|
DLSS 3.0 appears to be exclusive to 4000-series (and future, presumably) cards. The only mention of Turing and Ampere in the new DLSS page on their site is:quote:Meanwhile, gamers on all GeForce RTX graphics cards can continue to enjoy NVIDIA DLSS Super Resolution, which massively boosts performance in 216 games and apps by using AI to output higher resolution frames from a lower resolution input. I don't see companies bothering with both DLSS 2.0 and 3.0 profiles. efb
|
#
?
Sep 20, 2022 16:49
|
|
|
|
| # ? May 28, 2024 13:42 |
|
|
Omnicarus posted:1080Ti, you've gotten me through the years so far please don't give up the ghost anytime soon. Same. It's been doing really well
|
|
#
?
Sep 20, 2022 16:50
|
|
|
curried lamb of God posted:For someone who hasn't kept up with graphics cards in ages, does Nvidia typically announce the lower tier cards (xx50/60/70) at a later date? I get that they would release later than the flagships, but I'm surprised that they didn't even merit a mention today the RTX 3060 [non-Ti] was launched on Jan 12, 2021, some months after the RTX 3080, which launched on Sep 1, 2020
|
|
#
?
Sep 20, 2022 16:50
|
|
|
Agreed posted:My inclination is to say I would not buy that poo poo at gunpoint, gently caress DLSS 3.0 then Why are you mad that they made a new product? It doesn’t make your current GPU worse and you don’t gotta buy it dude.
|
|
#
?
Sep 20, 2022 16:50
|
|
|
Bloodplay it again posted:I don't see companies bothering with both DLSS 2.0 and 3.0 profiles. i would assume DLSS 3.0 integration is a superset of 2.0 (they'll still be doing the TAAU upscale in addition to frame interpolation) so 3.0 integrations should be able to gracefully fall back to 2.0
|
|
#
?
Sep 20, 2022 16:51
|
|
|
I expect the 4090s will get discounted fast
|
|
#
?
Sep 20, 2022 16:51
|
|
|
wibble posted:When is the remote control car game coming out!
|
|
#
?
Sep 20, 2022 16:52
|
|
|
Omnicarus posted:1080Ti, you've gotten me through the years so far please don't give up the ghost anytime soon.  With how they usually price their poo poo for Europe, a real 4080 is going to be $1500. gently caress off Jensen.
|
|
#
?
Sep 20, 2022 16:52
|
|
|
orcane posted:LMAO big The 4090 is 1676 or so pounds so yeah the 4080 is gonna be stupid for Europe
|
|
#
?
Sep 20, 2022 16:53
|
|
|
Cygni posted:Why are you mad that they made a new product? It doesn’t make your current GPU worse and you don’t gotta buy it dude. I don't see how a reply like this helps to do anything but rile up the person you're replying to more. Anyway, see you 50 series.
|
|
#
?
Sep 20, 2022 16:55
|
|
|
I just noticed that they're charging $900 for a GPU with a 192-bit bus. Wowza. That is a new low.
|
|
#
?
Sep 20, 2022 16:55
|
|
|
repiv posted:i would assume DLSS 3.0 integration is a superset of 2.0 (they'll still be doing the TAAU upscale in addition to frame interpolation) so 3.0 integrations should be able to gracefully fall back to 2.0 That's a fair point. Kinda like how if you're watching media with a dolby vision profile on a display without dv support, it can roll back to the HDR10 profile. I hadn't really considered that. I was thinking that 3.0 support would look for X hardware and since Ampere and older won't have it, it wouldn't even show up as an option.
|
|
#
?
Sep 20, 2022 16:56
|
|
|
Twerk from Home posted:I just noticed that they're charging $900 for a GPU with a 192-bit bus. Is there much difference between a 10G 320-bit bus and a 16G 256-bit bus for the 3080 v 4080
|
|
#
?
Sep 20, 2022 16:57
|
|
|
Bloodplay it again posted:That's a fair point. Kinda like how if you're watching media with a dolby vision profile on a display without dv support, it can roll back to the HDR10 profile. I hadn't really considered that. I was thinking that 3.0 support would look for X hardware and since Ampere and older won't have it, it wouldn't even show up as an option. i'll check the SDK when they release it, the sensible thing would be for the DLSS3 runtime to work on turing/ampere just without the new interpolation feature but we'll see
|
|
#
?
Sep 20, 2022 16:57
|
|
|
ijyt posted:I don't see how a reply like this helps to do anything but rile up the person you're replying to more. For real, I don't know why dude cares so much that I am not in favor of this news. Why single me out anyway Cygni, you're spoiled for choice at people surprised by the overall announcement. Rumors had the 4080 MSRP at $699-850 as recently as six days ago, I guess that was purely wishful thinking. To me the segmentation of the two 4080s reads more like 10GB 3080 vs 16GB 3090 from previous generation, with the top end 24GB card being more analogous to the 3090 TI.
|
|
#
?
Sep 20, 2022 16:58
|
|
|
So what's the smart upgrade look like these days for someone still using a 1070? I've got a 1440p monitor that I'd like to use with a new card. Before crypto poo poo everything up I'd been targeting a 3070, but was never willing to pay scalper prices. edit: budget sub-$400, ideally, $450 if it's the difference between garbage and something good. I'm just not going to spend anything like what the 80 and 90 cards cost.
|
|
#
?
Sep 20, 2022 16:58
|
|
|
Is that $900 for the neutered 4080 or the good one?
|
|
#
?
Sep 20, 2022 16:59
|
|
|
KillHour posted:Is that $900 for the neutered 4080 or the good one? It's for the 12GB, so not the good one
|
|
#
?
Sep 20, 2022 17:00
|
|
|
KillHour posted:Is that $900 for the neutered 4080 or the good one? The former, 4080 starts there
|
|
#
?
Sep 20, 2022 17:00
|
|
|
Cyrano4747 posted:So what's the smart upgrade look like these days for someone still using a 1070? Get a 30 series card while they’re on fire sale.
|
|
#
?
Sep 20, 2022 17:00
|
|
|
What the actual gently caress
|
|
#
?
Sep 20, 2022 17:00
|
|
|
i hate how little attention nvenc is getting at these events. is it just gonna be av1 with otherwise similar performance/quality for h264 ...Cyrano4747 posted:So what's the smart upgrade look like these days for someone still using a 1070? at the very least, wait and see what amd show us november 3.
|
|
#
?
Sep 20, 2022 17:01
|
|
|
Agreed posted:For real, I don't know why dude cares so much that I am not in favor of this news. Why single me out anyway Cygni, you're spoiled for choice at people surprised by the overall announcement. I replied to you once with two sentences, relax. Someone expresses similar sentiment most launches, and I just wanted to know if there was something specific that was making you upset about it.
|
|
#
?
Sep 20, 2022 17:04
|
|
|
I'm catching up on this poo poo. I guess I'll be waiting until Summer 2023 at least. Because like I'm going to let NVidia gently caress me in the rear end with another memory bump down the line, like with the 3080 10GB. Also, there's nothing special on the schedule. Starfield maybe, but I'm expecting this to be delayed another time, beyond "first half" of 2023.
|
|
#
?
Sep 20, 2022 17:05
|
|
|
If I were buying today I'd aim for a 12GB 3080 I think.
|
|
#
?
Sep 20, 2022 17:06
|
|
|
Cyrano4747 posted:So what's the smart upgrade look like these days for someone still using a 1070?
|
|
#
?
Sep 20, 2022 17:08
|
|
|
So it’s still DisplayPort 1.4a after 80Gbps usb4 2.0 was announced.
|
|
#
?
Sep 20, 2022 17:08
|
|
|
ijyt posted:Is there much difference between a 10G 320-bit bus and a 16G 256-bit bus for the 3080 v 4080 Big memory buses are really expensive, they are hard to design and use a ton of silicon space. GPU makers will do everything they can to avoid putting bigger buses on GPUs, like using GDDR6X which is also expensive and power hungry. Total memory bandwidth is basically bus size * memory clock speed, so if you use faster memory you can use a smaller bus. You'll see that in every single GPU generation, usually price goes down significantly as you get to the smaller buses, whether they're 160b, 128b, or whatever size. GPU makers are always looking for a way to get away with a smaller bus. Nvidia started using some type of compression with the 9xx generation, and AMD is now doing "Infinity Cache" which is basically a gigantic cache that means you don't have to go to memory as often. The last huge bus GPU that I remember was the 512-bit R9 290/290X, which was a monster that lasted a really long time and performed well for ages.
|
|
#
?
Sep 20, 2022 17:08
|
|
|
Calm down ffs. Existing 30 series and even below are well prepared to handle current titles, pushed to the capability of current display tech. You're probably a whale or high level enthusiast if you are eying high refresh 4k anyway and who cares if you drop a couple grand on your hobby? What if you were into restoring cars or some poo poo? Not to mention those earning money back through streaming and such. You're not obligated to buy. There are plenty of options out there for every budget
|
|
#
?
Sep 20, 2022 17:08
|
|
|
100$ increase over the 3090 to the 4090 isn't too scary... oof gonna be over 1700$ after tax though Jensen is a smooth salesman gotta give him that.
|
|
#
?
Sep 20, 2022 17:10
|
|
|
Skip down to the feature list, and  The columns are 40 series / 30 series / 20 series. So no listed change at all between Lovelace/Ampere with respect to HDMI or Displayport feature support. Did we know that already?
|
|
#
?
Sep 20, 2022 17:11
|
|
|
codo27 posted:Calm down ffs. Existing 30 series and even below are well prepared to handle current titles, pushed to the capability of current display tech. You're probably a whale or high level enthusiast if you are eying high refresh 4k anyway and who cares if you drop a couple grand on your hobby? What if you were into restoring cars or some poo poo? Not to mention those earning money back through streaming and such.
|
|
#
?
Sep 20, 2022 17:12
|
|
|
codo27 posted:You're not obligated to buy. There are plenty of options out there for every budget Are there? What's the $200-300 nvidia option?
|
|
#
?
Sep 20, 2022 17:12
|
|
|
I'm trying to chill tf out here and think. Ok: 3080 10GB had 82% as many CUDA cores as the 3090 16GB. 4080 12GB has 78% as many CUDA cores as the 4080 16GB. I think what's going on here is they're rejiggering the product tiers somewhat, although in doing so I feel they're being kinda miserly with VRAM. There was not an analogous "holy poo poo that's a fuckload more cores" product as the 4090, where the product below it has only 59% as many CUDA cores. The only step above the 3080 10GB at Ampere's launch was the 3090 and it had 18% more, not 41% more. That's a lot closer to the difference between the two 4080s. While $900 is fifty bucks higher than the high estimates in rumors, that's less egregious than my initial reaction - 3090 launch MSRP was $1500, so getting that analogous product (that is, the 4080 16GB) is actually lower MSRP than before. Though, you'd think there would be a VRAM upgrade generation to generation. And the MSRP for the big daddy, while high, at least comes along with a likely very large increase in processing power with that massive CUDA count. Cygni posted:I replied to you once with two sentences, relax. Someone expresses similar sentiment most launches, and I just wanted to know if there was something specific that was making you upset about it. My brother I am high strung, and I was hoping to pay $800 for a 16GB card, that's the whole story Agreed fucked around with this message at 17:22 on Sep 20, 2022 |
|
#
?
Sep 20, 2022 17:13
|
|
|
https://www.nvidia.com/en-us/geforce/news/portal-with-rtx-ray-tracing/ some clarity on what RTX Remix actually is - it's a fork of DXVK which intercepts DX8 and DX9 calls, and redirects them to a fancy modern renderer, with hooks to let you replace assets, add lights, etc so only old games need apply but it should be widely applicable authoring a remix requires an RTX GPU, but a packaged remix mod should work on any card that supports vulkan raytracing repiv fucked around with this message at 17:19 on Sep 20, 2022 |
|
#
?
Sep 20, 2022 17:15
|
|
|
the 4080 branding is dumb as gently caress. the 12gb version has less compute cores, too. it’s a 4070 not a 4080
|
|
#
?
Sep 20, 2022 17:18
|
|
|
Are DLSS and its competitors pretty kick-rear end for increasing resolution and framerate (consistency) for VR? Or is the latency penalty enough to make it not worthwhile?
|
|
#
?
Sep 20, 2022 17:18
|
|
|
pmchem posted:the 4080 branding is dumb as gently caress. the 12gb version has less compute cores, too. it’s a 4070 not a 4080 It's the 1060 3gb and 6gb all over again.
|
|
#
?
Sep 20, 2022 17:19
|
|
|
codo27 posted:Calm down ffs. Existing 30 series and even below are well prepared to handle current titles, pushed to the capability of current display tech. You're probably a whale or high level enthusiast if you are eying high refresh 4k anyway and who cares if you drop a couple grand on your hobby? What if you were into restoring cars or some poo poo? Not to mention those earning money back through streaming and such. People can still mock the whack rear end poo poo though.
|
|
#
?
Sep 20, 2022 17:19
|
|
|
|
| # ? May 28, 2024 13:42 |
|
|
repiv posted:https://www.nvidia.com/en-us/geforce/news/portal-with-rtx-ray-tracing/ Oh man, time to spit shine Dishonored
|
|
#
?
Sep 20, 2022 17:20
|
|
