NVIDIA GeForce RTX 5090 3DMark Performance Reveals Impressive Improvements

[GGforever]

The RTX 50-series gaming GPUs have the gaming community divided. While some appreciate the DLSS 4 and MFG technologies driving impressive improvements in FPS through AI wizardry, others are left disappointed by the seemingly poor improvements in raw performance. For instance, when DLSS and MFG are taken out of the equation, the RTX 5090, RTX 5080, and RTX 5070 are around 33%, 15%, and 20% faster than their predecessors respectively in gaming performance. That said, VideoCardz has tapped into its sources, and revealed the 3DMark scores for the RTX 5090 GPU, and the results certainly do appear to exceed expectations.

In the non-ray traced Steel Nomad test at 4K, the RTX 5090 managed to score around 14,133 points, putting it roughly 53% ahead of its predecessor. In the Port Royal test, which does utilize ray tracing, the RTX 5090 raked in 36,667 points - a 40% improvement over the RTX 4090. The results are much the same in the older Time Spy and Fire Strike tests as well, indicating at roughly a 31% and 38% jump in performance respectively. Moreover, according to the benchmarks, the RTX 5090 appears to be roughly twice as powerful as the RTX 4080 Super. Of course, synthetic benchmarks do not entirely dictate gaming performance, and VideoCardz clearly mentions that gaming performance (without MFG) will witness a substantially more modest improvement. There is no denying that Blackwell's vastly superior memory bandwidth is helping a lot with the synthetic tests, with the 33% extra shaders doing the rest of the work.



View at TechPowerUp Main Site | Source

 

[ErikG]

Impressive?? Where??

Don't forget:

"Across all 18 titles the GeForce RTX 4090 performed on average 77% faster than the GeForce RTX 3090, a 1.8X increase in 4K performance gen on gen."

 

[Daven]

From the Videocardz article:

"On average, gamers can expect about a 20% performance improvement over the RTX4090, according to reviewers we spoke with."

This is what I expect in tomorrow's TPU review. So for an inflated price of $3000 due to low supply, you get to dump 600W of dissipated heat right onto your CPU for a 20% fps improvement in raster.

 

[N/A]

Impressive?? Where??

Don't forget:

"Across all 18 titles the GeForce RTX 4090 performed on average 77% faster than the GeForce RTX 3090, a 1.8X increase in 4K performance gen on gen."

The 3090 was held back by Samsung's crappy node. 4090 being slowed down by the stagnating GDDR6X. fair enough.

 

[freeagent]

Impressive?? Where??

Looks like 4090 got destroyed to me

 

[Daven]

Looks like 4090 got destroyed to me

In synthetic benchmarks only. The actual gaming performance is closer to 20% faster than the 4090 according to the Videocardz article.

 

[freeagent]

The actual gaming performance is closer to 20% faster than the 4090 according to the Videocardz article.

With baby fresh drivers too, not bad. I'm sure that will improve..

 

[tpuuser256]

33% more supposedly yields 33% more performance (power increased by roughly 33% so I assume it's for the same clocks) the improved node is said to be 6% better at same power.
75% increase in bandwidth oughta give an additional 10% to the framerates.
I just wish I found a model with 2x600w just to squeeze around 10% more performance.
Apparently a titan blackwell gpu is in the making - using the full die - but the price is going to be egregious (if it isn't already)

 

[Daven]

33% more supposedly yields 33% more performance (power increased by roughly 33% so I assume it's for the same clocks) the improved node is said to be 6% better at same power.
75% increase in bandwidth oughta give an additional 10% to the framerates.
I just wish I found a model with 2x600w just to squeeze around 10% more performance.
Apparently a titan blackwell gpu is in the making - using the full die - but the price is going to be egregious (if it isn't already)

Don't forget that games are not well optimized to yield linear results at 4K and you get diminishing returns at lower resolutions.

 

[Macro Device]

Impressive?? Where??

It's impressive how unimpressive it is. New language is wild.

 

[Chomiq]

We're one week from reviews, does anyone really care about leaks at this point?

 

[N/A]

We're one week from reviews, does anyone really care about leaks at this point?

48 hours.

 

[Daven]

We're one week from reviews, does anyone really care about leaks at this point?

Reviews are tomorrow, January 23. The 5090 is available to buy on January 24. You might be thinking about the 5080.

48 hours.

I think the reviews are tomorrow so 24 hours.

 

[JustBenching]

Looks like 4090 got destroyed to me

Don't care about raster performance, I hope the 5090 is a good 40%+ faster in rt. Then I might be opening my wallet.

 

[renz496]

Impressive?? Where??

Don't forget:

"Across all 18 titles the GeForce RTX 4090 performed on average 77% faster than the GeForce RTX 3090, a 1.8X increase in 4K performance gen on gen."

if you did not find this impressive then you simply take all the advancement in GPU world for the last 30 years for the granted.

 

[Hecate91]

if you did not find this impressive then you simply take all the advancement in GPU world for the last 30 years for the granted.

GPU advancement for the last 30 years hasn't been marketed exclusively on fake frames and upscaling.

 

[Macro Device]

if you did not find this impressive then you simply take all the advancement in GPU world for the last 30 years for the granted.

Nah, the point was we used to always get +X % performance per dollar every new generation, with X being beyond 100 a couple decades ago. 4090 destroyed 3090. 3080 Ti destroyed 2080 Ti. 2080 not only was faster than 1080 Ti but it also had a couple new features. 1080 Ti left 980 Ti with no business.

Today, we get a more expensive GPU with even more ridiculous power draw that is about as much faster. FPS/$ doesn't change to the better so what is there to be impressed with? nVidia are taking our quid for granted, this is what happens. Sure, 5090 will make 4090 look pathetic at AI workloads but that's the only field where you don't need a FPS counter to tell the difference between them.

They CAN do better, there's no physical evidence they can't. We haven't run out of nanometres. We haven't run out of silicon. Just why try harder when you can be king even without releasing anything?

 

[BorisDG]

25-30%... "very impressive" and pretty much what most people already calculated via specs.

 

[lexluthermiester]

Looks like 4090 got destroyed to me

Yeah, that's what I'm seeing too. Looks exactly like I was expecting from historical gen-on-gen performance jump trends. Of course we need to wait for the official numbers..

35-40%...

Fixed that for you.

 

[PixelTech]

...The 5090 is available to buy on January 24...

The 5090 is available to buy on Jan 30th. Reviews for the FE come out Jan 23rd and the AIB reviews are out 24th.

 

[nguyen]

Oh well guess I will take the 30% higher raster, 40% higher RT and the improved DLSS Super Res, Ray Reconstruction and frame pacing (with MFG)

 

[arestavo]

Not really impressive. IIRC, that means it's 33% faster for 30% more money (if you're lucky!), and 20% more power consumption. (edit: 4090 to 5090)

That's not impressive at all. Maybe it's impressive marketing, ha!

 

[LittleBro]

RTX 5090 Ti (or TITAN) will be the card that totally destroys mighty RTX 4090. In mentnioned 3D Mark tests results, it's around +38% more perf. across the board for 5090 over 4090. With 5090 Ti, this might as well change to +55...60% improvement over 4090, just as we saw when 4090 came out after 3090. Sometimes I think RTX 5090 Ti was supposed to be real sucessor to RTX 4090, as CUDA cores increase from RTX 3090 to RTX 4090 was around 56%, while from 4090 to 5090 it is around 33%. 5090 Ti, assuming the rumors are true, will have exactly 50% more (24 576) CUDA cores than 4090 (16 384).

They CAN do better, there's no physical evidence they can't. We haven't run out of nanometres. We haven't run out of silicon. Just why try harder when you can be king even without releasing anything?

People talk about hitting a silicon wall, though no chip maker really talks about it. I keep hearing about hitting these walls and yet another nodes are always developed and released.

 

[dartuil]

what i care is native resolution + rt perf

 

[BorisDG]

Yeah, that's what I'm seeing too. Looks exactly like I was expecting from historical gen-on-gen performance jump trends. Of course we need to wait for the official numbers..


Fixed that for you.

I can read and it's not 40% for sure. The best scenario in the picture is 35%. Still pretty poor jump considering the bus, memory and more stuff in the die.