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.
132 Comments on NVIDIA GeForce RTX 5070 and RTX 5070 Ti Final Specifications Seemingly Confirmed
How does that make sense to anything I said?
It's the complete opposite. The patch isn't made by any company but people so that the users of other gpus below 12gb of vram is able to enable and use path tracing. It's that game developers (possibly with aid of nvidia) disabling it for anything below 12gb of vram.
Also proves it can work with less vram.
People act like you HAVE to twist your arm into the meat grinder because you HAVE to upgrade and buy the new shiny bing bing wahoo. That consumerist mindset is why the average american has over $7k in CC debt.It's not the first. Geforce 8000 series got 10 years. A decade has been pretty typical for Nvidia for awhile.
And frankly, who was still using Fermi on anything even remotely modern in 2019? Or Kepler in 2021? Even with the end of bugfixes, youtubers showed in 20/21 that fermis till worked in newer games, they were unplayably slow, but they still worked. it took another couple before the cards just flat out didnt work. Of course, they still worked fine for anything 2018 and earlier. Same with kepler.
I had 3 7900xtx and each one had issues so I'm not going AMD even if they had a product to suit me but Nvidia is looking equally bad for different reasons. It's a suck ass time to be a PC enthusiast.
Its also logically the best deal for Nvidia because it keeps their previous halo product 'quite halo ish' by letting the 5080 land just under it. They're not cannibalizing their own stack; but they do EOL it rapidly, creating persistent value for past gen's 4090 - but killing the value of anything under it and making the 5080, while just a meagre upgrade gen-to-gen, a wanted card for being a cheaper 4090.
1080 Ti was 35% faster than 1080, meanwhile 4090 beats 4080 by 31%.
2070 Super has 30% less CUDA Core and 10% less Bandwidth than 1080Ti, yet still match 1080 Ti in older DX11 games
TSMC 12nm (2070S) was only a tweaked node vs TSMC 16nm (1080 Ti), TSMC introduce them as same node just like their N5/N4nm nodes.
Nvidia has been very consistent in the past decade in making 104 die perform just as well as previous gen 102 die, but now we will see 103 die (GB103) carry that torch.
In some games there is a big difference in image quality on 4060 8GB and 16GB, or when compared with a 12GB equipped competitor. I have not seen too many reviewers point it out lately.
I couldn´t care less that e.g. the RTX 4060 has 10% higher framerate than the Arc B580 in some game, when image quality is worse. How long before we see the same thing happening in 12 vs 16GB VRAM scenarios....
LOD hacks and other tricks like it, is not fair. Players and benchers gets banned, rankings deleted and accounts disabled. Not to mention all the personal attacks and hate mail after numerous videos are made about them, calling them all sorts of stuff. Deliberate use of unfair advantages is cheating.
I'm also guessing they'd at least hit 10-15% IPC though but that's super hard to even measure with gpus outside of synthetic workloads. My guess is Pathtracing will be the best measuring stick as far as gaming goes.
At the end of the day raw performance without upscaling or frame generation is all that matters.
As much as do hate Nvidia, I have to give them credit to do that (still RIP to Fermi and Kepler, you were forgotten too early).Oh boy I miss them E6400 days...
With GTX 1070 / RX 5700 and give or take similar GPUs being that cheap I don't see that as an issue. Let it go.
Blackwell is supposed to make this graph even bigger, it'll be fun to update it once the products are available.I had also thought about this, but there is no GB204 chip so far, (04 were where the previous x70 models usually used), so seems like it just got downgraded, but the segmentation from the 70ti and 70 is not news.