Sunday, June 29th 2025
NVIDIA GeForce RTX 5070 Ti SUPER Planned with 24 GB of GDDR7 Memory
We have already covered that NVIDIA is planning a SUPER refresh of its GeForce RTX 50 series GPUs, with the most recent leak revealing the GeForce RTX 5070 SUPER, which is expected to feature a slight increase in CUDA core count and 18 GB of memory. However, NVIDIA is also overhauling its RTX 5070 Ti with the SUPER treatment, now enabling the RTX 5070 Ti SUPER to feature as much as 24 GB of GDDR7 memory with no change in CUDA core count. According to kopite7kimi, the RTX 5070 Ti SUPER will feature the same 8,960 CUDA core configuration as the GB203-350-A1 SKU, based on the PG147-SKU55 PCB. The only difference here is the VRAM surplus resulting from an additional 8 GB of GDDR7, which totals 24 GB capacity, compared to the 16 GB now on the regular RTX 5070 Ti SKU.
This also changes the TBP of the card, where the RTX 5070 Ti SUPER GPU is now rumored to feature a 350 W TBP, an upgrade from the non-SUPER's RTX 5070 Ti, which has a 300 W power rating. Perhaps the extra power is dedicated now not only to more VRAM (which is also a power consumer) but to higher clock speeds, too. It wouldn't be sufficient to call it a SUPER variant without some raw compute increase, so NVIDIA might be selecting better die bins for the SUPER SKU to enable higher default clock speeds at only a 50 W TBP increase. Nonetheless, we expect to hear more details in the coming weeks, especially since the wave of rumors has started to flood social media and leaks are becoming more frequent.
Source:
kopite7kimi
This also changes the TBP of the card, where the RTX 5070 Ti SUPER GPU is now rumored to feature a 350 W TBP, an upgrade from the non-SUPER's RTX 5070 Ti, which has a 300 W power rating. Perhaps the extra power is dedicated now not only to more VRAM (which is also a power consumer) but to higher clock speeds, too. It wouldn't be sufficient to call it a SUPER variant without some raw compute increase, so NVIDIA might be selecting better die bins for the SUPER SKU to enable higher default clock speeds at only a 50 W TBP increase. Nonetheless, we expect to hear more details in the coming weeks, especially since the wave of rumors has started to flood social media and leaks are becoming more frequent.


83 Comments on NVIDIA GeForce RTX 5070 Ti SUPER Planned with 24 GB of GDDR7 Memory
You were one of the defenders of the 5060 Ti 8GB.You said that generally that buffer is fine for that card.
Others were very critical of this stance.
I'm not going to take a side now, but by this logic if 8GB is enough for the 5060 Ti, wait let's make it less controversial, if I say that 8GB is enough for the 5060 instead (and you would agree) then what about the 2070 which also had an 8GB buffer?
There is a clear performance difference between these two, therefore one of them has more VRAM than needed (for gaming, in the most encountered scenarios) relative to its horsepower. Would you agree?
If you agree, and I use another example, let's say the 3090 and 4090, one of them has an oversized buffer relative to its horsepower. Which is it?So it seems that in fact we can identify with a reasonable degree of accuracy what sort of VRAM buffer a card should have. Otherwise how would the gaming community come to a consensus regarding the fact that the 5060 Ti 8GB has an undersized buffer?Okay the 1080 Ti had 11GB and 3060 12GB, if we agree that they had properly sized buffers, than a card with twice the performance should have twice the buffer, yes?
What about the 5090 then, relative to the 3090? Its twice as strong, it should have had a 48GB buffer? Meaning its current 32 is undersized? Also the 4090 with 3GB chips it could have 36GB. So if you argue that the 3090 has a properly sized buffer then that's the only way to make this make sense, the 4090 and 5090 both have less VRAM than they should have.Are they?
AI needs more VRAM, therefore the gamers pay extra for that VRAM even though generally the games don't need it.
I would say the AI people drew the long stick, they're getting gaming cards that are more than good enough for AI, they're far more affordable than Pro cards, the only drawback is they have smaller buffers than Pro cards, because, go figure, they're primarily gaming cards.
So now the AI people figured that if they advocate for more VRAM and get the gamers on that train, they have the chance to get the ideal products -> cheap (for them) gaming cards with good performance and adequate VRAM.
They're chasing their own interest, they're not doing gamers a favor.
/////
Again, I'm not advocating for less VRAM, I'm advocating for more horsepower for a particular sized buffer. Therefore I'm not saying cut some VRAM off the 5090, I'm saying make the core stronger. Unfortunately it is far easier and cheaper for a GPU maker to slap on more VRAM than it is to make more powerful cores. So I have to accept this reality, that everything is a compromise to a degree, some more than others.
Therefore the ideal product is very difficult to make, and it would only be ideal for a certain price bracket and for a limited amount of time.
For example:
A 5060 with 24GB of VRAM would not make the card any faster, it would ONLY provide for more data access LOCAL to the GPU. It can not and will not make the GPU itself any more powerful.
Contrariwise;
A 5090 with only 12GB of VRAM might limit certain aspects of performance IF the game/software did have proper memory access/scheduling routines(which devs most don't spend nearly enough time on). Such a card would be entirely doable with software optimized for it. That's the key point, optimized software.
Too much VRAM is a myth. Stop propagating it. Too little VRAM can be a problem with unoptimized code or for certain situations.
Just an FYI, a GTX980 4GB can do 4K gaming fine with it's 4GB of VRAM, within reasonable settings. Ask me how I know...
So to say a 5070 with 24GB of VRAM is too much VRAM is total and complete nonsense, just like saying a 5060 8GB is too little. It's all about the optimizations of the code, the settings and the system configuration.
The point is:
Should the people that bought the 5070 Ti be butthurt because they haven't waited for the Super?
Furthermore, should the people that will buy the 5070 Ti Super be butthurt because they haven't waited for the 6070 Ti (same buffer more compute)?But I know that since I've already expressed in one of my earlier posts on the previous page:However you cannot always trim very close to the ideal as you're limited by the design, therefore some cards end up with slightly oversized buffers while others undersized and then they get chips on the backside making the buffers slightly oversized and yeah... all that crap all over again.Doesn't that mean it's a waste to put that much on such a weak card? Rephrasing it results in that card having too much VRAM relative to its compute horsepower.But you have to stop somewhere, otherwise you're paying extra for what exactly? Again you will never be able to cover every scenario, and to just put very large amounts because it might help in a certain situation seems wasteful and dishonest.
I repeat, I'm not saying that cards should have less VRAM, I'm saying some cards should have more horsepower relative to their VRAM. I'm not the bad guy here (although that's my nickname, lolz) as I'm not trying to take anything away from anyone.
Which is more likely for gamers with older cards compared to new/current ones, to be bottlenecked by the VRAM amount or by the compute horsepower?Okay then what am I looking at here?
www.techpowerup.com/review/battlefield-6-performance-benchmark/4.html
www.techpowerup.com/review/silent-hill-f-performance-benchmark/4.html
www.techpowerup.com/review/dying-light-the-beast-performance-benchmark/4.html