Oxide: Star Swarm benchmark to be released on Steam this month!

[DaZeeMan]

Longtime Anand reader, first time poster.

I've been following the progress on the Oxide engine, mainly via the Stardock Forums (Brad Wardell, aka Frogboy, often posts interesting tidbits).

Recently, Brad posted some new comments about the Nitrous engine here:
http://forums.elementalgame.com/451041

And, also replied to some more technical questions about Nitrous, Mantle, and DX11 here, and also about the Star Swarm Benchmark:
http://forums.elementalgame.com/451041/get;3430478
See replies #24 and #25

Here's some quotes from above linked replies:

Oxide's engine, Nitrous biggest advantage is that it is completely asynchronous in its rendering. The engine is not tied to a thread so the more cores you have, the faster it gets.
The biggest advantage of Mantle is that unlike DirectX, Mantle is truly multicore aware. With DirectX 11, more cores don't buy you nearly as much. That said, Nitrous, even on DirectX, is still two orders of magnitude faster than say a typical DirectX 9 engine.
A Mantle optimized game can show a massive performance gain depending on how many cores the user has on their CPU. Contrary to what I read on some forums, most games remain CPU or video driver bound (i.e. the GPU is waiting to be fed). Mantle lets you get a lot more stuff onto the GPU.

And, in answer to the question 'Why hasn't any big developer done this as of yet?'

The reason is that it is non-trivial. Stardock internally couldn't build this and we have some extraordinary talent. The GalCiv III engine, written from scratch with DirectX 10/11 in mind, can't touch Nitrous and make no mistake, the GalCiv III engine is very good. But Nitrous is something I've never seen before in my career.
I know I'm not the only one who was dubious at the idea of having a completely core-neutral 3D engine. Who manages the jobs? How do you synchronize tasks? How do you take care of ordering?
In a normal modern 3D engine (and when I say modern I mean, an engine that would normally be written in 2013) you would have a dedicated thread for the job system with a bunch of worker threads that handle executing on that job system. That will get you pretty far and requires DirectX 10/11 to do because it means multiple threads will be touching the GPU.
But Nitrous has no concept of dedicated threads with specified jobs.
Being able to show thousands of units on screen at once is the thing the guys get geeked out at the most. But when it gets demoed, that's not even the thing that catches peoples eyes the most. Instead, it's how things get rendered. The Nitrous engine is basically a real-time version of renderman (i.e. the CGI you see in movies) so the way light and materials work is radically different.

And finally, because I KNOW y'all like benchmarks...

BTW, Star Swarm (the benchmark) is going to go up on Steam this month so all this admitted hyping I'm doing is something you will be able to verify first hand in a few weeks.

So, it looks like we will be able to see for ourselves, on our own 64 bit machines at least, a Nitrous implementation soon!


The asynchronous engine threading, and DX 10/11 performance supposedly being 2 orders of magnitude faster are comments I found quite fascinating, hence my need to share with the collective here... This sounds like great news all around.

Of course, as Brad also noted in his first post, Nitrous isn't anywhere near ready for 3rd party licensing, and I am not seeing any game titles mentioned in associated with the Nitrous engine yet, but nonetheless I'm pretty excited to see where this ends up.

 

[BallaTheFeared]

Interesting information, however it will be locked due to the Mantle thread.

 

[DaZeeMan]

Interesting information, however it will be locked due to the Mantle thread.

My main goal here was to give everyone a heads up about the benchmark coming to Steam soon in a few weeks, according to Brad at least.

And a good chunk of the info isn't Mantle specific, and focuses on DirectX, which applies to pretty much everyone. I figured the NVidia types around here would appreciate those tidbits as well.

Also, Brad's initial post is dated December 26th, which makes it fairly recent, and hence relevant.

 

[AnandThenMan]

Cool, looking forward to this.

 

[blastingcap]

Interesting information, however it will be locked due to the Mantle thread.

Not necessarily, this this is about the Nitrous engine and not Mantle per se. Also it's actual news, not talking about a press release from months ago. If the Nitrous engine produces a working benchmark, people can use it for benchmarks like people use 3DMark today. BF4 Mantle was supposed to be the first implementation, but it may not come out until after the Nitrous demo.

 

[xSneak]

So is the benchmark going to use DX11 or mantle?

Thanks for the info OP.

 

[BallaTheFeared]

So is the benchmark going to use DX11 or mantle?

Thanks for the info OP.

Sounds like maybe both, but at least DX11...

I'm looking forward to it, hopefully it will be like that first 3D mmo experience where you're just blown away coming from top down 2D :biggrin:

 

[DaZeeMan]

So is the benchmark going to use DX11 or mantle?

Thanks for the info OP.

Someone posed a similar question to Brad in that thread. If I see a relevant reply, I'll make sure to share it here.

 

[SiliconWars]

So I guess they'll really push the engine to bring systems to a grinding halt, otherwise it won't be much of a benchmark. Wonder how many ships that'll take, should be interesting!

 

[ShintaiDK]

So is the benchmark going to use DX11 or mantle?

Thanks for the info OP.

I guess DX11 only. Is there even mantle drivers out?

Also is this thread not just another bypass of the locked mantle thread?

 

[Imouto]

So is the benchmark going to use DX11 or mantle?

Thanks for the info OP.

It's safe to assume that it will work in both modes to remark the massive difference between them. i.e systems with GCN cards slaughtering non-GCN ones at impossible amounts of draw calls.

Just another benchmark to prove nothing like Unigine, Tessmark, 3DMark and such testing extreme scenarios that no game will ever feature.

This is getting really old already without the BF4 Mantle patch out and no actual numbers.

 

[krumme]

I have been saying 3 times in the Mantle thread that the oxide engine is a revolution because there is no lead thread (as we understand a lead thread) . Nobody seems to grasp the consequences of it. We have an rts and roleplay game revolution just in front of us and some prefers to piss all over the mantle thread because they are scared of their own little investment in a gfx card.

The Oxide engine is in fact a huge step in front of frostbite3 from dice in that it uses true mt. If you listened carefully to the q&a at apu13 you would have noticed Johan saying they are still limited by the lead thread and then gets corrected by one of the oxide guys - whoops

I hope the concept of true mt comes into fps also. Fb4 could use it to make far more and better levolution.

I think oxide engine is a perfect example of what happens when mantle give the devs the control. Who would have thought before it was even possible to avoid the lead thread?

What was considered impossible just 3 months ago is a reality today.

 

[Imouto]

Are we still talking about this benchmark or the Oxide engine or a far fetched Oxide based game?

This standalone benchmark will prove nothing like none out there do. Like not even built in benchmarks in games do.

 

[3DVagabond]

Interesting all these people certain what it will bring, but they know nothing about it.

 

[krumme]

Are we still talking about this benchmark or the Oxide engine or a far fetched Oxide based game?

This standalone benchmark will prove nothing like none out there do. Like not even built in benchmarks in games do.

What will happen with this bm is that when you go over draw call limit for dx, fps will plummet. Thats what one of the oxide guys told at their apu13 presentation when asked the question by the audience.

What that also goes to show is the new games using the engine, is only really pratically playable with gcn hardware. We can just be straight about it.

Oxide told in application performance was about 3 times as fast, 300%, using mantle compared to dx.

For drawcall they told the difference was 10x. And that was even before true mt.

 

[Carfax83]

I want to see how this engine stacks up against another proper DX11 engine like Firaxis's Lore Engine, which can do 15 to 20K draw calls in a frame with DX11 multithreading.

Then I want to see how good the Mantle path is :biggrin:

 

[krumme]

I want to see how this engine stacks up against another proper DX11 engine like Firaxis's Lore Engine, which can do 15 to 20K draw calls in a frame with DX11 multithreading.

Then I want to see how good the Mantle path is :biggrin:

Yes and they use a 6c 12 thread Intel i7 in a game engine they took 3 years to develop. Oxide used 2 months to get 400% as many drawcalls on a far less powerfull amd machine.

 

[Carfax83]

What that also goes to show is the new games using the engine, is only really pratically playable with gcn hardware. We can just be straight about it.

Yeah well I hope they sell enough copies to the 5% or 6% of GCN owners out there (according to Steam hardware survey at least) to keep afloat

Oxide told in application performance was about 3 times as fast, 300%, using mantle compared to dx.

For drawcall they told the difference was 10x. And that was even before true mt.

When this engine is used in an actual game, then they can talk...

 

[Imouto]

What will happen with this bm is that when you go over draw call limit for dx, fps will plummet. Thats what one of the oxide guys told at their apu13 presentation when asked the question by the audience.

What that also goes to show is the new games using the engine, is only really pratically playable with gcn hardware. We can just be straight about it.

Oxide told in application performance was about 3 times as fast, 300%, using mantle compared to dx.

For drawcall they told the difference was 10x. And that was even before true mt.

What will happen is that ppl will boast an insane amount of draw calls with overclocked machines meaningless for actual games leaving in the table the benefits of Mantle for lower end systems.

What's the actual number of draw calls or batches needed for a game? Is this benchmark going to tell you anything relevant than the difference between DX11 and Mantle? Is this benchmark optimized for DX11 so we can see the true difference?

Useless benchmark is useless.

 

[Carfax83]

Yes and they use a 6c 12 thread Intel i7 in a game engine they took 3 years to develop. Oxide used 2 months to get 400% as many drawcalls on a far less powerfull amd machine.

The 15 to 20K for the Lore engine is in the actual game though, and not in a demo or benchmark.

At any rate, the true comparison will be on the same rendering paths. The Nitrous engine on DX11 vs the Lore engine on DX11.

 

[Paul98]

Very cool, I am going to have to read up more on how this engine works. Sounds very interesting on the actual internal workings. Also a demo would be great and so quick will be even better. I would like to see this demo with better looking ships.

 

[Paul98]

Reading some of these posts seems obvious people didn't even bother to click on the links.

 

[Imouto]

Some people is really vague.

 

[krumme]

What will happen is that ppl will boast an insane amount of draw calls with overclocked machines meaningless for actual games leaving in the table the benefits of Mantle for lower end systems.

What's the actual number of draw calls or batches needed for a game? Is this benchmark going to tell you anything relevant than the difference between DX11 and Mantle? Is this benchmark optimized for DX11 so we can see the true difference?

Useless benchmark is useless.

I dont find any synthetic bm interesting - and agree this one is just as uninterestning. BM tools is just marketing. Its especially useless for gaming, when we can just game and judge the experience our selves and just meassure fps when we do so to add to that experience.

The drawcalls is limiting the RTC big time. I think the posters at the forum have plenty of examples.
http://forums.elementalgame.com/451041/page/1/#3430478

What about a Roman army of 80 men as we have today? With mantle using the same ressources for developing, and a similary strong quad/octo multicore machine, an estimated 800-1600 men is possible. That accounts for something.

Oxide predicts 3-300k drawcalls 2015 and 1M 2018. And it will just transform games because we can make games we couldnt before.

 

[dacostafilipe]

Is it me or does it sound like an event driven framework with an eventloop per core?

Parallelism and concurency?