Devs dont need to support every card out there. Only kepler, maxwell, gcn1, 2 and 3 have dx12 support anyway. Also as dan baker said, there are programmatical ways to optimize for different uarchs.they do one close box system where every using same specs.
There are to many pages.
In find this here.
http://www.neogaf.com/forum/showthread.php?t=1204574&page=1
There too many different card with too many different CPU with too many different software, drivers, hardware.Devs dont need to support every card out there. Only kepler, maxwell, gcn1, 2 and 3 have dx12 support anyway. Also as dan baker said, there are programmatical ways to optimize for different uarchs.
Not to even mention that that really is the job of middleware providers a la unreal and unity.
So yes, it is more complicated but more performant and easier to debug.
Sent from my SM-G930T using Tapatalk
Because they make GPUs for PCs that are actually capable of running games and have no need of the GPU to run the whole game.
Look at that even the i3-2100 hits 56FPS avg and that's with sli which is degrading performance.
And not only that but the ancient dual core only runs at 60% usage.
For anyone that complained about AI and degrading gameplay until now, the party is just starting, new games will be graphics pr0n with a minimum of gameplay.
Then just have to:There too many different card with too many different CPU with too many different software, drivers, hardware.
It can be around 1000s of different configuration.
I see, so you're saying that Nvidia makes GPUs that are capable of play games worse than the competition.
I guess when a game is GPU bottlenecked, Nvidia suffers worse than AMD now.
That's unfortunate, anything else you'd like to share with us?
Also, do you know where I can buy an i3-2100+GTX980TiSLI system? Sseems really useful.
which means that the game runs almost only on the GPU.
Nvidia makes GPUs that provide features that the CPU can not do or is very bad at,new games use the GPU as an CPU substitute (lower then 1,5Ghz console cores) figure it out.
I didn't start this thread, I saw the benchmark posted and I commented on it,a very weak dual core running sli gets 56FPS with ~60% usage which means that the game runs almost only on the GPU.
Looking at those graphs it appears that AMD is winning big time ... D:
DX12 is a true saviour for AMD ...
Are they even using async compute ? :|
After watching that video proof, I have to recant my OP about how broken the game is on all hardware. It's clearly not that broken on AMD.
There seems to be an issue with PresentMon here as well, because it suggests a lot of stutter, but it's not very noticeable on the 390 in that 60 fps video. On pcgameshardware.de's video with their 980, the stutter was obvious.
So basically it's Quantum Broken for NV.
As for whether it uses Async Compute, I am almost certain it does. All that fancy tech on the Xbone's weak hardware would be impossible without it.
DF mentions Volumetric Lighting settings tank NV performance, that is a clear indicator, as VI is a compute effect. TR and other console games run VI, SSAO, (& PureHair) in AC mode. There's got to be AC in play else the performance wouldn't tank so heavy on NV GPUs.
Broken for NV.
As well as almost every other DX12 game...
This is the devious plan of NV to make maxwell obsolete just like they did with Kepler. Guys like sontin or desperado will gladly shell out another $700 bucks for a gtx 1080. It's just like apple fanboys that like paying more for equal or worse hardware.
This is the devious plan of NV to make maxwell obsolete just like they did with Kepler. Guys like sontin or desperado will gladly shell out another $700 bucks for a gtx 1080. It's just like apple fanboys that like paying more for equal or worse hardware.
Which means the game was gpu bound during testing which means it's comparing only the gpus and not the cpus. If it was 100%, the game would be could bound the gpus would have the same performance and it would be dependent on cpu.Nvidia makes GPUs that provide features that the CPU can not do or is very bad at,new games use the GPU as an CPU substitute (lower then 1,5Ghz console cores) figure it out.
I didn't start this thread, I saw the benchmark posted and I commented on it,a very weak dual core running sli gets 56FPS with ~60% usage which means that the game runs almost only on the GPU.
I sure hope Pascal brings proper multi-engine hardware and handles Async Compute & DX12 well else its going to be a messy few years for pc gaming fragmenting.
I suspect that it will though. Even with some basic support so that it won't drop in performance when AC is used.
Pascal is an improvement on maxwell though so we expect that.Unless NV have added dedicated ACE-like logical units, which we don't see on P100 where it would make the most sense to put them, then the scheduling will still need to be emulated in software. This could work for games which aren't demanding on the CPU but would be quite disastrous for any games which are demanding CPU side.
Looks to me like Volta is where it's at for Async + graphics.
Get one of those,play a game on it and tell us how that goes...Considering what GPU compute is, you might want to tell everybody buying GPU based supercomputers that they're doing it wrong and should use CPUs instead. I bet that'd go over great and you'd get hired on the spot. The only difference is that games are mixing in compute with their graphics.
How much is it then? 30% 120% ?Having 4 threads at ~60% on that Dual Core + HT processor is not 60% CPU usage.
Nvidia has less compute, period.I have figured it out. Nvidia is struggling to handle the new compute features in dx12 that add new graphics features to games.
I really wish they would fix this since I like new graphics features and this is already hurting Nvidia.
Have people looked into the aots game to find out if Nvidia still have graphics features turned off in order to run well?
As someone in this thread mentioned volumetric lighting uses compute. I guess Nvidia doesnt want to support such graphics features now in dx12? You want them on the cpu?
Remember the days when we recommended Nvidia gpus for people with weak cpus because of amd driver overhead? Well, with theelfs advice, maybe Nvidia can go down the same path this time.
Sent from my C6833 using Tapatalk