AT has dx12 drivers?

[BFG10K]

All this is well but was it mentioned that DX12 will be available only with Windows 10..?.

http://www.hardware.fr/news/14073/directx-12-benchmarks-exclusivite-windows-10.html

That doesn't really matter given Windows 10 is free to 7 users (and up).

 

[Carfax83]

So its not even rendering the same scene.. lots of ships have no trails. Then explosions have no particles.

You can't look at the scene for differences. There's so much stuff happening so fast, there's no way you can track anything.

What you have to do is look at the statistics being generated. For example, if you look at the YouTube videos on YouTube so you can see them better, you will see there are differences in the number of units on screen between the GTX 980 and the R9 290x in DX12 mode throughout the entire video and that the number is in constant flux.

 

[Abwx]

That doesn't really matter given Windows 10 is free to 7 users (and up).

You re right, although according to hfr the following wont be eligible :

Windows 7/8/8.1 Entreprise

http://www.hardware.fr/news/14055/windows-10-gratuit-depuis-windows-7-ou-8-1.html

 

[Gloomy]

Email the developer and ask if there are outstanding bugs of the type you guys are seeing here.

For my piece, I know from experience things in Star Swarm can look different at different frame rates. At high frame rates everything is less aliased and more blurry, etc.

It's probably the motion blur that causes it. So I don't think it's weird at all that the 980 looks different since it has a much higher framerate.

Also, even before DX12 it was pretty obvious that whatever this stress test does heavily favors Nvidia GPUs. Hence why after those two or three optimization drivers the 780Ti in DX11 was performing somewhat faster (in averages) than the 290X in Mantle.

DX12 just makes it so Nvidia GPUs don't totally crater when the camera shows the whole battle (like it does for much of the RTS bench).

 

[escrow4]

So now if you are buying a new GPU you want one that supports full fat DX 12 and all extensions . . . . . . "Also absent for the moment is a definition for DirectX 12’s Feature Level 12_0 and DirectX 11’s 11_3".

 

[at80eighty]

I may have missed it; but what is the expected impact for gaming w/ref to integrated graphics, considering it might widen a minimum baseline for graphics / physics fidelity?

 

[Carfax83]

I may have missed it; but what is the expected impact for gaming w/ref to integrated graphics, considering it might widen a minimum baseline for graphics / physics fidelity?

DX12 will have a major impact for mobile gaming (both performance and power usage), but for regular desktop CPUs with integrated graphics, probably not much. Desktop CPUs with integrated graphics are already very likely to be completely GPU bound, unlike with discrete graphics cards.

 

[96Firebird]

I've always wondered how integrated graphics were handled on the chip as a whole... If CPU usage is decreased, can the extra performance be used for graphics? In other words, the chip is capable of 80% CPU/20% graphics, or 20% CPU/80% graphics, etc...? Or are they two separate entities with their own power requirements for each?

 

[SPBHM]

I've always wondered how integrated graphics were handled on the chip as a whole... If CPU usage is decreased, can the extra performance be used for graphics? In other words, the chip is capable of 80% CPU/20% graphics, or 20% CPU/80% graphics, etc...? Or are they two separate entities with their own power requirements for each?

they share the same TDP, if the CPU part is using less power, it means the GPU can use more.

specially for those 15W CPUs it should help a little

 

[krumme]

Nah, its still going to be the same. Haswell i5 is around 50% more powerful that kaveri in ST. Its about 50% more powerful in MT as well. The difference won't change, simply the kaveri will more up to playable fps (ie 30 vs. 45 fps on DX 11 vs. 60 vs. 90 fps on DX 12 assuming no GPU bottleneck - the main point is that kaveri is now getting 60 fps).

Wrong.
Its min fps deciding playability and compettiveness especially at fps like bf4 on huge maps with many players. I can testament that even a i5 quad have plenty of benefit from mantle here for min fps and smoothness. For a kaveri its a different world. You still want your 60fps consistently and at all times. Its not about some stupid average fps on a single player map, who gives.
Why is this so difficult to accept? - look at the consoles, there is no way 6 cores jaguar could
pull bf4 off without a thin driver. Its also a testament to how effective a thin driver is even for a fps - and even in a first generation engine. I think next gen dice engine is an even better fit - they said that themselves as i recall.

 

[at80eighty]

DX12 will have a major impact for mobile gaming (both performance and power usage), but for regular desktop CPUs with integrated graphics, probably not much. Desktop CPUs with integrated graphics are already very likely to be completely GPU bound, unlike with discrete graphics cards.

Gotcha, thanks. Is your observation based on current architectures or any idea if Skylake might change things?

I'm essentially curious if integrated graphics (now being the lowest common denominator, as consoles have 'upgraded' recently) achieve a reasonable minimum performance; how do developers take advantage of that

If that question makes no sense, you may safely ignore it

 

[krumme]

they share the same TDP, if the CPU part is using less power, it means the GPU can use more.

specially for those 15W CPUs it should help a little

I think the interesting part about the thin drivers is that they enable more parallization of task what was prior more single thread. And what the implications on the market is then.

Eg.:
It moves complexity to the software side, lessening the demands on single thread perf. As we know from the lead programmer of the starswarm demo, even the main history thread in games can go away in some future games !

Look eg at how nv designed Denver. 7 wide in order design. Small building block. Its a brilliant design in the way that is leverage their work on smaller arch, and then uses their competences on the compiler - software - side. Conceptually its brilliant.

There is no way anyone can compete with Intel when task is held single threaded. And everything Intel does is to protect that market. What they in anyway will not want to happen is the paralellization of tasks, because then the efficiency of their single thread cpu will be of relatively minor importance, and instead their gpu perf will be of more importance. And they dont hold the same advantages here.

 

[2is]

This is the type of review I've come to expect from AT over the years. Glad they still have it in them. Really excited for DX12

 

[Enigmoid]

Wrong.
Its min fps deciding playability and compettiveness especially at fps like bf4 on huge maps with many players. I can testament that even a i5 quad have plenty of benefit from mantle here for min fps and smoothness. For a kaveri its a different world. You still want your 60fps consistently and at all times. Its not about some stupid average fps on a single player map, who gives.
Why is this so difficult to accept? - look at the consoles, there is no way 6 cores jaguar could
pull bf4 off without a thin driver. Its also a testament to how effective a thin driver is even for a fps - and even in a first generation engine. I think next gen dice engine is an even better fit - they said that themselves as i recall.

You are agreeing with me.

DX12 decreases the amount of power needed to hit a certain framerate. If one system has more CPU power it will still be able to put out higher framerates (if not GPU limited). However, exactly as you are saying the weaker CPU will hit 60 and so for vsync purposes is perfectly sufficient. The i5 is still capable of more but kaveri is perfectly adequate now.

 

[Final8ty]

Just looked at the CPU used and again a very expensive high end Core i7-4960X, there is more to it then just threads, IPS, cache and other aspects so no just reducing how many cores are used and clock speeds is not accurate of emulating other CPU scaling at all, they should of least used a AMD CPU as well because you simply can not use a Intel CPU to represent it..

 

[zlatan]

So its not even rendering the same scene.. lots of ships have no trails. Then explosions have no particles.

StarSwarm is not designed to accurately compare different GPUs. It's designed to represent the efficiency of the new APIs. Nothing more, nothing less. It's certainly not the best program for benchmarking, but useful in other ways.
The StarSwarm program is probably an internal test scene for Nitrous, the devs just release it to prove their visions about the future.

 

[PPB]

I think the page editor should add what zlatan said as a disclaimer, lots of people going into absurd conclusions by what was presented as first as abenchmatk when it id clearly not the case

 

[frozentundra123456]

DX12 will have a major impact for mobile gaming (both performance and power usage), but for regular desktop CPUs with integrated graphics, probably not much. Desktop CPUs with integrated graphics are already very likely to be completely GPU bound, unlike with discrete graphics cards.

Aren't igps actually bandwidth limited? I have never been able to find a statement of how mantle or DX12 will affect this. The results I recall from Kaveri on the desktop, if I recall correctly, dont show a lot of benefit. As for mobile, if it can reduce cpu usage and free up more of the tdp for the igp, perhaps it will show more benefit.

Overall though, i think it will be a long time before we see how these APIs shake out. Hopefully we will see some good advances, but after seeing recent console ports, i have reservations. It also remains to be seen how quickly DX12 will be adopted, what will happen to mantle, and as well there will still be a lot of older games tht dont use it. It seems like this would make an ideal API for RTS games, but those have almost died out entirely, and i dont know if they will come back.

 

[frozentundra123456]

You are agreeing with me.

DX12 decreases the amount of power needed to hit a certain framerate. If one system has more CPU power it will still be able to put out higher framerates (if not GPU limited). However, exactly as you are saying the weaker CPU will hit 60 and so for vsync purposes is perfectly sufficient. The i5 is still capable of more but kaveri is perfectly adequate now.

Well, the key is the gpu limitation. DX12 should shift the limitation from the cpu to the gpu, making AMD cpus more competitive. And mantle at least, and i would assume DX12, allows more efficient use of multiple cores, so we should see fewer of the highly single threaded games in which the FX really tanks. However, i still expect intel to be the overall more well rounded and efficient platform. especially since it will take at least a couple of years for DX12 to become widely adopted. Not to mention that by that time Skylake will be out, and 10nm could be on the horizon.

 

[NTMBK]

Aren't igps actually bandwidth limited? I have never been able to find a statement of how mantle or DX12 will affect this. The results I recall from Kaveri on the desktop, if I recall correctly, dont show a lot of benefit. As for mobile, if it can reduce cpu usage and free up more of the tdp for the igp, perhaps it will show more benefit.

You'd hope that the low-level access would make it easier to share memory between the CPU and GPU, so you wouldn't get so many unnecessary memcpy's between CPU and GPU memory partitions to "upload" textures- which would theoretically improve bandwidth utilisation a little. Not a game developer though, so I don't know how much impact it would have.

 

[SPBHM]

Aren't igps actually bandwidth limited? I have never been able to find a statement of how mantle or DX12 will affect this. The results I recall from Kaveri on the desktop, if I recall correctly, dont show a lot of benefit. As for mobile, if it can reduce cpu usage and free up more of the tdp for the igp, perhaps it will show more benefit.

Overall though, i think it will be a long time before we see how these APIs shake out. Hopefully we will see some good advances, but after seeing recent console ports, i have reservations. It also remains to be seen how quickly DX12 will be adopted, what will happen to mantle, and as well there will still be a lot of older games tht dont use it. It seems like this would make an ideal API for RTS games, but those have almost died out entirely, and i dont know if they will come back.

mobile Intel IGPs are very TDP limited, specially on 15W and lower... they can't sustain the maximum GPU turbo clock for long once the CPU cores are loaded, DX12 loading the CPU a lot less should give extra room for the IGP to run at higher clocks during games.

 

[ShintaiDK]

You re right, although according to hfr the following wont be eligible :



http://www.hardware.fr/news/14055/windows-10-gratuit-depuis-windows-7-ou-8-1.html

I dont even think you can buy those as standalone. The corporations have been forced to use a leasing model for quite a long time. So its free upgrades.

 

[mizzou]

I am VERY excited for this. I will obviously have to upgrade my old video card and CPU, but once we get some mid-range hardware that takes advantage of low-level API, this will be a fantastic increase in gaming performance (hopefully). We've gone years and years with multiple core technology being basically wasted by having to cater to the lowest denominator.

 

[Pottuvoi]

Either way the real performance killer in star swarm is motion blur, nothing else will give you extremely big performance differences. Motion blur will drop your perf in half.

They seem to use old good accumlation buffer method.
Render the scene several times with different timesteps and accumlate the result together. (they only shade once due to their object space shading, but rendering object ~5 times for frame is not cheap.)

 

[psolord]

Can't wait for DX12 Supreme Commander 3 with 2000 unit limit per team!

If it comes that is and if it doesn't get supremely consolised!