They'd sell off their entire inventory at $2, Nvidia would buy it all, and resell it and take AMD out and keep themselves....
Dumping your inventory first off is illegal like that.
Secondly, it's been proven to be a completely horrendous strategy so really don't see that.
If they were a sitting duck they'd sell off their company... lol....
Company's are in the business of making a profit, not "taking another company with them".
While I agree with you that about open source, etc.. I think your politics suck. You might want to leave them out of VC&G. Not only isn't this the place, but it could bias people's opinion of you, unnecessarily.
Nvidia does not really need to pay them anything, if its a feature that only works on AMD then no game will gona use it, we already seem that happening with Deferred Contexts on DX11, and thats was on Nvidia side that has most of the market. Same reason of why Mantle was DOA no matter if it was good or not. Same reason of why nvidia has to pay for them to use Gameworks, etc.
Nvidia does not really need to pay them anything, if its a feature that only works on AMD then no game will gona use it, we already seem that happening with Deferred Contexts on DX11, and thats was on Nvidia side that has most of the market. Same reason of why Mantle was DOA no matter if it was good or not. Same reason of why nvidia has to pay for them to use Gameworks, etc.
NVIDIA claims "full support" for DX12, but conveniently ignores that Maxwell is utterly incapable of performing asynchronous compute without heavy reliance on slow context switching.
GCN has supported async shading since its inception, and it did so because we hoped and expected that gaming would lean into these workloads heavily. Mantle, Vulkan and DX12 all do. The consoles do (with gusto). PC games are chock full of compute-driven effects.
Nvidia does not really need to pay them anything, if its a feature that only works on AMD then no game will gona use it, we already seem that happening with Deferred Contexts on DX11, and thats was on Nvidia side that has most of the market. Same reason of why Mantle was DOA no matter if it was good or not. Same reason of why nvidia has to pay for them to use Gameworks, etc.
How about games (and game engines) with Nvidia logos everywhere are those too going to be discounted as a fair assessment of hardware speed/capability?Ultimately, we have very little data and a lot of wishful thinking. The real games of dx12 will be all that matters, not some alpha benchmark that starts up with an AMD logo on it
As a developer if you had to make a cross-platform game, you would design your game engine around the abilities of the important market, consoles. You would therefore try to find ways to maximize their performance due to their weaker APUs. How would you achieve that? Well, the ACEs are there to be used, so you offload all your lighting, shadow processing and physics compute to operate in Async Mode.
Now when you port it over to your PC, you find that AMD runs it well, NV runs it less well. What option would you take that is the fastest and easiest?
Assuming:
1. Neutral title
2. AMD sponsored title
3. NV sponsored title
Btw, some food for thought, there are games where AMD runs better. Such as SOM, Civ, Hitman, Sleeping Dogs etc. It did not hurt their sales. Do you think the average gamers blame devs or NV/drivers when a game runs poorly on NV GPUs?
Is that program from Sebbi? Since he said he was thinking of doing something like that but was too busy before.
If anyones curious (and assuming this wasn't posted already) a guy from the beyond3D forums wrote a small program that basically runs a compute job alone, a graphics job alone, and then compute and graphics together ([URL="[/URL]).
Basically if your GPU supports async compute, then the time taken for compute and graphics together should be less than the sum of them run alone (but not less than the longest of the two, since they can be run in parallel with async compute). Without async compute the time taken to run compute and graphics together should be exactly equal to the sum of them run separately (since without async compute, they are run serially).
Early tests (from the linked beyond3D thread) indicate that Maxwell 2 (GTX 960 and GTX 970 have been tested) does in fact not support async compute (as far as this test is concerned anyway), since the time taken for compute+graphics, is basically the same as the sum of the two run separately.
Of course this is not an actual game, and there might be any number of bugs in it and/or unintended behavior, so take it with the usual grain of salt
No, he said he was too busy.
Do we know yet that the test is actually designed in a way that's conducive for async compute to improve performance?
It doesn't look like it considering none of the cards seem to benefit(Unless I'm reading the results wrong).
GCN completes the async (compute + graphics) task in the same amount of time as if it was running either graphics or compute only.
NV completes the asyc task in the time it takes for the combined graphics + compute, meaning it doesn't do it asynchronously, it does it in serial, one then the other. The time is added up.
That's about all we can say without knowing more about the software and what the actual task is.
CIV5 as far i recall was running best on NV because of deffered contexts im not sure if i missed anything here. The rest are barely a game, specially if it comes from ubisoft.
xone runs DX12, but the PS4 do not, not matter how similar it is, it needs for a developer to put in money, no company ever do that, specially howdays.
I was the same argument of why everyone whould use Mantle(GCN is on Consoles!!!!!!) and it did NOT happen! and it will not happen now.
Just stop for a moment and think, if a company has to put in money for it and it only works for a portion of the target market, they will not do it, i mean wake up, games are getting released before their development is complete.
If GameWorks where Nvidia only, no one whould use it, apart from the ones getting paid.
CIV5 as far i recall was running best on NV because of deffered contexts im not sure if i missed anything here. The rest are barely a game, specially if it comes from ubisoft.
So to resume, the problem is that nvidia cards cant do compute and graphics at the same time? There is anything else beyond that?
Oh right. Didn't notice. Still wondering why the times increase for each iteration.
The time doesn't actually increase for each iteration (thread), but rather it increases every 32 threads:
This is on a GTX 980 Ti. I don't know exactly what's going on, but I would imagine that it might have something to do with Nvidia's HyperQ implementation "only" being able to handle 32 threads at a time, whereas the 8 ACEs in GCN 1.1/1.2 can handle up to 64 threads at a time (I don't have a graph from an AMD GPU to compare to though).
Of course in actual games I doubt you would ever see anywhere near this many threads.
Civ BE I meant. The other games aren't Ubisoft. Get facts straight please.
My point was console devs will take the easiest route when they port to PC. It's far easier to enable/disable a certain feature than to write an alternate path that runs well on NV. For that to happen, NV would have to pay them, ie. GameWorks sponsorship.
Do you realise that is not a big deal right? how much OpenCL/DirectCompute is used in gaming? those has been around a long time.
So @Shivansps says Compute is no big deal in games.. devs say it is, thats why they wanted next-gen APIs.
Sony defnitely thinks its important, thats why they help AMD designed GCN with 8 ACEs, and PS4 games have already started to use Async Compute to give some some performance gains.
As said, when working with weaker hardware on consoles, devs will have to find methods to extract peak performance out of it, so ofc the statement from Oxide that console devs are pushing the console's ACEs in next-gen games makes perfect sense.
The question is what will be done for those heavy async compute console games when they get ported to PC? This is where you are free to speculate. But to say consoles don't use async compute, is wrong.