The rumor mill is abuzz the moment again as Nvidia’s next-gen Ada Lovelace GPUs draw nearer. Notable tipster @kopite7kimi is giving new particulars in the Twittersphere on what Nvidia has in retail store for gamers in the tumble. This time all over they’re shedding some new gentle on leading three GPUs in the product stack, which will very likely be the first to start.
In a series of new tweets, @kopite7kimi specifics the specs for these forthcoming GPUs. As has been claimed earlier, the flagship die will be Advert102. Apparently, the top-shelf “Lovelace” GPU will be outfitted with 24GB of GDDR6X functioning at 21Gb/s. This is the precise exact configuration as the present RTX 3090 Ti. It should be famous that the previous era Turing-primarily based Titan RTX also had 24GB of GDDR6 memory. It seems that for now, this is as considerably as Nvidia wishes to just take memory allocation on its significant-finish SKUs. There have been preceding rumors that Advertisement102 may come equipped with GDDR7 memory running at 24Gb/s, but people might be incorrect. What is seemingly firmer at this stage is the GPU’s Complete Board Energy (TBP). It is the moment once again listed as 600W, which appears like a offered due to the fact Nvidia would like to force this card to the absolute maximum stage of efficiency. It may possibly have to in purchase to fend off AMD for this spherical.
Going down the stack we occur to the RTX 4080. Even though the Ampere-centered RTX 3080 applied a reduce-down model of the flagship die, GA102, that won’t be the situation for the 4080. In its place of employing the massive Ad102 die, or the smaller 104 edition like with Turing and former generations, it is receiving its very own slice of silicon. It’ll be an Advertisement103 chip, slotting in concerning the large chip on the 4090 and the lesser chip on the 4070.
This card is rumored to offer 16GB of GDDR6, which will be a welcome update for people who considered the existing-gen’s 10/12GB was inadequate. Even so, it seems Nvidia will be downgrading the memory from GDDR6X to slower GDDR6. For comparison, the GDDR6X in the present RTX 3080 operates at 19Gb/s, whilst GDDR6 can hit about 14Gb/s. It is feasible Nvidia will compensate for this with a wider memory bus, or even added L2 cache. This GPU will supposedly have the exact TBP as the best-shelf GA102 die, which is 450W in the RTX 3090 Ti.
Lastly, we occur to the RTX 4070. This “tween” GPU could be the sweet location in conditions of general performance and energy usage for the Lovelace architecture. It’s noteworthy that for this era it’ll be having the Advertisement104 die, which is more substantial than what’s been reserved for this course earlier. With Turing and earlier generations, the 104 die was for the x80 GPU, and the more compact 106 die was for the x70 cards. In addition to a more substantial die it is also acquiring a major up grade in the memory division. It’ll be outfitted (reportedly) with 12GB of GDDR6 functioning at 18Gb/s. That’s around the same as the boosted RTX 3070 Ti, which capabilities 19Gb/s GDDRX memory. Each 3070 playing cards only have 8GB of memory nevertheless. Complete board energy is a quaint 300W. That made use of to appear like a large amount, extended in the past. Now it’s pretty much lovable. The leaker notes that neither the RTX 4080 nor the 4070 have started the screening stage. This does indicate the 4090 is presently at that phase.
This era of GPUs claims to be very exciting, as both equally Nvidia and AMD are rumored to be unleashing every thing they have in their arsenals. We know that does not seem any various from earlier eras considering the fact that these two firms have been battling for ages. However, Nvidia going up to 600W is absolutely a new issue. It implies it anticipates some major resistance from the AMD and its chiplet-based RDNA3 GPUs.