digitalwanderer
Legend
Do you need a 5000 series to use it?
NVIDIA discussion [2025]
- Thread starter SlmDnk
- Start date
you only need a 50 series for multi frame generation. transformer super resolution, ray reconstruction, and dlaa works on all rtx cards
My understanding is that @ILikeFeet is correct. Only MFA and maybe neural materials are locked to Blackwell. Not sure about neural materials.
Last edited:
What res monitor you got? I've found it to be way, way more effective at 1440p than on my old 1080p monitor. But even on the 1080p monitor DLAA (native) looked good.
Yea you'd benefit hugely from DLSS, even Quality mode. Though I don't know what the pixel density is like on a huge superwide monitor like that. Seems it's more useful with higher pixel densities.
There's zero life necessity for a halo video card; in fact there's zero life necessity for ANY video card. If prices get too high, people will stop buying them. And do you know what you call a company who makes things too expensive and literally nobody buys their stuff? You call them bankrupt.
NVIDIA is not a charity, nor is AMD, or Intel, or Asus or Gigabyte or ASRock or MSI or... These are for-profit enterprises, and their sole purpose in life is to make money. Here's the cool part: they only make money if people actually buy their stuff! If people stop buying their stuff, then they're all (as a collective) smart enough to figure out either the price or the product is wrong, and sometimes even both. Here's the antithetical version: if it becomes unprofitable to sustain the uppermost echelon of PC performance, then that segment will simply cease to exist.
Yes, I too would like a 5090 to be the same price as a 1080 Ti was eight years ago. What's cool is, 12 billion transistors and 11GB of RAM (the constituent components of a 1080 Ti) are notably cheaper now; I suspect the RTX 5070 is likely to be at least 50% more transistors than the 1080Ti, with another gigabyte of radically faster RAM, running on less power, and costing $150 less (a 22% discount) to boot! Whether you like it or not, whether you realize it or not, you're getting your savings. It just turns out the high end is insanely higher than it ever was, and those sorts of things mean customers are willing to pay more to achieve it.
DegustatoR
Legend
And "RTX Hair" since that requires new geometry primitive which only Blackwell RT cores support.
GTX1080TI used a cut down GP102 with 471mm^2. The 5070 TI for $749 has a cut down GB203 with ~410mm^2. There is not so much difference, especially that nVidia can deliver higher bandwidth with a 256bit memory interface.
NVIDIA has had a fascination with hair for a long time. It's a long standing problem in games, but they're overthinking it. This could easily be solved by making all the characters bald. RTX Bald could run on anything.
DegustatoR
Legend
It is a bit different this time though, and I don't get why they even call it "RTX Hair" at all as it's just a new acceleration primitive for ray tracing. One would think that games may have other geometry suitable to be represented by spheres aside from hair and calling it this way kinda diminish what the feature actually is and capable of.
Yet I didn't even mention the 5070 Ti, did I?
The 1080Ti was $699 MSRP, the 5070 is $549 MSRP. The latter comes with way more than double the transistors, one gigabyte more memory, more than a 50% increase in memory bandwidth, more than double the FP performance (depending on which FP format we're talking about), about 30% more pixel and texel fillrate, while using roughly the same power and a slightly smaller footprint. Best of all for the folks decrying ever-inflating prices, the 5070 comes at a 22% discount compared to the 1080Ti despite spanking it in every measurable way.
Last edited:
Hehe, you know, I wasn't even gonna go there, but since you mentioned it... https://www.usinflationcalculator.com/ says the inflation-adjusted price for a $699 video card in 2017 comes out to near-exactly $899 today. I'm assuming US dollars because, in most ways, the US came out better than most in the big global inflation situation we've had over the past few years. As such, this is a very conservative estimate compared to what our friends and colleagues outside the US have to deal with.
By the math above, the 5070 MSRP is within rounding error of a 40% discount from the 1080 Ti MSRP.
"But cards are so expensive now!" /s
DuckThor Evil
Legend
I fail to see what is the point of comparing the 1080ti from 2017 to the 5070 again? 5070 would be even more awesome if it is compared to a GTX 280...
That said, the 5000 series is looking pretty good to me, although personally the 3GB modules are needed here. 5080 with 24GB or 5070 with 18GB would make those much more appealing to me. Currently I'm happy with my 4070 Super, but we'll see.
That said, the 5000 series is looking pretty good to me, although personally the 3GB modules are needed here. 5080 with 24GB or 5070 with 18GB would make those much more appealing to me. Currently I'm happy with my 4070 Super, but we'll see.
The argument was somehow cost is going up without checks or balances. As it turns out, costs are going down in comparable hardware, however the absolute ceiling on performance is an order of magnitude higher now. I also chose the 1080 Ti purposefully as it is continually regarded as the last of the "best performance per dollar" cards, yet somehow it underperforms and is overpriced by modern equipment. Once again proving costs have continued to reduce in both absolute and relative terms, yet the halo series still costs more not because it's necessary, but because people are willing and able to pay for it.
It really shouldn't be surprising to anyone, but somehow people still miss it.
But did we though? There's another conversation somewhere on this forum about how high AA and high resolution and high framerates really never happened in the top AAA titles because there still wasn't enough horsepower then, either. It always ends up being a circular argument (as you well know and alluded to in your reply) because one person will have a story about how this one game could be maxxed out and they remember it fully and blah blah, but then another person remembers this other game couldn't be maxxed out... But then we devolve into whether "max" graphical options mean anything, and then who would even notice, inevitably by the same people with strong opinions on how "fake frames" have no place in high quality graphics output.
My statement remains: we are getting more and better hardware for our dollar, even though the absolute maximum prices continue to go up. Why do they continue to go up? Because no amount of money RIGHT NOW could ever buy you the performance you're about to get from a 5090. Does that mean everyone needs a 5090? Does that mean ANYONE needs a 5090? Nope. But they're gonna make a ton of them, and they're gonna sell them all, because people will want them and will be willing to pay for them.
Last edited:
DuckThor Evil
Legend
I understand and personally I'm at peace with these companies pricing their products, but Imo it's a fact that nVidia has become much more "efficient" in pricing their lineup and consumers has been on the losing side of this tug of war. GTX 570 was 94% of the biggest die enabled, 5070 is something else. 1080Ti in 2017 was also an epic performance monster and nobody was worrying about it having enough memory for games. Back then nVidia left money on the table, these days they are a bit better at not doing it. But yeah the 5000 looks reasonable to me, I just want to see those 3GB modules.
Similar threads
