As we inch closer to the launch of the RTX 40-series, every day a new leak is coming out contradicting the previous one. Just a couple of days ago, I covered a rumor going over the leaked power limits of the upcoming Ada Lovelace GPUs. Today, we have a follow-up to that of sorts.
The rumor is coming from the same leaker, @koptite7kimi, who has taken to Twitter to yet again to share the updated specs of the RTX 40-series graphics cards. Since day one, the exact specifications of these GPUs have been in contention with updates flying around all the time. Let’s take a look at what Kopite has to say this time.
And TBP, 450/420?/300W.
— kopite7kimi (@kopite7kimi) June 23, 2022
In his tweet attached above, you can see that he goes over the specs of the initial launch lineup of the RTX 40-series GPUs. That includes the RTX 4090, the RTX 4080, and the RTX 4070. Starting with the top-of-the-line SKU, we see only one update to its spec-sheet.
RTX 4090 updated look
The RTX 4090 has been rumored to feature 16128 CUDA Cores for a while now. The card will use the flagship AD102 GPU, but not the full die, that is reserved for the RTX 4090 Ti (and possibly a new TITAN). The new, updated specs suggest that the RTX 4090 will instead rock 16384 CUDA Cores.
That takes the Streaming Processor count to 128, from 126. As mentioned, the full AD102 die is much more capable, at 144 SMs. Regardless, rest of the RTX 4090 remains unchanged. It is reported to still come with 24GB of GDDR6X memory clocked in at 21Gbps, running across a 384-bit wide bus. The TDP is also the same at 450W.
RTX 4080 updated look
Moving on to the RTX 4080, this is the next-in-line to the launch flagship of the Ada Lovelace lineup. Kopite’s updated specs actually provide us with our first look at the RTX 4080 and RTX 4070’s internal configurations. Previously, we really only knew what GPU they would be using.
So, the RTX 4080 is based on NVIDIA’s AD103 GPU and the card is most likely not utilizing the full die, due to the prospect of unlocking that potential in a future RTX 3080 Ti down the line. The RTX 4080 will reportedly feature 10240 CUDA Cores.
As for the memory, we’re looking at 16GB of GDDR6 spread across a 256-bit wide memory bus. The memory is clocked at 18Gbps. That would make it slower than the RTX 3080, which has 19Gbps memory due to be using faster modules.
Unlike the RTX 4090 (and majority of the current 30-series), the RTX 4080 is apparently not going to be using GDDR6X memory. This is most likely going to not stay this away as more leaks come to surface in the coming weeks.
For those who don’t know, GDDR6X is NVIDIA’s special flavor of the standard GDDR6, developed hand-in-hand with Micron, and exclusively for its GPUs. It runs hotter, but is supposedly faster and the trade-off is worth it in NVIDIA’s eyes.
Anyways, even if the RTX 4080 ships will slower memory than its predecessor, it will only be so on paper. Because the Ada Lovelace GPUs have larger L2 caches, which make them a lot more efficient, which in turn can make the GPU just as fast, or even faster than the GDDR6X variant in terms of memory.
RTX 4070 updated look
Lastly, we have the RTX 4070, the card that will arguably sell the best at launch out of all the 40-series offerings. Here, the only update to the specs is in the memory capacity. The card is now expected to sport 10GB of GDDR6 memory instead of the previously-rumored 12GB.
This 2GB-cut would result in a narrowing of the memory bus from 192-bit originally to, now, 160-bit. Apart from that, the CUDA Core count is reportedly 7168, which is 512 less than what the full AD104 GPU has. Again, an RTX 4070 Ti is where we’ll likely the see the fully unlocked AD104 put to use.
The next generation
With that, we conclude this breakdown of the updated look at the next generation of NVIDIA’s GPUs. At the tail end of Kopite’s tweet, he mentions (or rather warns) to not expect lower MSRPs. We’re already paying more than the suggested retail prices (bar the recent crypto crash) so a further increase in prices could be a real upset.
TSMC has reportedly increased the price of their wafers by around 10-20%, according to DigiTimes. That mixed with the on-going hyperinflation being experienced around the world, costs are bound to go up and customers are expected to eat up the consequences.
All in all, the RTX 40-series looks promising. With higher power requirements and more boastful spec-sheets than ever, this is going to be a rather radical upgrade. AMD is also ready to unleash their RDNA 3 GPUs to the world very soon, a lineup that is perhaps even more exciting that NVIDIA’s given the hill AMD is currently standing on.
RDNA 2 came so close to dethroning the Green Team for the superior graphics card, but RDNA 3 is expected to be the nail in the head. Not only that, Intel is also gearing up its Arc A-Series GPUs, which can present another unforeseen threat to NVIDIA’s dominance.
That battle for customers’ pockets between AMD, Intel, and NVIDIA is going to be an interesting one and perhaps there will be no one clear winner. Whether all of these companies can prevail and find their own footings in the rapidly-changing GPU market with next-generation products remains yet to be seen.
We don’t have to wait long to see this all unfold as RDNA 3 GPUs are expected to launch sometime in November, while Intel Arc GPUs are already starting to hit the shelves. NVIDIA’s RTX 40-series is rumored to release in Q2/Q4 of 2022 with each of the aforementioned three GPUs launching one month apart from each other.