Wednesday, December 25th 2024
NVIDIA GeForce RTX 5070 and RTX 5070 Ti Final Specifications Seemingly Confirmed
Thanks to kopite7kimi, we are able to finalize the leaked specifications of NVIDIA's upcoming GeForce RTX 5070 and RTX 5070 Ti graphics cards.
Starting off with RTX 5070 Ti, it will feature 8,960 CUDA cores and come equipped with 16 GB GDDR7 memory on a 256-bit memory bus, offering 896 GB/s bandwidth. The card is reportedly designed with a total board power (TBP) of 300 W. The Ti variant appears to use the PG147-SKU60 board design with a GB203-300-A1 GPU. The standard RTX 5070 is positioned as a more power-efficient option, with specifications pointing to 6,144 CUDA cores and 12 GB of GDDR7 memory on a 192-bit bus, with 627 GB/s memory bandwidth. This model is expected to operate at a slightly lower 250 W TBP.
Interestingly, the non-Ti RTX 5070 card will be available in two board variants, PG146 and PG147, both utilizing the GB205-300-A1 GPU. While we don't know what the pricing structure looks like, we see that NVIDIA has chosen to make more considerable differentiating factors between its SKUs. The Ti variant not only gets an extra four GB of GDDR7 memory, but it also gets a whopping 45% increase in CUDA core count, going from 6,144 to 8,960 cores. While we wait for the CES to see the initial wave of GeForce RTX 50 series cards, the GeForce RTX 5070 and RTX 5070 Ti are expected to arrive later, possibly after RTX 5080 and RTX 5090 GPUs.
Sources:
@kopite7kimi #1, @kopite7kimi #2
Starting off with RTX 5070 Ti, it will feature 8,960 CUDA cores and come equipped with 16 GB GDDR7 memory on a 256-bit memory bus, offering 896 GB/s bandwidth. The card is reportedly designed with a total board power (TBP) of 300 W. The Ti variant appears to use the PG147-SKU60 board design with a GB203-300-A1 GPU. The standard RTX 5070 is positioned as a more power-efficient option, with specifications pointing to 6,144 CUDA cores and 12 GB of GDDR7 memory on a 192-bit bus, with 627 GB/s memory bandwidth. This model is expected to operate at a slightly lower 250 W TBP.
Interestingly, the non-Ti RTX 5070 card will be available in two board variants, PG146 and PG147, both utilizing the GB205-300-A1 GPU. While we don't know what the pricing structure looks like, we see that NVIDIA has chosen to make more considerable differentiating factors between its SKUs. The Ti variant not only gets an extra four GB of GDDR7 memory, but it also gets a whopping 45% increase in CUDA core count, going from 6,144 to 8,960 cores. While we wait for the CES to see the initial wave of GeForce RTX 50 series cards, the GeForce RTX 5070 and RTX 5070 Ti are expected to arrive later, possibly after RTX 5080 and RTX 5090 GPUs.
110 Comments on NVIDIA GeForce RTX 5070 and RTX 5070 Ti Final Specifications Seemingly Confirmed
In any case - meh. I find it hard to care. I play less and less games that are demanding these days. Playing the constant “keep up with upgrades so that the industry can release more poorly made slop” is a fools game. I’ll probably buy some cheap lower powered GPU next year just to get modern decode so that YouTube stops hammering my CPU with AV1 and then go play modded Morrowind or HoMM3 or whatever.
It looks like all large dies are eaten by
Has been for several decades now. Almost every Nvidia card is hard capped by VRAM.
What stings extra today is that its Nvidia itself pioneering all sorts of tech in their own GPUs that requires additional VRAM (and selling you that tech for a premium). Its a typical case of corporate strategy, create a demand where there wasn't one or shouldn't be one, own that segment, and cash in by giving consumers just slightly below what they want.
And yet still today we have idiots that keep telling themselves 'the card is out of juice anyway, no need for more GBs'. Talk about shooting yourself in the foot... and only to not have to admit that you bought a sub par product. Gosh, why isn't Nvidia giving the 5070 16GB now, I wonder...
Lolno. There’s almost a 50% performance delta between the 4070Ti Super and the 4090. 4090 is the definition of a halo card - it’s a monster, far more than the 3090 was. Nothing in these specs suggests a 40-50% generational uplift, especially not on the same node process. It’s THEORETICALLY possible, I guess, but that would even beat out Maxwell in how impressive that would be and, well, seeing the trends of last few gens and the fact that extracting more GPU performance seems to be harder with every generation… Yeah, I would say that’s not happening.
Then I also realized I could just watch AAA slop on Youtube for free and miss out pretty much nothing since the gameplay is all the same.
And people are asking why we call nVidia, nGreedia.
Proof. ;)
This is why we are in this situation.Blasphemy!
How dare you ignore new games with the magnificent RT and now PT!
/s
Same here, i’m playing old games and only buy when those old games drop low enough.Weird, I found a nice 7900xtx for 899 and 2 games worth 170 and its not a Ngreedia gpu.
AMD.. not even on the radar, again. I thought they would have had something decent.
Maybe after they get used to making AI stuff for awhile they will be back.
I believe Nvidia would charge 5060 for $299, and probably they will say, "It has memory compression with GDDR7, so buy it for $299. We just want you to save (the more you buy, the more you save)."
Just look at 1080Ti vs 2070Super, the 2070Super has fewer CUDA Cores but the die size is even bigger than 1080Ti and they perform roughly the same with old games (2070Super pull ahead on modern DX12 games)
5070Ti will be using GB103 die, if the GB102 were ~750mm2 then GB103 die can be ~600mm2
Funny how PS4's CPU side is a total joke, but the GCN "HD 7860" is still somewhat capable.
Is 2x8 pin on 5070 Ti.
It won’t happen,
But one can still dream…
I also enjoy science fiction, but no, that’s not happening. The GA102 was already essentially as big and dense as the process allowed. Hence the poor yields. And comparing 4090 to the 1080ti vs 2070 situation is silly - the 1080ti was absolutely not as far removed from the rest of the stack as the 4090 is. Not to mention that Turing and onwards the die sizes are not comparable to previous architectures - those chips didn’t have RT and Tensor cores. And that was also on a new node. None of those factors will be in play here. Can the 5070 Ti come close, like in a 10-15% range. Potentially. Being faster? Not happening. Again, we’re not getting 50% generational uplifts anymore.
tl:dr The 5070Ti will be 4080S level or a bit faster.
Anyway, 1080 Ti is the best card NV has ever made. Loved mine when I had one, still a dope card for 1080p gaming except for some new AAA titles.