Armenius
Extremely [H]
- Joined
- Jan 28, 2014
- Messages
- 53,307
Considering the game runs at 10-15 FPS with full path tracing and no scaling on a 4090 I'd say 30 FPS is a huge improvement. That is double your performance right there.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
Closed.
- Thread starter Blackstone
- Start date
- Status
Considering the game runs at 10-15 FPS with full path tracing and no scaling on a 4090 I'd say 30 FPS is a huge improvement. That is double your performance right there.
- Joined
- Jul 5, 2012
- Messages
- 4,612
For sure! But still not fast enough to do it at native 4K, well not with what I'd consider to be playable fps at least.
- Joined
- Apr 10, 2005
- Messages
- 17,145
I wonder how many of those cards are in containers on ships rn.. whewww..
deceptXS
Limp Gawd
- Joined
- Aug 8, 2024
- Messages
- 164
It runs at around 18-22 FPS on a 4090 if I recall.
I think the 5090 will be around 40% faster than the 4090.
- Joined
- Apr 10, 2005
- Messages
- 17,145
Well… looks like a decent upgrade cycle overall.. can’t wait for the benches.
StoleMyOwnCar
2[H]4U
- Joined
- Sep 30, 2013
- Messages
- 3,472
Is this supposedly 2x Stable Diffusion speed? Yeah might be interested, then. But not sure if it's really worth it tbh. Gonna wait for the benchmarks. if somehow the two Microcenters near me run out of stock entirely before/if I end up ready to buy, then that's just how the cards fell. *Shrug*
Overall, the benchmarks are kind of unimpressive otherwise. Or rather they make me feel a bit conflicted. The speed improvements this time look like they mostly rest on frame gen and dlss. Otherwise it's... okay? But otherwise exactly what I said, if the 4k native FPS is so low to begin with, is even a 2x increase worth much of anything? Oh well. Looks like that's where Nvidia's money mostly went this time. Instead of brute force, they worked on the AI side of things. Different future of gaming from what most people probably wanted.
- Joined
- Jun 9, 2003
- Messages
- 7,023
For $2000 I want to see raw performance numbers and comparisons at 4K minimum, not DLSS and certainly not DLSS frame generation. That's an extra feature but when you're asking for something that is minimum $400+ over the equivalent card last generation, I want raw performance comparisons not "Oh its 2X frames when using our magical interpolation at 1080P and frame insertion" for most of it. I know it will come in time but the idea that all of our benchmarking should be using DLSS or whatever the hell or worse, that we should expect games to be virtually unplayable with all the bits and bobs turned on, even with a $2000 card, unless upscaling is enabled is frustrating. As far as I'm concerned, in the slides posted above the only game that shows something useful to deciding abouit the card is FarCry 6, and the generativeAI raw performance bit if that is of interest.
- Joined
- Dec 2, 2004
- Messages
- 32,715
Huh? 5080 is $999. Not 1199.
patience I have old man brain at night, it's edited.
StoleMyOwnCar
2[H]4U
- Joined
- Sep 30, 2013
- Messages
- 3,472
That's the thing, and I don't think you're going to like it: I think that a majority of their focus this time (and probably going forward) will be AI featureset, not brute force. In other words they're saying "screw it, we can't keep up with game visual fidelity demands at all just by raw hardware needs. We're just going to increase AI tech till it gets to the point where it's perfect enough." And that might just be where the future of gaming is going (or rather, has been going for quite a while).
Although it's kind of a weird idea, isn't it? Developers keep making higher and higher fidelity games with higher and higher fidelity assets, only for those assets to get reinterpreted and re-imagined (loosely speaking) via the AI engines on modern GPUs, just so that the game can actually be playable at some level of visual fidelity within that region. At some point it will come down to exactly what I said a long time ago, the middleman will be cut out and games and their textures can be generated entirely via prompts. Soon we'll have completely procedurally generated experiences. Anyway that's a tangent and a scary thought.
It also explains why just raw RT growth numbers look like crap, but the generative AI performance is like 2x. That's basically where all of their R&D went. Nvidia is an AI first company, but instead of abandoning gaming, they simply figured out how to make AI and gaming align better.
Last edited:
AMD 45% improvement supposedly in RT, Nvidia 100%. Hmmm, AMD will be further behind in RT, lol. Have to wait for real benches in the end.
Now did I get that right from Nvidia, showing 28fps and then over 100fps using 4x FG? If so , those will be the most laggy 100fps ever. WTH? FPS does not always indicate gaming experience.
Price looks good if real, do like Nvidia design. If it cools well and does not explode due to power and size, it appears to be a winner. AI engineering design cooler?
- Joined
- Jul 5, 2012
- Messages
- 4,612
No they are using super resolution plus FG. So the base frame rate will be much higher than with DLSS completely off. Performance mode at 4K means 1080p internal resolution so the base frame rate is over 60 to start before frame generation.
I actually like standard frame gen with a controller. Once I use a mouse and things are moving quickly I don't like the way it looks. I'm not sure i have a monitor fast enough to bother with 4x frame gen. I also don't have high hopes for it looking great in motion. The rest of the new DLSS features seem great.
Vengance_01
Supreme [H]ardness
- Joined
- Dec 23, 2001
- Messages
- 7,702
man if we could have had 16gb on the 5070 gg AMD. I will skip this gen and wait. 4070 super at 1440p should be good for another 1-2 years
- Joined
- Jul 5, 2012
- Messages
- 4,612
The 5090 is a 575W GPU though. So with the FE cooler only being 2 slots it's either not going to be very quiet under load, or it's going to run on the hot side. But who knows maybe Nvidia has worked some black magic to make that 2 slot cooler work well with 575W! I absolutely love the FE cooler on my 4090 but it's quite a big larger and the 4090 is a 450W GPU not 575W so that ensures it runs cool and quiet. I just have a hard time believing they can achieve the same noise levels and temps while shrinking down to a 2 slot cooler.
- Joined
- Apr 10, 2005
- Messages
- 17,145
I feel like the 5090 will be a good candidate for an AIO water cooling situation…
It's strange, maybe 575w is total BS for gaming scenarios, or they don't want the FE to be as valuable and sought after as it was previously and are happy to let it run hot and loud.
That would save them a few $ as well I guess, the 4090's cooler was after all overkill (functionally speaking) since designed for 600w. For the end user it meant quiet operation and plenty of headroom for temperatures though
Last edited:
I am currently using a 1000W PSU, hopefully that will be enough. I have a brand new EVGA 1300W one sitting here from a return I planned on selling, now I might hold onto it for a while.
Edit- Will my Vetroo 1000W PS be adequate? I just noticed it has PCI-E 5.0 16pin (12+4) and not the new standard.
Any idea when the aftermarket cards will become available? I still don't understand why it's so difficult to find a model with two HDMI ports. If the only cards with two HDMI are a lot more expensive I'll just get Nvidia's card.
Edit- Will my Vetroo 1000W PS be adequate? I just noticed it has PCI-E 5.0 16pin (12+4) and not the new standard.
Any idea when the aftermarket cards will become available? I still don't understand why it's so difficult to find a model with two HDMI ports. If the only cards with two HDMI are a lot more expensive I'll just get Nvidia's card.
Last edited:
deceptXS
Limp Gawd
- Joined
- Aug 8, 2024
- Messages
- 164
The Flux.dev performance is like the only selling point for me, so this better be true. But to answer your question, yes this performance should be nearly the same increase in SD since they share a lot in common and seeing how the 5080 receives the same uplift (2x) as the 5090 means is not just down to bandwidth. Flux uses ~4x the VRAM of SD and will use a 4090s entire VRAM pool, so I was surprised when the 5080 also received a 2x uplift with 16GB. So to me these graphs are kind of trash for gamers, but looks very positive for us who are do generative AI. If this translates to a 2x uplift in training performance that would be amazing for me, can't imagine training a checkpoint in 1.5 days instead of 3 days, or a LoRA in 1.5 hours instead of 3 hours.
Armenius
Extremely [H]
- Joined
- Jan 28, 2014
- Messages
- 53,307
That is probably what Reflex 2 is for since the first thing I thought when they said predicting 3 frames ahead was that the input latency is going to be massive. Reflex isn't some magical bullet, either, as input can definitely feel weird in some games, especially with Boost turned on. I don't think AI no matter how good it is can ever accurately predict user input.
Power will depend on what kind of CPU you're using. 1000W probably isn't enough with an Intel CPU and it will probably be pushing it with a standard AMD CPU. An AMD X3D CPU will probably do fine. I posted earlier that my PC in sig pulls 700-750W from the wall while gaming. That would theoretically go up to 825-875W with a 5090.
The power standard is still the same, it's just the cable that changed. The sense pins on the new cable are shorter and the conducting terminals are longer. You can probably contact the manufacturer of your PSU for a 12V-2x6 cable, but a 12VHPWR cable will still work.
HDMI has royalty fees attached to every product. The more HDMI outputs on a product, the more the manufacturer has to pay.
Armenius
Extremely [H]
- Joined
- Jan 28, 2014
- Messages
- 53,307
Only reviews I care about are The FPS Review and TechPowerUp.
Armenius
Extremely [H]
- Joined
- Jan 28, 2014
- Messages
- 53,307
NVIDIA recommends a 1000W power supply for the 5090.
LittleBuddy
Gawd
- Joined
- Jan 3, 2023
- Messages
- 525
Other than the AI performance, the most interesting thing is Reflex 2 IMO. I have no interest in the framegen 4x mode. But if we can do 2x mode with Reflex 2 the latency should be minimal.
- Joined
- Apr 25, 2001
- Messages
- 15,664
My $2K is ready to leave my wallet. Can't wait to get my 5090.
LittleBuddy
Gawd
- Joined
- Jan 3, 2023
- Messages
- 525
Flux.dev being a 2x increase means SD will also see a 2x increase, the only difference really is Flux uses Flow Matching, while SD uses LDM. I'm actually kind of surprised it is only 2x improvement with the AI tops jump from 4090 (~1300) to a 5090 (3400) it should be a 2.5x improvement, which might be seen in SD since SD isn't bandwidth limited, but Flux.dev is.
Edit:
They used FP4 on the 5090 and FP8 on the 4090. This Flux.dev benchmark is useless. Actually the entire benchmark slides released are useless. The 5090 might not even be 40% faster than the 4090 in games, could be closer to 20-30%.
Edit:
They used FP4 on the 5090 and FP8 on the 4090. This Flux.dev benchmark is useless. Actually the entire benchmark slides released are useless. The 5090 might not even be 40% faster than the 4090 in games, could be closer to 20-30%.
Last edited:
LittleBuddy
Gawd
- Joined
- Jan 3, 2023
- Messages
- 525
I have a feeling the 5070 might be 10-15% faster than the 4070 at most. I don't think the upgrade is worth it, definitely wait for real benchmarks before buying anything.
leSLIe
Fully [H]
- Joined
- Oct 18, 2004
- Messages
- 17,347
Guess its time to upgrade from the old and trusty Trident 9680 
LittleBuddy
Gawd
- Joined
- Jan 3, 2023
- Messages
- 525
Looking at available information through the slides and specs...
5090: ~35% faster than 4090 (raster) | possibly 50% faster in RT
5080: ~25% faster than 4080 (raster) | possibly 40% faster in RT
5070: ~15% faster than 4070 (raster) | possibly 20% faster in RT
Nvidia is hiding everything behind DLSS tech differences, no example was a 1:1 comparison other than maybe FC6. I think real benchmarks will expose this.
I hope I'm wrong, but not very optimistic looking at the shenanigans in the slides. On paper the 5090 should be around 35%-40% faster than the 4090, but the 5090 has a lower boost clock as well to account for.
During the presentation Jensen Huang says the 5090 is double the performance of the 4090, I think he considers DLSS Framegen "performance" otherwise his claims don't align with the specifications and slides shown. He's a very good salesman, but I'm seeing right through his nonsense. It's not even going to be 2x faster in AI. Hopefully I'm not killing anyone's hype here. But if my thoughts are wrong on this someone correct me.
5090: ~35% faster than 4090 (raster) | possibly 50% faster in RT
5080: ~25% faster than 4080 (raster) | possibly 40% faster in RT
5070: ~15% faster than 4070 (raster) | possibly 20% faster in RT
Nvidia is hiding everything behind DLSS tech differences, no example was a 1:1 comparison other than maybe FC6. I think real benchmarks will expose this.
I hope I'm wrong, but not very optimistic looking at the shenanigans in the slides. On paper the 5090 should be around 35%-40% faster than the 4090, but the 5090 has a lower boost clock as well to account for.
During the presentation Jensen Huang says the 5090 is double the performance of the 4090, I think he considers DLSS Framegen "performance" otherwise his claims don't align with the specifications and slides shown. He's a very good salesman, but I'm seeing right through his nonsense. It's not even going to be 2x faster in AI. Hopefully I'm not killing anyone's hype here. But if my thoughts are wrong on this someone correct me.
- Joined
- Jul 5, 2012
- Messages
- 4,612
I've always expected that if you just count raw raster without DLSS then the 5090 wouldn't be some crazy 70% faster than a 4090 as we've pretty much hit a wall with traditional rendering as proven by the fact that pure raster games today look almost the same as pure raster games from 10 freaking years ago. This is a hot take from me but I'm all onboard with new rendering techniques purely relying on AI if the end result that my eyeballs are going to see is better looking games running at higher frame rates than ever before even though tons of people are calling this out as "gimmicks". I personally love DLSS and use both SR and FG in all my games and seeing the image quality continue to improve in this area has me stoked. If you're a DLSS hater then the 50 series (and probably every single future GPU release from now on) isn't for you. But I've come to embrace DLSS so if all the future image quality improvements come from AI/machine learning then I'm totally ok with that. I personally feel like trying to do nothing else but brute force rasterization performance at this point is just living in the past and won't get us better looking games than that what we've had since 2013 with Crysis 3 but again this is just my hot take and I'm sure all the DLSS haters are going to be triggered by this lol.
Completely
- Joined
- Mar 19, 2003
- Messages
- 11,071
For 4K 240 it doesn’t matter whether card is 20% or 50% faster as it needs to be faster. Once I hit 4K 240 in games I play, maybe I can think about taking a breather from top tier. Until that point I will use all tricks to try to hit that goal and 5090 offers flexibility to use all types of tricks to get there. I am also excited about multi frame DLSS as long as it has a quality setting that doesn’t suck.
- Joined
- Apr 10, 2005
- Messages
- 17,145
yeah seems like a downgrade from the previous iterations.. vents all of the heat into the PC case now..
- Joined
- Jul 5, 2012
- Messages
- 4,612
Definitely aiming to get one of those chonky AIB cards this time around. Gonna need a big boy to handle that 575 watts!
https://www.moddiy.com/products/ATX...2V%2d2X6-675W-12VHPWR-16-Pin-Power-Cable.html
I think this is the cable I will need for my Vetroo GV1000 PSU. It comes with one 16pin ATX 3.0 cable.
I think this is the cable I will need for my Vetroo GV1000 PSU. It comes with one 16pin ATX 3.0 cable.
- Joined
- Apr 10, 2005
- Messages
- 17,145
the cooler on my MSI seems effective and is very quiet... it's not even the vapor chamber version. I'll probably go for another MSI variant again.
https://www.msi.com/Graphics-Card/GeForce-RTX-5090-32G-SUPRIM-SOC
edit: holy crap.. look at this big boy!
https://www.msi.com/Graphics-Card/GeForce-RTX-5090-32G-VANGUARD-SOC-LAUNCH-EDITION
It's got the chonk!
LOL
Yup… how dare we even suggest that price will be important to decision making.
- Status