[FREE] Pure RayTracing and Non-RT benchmarks

[miguel1900]

Hi all!

I have created a new GPU benchmark, designed to know how powerful our GPUs are, and great to find the best overclock. It's very easy and fast to try it: download, decompress and execute, without installing anything.

Currently, there are two independent versions of the benchmark: one with Ray Tracing enabled, and another one, without Ray Tracing, as a 'standard' benchmark, so any user, even those without a RT capable GPU, can use it in a usefu way. In a future realease, both benchmarks will be integrated into only one, simply being different options to be selected in an unified benchmarking software.

You can find more information and the free download link at the project's page:
- Ray Tracing edition (Pure RayTracing benchmark): https://marvizer.itch.io/pure-raytracing-benchmark
- 'Standard' edition (Pure Non-RT benchmark): https://marvizer.itch.io/pure-non-rt-benchmark

In addition, if you have doubts about your PC being Ray Tracing ready, I have created this very simple and ultra fast tool to check if your PC is able to run the Ray Tracing technology: https://marvizer.itch.io/ray-tracing-compatibility-checker

I would be very happy if you could try it and share your results here! (If only passing one benchmark, I would recommend the Common Settings at QHD).

Updated scores: at the same itch.io project's page.

Thank you very much and best regards!

 

[londiste]

Yours is content and it is based on newest Unreal Engine?
Would be nicer if it didn't automatically initialize VR headset at start

 

[sepheronx]

wow, that is super cool!

I am impressed with your work.

 

[miguel1900]

Thank you very much!

It's made with latest UE4 version. And yes, all mine except 3D models; I used some marketplace packs for models, as my goal was to focus in lighting, ray tracing capabilities and benchmarking measurements (and the rest, of course, like camera effects, ambience, gameplay...).

Didn't realise about the VR headset. I didn't try it with a HMD connected , but will keep an eye on it for an upcomming update!

I'm working too in an option to use it without Ray Tracing, so all kind of users will be able to pass it as a "normal" benchmark too.

Best regards!

 

[agent_x007]

@miguel1900 System requirements ?
1) Windows version Win7/Win10 and if Win10, does it need RayTracing enabled build ?
2) DirectX or Vulkan ?
3) Does it use software RT or is it RTX only ?
4) Are RayTracing 1.0 cards (ie. Pascal) "good enough" ?
5) How much RAM/VRAM is needed (and when you will "run out") ?
6) Do you need special CPU instructions to run it (AVX/SSE4.2/FMA, etc.)
7) What can be called a "recommended CPU" (do you need more cores or is it single threaded benchmark) ?

 

[miguel1900]

@miguel1900 System requirements ?
1) Windows version Win7/Win10 and if Win10, does it need RayTracing enabled build ?
2) DirectX or Vulkan ?
3) Does it use software RT or is it RTX only ?
4) Are RayTracing 1.0 cards (ie. Pascal) "good enough" ?
5) How much RAM/VRAM is needed (and when you will "run out") ?
6) Do you need special CPU instructions to run it (AVX/SSE4.2/FMA, etc.)
7) What can be called a "recommended CPU" (do you need more cores or is it single threaded benchmark) ?

Hi @agent_x007 , good questions! Will try to clarify them (but, basically, the same system requirements to execute RT, in general):
1) Windows version Win7/Win10 and if Win10, does it need RayTracing enabled build ? As for RT in general, Nvidia GPU (currently) with drivers 425.31 or higher, Windows 10 v1809 or higher.
2) DirectX or Vulkan ? DirectX 12
3) Does it use software RT or is it RTX only ? Hmmm not sure about this meaning. It's "real" Ray Tracing, but even an AMD could run it, but RT wouldn't be working, if it helps to your question in any sense.
4) Are RayTracing 1.0 cards (ie. Pascal) "good enough" ? They can run it, but the performance gap is the well known huge gap. They will run it at a very low FPS.
5) How much RAM/VRAM is needed (and when you will "run out") ? Not sure. I will need more testers, but 6gb of Vram seems to be enough, as even my old GTX 1060 6gb ran it (but at 6fps).
6) Do you need special CPU instructions to run it (AVX/SSE4.2/FMA, etc.). You don't, as far as I know.
7) What can be called a "recommended CPU" (do you need more cores or is it single threaded benchmark)? If it isn't a bottleneck for your GPU, any kind of CPU should run smoothly. CPU is almost "sleeping", being only a high-demandant GPU benchmark.

Thank you!

 

[Ibotibo01]

Nice benchmark, good work.
GTX 1660 gets 11.49 FPS at 1080p Common Settings.

 

[miguel1900]

Nice benchmark, good work.
GTX 1660 gets 11.49 FPS at 1080p Common Settings.

Thanks @Ibotibo01 !

Here my slightly OCed 2080 Ti, @ QHD Common Settings:

FHD Common Settings:

 

[Toothless]

I got a wonderful 6-7fps on pro settings/4k. I miss Microsoft Powerpoint.

 

[miguel1900]

Hi @Toothless , with which GPU? Were you sure your PC was executing Ray Tracing? An already tested RTX 2070 gets around 6fps at 4k/Pro.

Regards!

 

[mouacyk]

I got a wonderful 6-7fps on pro settings/4k. I miss Microsoft Powerpoint.

Wasn't aware MS Powerpoint did ray-tracing. I'll have to check it out.

 

[Toothless]

Hi @Toothless , with which GPU? Were you sure your PC was executing Ray Tracing? An already tested RTX 2070 gets around 6fps at 4k/Pro.

Regards!

I never bought into the ray tracing gimmick, and system specs are current if that helps.

Wasn't aware MS Powerpoint did ray-tracing. I'll have to check it out.

The best .png ray traced pictures ever.

 

[sil3ntearth]

Results from a 2060 Super.

 

[miguel1900]

Hi @Toothless

I'm not sure about understanding your last comment. I'm Spanish, sorry!

Thank you @sil3ntearth ! I have updated the scores chart on the itch.io project's page, but here you are a current copy&paste:

FHD Common Settings: 2080 Ti (+100, +900) 2080 Ti 2080 mobile 2070 2070 2060 Super 2060 Super 1080 Ti 1660	68 61 41.5 41.5 40 38 36.5 15 11.5
FHD High Settings: 2080 Ti (+100, +900) 2080 mobile 2060 Super	46.5 28.5 25
FHD Pro Settings: 2080 mobile 2060 Super	22 18.5
QHD Common Settings: 3090 2080 Ti (+100, +900) 2080 Ti 2080 2080 mobile 2070 2070 1080 Ti	67 41.5 37.5 29 25 25 24 9
QHD High Settings: 2080 mobile	17
4k Pro Settings: 3090 2080 mobile 2070	17 6 5.5

Regards!

 

[basco]

nv441.87 win10 1903
1080ti@1900mhz \ 5960x@4300mhz
1920x1080 common settings = 14,48 avg fps

 

[pantherx12]

Downloaded without a donation for now as I want to check it out first.

Will update later with performance and potentially send over a donation.

 

[miguel1900]

Downloaded without a donation for now as I want to check it out first.

Will update later with performance and potentially send over a donation.

Of course, don't worry!

Thank you for downloading and using The most important thing is to make it useful.

 

[pantherx12]

Average frame rate of 32.4 fps on my hardware ( see system specs) @ 1080p

Couple of suggestions, make it output results at the end. the more data the better, I imagine hardware reviewers will like this little benchmark.

More dynamic lighting moments, for example some explosions here and there, or perhaps sparks from some wiring.

Because whilst it's beautiful in its on way, it's also quite sparse.

I sent along a small amount of money, enough for a coffee or beer.

 

[hardcore_gamer]

Amazing work!

By pure ray tracing, you mean fully path traced (including the GI) and little to no rasterization?

 

[Hardi]

2080 super with a little undervolt and +800 to mem

 

[miguel1900]

Average frame rate of 32.4 fps on my hardware ( see system specs) @ 1080p

Couple of suggestions, make it output results at the end. the more data the better, I imagine hardware reviewers will like this little benchmark.

More dynamic lighting moments, for example some explosions here and there, or perhaps sparks from some wiring.

Because whilst it's beautiful in its on way, it's also quite sparse.

I sent along a small amount of money, enough for a coffee or beer.

Thank you very much! Also for the donation.

About "at the end", do you mean when you pass to control the character? In that case, final results remain at the screen. When you exit pressing escape, you get also the last benchmark result on the menu screen.

I thought about more dynamic moments too, at the beginning, but finally didn't add them because of some things, like I had no enough time during the weekend I had free time to work on it and whith the release of the new cards, and because I wanted to prevent little random things, that could vary the benchmark result (the explosions should have some random components, but also the sparks). Anyway, I love your idea, but I needed to decide then. It would be less boring, but simply add more effects to harm performance, which is quite under load, already. Right now, probably, I shouldn't make more changes, because people are already using the benchmark and showing their results. But I will kept it in mind. Maybe, if some reviewer are going to start testing it, make it more popular, and they tell me before reviewing, it could be a good moment to make new changes and make a new "beginning".

PS: I have tried writing some hardware webpages but usually with no replies. If you know someone, all help is welcome!

Amazing work!

By pure ray tracing, you mean fully path traced (including the GI) and little to no rasterization?

Thank you! Yeah, all lighting and effects are fully RT dynamic, except for the Common Settings, without GI and with Sky lighting backed, and the High Settings, with GI but still with the Skylight backed. Pro Settings has the Skylight dynamic. It's a little more explained at the itch.io project's page.

2080 super with a little undervolt and +800 to mem

Thanks for sharing!

 

[londiste]

RTX2080, Gainward Phoenix GS - Boost clock 1815 MHz.
Majority of the time Common runs at 1860MHz, High at 1845MHz and Pro at 1830MHz, basically each more taxing setting one runs at one clock bin lower.

Spoiler

1440p Common - 29.56 FPS
1440p High - 19.87 FPS
1440p Pro - 15.33 FPS

Spoiler

1080p Common - 46.81 FPS
1080p High - 32.51 FPS
1080p Pro - 25.85 FPS

By pure ray tracing, you mean fully path traced (including the GI) and little to no rasterization?

I bet it isn't. This is still Unreal Engine and the hybrid RT approach, just heavy with RT effects. RT Reflections, RT GI and RT shadows, I assume. @miguel1900, anything else?

Do you take feature requests? Especially if you'd like this to take off as a benchmark - batchs runs please.
Queue up tests with different settings with x amount of runs each, logging at least the results (if not in more details - min, max, avg, maybe frametimes, maybe monitoring data)
More details about what settings exactly are changed for High-Pro would be nice as well.

 

[miguel1900]

View attachment 170274 View attachment 170275 View attachment 170279

RTX2080, Gainward Phoenix GS - Boost clock 1815 MHz.
Majority of the time Common runs at 1845-1860MHz, High at one bin (15MHz) lower and Pro at two bins lower.

1440p Common - 29.56 FPS
1440p High - 19.87 FPS
1440p Pro - 15.33 FPS

I bet it isn't. This is still Unreal engine and the hybrid RT approach, just heavy with RT effects. RT Reflections, RT GI and RT shadows, I assume. @miguel1900, anything else?

Do you take feature requests? Especially if you'd like this to take off as a benchmark - batchs runs please. Queue up tests with different settings with x amount of runs each, logging at least the results (if not in more details - min, max, avg, maybe frametimes, maybe monitoring data)

Thank you!

Already added to the table of scores.

Well, it's not "path traced", but it's "ray traced". You are totally right, simply adding the RT Sky on the Pro test. On Pro, there are no rasterization nor backed lighting at all. How would you define your hybrid term? (Maybe it's the correct term!)

Of course! Suggestions are totally welcome. I have already added (but not published yet) a looping mode, and it shows the number of current loop, but didn't thought to allow the user to fix a cretain loop number. About the rest, it could be interesting too! but a little time consuming. The more used it is, the more time I will dedicate.

Thanks again!

 

[londiste]

Well, it's not "path traced", but it's "ray traced". You are totally right, simply adding the RT Sky on the Pro test. On Pro, there are no rasterization nor backed lighting at all. How would you define your hybrid term? (Maybe it's the correct term!)

Large part of the scene is still done with rasterization and other more common methods. There are certain effects or buffers (or parts of them as with your example of Skybox lighting getting raytraced at higher settings) that are raytraced (as a general term, whether the actual algorithm/method used is pathtracing, some hybrid and/or the nice postprocessing like inevitable denoising) that are then composited together. Reflections, shadows, GI/AO are the main ones with UE as far as I remember.

UE does have some type of full pathtracing renderer but to the best of my knowledge that is essentially for reference renders, not real-time usage. From what I have seen or heard from its performance, this is absolutely definitely not what your benchmark uses

Edit:
UE RT overview in their own documentation:

Hardware Ray Tracing Tips and Tricks in Unreal Engine | Unreal Engine 5.4 Documentation | Epic Developer Community

A collection of topics to aid with development of projects using hardware ray tracing features in Unreal Engine.

docs.unrealengine.com

Translucency is the one I forgot. GI and AO are separate as well.

 

[JalleR]

hmmm well i think a 3080 will be an upgrade even with the 10GB of memory