Nvidia might have accidentally revealed the specs of the RTX 4090 Ti

Yes, we’re all excited about the GeForce RTX 4090. Yes, it’s going to be available in just a few weeks. Yes, it’s all but guaranteed to be an absolute monster of a GPU, taking us to new 8K gaming heights (with an appropriately ludicrous price tag).

But there’s another, as-of-yet-unannounced graphics card that I’m more interested in. Well, technically there are two – I still want to see a more wallet-friendly RTX 4070 – but here I’m going to be talking about the potential new flagship Nvidia could have in the works: the RTX 4090 Ti.

See, the RTX 4090 has been confirmed to use Nvidia’s ‘AD102’ GPU chip, while the two models of the RTX 4080 will use the AD103 and AD104 chips. But the AD102 chip powering the RTX 4090 isn’t the full-fat version of the AD102; meaning that it doesn’t have access to the full suite of CUDA cores that the chip is capable of using.

Now, this is common practice for chipmakers; it’s cheaper to mass-produce one chip for two different GPUs than produce two completely different chips, after all. But now that we know the AD102 chip’s specs, Nvidia may have unintentionally given us a look at how powerful an RTX 4090 Ti using the AD102 could be.

A rendering of the Nvidia Lovelace GPU

(Image credit: Nvidia)

Analysis: Generationally, the RTX 4090 Ti could blow its predecessor out of the water

Let’s break down the numbers real quick. The RTX 4090 will have 16,384 CUDA cores, while the AD102 chip has a potential maximum of 18,434 – that’s 71% more cores than the GA102 chip found inside the current-generation RTX 3090 Ti. There are more Tensor and ray-tracing cores too, naturally.

Since the RTX 4090 does have the full 24GB of VRAM that the AD102 uses, this means we could be looking at around a 12.5% increase in performance if nothing else is tweaked. That sounds reasonable enough to me; the RTX 3090 Ti offered around 10-15% better performance than the regular 3090 in most areas, so this would be consistent with previous generations.

The RTX 4090 does use the AD102’s listed maximum boost clock of 2.52GHz, so don’t expect the 4090 Ti to have higher clock speeds – unless Nvidia has a secret AD101 chip in the works, of course. Still, a 4090 Ti equipped with a full-fat AD102 chip will no doubt be one of the best graphics cards ever made.

When could this potential GPU release? I’ve got no clue – Nvidia has been playing its cards close to the chest, only giving us official reveals and release dates for the RTX 4090 and 4080 so far. Still, RTX 3000 series pricing is dropping, and those cards will probably get even cheaper once RTX 4000 goes on sale, so if you need to upgrade and want to snap up a great graphics card deal, keep your eyes peeled.

Go to Source