Support NeoGAF

[T4keD0wN]

Can someone tl dr me without DLSS and AI bullshit,
what is raw performance improvement:
4080 to 5080 ?

Or better yet. Someone calculate 3080 vs 5070ti vs 5080.

Look at Far Cry 6: green is 5080, grey 4080. Its with RT and the RT cores are better (different amounts further complicate it) so the true raw performance improvement should be even lower than this, so its barely an improvement in that regard, its all about AI bullshit now.

4080 is about 149% of a 3080. 4080 to 5080 is a smaller improvement. ~35% with RT and less than 30% without RT probably, so 5080 should be slightly less than double of 3080, but without vram possibly holding it back, with RT you should be looking at 200%+ going from 3080 to 5080.

Worth noting that Far Cry 6 ray tracing is very light so it should not play a big role here, theres also a possibility of a CPU bottleneck since Dunia Engine does not have the best multithreading. My pessimistic estimate for raster 3080 to 5080 (if i presume better RT cores affect the chart) is about ~1.90x, optimistic ~2.00x if they dont.

 

[rofif]

There are other techniques in effect to lower input lag and they are the reason why Nvidia frame gen works well. If there wasn’t it would be pretty worthless.

Wrong. Thats what reflex is for and now with reflex 2 it will be even lower

if you start with 30fps, the input lag is 30fps + something.
Reflex only delivers the frames as they are ready without double or triple buffering + some reduction.

 

[proandrad]

if you start with 30fps, the input lag is 30fps + something.
Reflex only delivers the frames as they are ready without double or triple buffering + some reduction.

Im not an expert, but that sounds right. Frame gen is best used if you are already well over 60fps. Nvidia’s video isn’t just using frame gen to go from 20fps to 200. It’s using dlss to upscale and then using frame gen. Dlss upscaling probably gets it to above 60fps before the frame gen.

 

[Fess]

I said the same thing about STALKER 2 and Cyberpunk recently and people gaslit me saying it was a memory leak vs my card tapping out at 4K, it's why I was hoping AMD was going to make due on their promises

In Indiana Jones they show how much vram that is in use and also give suggestions on which settings to lower. So no reason to gaslit there. But to be in a lowering settings scenario on a new graphics card isn’t optimal, so I wonder if they have some other tech in place to sort it out. The only card here with more than 16GB is the 5090. Board reviewers should test the cards on Indiana Jones.

 

[LavitzSlambert]

Well, yes, and so what? Should I forget that $2 chip in my 1080p pre-OLED TV could inflate frames just fine?


The increased lag is not a side effect due to inferior tech, it's an inherent part of frame inflation.
There is no future in which it starts to make sense, bar filthy marketing purposes.

I'm agreeing with you on both points, just chipping in my two cents. I don't think all of these advancements can co-exist and still be a good experience for gaming, when the input lag cost is so high.

 

[Rocco Schiavone]

 

[dcx4610]

Didn't expect those prices. Good on Nvidia. I mean they aren't CHEAP but they are better than I thought they'd be.

Good luck finding a 5070...

I hope it's not a repeat of the 4000 series where you can't even buy the card before the 6000 series releases

 

[asdasdasdbb]

Didn't expect those prices. Good on Nvidia. I mean they aren't CHEAP but they are better than I thought they'd be.

There's a reason for that.

 

[scydrex]

Didn't expect those prices. Good on Nvidia. I mean they aren't CHEAP but they are better than I thought they'd be.

I think the rumors of higher prices was on purpose.

 

[analog_future]

Man, I really hope I can nab the 5090 FE. Love the aesthetics of it:

Realistically I will take whatever I can get though.

 

[Astray]

These prices (and DLSS enhancements for past-gen cards) are an absolute killshot to AMD, even for the (upper) midrange cards they're supposedly focusing on, these new cards will not only have to compete with new-gen Nvidia cards, but also with the old cards that just got new life.

If I'm Intel, I'm absolutely salivating at the prospect of fighting it out with AMD, especially given how much weakness they signaled in the lead up to this CES. They are absolutely there for the taking when it comes to GPUs.

Initial hype gone, new hype in 2 years. Probably hang on to my 4090 and get a 6090. I think.

Yeah I think that I don't really regret my 4090 purchase, especially given that I got it on a good discount.

 

[dcx4610]

I have a 2070 so I'm definitely down for a 5070. I wonder if it's worth the extra cash for the Ti to get 16GB of memory though...

 

[asdasdasdbb]

I think the rumors of higher prices was on purpose.

Disagree. I think NV messed up so bad that even for them they realized they had to cut prices. And then ripped off Lossless Scaling to double down on Fake Frames as a marketing tool.

 

[Celcius]

Man, I really hope I can nab the 5090 FE. Love the aesthetics of it:

Realistically I will take whatever I can get though.

We need an RTX 5090 wallpaper for https://www.nvidia.com/en-us/geforce/community/wallpapers/
If only https://gamingbolt.com/wp-content/uploads/2025/01/geforce-rtx-5090.jpg were available in 4K

edit: 4K is here: https://www.filemail.com/d/iredskxbriixhrt

 

[poppabk]

I have a 2070 so I'm definitely down for a 5070. I wonder if it's worth the extra cash for the Ti to get 16GB of memory though...

That's the goal. It's always worth the extra cash for the next model up.

 

[Puscifer]

One would be truly lucky to get a 5080 for under $1000.

What a bunch of BS.

Give it a couple months, I got my 4080 2 months after launch directly from Nvidia

 

[R3TRO]

I wonder how many aio models will be available at launch.

 

[Joelenheimer]

Bought a 4080 super not to long ago

Ffffffffffffffffffff

I’m fine
I’m fine

 

[Big Al]

We need an RTX 5090 wallpaper for https://www.nvidia.com/en-us/geforce/community/wallpapers/
If only https://gamingbolt.com/wp-content/uploads/2025/01/geforce-rtx-5090.jpg were available in 4K

Real 4K or fake AI special sauce 4K?

 

[TheRedRiders]

I know people are complaining about the jump in raw shader performance and rasterisation, but just going back to what Mark Cerny said last month and what many others have been saying for a while now is that going forward in the future we'll likely see more advanced use of silicon space rather than just increasing the shader and core count. As Nvidia is doing here, and as they have been doing steadily now we'll see a bigger focus on machine learning and ray-tracing, a major area will be ML rendered geometry and textures.

Personally I welcome it.

 

[asdasdasdbb]

Bought a 4080 super not to long ago

Eh, unless you love Fake Frames or paid way more than what people end up paying for the 5080, it's not that big of a deal. The 5080 is going to be nowhere near the 4090.

 

[ThatOneGrunt]

Eh. Great prices, but I think I'm going to stick with my 3070ti for a little longer. Not a fan of how they're flexing performance by using AI.

 

[Delt31]

Catching up on all of this. Here is my only question. Where can I find the difference in results of REAL FPS? I see these crazy high FPS based on AI/inserted/fake frames. What about the real raw performance of 4090 vs 5090. Any intel on that yet?

 

[Lethal01]

Great to see Nvidia fixing the worst part of online games, the other players.

 

[dreamstation]

Bought a 4080 super not to long ago

Ffffffffffffffffffff

I’m fine
I’m fine

Realistically how much better are the new cards over a 4080 super anyway? I wouldn't be worried mate

 

[FingerBang]

Catching up on all of this. Here is my only question. Where can I find the difference in results of REAL FPS? I see these crazy high FPS based on AI/inserted/fake frames. What about the real raw performance of 4090 vs 5090. Any intel on that yet?

You cannot. We have not been given that information. All the comparisons were done using DLSS4 + RT + Frame Generation. The estimate seems to be around 25 to 35%, though I'd expect it to be bigger for the 5090.

 

[Rockman33]

You have to spend 2k to get anything more than 16gb vram and everyone is happy/okay with that?

 

[HeisenbergFX4]

You have to spend 2k to get anything more than 16gb vram and everyone is happy/okay with that?

By far the vast majority are not ok with it

 

[Unknown Soldier]

By far the vast majority are not ok with it

Sales of the product will not reflect this sentiment

 

[Shane89]

yeah, let the developers release even more unoptimized shitty games, AI will do the rest.

 

[OverHeat]

You have to spend 2k to get anything more than 16gb vram and everyone is happy/okay with that?

I’m not happy but I’m weak hahahaah

 

[Mortisfacio]

By far the vast majority are not ok with it

I think most aren't impressed with it, but most are accepting this as our reality and buying anyways. Short term, given the new pipeline reduces VRAM utilization, I expect to see the VRAM bottleneck be alleviated. We'll just revisit this issue again with more and more games a few years down the road.

 

[JORMBO]

Not really worth upgrading my 4070ti S. But that is what I expected. I have been doing every other cycle.

 

[Topher]

I’m not happy but I’m weak hahahaah

OverHeat being "not happy" about spending $2k on a GPU

 

[HeisenbergFX4]

Sales of the product will not reflect this sentiment

I think most aren't impressed with it, but most are accepting this as our reality and buying anyways. Short term, given the new pipeline reduces VRAM utilization, I expect to see the VRAM bottleneck be alleviated. We'll just revisit this issue again with more and more games a few years down the road.

The vast majority will not spend the $2k for a GPU

 

[Unknown Soldier]

The vast majority will not spend the $2k for a GPU

The vast majority don't matter, and the $2k GPU will sell out anyways because supplies will be limited

 

[carsar]

Good luck finding a 5070...

5070 would be just 4070 super +x4 FG and still 12gb. Why anybody should be excited to buy such gpu?

 

[thuGG_pl]

Yeah i think people forget just how big of a jump the 4090 was.
I actually wouldnt be surprised if the 4090 beats the 5080 in some games.

It will beat it in all games when not using frame gen.

 

[HeisenbergFX4]

The vast majority don't matter, and the $2k GPU will sell out anyways because supplies will be limited

I was just answering his question asking if everyone was ok with spending 2k for above 16gb of vram

The 5090 will 100% sell out though

 

[FIREKNIGHT2029]

Oof! Wake me up when we get to the 6090. 28fps at native 4K isn’t the jump I expect for a $2000 GPU. 5090 will be amazing with DLSS but I expect more of raw performance increase between card generations. Hopefully I see that real increase when we get to the 6090 and 7090 which is when I’ll be ready to upgrade my 4090.

 

[keefged4]

Bought a 4080 super not to long ago

Ffffffffffffffffffff

I’m fine
I’m fine

You'll be fine. 30fps framegend' to fake 120fps is still 30fps in latency.

 

[StereoVsn]

I was just answering his question asking if everyone was ok with spending 2k for above 16gb of vram

The 5090 will 100% sell out though

Thing is I can see a lot of folks (relatively speaking) grabbing 5090 as an AI workload card. 32GB isn’t too shabby and $2K is a lot better vs $6-8K.

Edit: Not talking about MS putting these in Azure, but for AI devs and data scientists. We have been talking to some of our engineers and our procurement is discussing things with a distributor.

You will see quite a lot of that. So a lot of those sales might go outside of gaming.

 

[Zathalus]

if you start with 30fps, the input lag is 30fps + something.
Reflex only delivers the frames as they are ready without double or triple buffering + some reduction.

Don’t start with 30fps, even Nvidia recommends against it. Base fps (with upscaling but not frame generation) should be higher than 40fps, ideally higher than 50fps. With Reflex 2 latency will be under a regular 60fps game and frames would be whatever multiplier you select.

 

[HeisenbergFX4]

Oof! Wake me up when we get to the 6090. 28fps at native 4K isn’t the jump I expect for a $2000 GPU. 5090 will be amazing with DLSS but I expect more of raw performance increase between card generations. Hopefully I see that real increase when we get to the 6090 and 7090 which is when I’ll be ready to upgrade my 4090.

28 FPS standing still and no combat

 

[thuGG_pl]

Catching up on all of this. Here is my only question. Where can I find the difference in results of REAL FPS? I see these crazy high FPS based on AI/inserted/fake frames. What about the real raw performance of 4090 vs 5090. Any intel on that yet?

Nowhere, you need to wait for reviews, probably at the end of the month.

 

[Puscifer]

Bought a 4080 super not to long ago

Ffffffffffffffffffff

I’m fine
I’m fine

Wait till raster results come. I'm not even sure I wanna get a 5070ti anyone because their results are from DLSS performance, they didn't even bother with quality let alone balanced. Wait till the dust settles

 

[Zathalus]

For those expecting a larger performance uplift for the 90 series… how exactly? It’s GDDR7 with a 512bit bus and a 100 billion transistor GPU on TSMC 4nm. All that and a 575w TDP.

Silicon only has so much to give at this point, don’t expect massive leaps to the 6090 either. Not unless something fundamentally changes.

 

[HeisenbergFX4]

Best Buy starting to show some prebuilds that are coming

 

[Rocco Schiavone]

 

[iQuasarLV]

Catching up on all of this. Here is my only question. Where can I find the difference in results of REAL FPS? I see these crazy high FPS based on AI/inserted/fake frames. What about the real raw performance of 4090 vs 5090. Any intel on that yet?

Look at the DF x2/x3/x4 comparison. The x2 latency hit is based off what the initial frame rate was. The starting frame rate in Cyberpunk 2077 was 20fps based on the x2 latency hit. if you were to take the x4 hit its running a 17fps latency. Bleck!

FPS to MS converter