UL Solutions Previews Upcoming 3DMark Steel Nomad Benchmark

[GFreeman]

Thank you to the 3DMark community - the gamers, overclockers, hardware reviewers, tech-heads and those in the industry using our benchmarks, who have joined us in discovering what the cutting edge of PC hardware can do over this last quarter of a century. Looking back, it's amazing how far graphics have come, and we're very excited to see what the next 25 years bring.

After looking back, it's time to share a sneak peek of what's coming next. Here are some preview screenshots for 3DMark Steel Nomad, our successor to 3DMark Time Spy. It's been more than seven years since we launched Time Spy, and after more than 42 million submitted results, we think it's time for a new heavy non-ray tracing benchmark. Steel Nomad will be our most demanding non-ray tracing benchmark and will not only support Windows using DirectX 12, but also macOS and iOS using Metal, Android using Vulkan, and Linux using Vulkan for Enterprise and reviewers. To celebrate 3DMark's 25th year, the scene will feature some callbacks to many of our previous benchmarks. We hope you have fun finding them all!



View at TechPowerUp Main Site

 

[dampflokfreund]

Looks good, but far from impressive. Just like the other benchmarks they've been producing in the last few years.

Back in the day, 3DMark always set the new standard for graphics, but now they just kinda exist. Now regular games look far better than what they are releasing. Fire Strike was the last truly impressive benchmark imo. I wonder if its a budget issue?

What I would really like to see is them chaging the art style and produce a photorealistic benchmark that truly looks like CGI, using all tech available (Raytracing, Mesh Shaders, Sampler Feedback etc), in the vein of Alan Wake 2 but looking even better, truly photorealistic I mean.

 

[TumbleGeorge]

and we're very excited to see what the next 25 years bring

If we don't count the introduction of real holography in games, I don't see the point of other improvements. The picture quality is now completely satisfactory for average Joe in terms of colors, precision, physical effects and completely different from what existed 25 years ago and which forced some people to complain. Perhaps only finally 8K resolution for gaming, as a (last) step forward, already available in development plans drawn up 7-8 years ago, but not yet happened on a large scale.

 

[VulkanBros]

Just give me back my Matrox Mystique and 3DFx Voodoo cards - and I will pulverize these benchmarks

 

[Battler624]

If we don't count the introduction of real holography in games, I don't see the point of other improvements. The picture quality is now completely satisfactory for average Joe in terms of colors, precision, physical effects and completely different from what existed 25 years ago and which forced some people to complain. Perhaps only finally 8K resolution for gaming, as a (last) step forward, already available in development plans drawn up 7-8 years ago, but not yet happened on a large scale.

Same thing was said in 2010 and in 2020.

And yet to continue to see better visuals, bridging the gap. I'd say it'll be enough by 2050.

 

[gmn 17]

Yay hope its a free update for this BM

 

[HOkay]

Looks good, but far from impressive. Just like the other benchmarks they've been producing in the last few years.

Back in the day, 3DMark always set the new standard for graphics, but now they just kinda exist. Now regular games look far better than what they are releasing. Fire Strike was the last truly impressive benchmark imo. I wonder if its a budget issue?

What I would really like to see is them chaging the art style and produce a photorealistic benchmark that truly looks like CGI, using all tech available (Raytracing, Mesh Shaders, Sampler Feedback etc), in the vein of Alan Wake 2 but looking even better, truly photorealistic I mean.

I know what you mean, but it's not the job of a benchmark to be the absolute best looking, it's just got to be demanding & produce a very repeatable load. It would be cooler if it was also the best looking thing around, but it's not required.

 

[MentalAcetylide]

Just give me back my Matrox Mystique and 3DFx Voodoo cards - and I will pulverize these benchmarks

More like the benchmark attempts with one will pulverize the cards. They were great in their day, but games & such nowadays are written more towards fluff and eye candy rather than optimization. Graphics cards now are simply fueling the whole PC market and can serve multiple purposes rather than just displaying graphics/gaming. You can't just simply buy any graphics card and throw it in your system for a huge performance boost anymore. Now you have to be concerned with CPU, motherboard, cooling, PSU, RAM, etc.

 

[agent_x007]

Make a benchmark that can disable functions as required for testing each DX12 capable GPU.
Making one that works from Fermi's DX12 patch job up to Ada would be amazing to have.
Throwing in Mesh Shaders, Sampler Feedback, and other stuff that can increase performance or image quality as options would be even better (assuming card support it).

 

[AusWolf]

That water reflection looks pretty impressive. Makes one wonder why we need RT at all.

 

[Semik]

Will it be free like Basic Time Spy? Otherwise, it has no chance to spread like that ...

 

[napata]

That water reflection looks pretty impressive. Makes one wonder why we need RT at all.

In regards to reflections: because it's the most computionally efficient way to do accurate reflections, outside of baking reflections. Baking isn't really possible in anything but static linear games and even then it's very time consuming for devs so only limited to a handful of scenes. UC4 used a lot of baking for example for certain static scenes. It's not a realistic method in dynamic games with a lot of moving parts so any open world game. It's possible in a demo like this though.

Then there's the accurate raster ways like planar, which is fine for 1 reflective surface but scales terribly with multiple. Take Spider-Man: it has decent RT reflections on cars & windows, even on a PS5. If you'd use planar for this it'd make turning on RT reflections look like turning on AFx16 instead of AFx8. I truly wonder what the performance would be, but it's probably be 3 fps at 1080p on a 4090 levels of bad.

RT is the future for a reason as we hit the limitations of what's possible with raster in an efficient way. Raster is all about approximation and fake it until you make it but if the faking takes more effort than the real thing then that no longer makes sense.