunching and subsequent rebadging and repricing of the RTX 4080 12GB was the very best factor to occur to this third-tier Ada GPU. Now and eternally to be often called the RTX 4070 Ti, that is the cardboard that now makes it unimaginable to suggest AMD’s RX 7900 XT (opens in new tab), making that dear RDNA 3 providing much more unpalatable than it already was.
If, because the RTX 4080 12GB, it had launched forward of AMD’s first chiplet GPUs for the unique value it will have gotten a hammering, however now it is each cheaper and virtually as performant. And in some situations extra so.Â
It is had the identical detrimental impression on how I really feel concerning the $1,200+ RTX 4080 16GB (opens in new tab) card, too. It is on common some 21% slower on the 4K stage, a lot nearer at 1440p, however with a sticker value that is at the least 33% cheaper. I do not understand how you suggest somebody spend the form of cash retailers are asking for the RTX 4080 while you’ve obtained both a much more highly effective card within the RTX 4090 (opens in new tab) or a far cheaper one now within the RTX 4070 Ti bookending it.
I actually really feel like each these unloved AMD and Nvidia playing cards are due a value correction sooner reasonably than later. At the very least the inexperienced staff has discovered one thing else to do with the undesirable RTX 4080 silicon due to the brand new GeForce Now Final tier (opens in new tab).
In gaming phrases, the 4K efficiency of the RTX 4070 Ti is spectacular even with out upscaling, and is reasonably astounding with it. But it surely’s the 1440p efficiency the place it actually makes its mark, most particularly towards the opposite GPUs round it. The hole between it and the RTX 4080 slims down and it could possibly even publish larger body charges than each AMD’s prime RDNA 3 GPU at 1440p the place the Radeons’ further 20GB and 24GB reminiscence swimming pools can not help them out.
I would have perhaps hoped for a higher efficiency delta in contrast with the RTX 3080 (opens in new tab), the cardboard it is actually changing from the Ampere stack, particularly on condition that it is priced $100 larger. The retrograde comparative specs checklist can be disappointing—one thing that irked us when it was initially unveiled beneath its RTX 4080 12GB guise (opens in new tab)—and it actually appears like that 192-bit reminiscence configuration is Scrooge-level miserly.
However nonetheless, placing DLSS and Body Era apart, the RTX 3090-esque efficiency ranges of the RTX 4070 Ti are exhausting to disregard. That was a monster GPU this time final 12 months, and right here we’ve got a card that is virtually half the worth with basically the identical gaming chops.
Contemplating the present state of the market, that is the graphics card that I would need above just about every other on the market other than the RTX 4090.
Nvidia RTX 4070 Ti specs
What are the RTX 4070 Ti specs?
The RTX 4070 Ti is utilizing a wholly new Ada Lovelace GPU, the AD104. This in itself was one of many predominant causes for the consternation about it being known as one other RTX 4080 regardless of having a totally completely different chip at its coronary heart. It is a much smaller GPU, coming in at 294.5mm2 in comparison with the ~378.6mm2 of the AD103 within the precise RTX 4080. It is also sporting the total AD104 core (just a few decoders apart), the place the one AD103 we have seen up to now has been a 95% chip lacking the total complement of CUDA cores.
Housed inside this full AD104 GPU are 7,680 CUDA cores arrayed over 60 streaming multiprocessors (SMs), with 80 ROPs, 240 fourth-gen Tensor Cores, and 60 third-gen RT Cores.Â
Header Cell – Column 0 | RTX 4070 Ti | RTX 3080 12GB |
---|---|---|
GPU | AD104 | GA102 |
Structure | Ada Lovelace | Ampere |
Lithography | TSMC 4N | Samung 8N |
CUDA cores | 7680 | 8960 |
SMs | 60 | 70 |
RT Cores | 60 | 70 |
Tensor Cores | 240 | 280 |
ROPs | 80 | 96 |
L2 cache | 49,152 KB | 6,144 KB |
Enhance clock | 2,610MHz | 1710MHz |
Reminiscence | 12GB GDDR6X | 12GB GDDR6X |
Reminiscence interface | 192-bit | 320-bit |
Reminiscence bandwidth | 504 GB/s | 912 GB/s |
Die dimension | 294.5mm² | 628.4mm² |
Transistors | 35.8B | 28.3B |
TGP | 285W | 350W |
Worth | $799 | $799 |
That is a reasonably respectable stability sheet, however not the form of gen-on-gen increase that we have seen from the mighty RTX 4090 in contrast with its personal high-end predecessor. Actually, towards both the unique RTX 3080 or the mildly tweaked (and value comparable) 12GB RTX 3080, you are taking a look at a particular retrograde step by way of GPU configuration.
There are 1,280 fewer CUDA cores on this RTX 4070 Ti, and whereas Nvidia will probably now want to retcon this card as one which now should not be in contrast with the xx80-class GPUs of the earlier technology, it is tough to not see that as a step backwards. In fact the Ada and Ampere architectures are completely different, however not a lot that the precise CUDA cores are completely different in themselves.
So, the place does Ada’s further efficiency truly come from? The true change from Ampere to Ada stems from the TSMC 4N lithography enabling critically excessive clock speeds from even the largest Ada GPUs, and the large improve in L2 cache Nvidia has them jammed into every chip.
In contrast with the RTX 3080 we’re speaking a few GPU that is working over 1GHz quicker on common, and one which’s sporting virtually 10x the cache reminiscence inside it. I’ve beforehand gone into depth on the total Ada Lovelace GPU structure (opens in new tab) in my RTX 4090 evaluate, however in actual phrases it is not a basic change within the cores that delivers the RTX 40-series its efficiency, however the mixture of excessive clocks and cache.
After which, while you throw within the rising variety of video games with ray-traced lighting options, the additional efficiency of the up to date Ada third-gen RT Cores pushes body charges even additional.
The CUDA core depend is not the one generational downgrade both. The reminiscence system of the RTX 4070 Ti is extra akin to that of the RTX 3060 12GB (opens in new tab) than the RTX 3080 12GB. Certain, it is obtained the identical quantity of precise VRAM, but it surely’s working on an aggregated 192-bit reminiscence bus made up of six 32-bit reminiscence controllers, as an alternative of the 320-bit bus of the older card. You might be getting RTX 3070 ranges of bandwidth, nonetheless, due to the actual fact it is utilizing GDDR6X reminiscence working at 21Gbps, however that does all add as much as a sense that it is a GPU nicely down the ladder on the subject of general specs.
However Nvidia is not structuring its GPUs by underlying core configurations, it is wanting on the uncooked efficiency metrics and badging playing cards up based on body charges not specs. Which I assume is how you find yourself within the form of mess the place it proposes a pair of RTX 4080 playing cards with wildly completely different graphics chips inside them.
Nvidia RTX 4070 Ti efficiency
How does the RTX 4070 Ti carry out?
Presumably probably the most spectacular factor to say concerning the RTX 4070 Ti is that very usually it is stage to, or quicker than an RTX 3090. Whenever you assume that is the $1,500 GPU of the final technology that appears like an incredible gen-on-gen uptick in efficiency, particularly when that is on the prime 4K decision.
What’s perhaps much less thrilling is that, while you’re simply speaking in straight rasterized gaming phrases, it is not an entire lot quicker than the previous, cheaper RTX 3080 10GB at 4K. It is quicker, most particularly while you convey these third gen RT Cores into the equation, but it surely’s clear the upper clocks and heftier L2 cache is having to work exhausting to provide it the lead in uncooked body fee phrases over the older Ampere card.
Simply think about how good this card would have been if it was utilizing a cut-down AD103 GPU with the identical core depend because the RTX 3080, and with the identical reminiscence bus.Â
The place it seems much more optimistic is up towards the brand new AMD RDNA 3 playing cards, the RX 7900 XTX (opens in new tab) and RX 7900 XT (opens in new tab). It’s usually slower than the highest Radeon GPU, however towards the nonetheless dearer RX 7900 XT the RTX 4070 Ti usually posts larger 4K efficiency.Â
4K gaming efficiency
That Nvidia lead is extra constant on the 1440p stage, the place the additional Radeon video reminiscence cannot give RDNA 3 GPUs the sting and even the AMD RX 7900 XTX generally finds itself behind the third-tier GeForce card. That is an actual feather within the techie cap of the Nvidia structure, and that is earlier than we get into any of the upscaling shenanigans that both facet of the GPU divide can provide over and above the bottom graphics efficiency.
1440p gaming efficiency
The AD104 GPU as soon as extra highlights simply how environment friendly the nominally 4nm Ada Lovelace structure is. In common gaming phrases the RTX 4070 Ti runs nicely under the 300W mark and presents larger efficiency per watt metrics than all however the RTX 4080. That’s one place that it’s comfortably forward of the RTX 3080 and the Ampere structure as an entire.
In thermal phrases, too, regardless of being a dual-slot reasonably than a triple-slot card, this Gigabyte triple-fan RTX 4070 Ti runs impressively cool, too. I did not see it getting above 60°C in any of my testing. I imply, it could be ‘simply’ a dual-slot cooler, however that does nonetheless imply you are getting an actual mass of heatsink connected to the tiny PCB the RTX 4070 Ti it sits atop, and in real-terms it is not that rather more slight than the chonky RTX 4080 Founders Version.
System efficiency
So sure, with none of the DLSS 3/Body Gen stuff in attendance the RTX 4070 Ti is a really succesful performer, however as soon as once more the ability of Nvidia’s upscaling tech is preposterously good. I preserve making an attempt to see the place the Body Era know-how fails however I am unable to do it. Each time I am like ‘aha, there it’s, the tell-tale artefact of faux AI frames!’ I then try the native rendering and it seems precisely the identical. If not worse.
And with the additional real efficiency of the upscaled frames and the interpolated smoothness of the AI-generated frames, the efficiency enchancment is spectacular the place DLSS 3 is obtainable. Which must be increasingly more typically, with Nvidia’s Streamline SDK providing devs a one-stop possibility for enabling it and different distributors’ upscaling tech too.
DLSS efficiency
At its highest high quality setting DLSS 3 with Body Era reveals greater than double the 4K body fee of Cyberpunk 2077, Microsoft Flight Sim, and F1 22. For me, that is nonetheless the very best function of Nvidia’s next-gen structure, and I actually hope Body Gen will get opened as much as work on Ampere at the least down the road.Â
Nvidia RTX 4070 Ti evaluation
What does the RTX 4070 Ti imply for avid gamers?
I nonetheless do not actually understand how to be ok with yet one more extremely enthusiast-priced graphics card launching into the wan gentle of a brand new 12 months the place world financial considerations are entrance and centre in all our information media. Even on TikTok. Yeah, down with the youngsters, me.Â
We’re now 5 GPUs into this new technology of PC graphics playing cards, and the least offensively priced one remains to be $799. And that is solely due to a climbdown from Nvidia’s earlier place that this was an $899 GPU. That is all greatest case state of affairs, too, as a result of we have seen some AIBs nonetheless slapping $900+ value tags on their overclocked variations of the RTX 4070 Ti.
It is clear then we’re nonetheless a good distance from mainstream Ada Lovelace graphics playing cards with Nvidia’s Jeff Fisher fronting its detail-light CES 2023 occasion explicitly stating “the RTX 30-series continues to be the very best GPU for mainstream avid gamers.” Even when, in actuality AMD’s RX 6000-series GPUs are those you have to be dropping greenback on in the event you’re searching for the very best bang in your buck on the extra affordable finish of the graphics card market.
However there’s undoubtedly an argument to be made that in its personal value bracket the RTX 4070 Ti is a greater possibility for anybody trying to spend the very best a part of a grand on their new GPU than both the AMD RDNA 3 playing cards or the prohibitively priced RTX 4080.
And, actually, after I began testing the cardboard within the workplace pre-Christmas, I used to be shocked on the stage of efficiency it presents. I would seen all the unique benchmarks posted about it because the RTX 4080 12GB, and knew its inflated numbers had been right down to DLSS 3 and Body Era. But it surely’s how nicely it does with out all that which makes it such a worthwhile GPU funding for high-end avid gamers.
If, nonetheless, Nvidia had pushed it out on the similar $699 value as the unique RTX 3080, with the additional gaming efficiency that it presents I would have been much more effusive in my reward of it. And it will have completely castrated the AMD RDNA 3 playing cards, too. At $799 it is solely simply far sufficient forward of the last-gen GPU it is ostensibly changing.
However it’s forward, and usually sufficient performs on the similar stage because the $1,500 RTX 3090 from the Ampere lineup to nonetheless really feel like an incredible improve for anybody on a excessive finish card from both the 10- or 20-series GeForce generations.Â
The one considerations now are what the RTX 4070 Ti means for the decrease spec GPUs which is able to hopefully comply with quickly. As I discussed earlier, this AD104 chip has been pared again, most particularly on the subject of the reminiscence configurations, in contrast with the equal Ampere playing cards. Fingers crossed we’ll see RTX 4070 and RTX 4060 playing cards utilizing the identical GPU, with just a few SMs disabled, however utilizing the identical 192-bit bus. Slicing that any extra aggressively might actually hobble the extra mainstream Ada playing cards.