I think DX12 has done far more harm than good and has actually held gaming back.
When DX12 was announced there were plenty of predictions about the benefits but in practice we have seen very little.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD VEGA confirmed for 2017 H1
- Thread starter Doom112
- Start date
- Status
I think DX12 has done far more harm than good and has actually held gaming back.
When DX12 was announced there were plenty of predictions about the benefits but in practice we have seen very little.
I fully agree that in the past it did not make sense, because as Intel and Nvidia proved there was a LOT more performance left to extract from a single core and a clean pipeline. AMD was going wide as if there was no more depth left, but there was tons of it.
AMD was stuck with Bulldozer architecture so really it was sort of coming in from both ways: low-clock cores with crap IPC to fit the small power envelope, just lots of them. Add async capability to the GPU and voila! Here's your console.
Meanwhile Intel was increasing IPC, using newer processes to clock higher, adding a second thread on the same core rather than an extra core... AMD got stuck at 32nm and it was game over for them.
But NOW, now it may actually make sense. Everyone admits we will get stuck at 7nm for years and years, so the free lunch is over. Just as AMD had to go wide, so will everyone. That's why you have DX12 now. That's why AMD is working towards Navi. That's why even Intel says the future is more cores, not higher clocks.
So I wouldn't count that out. Especially with maturing game engines, devs will no longer care about the complexity of the pipeline so much. Furthermore they will learn and adjust. They always do.
One of the problems is - there are relatively few purely console game development studios and a lot of the revenue is driven by a small number of headline titles. while conversely on the PC there is a larger number of developers many of whom never produce games for console and the money comes from a wider variety of titles. Partly helped by games like Overwatch the PC re-emerged as a gaming platform over 2016 and looks like potentially continuing to grow in the future.
Problem is there is only so wide you can go with games - even forced to pay more attention to threaded functionality, etc. there is a lot of game logic and so on that can ever only work in serial. I think developers are going to be forced to innovate more than anything else and pay a lot more attention to the efficiency of their implementation from the word go rather than going back and optimising as much as possible once the game is emerging from beta state.
EDIT: IMO explicit multi adapter is going to be the next big thing - there is so much potential for farming out tasks in that way that you simply can't do on one GPU but for some reason developers seem reluctant to touch it - even as simple as rendering the UI on one GPU while rendering the rest of the game on another would make quite a difference.
Last edited:
I'm pretty sure AMD aren't that happy with RX480 sales, The 1060 is outselling it 4 to 1:
http://www.game-debate.com/news/220...060-outsells-rx-480-4-to-1-shows-steam-survey
Agreed. The move to lower level API on the PC is a reversal of the trends in software engineering of the last 50 years.Consoles are unique in that they are relatively under powered, have a fixed hardware over the life of the produce and will stay that way for 4-6 years so low level optimizations are much easier and much more beneficial. Even so most console games have absolutely zero benefits going to a lower level API, its really soemthign restricted to the bigger studios on some AAA games and engines.
The problem with DX12 on the PC is it after doing a lot of work re-writing your back-end you will almost certainly find it slower than DX11. Only by very careful optimization will you bring back performance. And the large potential gains only exists in certain scenarios. If your engine is designed around the known limitations DX11 then you don't see many gains going to DX12.You also have a lot of ways to screw up Even AMD's poster child DX12 games show lots of issues when changing architectures performance varies wildly in Ashes etc.
DX and OGL definitely needed a complete face lift but going low level was really not helping many developers.
I know someone who bought a 1060 over a 480 due to it being faster in some DX11 games like GTA5. He was not too happy when he saw the 480 beat it in newer games like Titanfall2 and COD.
RX480 sales are ok even with the 1060 pulling a lot of sales that could have gone AMD's way but they certainly could have made up more ground. They are kind of treading water healthy though rather than pulling in the money you need to invest going forward to keep growing.
I do a little dabbling - one thing I quite like is space type games and I've a little engine for rendering lots of asteroids, etc. where DX12 actually could provide a huge performance increase like 9-10x the performance - but I pretty much face palmed when I looked into what I needed to do to unlock that :|
Last edited:
I suspect a mod will clean up this thread due to again D.P does off topic derails again.
Not sure why he isnt banned for 6 months or so for this repeated behavior.
Vega didnt run max clocks to beat the 1080 in battlefront or Doom.
Its run on un-mature drivers still.
It has with the 2 weeks old card demoed in CES shown AMD has a huge winner on their hands.
So, Vega beats Nvidias cards easily.
Starbound power coming to you.
Time to sell your 1080 and buy the real deal.
Not sure why he isnt banned for 6 months or so for this repeated behavior.
Vega didnt run max clocks to beat the 1080 in battlefront or Doom.
Its run on un-mature drivers still.
It has with the 2 weeks old card demoed in CES shown AMD has a huge winner on their hands.
So, Vega beats Nvidias cards easily.
Starbound power coming to you.
Time to sell your 1080 and buy the real deal.
The thing about low level APIs is that they are quite thin and leave all options open. The first thing you do when faced with such an API is write wrappers that make sense. Once someone gets it 'right' the others just ...build on top of that.
Game engines will provide the framework for developers to start writing quick and dirty code. You will have 'flavours' of DX12 used by different studios, which a small team of core devs evolving the engine and everyone else will be like... scripting on top of Unity.
It's easier to build on top of a low-level API, than to dissect a coarse one and code around its hidden internals to get the most out of it.
So I think of it as the beginning of the face lift: let's go back to the metal and start building. We'll see who makes the most reusable easy engine to write for.
Cos it's G-force mate. It doesn't matter how good anything else is...G-force.
That must be Gregster installing 1060s in every build he's asked to advise on!
The last 2 builds I did used a 470 and 390. Not built anything in a while though but they were the best choices for budgets.
The 470 is so underestimated. Overclocks ok for an AMD card too.
From what I've seen though no one seems that interested in wrapping either DX12 or Vulkan extensively. Ironically it seems nVidia is leading the way on wrapping Vulkan but that is probably because by doing so they can lead developers in terms of techniques used that play into their hands as it is an area they are on the back foot otherwise.
Permabanned
- Joined
- 12 Sep 2013
- Posts
- 9,221
- Location
- Knowhere
That's to be expected, Has Just Cause 3 got it's DX12 patch yet? I read that Nvidia were involved with developing it and it was going down the rasterizer route which is missing from AMD cards so it was a case of DX12 but only the features we have not the ones we don't, leading to yet more fragmenting of the market. I was glad to read that that was being added to Vega.
Or maybe it's because of the whole power-debacle and because the GTX 1060 is still cheaper (in a lot of countries), and outperformed the RX 480 in the most popular games until recently?
TBH steam charts doesn't worth ****.
If 1060/70/80 vs RX470/80 sales would be like steam shows, NV would have been grabbed back a big chunk of market share in Q3....but they didn't...they grabbed 0.8% compared to Q2.
As I demonstrated before that doesn't have to happen - if a good proportion of people moving to the 1060 are pre-existing nVidia customers (and the decrease in some other GPUs in the steam survey tentatively supports that) and/or the uptake of new customers that aren't pre-existing AMD or nVidia customers also roughly splits as before it doesn't affect overall market share.
Steam's survey might not be the most accurate - but when people like JPR have published numbers that are possible to compare before they've been broadly right - people arguing against it tend to be people who don't like what it says :S
Last edited:
Market share is based on GPU shipments. Things you mention (people moving from existing NV cards to newer) doesn't affect it.
Err, he didn't actually say that. From the transcript -
http://wccftech.com/amd-confirms-vega-10-doom-4k60fps-vulkan-demo-outperform-gtx-1080-opengl/
Please, lets keep the Hypetrain in the siding
- Status