NVIDIA has released a set of benchmarks that provide an early idea of how its GeForce RTX 4080 graphics cards might differ from one another. According to the metrics, which originate from yesterday’s article confirming this month’s DLSS 3 games, the GeForce RTX 4080 (12 GB) could be up to 30% slower than the GeForce RTX 4080 (16 GB). This is suggested by the metrics that NVIDIA shared for A Plague Tale: Requiem, which show the 16 GB model being ~21% faster than the 12 GB model with DLSS 3 enabled, and ~30% faster with DLSS 3 disabled.
Here’s what the performance of the GeForce RTX 4080 (12 GB) looks like relative to other models in the GeForce RTX 30 and 40 Series, according to a chart from VideoCardz:
From a VideoCardz report:
Here’s a reminder that RTX 4080 16GB features AD103 GPU with 9728 CUDA cores while RTX 4080 is equipped with AD104 processor with 7680 cores. That’s 26% more cores on the 16GB variant. However, major changes can also be seen in memory bus width (256-bit vs. 192-bit) and TDP range (320W vs. 285W).
It also looks like RTX 4080 12GB won’t be able to outperform RTX 3090 Ti in benchmarks, but it at least it is 10-20% faster than original RTX 3080. Both models are to launch in November, but NVIDIA is yet to confirm the exact date.
The NVIDIA GeForce RTX 4080 (16 GB) and GeForce RTX 4080 (12 GB) start at $1,199.99 and $899, respectively. Critics have argued that the 12 GB model is actually a rebranded GeForce RTX 4070.
Discussion (19 replies)
Join Discussion →Ok so the graph is ... interesting.
I'm going to go out on a limb and say the lesser 4080 roughly matches the current 3090 (non-Ti).
I can find a 3090 right now for less than the $899 MSRP of the 4080 Lesser Edition, and I can get that right now. Sure, you don't get DLSS 3 (or will you?), and I bet the Ada will probably still raytrace a bit better, but you do get double the VRAM...
Trying to decide how I feel about that. On one hand, it's nice that the performance has jumped as it has. On the other... the prices jumped right along with it... so I don't feel like we really gained anything on the price/performance curve on this generation - just performance.
And sure, that doesn't sound like such a bad tradeoff, especially if you agree that Moore's Law is dead and we should be seeing the prices go up; that price vs performance is a meaningless metric. But I would say - what about the mid and low tier cards that just didn't exist last generation. If the new 4070 starts out at $899, and that used to be a $350 tier relative to performance in a given generation; have we just priced out a large part of our already niche marketplace?
If you want to stream or if that enters a value proposition for you full av1 encoders are nice on the Nvidia cards.
Yeah, but you can get one of those on an Arc for ... considerably less, and just use that as a second card just for compression. If you trust Intel drivers to work, that is.
If I were to dare (well, I use to Heli-skiing and bungee jump IRL before the family ?) I would say that Zotac is the next EVGA-like forced out. ?
Might be cheaper to get an AMD AM5 CPU... I'd say wait for 13th-gen, but that isn't going to get new IGP tech. Maybe Intel will have it for 14th-gen.
Pretty much coming to this conclusion as well. I do think that Intel's transcoding drivers work, at least if they work as well as they do on the IGPs.
I was actually looking into buying a 3090 just now, but decided I don't really need it. For $870 I'd probably change my mind but it still costs significantly more here in Europe.
Yep, I posted about this one in another thread on the 1st day of prime early access. Yesterday I even saw a Zotac Amp Extreme 3090 Ti for $999.
The lesser model really should have been a 4070.
It's moronic of them to muddy the waters with these stupid confusing model sames.
But then they'd be selling a x070-series card for US$899, so... :)
They have shown that they have absolutely no shame when it comes to pricing, so I don't see why that is a problem :p
I disagree, the lesser model is a 4060ti and the top one is a 4070 in disguise.
Seems they left a lot of room for better 4080's, as to be able to react to RDNA3 or something.
4070 in disguise, well that's what you are...
A rasterization pie with your brand new card
Reflections processors but not as many as you should
4070 in disguise, and overpriced...
Another part of that is:
I don't know that nVidia wanted to position a x70 as roughly equivalent to the previous generation x90 (or what was formerly known as Titan).
It was a big jump in performance, and nVidia wants the naming to reflect the relative pricing, not performance inside the generation. Unfortunately, that means the price/performance ratio delta between generations we are used to seeing just got thrown out the window. My fear as that the lower end of the spectrum has already been entirely eliminated from the RTX spectrum, and if they are shifting all the naming up, we may see that middle tier completely dissolve as well.
Yea... I didn't pounce on a 4090 but I don't expect availability to be as tight as covid years. Plus I want to wait to see the competition though with the leap... yea I don't have a bunch of high hopes. I just know if I get a 4090 that will necessitate a 4k monitor to get within performance specs.
They are going to rename the 12GB variant https://www.nvidia.com/en-us/geforce/news/12gb-4080-unlaunch/