NVIDIA RTX 4080 SUPER Sticks with AD103 Silicon, 16GB of 256-bit Memory

[gmn17]

You're not taking into account the GPU limitations. Of course, if you change all the specs, sure, it would work I guess. But for this specific references, it won't. Look at the 4060 TI. It doesn't benefit from the extra VRAM. As Effikan pointed out above, take the 4070 with only 12Gb as a comparison; the performance difference increases with the 4060 TI 16Gb as you raise the resolution.

Also an ad102 die with the 20gb card is a possibility

 

[Dahita]

Also an ad102 die with the 20gb card is a possibility

A 4095 with 30GB and an automatic coffeemaker integrated is also possible, if you are talking about adding 4th reference...

 

[gmn17]

A 4095 with 30GB and an automatic coffeemaker integrated is also possible, if you are talking about adding 4th reference...

This is from TPU GPU DB

 

[Dahita]

This is from TPU GPU DB

It's in the title: "NVIDIA RTX 4080 SUPER Sticks with AD103 Silicon". No Ti in sight.

 

[Macro Device]

Also an ad102 die with the 20gb card is a possibility

Technically yes but practically it would damage nVidia's profits, unless marketed correctly but they don't wanna spend much more time adjusting their marketing. That's why we won't see 20 GB Ada GPUs.

 

[gmn17]

Technically yes but practically it would damage nVidia's profits, unless marketed correctly but they don't wanna spend much more time adjusting their marketing. That's why we won't see 20 GB Ada GPUs.

I agree it's in dreamland for now.

 

[f0ssile]

I did not. It's incorrect to compare raw VRAM bandwidth of GPUs of different architectures. Both GPUs handle this resource differently and I can't say which GPU does this job better. When I don't know what is right I decide not to spread fallacy but rather keep myself silent.

Because I don't know how they work and how Ada VS RDNA3 fares in this department.

Yes, I'm biased towards balance and value. RX 7800 XT lacks the former, yet offers the latter. 4070 offers both. You're talking like I am a huge fan of naming what otherwise would be a 4060 Ti at best, a 4070 and at 600 USD before taxes, granting "godlike" 12 GB. No, I'm not, this GPU is an awful rip off.

And 7800 XT is a very great GPU in isolation. Fast, furious, great overclocker, loads of VRAM for its $, but... it's not enough because 4070 is just a smarter product. Just like a 7700 XT is a better item than a 4060 Ti. Just like buying a 6800 non-XT is a no-brainer at <400 USD if obtainable brand new.

Why do I need to? One GPU dies of insufficient VRAM, another dies of insufficient RT performance and no one can assure you the latter won't be more severe. Just check it out: https://www.techpowerup.com/review/avatar-fop-performance-benchmark/5.html

I don't say things will go exactly the way of increased RT loads in newer games. What I say is this is more than possible. That's a gamble but at this win probability, I'd place my bet on a 4070.

But did you re-read what I quoted?
It seemed like the 7800 XT was hopeless...
the pros of him were hidden by the cons, including stuff that has nothing to do with gaming...
...if you only use CUDA, get the 4070 even if it's half as good, right?
The 6900 XT sucks even more, it is the epitome of crap.

Before, bandwidth mattered more than quantity (they are different things, but you only look for diversity when you want it)...
Then both count for nothing, the important thing is to have CUDA and 50 watts less consumption.
Who knows if at the time of the 3080 consumption was so important, here the past would become important...

I bet that the average 12% more RT will be important, no fortune tellers needed there, eh...
I repeat that the 4070 is worth it as long as it doesn't saturate the VRAM...
So we could get to the 4070 which is less successful due to its own fault, think what a prediction.

The test is nice and coincidentally very favorable to Nvidia, in which you can also see what happens at 4K.
In the future it could also happen under 4K, but what can I tell you...
In your opinion, why does the 4070 drop below 4K from its clear advantage at 1440p?
Is it because of the CUDAs who are mischievous at 4K, or maybe the quantity and the gang are taking part in the dance?

Interesting to digress on differences in VRAM and/or normal map occupancy. Would you like to get to 16GB from 12? It's worse than I thought...
But you make a 5-year prediction of 4070 with 12GB more than enough, there you go too far, eh...
Hasn't history already said that barely enough VRAM won't last 5 years?
Do textures and normal maps become sub-optimal when the 4070 doesn't arrive?
But I bet you will play with low textures just to activate the RT, you know what a luxury.

Reread my post, it's still good, in fact, it's better than before...

PS: Guess which GPU I'm looking at the screen from.
PS2: In the meantime, I saw an even more interesting guy...

 

[efikkan]

But you make a 5-year prediction of 4070 with 12GB more than enough, there you go too far, eh...
Hasn't history already said that barely enough VRAM won't last 5 years?

May I ask, where is the evidence that RTX 4070 has "barely enough" VRAM?
If anything, the evidence points to it running out of steam long before VRAM, just look at the scaling in 4K vs. 1440p, even in RT, RTX 4070 and 4070 Ti continues to scale pretty much the same. If 4070 Ti continues to scale with only 12 GB, there is no reason to think the little brother wouldn't.

I've participated in discussions about extra VRAM for "future proofing" for probably close to 15 years, and if the past predicts the future, then "future proofing" with more VRAM is probably going to be irrelevant in >95% of cases. Yet we often see the claim repeated, even some video reviewers keep repeating the anecdote about extra VRAM. But you will only find some fringe edge-cases where it really matters, and the explanation is simple; new more advanced graphics tends to be more taxing on computational performance first, then bandwidth secondary, even before VRAM capacity becomes an issue. There are also the physical limitations; like with e.g. 504 GB/s, and a target FPS of >60, the game wouldn't have enough time to access more than 8.4 GB total in theory (in practice it's much less than this), so having 16+GB of VRAM without lost of more bandwidth and good resource management isn't necessarily going to be truly accessible. We've heard this complaint now with pretty much every Nvidia generation going back to the 400-series or before, but I've yet to notice any upper mid-range or high-end GPUs that turned out to be bad performers long-term due to lack of VRAM.

And do you know which team is known for dropping driver support first? Team red. Long term driver support is actually important if you're looking to get the most mileage out of the purchase, and AMD usually drops their normal driver support before 5 years. And don't trick yourself into thinking a card with much more (sometimes slower) VRAM is going to enable you to play the most demanding games 5+ years from now at the higher settings (assuming games continues to advance).

So the bigger question is really, what are you giving up to get some more VRAM? (generally speaking)
Are you picking an alternative which is either more expensive, sacrificing some performance or features, or complaining just to find something to criticize?

Do textures and normal maps become sub-optimal when the 4070 doesn't arrive?
But I bet you will play with low textures just to activate the RT, you know what a luxury.

Normal maps, along with most other types of maps (incl. reflectiveness etc. used with RT) are often easily compressed, and most buffers (stencil buffers, etc.) are even moreso.
Whenever we are talking about memory compression in GPUs, we're not talking about the files stored on your PC, but rather the lossless and transparent compression done by the memory controller of the GPU.

In addition, modern GPUs are incredibly advanced when it comes to resource management compared to back when I first learned OpenGL and DirectX in the early 2000s. Now with dynamic tiled rendering, clever scheduling, memory compression and caches, they can achieve much more throughput than the specs alone will lead you to believe.
As I always say, the truth is revealed in good benchmarking, not "good specs".

 

[f0ssile]

You're mixing too many things, they all seem related to you, except the ugly duckling amount of VRAM.
You are trying to weigh the value of the quantity, weighing it partially as you can do with RAM between 16GB and 32GB, is inappropriate to say the least.
Would you make the same argument as though in that case? If not, why?

Although 5 minutes are not suitable for weighing differences in quantity, whether the VRAM should saturate immediately is your guess to take.
I feel the Deja-Vu of the conversation with the other: The facts are his, the other has to explain the origin of the universe to demonstrate it...

In 25 years have you never seen future proof limits on the amount of VRAM? Are we in the same universe as above?
I, on the other hand, remember for more than 20 years conversations calibrated on the quantity that is useless without power, as if more resolute textures had to weigh on the GPU.
Coincidentally, the Nvidia VRAM seems optimal for you too, on closer inspection also for the 3080, so much so that it also rectifies the other user's strange statement.

Allocation that is always useless is another constant, as if it cannot predict an optimized use of resources.
We seem to hear about the uselessness of the cache in general, even on Win etc...

As with RAM, quantity has very little to do with frequency/bandwidth.
When you need the quantity if you don't arrive you choke. It's ridiculous just to think about mixing it up like it's nothing.
Normal maps were an example of the other that takes up space, also related to textures.
What does compressibility mean? If complexity grows, employment grows, the concept seemed simple to me...

Lossless color compression is another thing that has little to do with it, yet another excuse to digress.
With 8GB of VRAM there are cases of heavy stuttering at 1080p. Are you saying that you have to become a fortune teller to predict the end of 4070?

I remember well the 770 2GB, which I think was medium-high range, or am I wrong?
From a future proof perspective, didn't the 4GB one last just a hair longer?
No, what am I making up, it wasn't like the 2GB mid-range 960, the 4GB one can still be seen in the minimums...

Wow, we missed the re-recycled history of AMD drivers, we needed the astral peak of stereotypes.
Besides, you didn't last long before changing sides, first it was CUDA, now the drivers...

I thought I saw fresh drivers for the old r480, which a friend of mine still uses effectively:

https://www.amd.com/en/support/graphics/radeon-400-series/radeon-rx-400-series/radeon-rx-480

Nothing, AMD won't even get there in 5 years, eh, there you did the quadratic mean in an isolated case, I imagine.

Who knows if it will still be able to raise the quality of the textures at 1080p?
Is that future proof, or inconvenient past because it's not convenient? In the aftermath...
It should also serve to make it clear that the various optimizations cannot squeeze the memory too much. But 8GB and 12GB are always current, eh...

There's no point in digressing about the rest, you seem to muddy the water to make it seem deep (Cit. 1)
The same old story, the eternal return of the same (Cit. 2).
You think I don't know what you're talking about, but if I know, let's change the subject, it's better...

Ah, perhaps I had asked too much by asking you to guess...

Ask me whenever you want, maybe you are capable of thinking that I will do it outside the home.
I've always had both AMD and Nvidia (often at the same time), I don't think I'm the one who sees substance badly.

 

[gffermari]

NVIDIA RTX 4080 SUPER reportedly costs $999, RTX 4070 Ti/4070 SUPER at $799/$599 - VideoCardz.com

 

[theouto]

If the 4070 super is replacing the current 4070 in its price bracket as it says there, and they change the pricing of the 4070 accordingly, maybe this launch won't be so worthless after all!