Quote from: LL on Yesterday at 23:23:16You should not use TDP values.

Because using them for analysing exhibits fake progress sold by Nvidia to us as real progress?

Quote from: KS9 on Today at 03:01:12I doubt this. I have a 9800X3D and 4090 and hit 60-70 fps with RT, Ultra at 4K in Cyberpunk. Unless it's being implied that the 5090 will be worse than the 4090.

Have the same configuration. It's 26 - 27 fps pure, 60 - 70 using DLSS Quality preset and FG plus it goes up to 110 fps when looking to the ground or in less detailed areas.

5090 will do 28 fps pure, around 110 fps using DLSS Quality and 280 or whatever fps using MFG x4.

I doubt this. I have a 9800X3D and 4090 and hit 60-70 fps with RT, Ultra at 4K in Cyberpunk. Unless it's being implied that the 5090 will be worse than the 4090.

Quote from: RobertJasiek on Yesterday at 17:13:22

Quote from: LL on Yesterday at 16:04:52Score that just appeared in Blender Open data, note that this is was achieved with Blender 3.6 an old version. Note the more recent 4.3.

NVIDIA GeForce RTX 5090   17822.17        1
NVIDIA GeForce RTX 4090 13063.74   1450 (number of benchmarks)

So allegedly,
RTX 5090 = >136% at 575W = 127,77%
RTX 4090 = 100% at 450W = 100%

Hence, relativ speed normalised per watt:
RTX 5090 = 107%
RTX 4090 = 100%

Even the 5090 is a scam! (Except for its cooling.)

You should not use TDP values.

NVIDIA itself has admitted it is no longer a consumer based company. So anyone buying a non consumer based product for consumer based applications is enabling them to develop and take out the competition across the board.

12vhpwr was an improperly manufactured plug. It was revised to fix the problems not created by costumers then they blamed the customers and were allowed to get away with because the costumers didn't want to believe the fact that the company that was once the best gpu manufacturer is no longer even manufacturing a product for them. So stop living in the past. Presently NVIDIA isn't even making a good product line. 5090 30% better than 4090. 5080 15% better than the 4080. Lower teir cards are all 20% better. So that literally means they purposefully didn't make the 5080 25% better than the 4080. That's 10% less of performance than the customer should have gotten for the same money.

Quote from: Tkun on Yesterday at 16:43:306 years since NVIDIA first introduced ray-tracing hardware onto the market and people still don't understand how intensive path-tracing is on hardware. The fact that it's able to get close to 30fps...

Thing is, they sell you on something for 6 years and up till now it is not really ray tracing. And i don't think it's saving the devs all the promised time by not having to worry about baked in lighting at all. So it's still not really ready for all consumers.

I remember ray tracing that shiny newton's cradle above a checkerboard floor back in the day, and it used to take hours to render one frame at low resolution. So fully aware of where we have got to. But still not quite convinced!

*keep track of both

The fact is that Moore's Law is slowing down and new nodes are getting insanely expensive and it's just not possible to make GPUs with huge raw performance gains each generation anymore and still keep the price even somewhat reasonable.

Are AI solutions perfect? No. But DLSS is apparently getting really good. Frame gen has the latency issue but technologies like Reflex 2 will help with that somewhat.

Going forward, we'll need to keep of both frame rate with and without frame-gen. For some games where latency is more of a concern, people will need to turn down settings to keep non-frame-gen framerate high enough for responsiveness. In titles like third person games where responsiveness doesn't need to be as good, then people can afford to turn up settings more and still enjoy the smoothness of frame-gen, assuming DLSS 4 is as good as I've heard.

Nvidia shouldn't try to sell us the idea that AI generated frames are equivalent to traditionally rendered ones, but DLSS 4 looks to be a game changing technology nonetheless.

Quote from: TruthIsThere on Yesterday at 15:12:41Four-generations later, heck, there's even a RTX2080Ti CP2077 so-called special edition, and NVIDIA still fails at a stable performance level for this title with everything at max setting without the fake frames.

Blackwell is a sad joke - it cost ridiculously more, much more power hungry, ect. and that's without any users' feedback for this thing yet.

Or maybe the engine is a joke and a huge performance bottleneck and that is why CDOR left it for Unreal Engine 5 and you have no idea what you're talking about. I'd bet on that more than anything.

Quote from: Tkun on Yesterday at 16:43:306 years since NVIDIA first introduced ray-tracing hardware onto the market and people still don't understand how intensive path-tracing is on hardware. The fact that it's able to get close to 30fps is an insane technical feat. Kids are spoiled. Try getting close to that on any AMD GPU at 4K with path-tracing enabled and no upscaling or frame generation. You're getting single digit FPS numbers. It doesn't have to do with "lack of optimization" or hardware being trash, path tracing on a MODERN GAME is just absurdly intensive. If you don't realize this then you're too uninformed to take part in discussion about it.

Exactly this. Jensen himself even said this is impossible without DLSS. They didn't try to hide anything. You just simply cannot calculate path tracing in real time at any playable frame rate. They invented a way to make it passable and beautiful and the mouth breathers still yell "optimize!"

Quote from: Tkun on Yesterday at 16:43:306 years since NVIDIA first introduced ray-tracing hardware onto the market and people still don't understand how intensive path-tracing is on hardware. The fact that it's able to get close to 30fps is an insane technical feat. Kids are spoiled. Try getting close to that on any AMD GPU at 4K with path-tracing enabled and no upscaling or frame generation. You're getting single digit FPS numbers. It doesn't have to do with "lack of optimization" or hardware being trash, path tracing on a MODERN GAME is just absurdly intensive. If you don't realize this then you're too uninformed to take part in discussion about it.

100 percent agreed. The general public and this whole youtuber generation is clueless and spoiled.

Quote from: LL on Yesterday at 16:04:52Score that just appeared in Blender Open data, note that this is was achieved with Blender 3.6 an old version. Note the more recent 4.3.

NVIDIA GeForce RTX 5090   17822.17        1
NVIDIA GeForce RTX 4090 13063.74   1450 (number of benchmarks)

So allegedly,
RTX 5090 = >136% at 575W = 127,77%
RTX 4090 = 100% at 450W = 100%

Hence, relativ speed normalised per watt:
RTX 5090 = 107%
RTX 4090 = 100%

Even the 5090 is a scam! (Except for its cooling.)

6 years since NVIDIA first introduced ray-tracing hardware onto the market and people still don't understand how intensive path-tracing is on hardware. The fact that it's able to get close to 30fps is an insane technical feat. Kids are spoiled. Try getting close to that on any AMD GPU at 4K with path-tracing enabled and no upscaling or frame generation. You're getting single digit FPS numbers. It doesn't have to do with "lack of optimization" or hardware being trash, path tracing on a MODERN GAME is just absurdly intensive. If you don't realize this then you're too uninformed to take part in discussion about it.

Score that just appeared in Blender Open data, note that this is was achieved with Blender 3.6 an old version. Note the more recent 4.3.

NVIDIA GeForce RTX 5090   17822.17        1
NVIDIA GeForce RTX 4090 13063.74   1450 (number of benchmarks)

Four-generations later, heck, there's even a RTX2080Ti CP2077 so-called special edition, and NVIDIA still fails at a stable performance level for this title with everything at max setting without the fake frames.

Blackwell is a sad joke - it cost ridiculously more, much more power hungry, ect. and that's without any users' feedback for this thing yet.