Nature test is hefty enough both for CPU and GPU. And yeah, all due to high geometry count and lots of animations for vegetation, which are not optimized.

I haven't investigated this thoroughly, but it always seemed to me like, past a certain point (i.e. with fairly modern graphics cards), it's heavily CPU bound.
It's most likely single threaded, so it probably makes sense given the scene complexity and, yeah, lack of optimization.

Thanks, I didn't thought about the CPU utilization cause being a self running scripted benchmark I'd not expect serious CPU usage was involved on a Pixel/Vertex Shaders 2.0 test like that where I'd expect everything were heavy mostly on the GPU.
Sure beside the nice water shader effect it's far from good looking as the 2005 Nature test that when released impressed me a lot for its gfx/sound good result and that's too a still very heavy test but less demanding I suppose on a wattage point of view.

No problem. Just tested right now on my Ryzen 5 3600X / RTX 2060, without a doubt it's a CPU bottleneck.
Even at 2560 x 1440, the GPU usage drops to 51 - 52% when the camera tilts and captures the whole forrest. And, yeah, only one CPU core is used and it's maxed out.

Thanks! I like to see modern hw how performs in older bench. Still 350fps is quite a lot in that point. From the wide camera angle and the water only effect there is like a 90% speed difference on the cards I test usually.. 😁
Just curiosity how fast does that config run 3DMark05/06 Nature test?

I got a 51064 score in 3DMark06 v1.1.0 (2560 x 1440 resolution) - also heavily CPU bound, especially the first benchmark test.
Furthermore, it kept complaining during installation that my CPU "doesn't have SSE instructions". 😁
So... who knows what arbitrary non-SSE code it's running, which is very inefficient and untested. 😀

Anyway, as Futuremark/UL states on their page, these benchmarks should not be used with modern hardware, they were never meant to run on it. Nonetheless, it's fun to do so. 😀

bloodem wrote on 2021-08-20, 08:02:

I got a 51064 score in 3DMark06 v1.1.0 (2560 x 1440 resolution) - also heavily CPU bound, especially the first benchmark test. F […]
Show full quote

I got a 51064 score in 3DMark06 v1.1.0 (2560 x 1440 resolution) - also heavily CPU bound, especially the first benchmark test.
Furthermore, it kept complaining during installation that my CPU "doesn't have SSE instructions". 😁
So... who knows what arbitrary non-SSE code it's running, which is very inefficient and untested. 😀

Anyway, as Futuremark/UL states on their page, these benchmarks should not be used with modern hardware, they were never meant to run on it. Nonetheless, it's fun to do so. 😀

Thanks. The EDIT: Firefly Forest tests at which average high/low frame rate run? So even in those ones the GPU is only half used? Anyway even if CPU limited is impressive how complex those old bench were and somehow looks even nowdays. Low end cards of almost a decade later can't run them not nearly fast enough. 😉

Attached another screenshot, taken at the end of the 3DMark03 Nature benchmark @ 2560 x 1440, with Avg (green FPS), 1% lows (orange FPS) and 0.1% lows (red FPS)

bloodem wrote on 2021-08-20, 17:20:

Attached another screenshot, taken at the end of the 3DMark03 Nature benchmark @ 2560 x 1440, with Avg (green FPS), 1% lows (orange FPS) and 0.1% lows (red FPS)

Thanks. If someday it happens to run 3DMark05 and 06 it'd be interesting to know how it performs there in their heavy Firefly Forest and in the Deep Freeze tests. 😉

386SX wrote on 2021-08-20, 18:28:

bloodem wrote on 2021-08-20, 17:20:

Attached another screenshot, taken at the end of the 3DMark03 Nature benchmark @ 2560 x 1440, with Avg (green FPS), 1% lows (orange FPS) and 0.1% lows (red FPS)

Thanks. If someday it happens to run 3DMark05 and 06 it'd be interesting to know how it performs there in their heavy Firefly Forest and in the Deep Freeze tests. 😉

Here are my 3DMark03 and 05 scores from a few years ago, when I still cared about those old versions of 3DMark. 😜

These are from my old 4.6 GHz i7-4930K machine. I still use that machine as a secondary rig (though, it now has a GTX 1660 Ti and actually scores lower than my highly overclocked GTX 970. I blame that on the much newer driver and possibly Win10).

It's been awhile, but I actually remember 3DMark03 as being fairly GPU bound. Increasing the resolution to 2560x1440 brought my score way down to around 96K, whereas 3D'05 pretty much stayed the same. I also remember the absolutely terrifying coil whine that seemingly only 3DMark03 could pull off.

Standard Def Steve wrote on 2021-08-21, 07:35:

Here are my 3DMark03 and 05 scores from a few years ago, when I still cared about those old versions of 3DMark. :P […]
Show full quote

386SX wrote on 2021-08-20, 18:28:

bloodem wrote on 2021-08-20, 17:20:

Attached another screenshot, taken at the end of the 3DMark03 Nature benchmark @ 2560 x 1440, with Avg (green FPS), 1% lows (orange FPS) and 0.1% lows (red FPS)

Thanks. If someday it happens to run 3DMark05 and 06 it'd be interesting to know how it performs there in their heavy Firefly Forest and in the Deep Freeze tests. 😉

Here are my 3DMark03 and 05 scores from a few years ago, when I still cared about those old versions of 3DMark. 😜

These are from my old 4.6 GHz i7-4930K machine. I still use that machine as a secondary rig (though, it now has a GTX 1660 Ti and actually scores lower than my highly overclocked GTX 970. I blame that on the much newer driver and possibly Win10).

It's been awhile, but I actually remember 3DMark03 as being fairly GPU bound. Increasing the resolution to 2560x1440 brought my score way down to around 96K, whereas 3D'05 pretty much stayed the same. I also remember the absolutely terrifying coil whine that seemingly only 3DMark03 could pull off.3D05-4.6-GTX970-W7.png
3D03-4.6-GTX970-W7.png

Thanks! Well 3DMark03 seems to fly there but 3DMark05 while indeed reach high fps still seems to be heavy for such high end (from my usual config point of view) system. Firefly Forest was indeed an heavy test. Also the latest Deep Freeze of 3DMark06 I remember was quite heavy. It's interesting how a decade later low end GPUs still didn't make miracles in such old benchmarks. 😉
I suppose GPUs did a long way in the high end market but the low end has been left to HTPC only oriented cards. In the past usually newer low ends could surpass older generation middle end or even high end cards.