Everything you need to know about directx 12 (we include benchmark)

First of all, there are two important things that have been discussed a lot lately, which are their “ feature level ” and the “ tier ” to which they belong. How does this affect us?

In order to guarantee that Dx12 would work on a wide range of equipment, Microsoft and its partners agreed to divide the support levels for resource-binding into 3, that is, the resource model, thus:
- Tier 1: Intel Haswell, Broadwell and Nvidia Fermi. Tier2: Nvidia Kepler, Maxwell 1.0 and Maxwell 2.0. Tier3: AMD GCN 1.0, GCN 1.1 and GCN 1.2.
Each level is the super-set of the previous one, that is, Tier 1 hardware comes with the strongest constraints on the resource model, and Tier 3 by contrast has no limitations, while Tier 2 represents intermediate level of constrictions. How easy it would have been if it had been the other way around, right? Tier1 the one who has everything, and so on, but not ... Complicating life is his fate. Therefore and summing up, Amd Tier3 is the one with no restrictions, Tier2 the one with certain limitations and Tier1 because that, the one that more or better said the one with the most "basic" support.

Lately there is a lot of talk on the Internet if they all support everything (like the Maxwells) or if Amd supports or not all the features of Dx12 and when it was said yes, it was misinterpreted, meaning that it has no limitation in what has been seen so far but that very different are the "feature level", and now we are going to see what the feature level are like since architectures or cards support it ... Why wouldn't you give us poetry huh ?.
Continuing with the theme, in addition to the Tiers, Dx12 has different “feature levels”, that is, levels of operation, and there are four to date, which also each have different characteristics and hardware that needs support. These “feature levels” do not have to be linked to the Tiers and they have, as it were, a more secondary role than what has been seen above, which have the important and main rendering characteristics.

Some of these "feature levels" are not covered by even the highest Tier3, so this makes it an individual feature, with hardware (the graphics card in question) being the determining factor.
How do we know what “feature levels” each hardware has? We identify them like this:
- Feature Level 11 -> Nvidia Fermi, Kepler, Maxwell 1.0. Feature Level 11.1 -> AMD GCN 1.0, Intel Haswell and Broadwell. Feature Level 12.0 -> AMD GCN 1.1 and 1.2 GCN. Feature Leve 12.1 -> Nvidia Maxwell 2.0
We have messed with you right? It is not for less, we have in mind Tiers, Feature Leves and different graphics and no games ... Great! How do we see it? Very simple, we are going to list first which cards correspond to which architecture.
- Nvidia Fermi: All those that carry the chip at the beginning of their model, "GF", such as GF117, 110, 100 and all those in the middle that, as you well know and speaking of the most popular, would be GT450, GTX460, 470, 560 and 580 among others.
- Nvidia Kepler: Like the GF in this case they are called GK, if it is like thinking of "Gpu Kepler". Not all the 600 or 700 series of Nvidia are, there are some that are refried from GF that is to say from Fermi, therefore it is convenient to make sure but as an example, we will tell you that they include among others the popular GTX650, 660, 670, 680, 760, 770, 780 and Ti.
- Nvidia Maxwell and Maxwell 2. 0: Here the list is shorter, Maxwell 1.0 was born with the GTX750 and 750Ti, which as you can see belongs to the 700 series without being Kepler, and they are identified with GM107 and 108. In Maxwell 2.0 there are fewer cards that there are, starting from the new GTX950 and we go from this to the 960, 970, 980 and Ti as well as the Titan X and later.
- AMD GCN 1.0: Being a little easier to know which affects, the AMD 7000 series from 7350 onwards to 7990 are GCN 1.0 architecture (except 7790 which is 1.1). You have to be careful, because in the following series like the R3, R7 and R9 there are "intruders" or rather rehash, who have this architecture, such as the 270, 280X etc. They are based on the Tahiti, Pitcairn, Curacao, Cape Verde chip…
- AMD GCN 1.1 and 1.2 : These are supported by the next generation which is more modern, such as the R7 260 and 260X which are 1.1, the 7790, and those based in Hawaii such as the 290, 290X and Apus Kaveri, which are based on the Sea Islands architecture. The 1.2 are more rare that are based on the Volcanic Islands architecture, such as 285 or 380, and the new Fury based in Fiji. The 300 series, many of them are 1.0 and 1.1, you should be careful not to confuse them, such as the 390 and 390x, which are 1.1 or the 370, which is 1.0. They couldn't have done better (sarcasm).
Well, we already have the cards included and positioned with their specific support, but in What really improves Dx12 ?, Let's summarize it clearly and easily.
- Reduce the bottleneck in the cpus, a feature that in Dx11 is really saturated. Increase the scaling by having a greater number of cores in the CPU, finally. Greater control for the developer. Api efficiency similar to a console, that is to say, which will have a wider and closer control hardware - software (games). All the functionalities of Dx11 conserve them.
This is let's say the main feature or base of Dx12, therefore, if we have a card that does not support 100 & minor features, can we use Dx12? Yeah But ... no buts, Si es Si. The differences will be found when as time passes games that support these cards arrive, meanwhile, it can be emulated by software. The safest and most plausible is that the games that arrive this 2015 and the following 2016 are based and stick to the base of Dx12 since many of them will be post-launch patches being ported to the new Api, similar to what was experienced with Battlefield 4 and Mantle, who released his support a few months later.
How do we measure the real difference between Dx11 and 12?
What card do I buy?
And what games come to us?

With the arrival of Windows 10 and as has been usual in Microsoft with each operating system, comes its latest graphics API that gives rise to a new generation of graphics cards, and especially games. Directx12 is the new Api, this time low-level and with a new environment, as well as Vulkan - which we'll talk about later - will invade our world.

Probably you will find other technical articles, of great value but difficult to understand because many things we will not get to understand them and ordinary people need to know in a simple and clear way, if their team or whoever needs to update, can with these new Apis. If you feel like this, this is your place! Do not miss it!.

First of all, there are two important things that have been discussed a lot lately, which are their “ feature level ” and the “ tier ” to which they belong. How does this affect us?

In order to guarantee that Dx12 would work on a wide range of equipment, Microsoft and its partners agreed to divide the support levels for resource-binding into 3, that is, the resource model, thus:

Tier 1: Intel Haswell, Broadwell and Nvidia Fermi. Tier2: Nvidia Kepler, Maxwell 1.0 and Maxwell 2.0. Tier3: AMD GCN 1.0, GCN 1.1 and GCN 1.2.

Each level is the super-set of the previous one, that is, Tier 1 hardware comes with the strongest constraints on the resource model, and Tier 3 by contrast has no limitations, while Tier 2 represents intermediate level of constrictions. How easy it would have been if it had been the other way around, right? Tier1 the one who has everything, and so on, but not… Complicating life is his fate. Therefore and summing up, Amd Tier3 is the one with no restrictions, Tier2 the one with certain limitations and Tier1 because that, the one that more or better said the one with the most "basic" support.

Lately there is a lot of talk on the Internet if they all support everything (like the Maxwells) or if Amd supports or not all the features of Dx12 and when it was said yes, it was misinterpreted, meaning that it has no limitation in what has been seen so far but that very different are the "feature level", and now we are going to see what the feature level are like since architectures or cards support it… Why wouldn't you give us poetry huh ?.

Continuing with the theme, in addition to the Tiers, Dx12 has different “feature levels”, that is, levels of operation, and there are four to date, which also each have different characteristics and hardware that needs support. These “feature levels” do not have to be linked to the Tiers and they have, as it were, a more secondary role than what has been seen above, which have the important and main rendering characteristics.

Some of these "feature levels" are not covered by even the highest Tier3, so this makes it an individual feature, with hardware (the graphics card in question) being the determining factor.

How do we know what “feature levels” each hardware has? We identify them like this:

Feature Level 11 -> Nvidia Fermi, Kepler, Maxwell 1.0. Feature Level 11.1 -> AMD GCN 1.0, Intel Haswell and Broadwell. Feature Level 12.0 -> AMD GCN 1.1 and 1.2 GCN. Feature Leve 12.1 -> Nvidia Maxwell 2.0

We have messed with you right? It is not for less, we have in mind Tiers, Feature Leves and different graphics and no games… Great! How do we see it? Very simple, we are going to list first which cards correspond to which architecture.

- Nvidia Fermi: All those that carry the chip at the beginning of their model, "GF", such as GF117, 110, 100 and all those in the middle that, as you well know and speaking of the most popular, would be GT450, GTX460, 470, 560 and 580 among others.

- Nvidia Kepler: Like the GF in this case they are called GK, if it is like thinking of "Gpu Kepler". Not all the 600 or 700 series of Nvidia are, there are some that are refried from GF that is to say from Fermi, therefore it is convenient to make sure but as an example, we will tell you that they include among others the popular GTX650, 660, 670, 680, 760, 770, 780 and Ti.

- Nvidia Maxwell and Maxwell 2. 0: Here the list is shorter, Maxwell 1.0 was born with the GTX750 and 750Ti, which as you can see belongs to the 700 series without being Kepler, and they are identified with GM107 and 108. In Maxwell 2.0 there are fewer cards that there are, starting from the new GTX950 and we go from this to the 960, 970, 980 and Ti as well as the Titan X and later.

- AMD GCN 1.0: Being a little easier to know which affects, the AMD 7000 series from 7350 onwards to 7990 are GCN 1.0 architecture (except 7790 which is 1.1). You have to be careful, because in the following series like the R3, R7 and R9 there are "intruders" or rather rehash, who have this architecture, such as the 270, 280X etc. They are based on the Tahiti, Pitcairn, Curacao, Cape Verde chip…

- AMD GCN 1.1 and 1.2: These are supported by the next generation which is more modern, such as the R7 260 and 260X which are 1.1, the 7790, and those based in Hawaii such as the 290, 290X and Apus Kaveri, which are based on the Sea Islands architecture. The 1.2 are more rare that are based on the Volcanic Islands architecture, such as 285 or 380, and the new Fury based in Fiji. The 300 series, many of them are 1.0 and 1.1, you should be careful not to confuse them, such as the 390 and 390x, which are 1.1 or the 370, which is 1.0. They couldn't have done better (sarcasm).

Well, we already have the cards included and positioned with their specific support, but in What really improves Dx12 ?, Let's summarize it clearly and easily.

Reduce the bottleneck in the cpus, a feature that in Dx11 is really saturated. Increase the scaling by having a greater number of cores in the CPU, finally. Greater control for the developer. Api efficiency similar to a console, that is to say, which will have a wider and closer control hardware - software (games). All the functionalities of Dx11 conserve them.

This is let's say the main feature or base of Dx12, therefore, if we have a card that does not support 100 & minor features, can we use Dx12? Yeah But… no buts, Si es Si. The differences will be found when as time passes games that support these cards arrive, meanwhile, it can be emulated by software. The safest and most plausible is that the games that arrive this 2015 and the following 2016 are based and stick to the base of Dx12 since many of them will be post-launch patches being ported to the new Api, similar to what was experienced with Battlefield 4 and Mantle, who released his support a few months later.

How do we measure the real difference between Dx11 and 12?

I know that many of you may want to talk about each “feature level” but since there are still no games that describe them or details of what game will be implemented, we are going to focus on the main improvements that are the enormous release of the neck of bottle that can generate the gpu, the cpu, which is where Microsoft is working the most, to make it more flexible and take advantage of both the processor and the graphics as a whole. Also, this is how this article is, a quick and easy look to get ready.

To do this we have prepared a table with the new game (still in a pre-beta state) Ashes of the Singularity to see what benefit we get by going from Directx 11 to 12, and the 3DMark Vantage the number of drawcalls or "calls" that can be made the cpu to the graphics card.

Although it is being criticized a lot (and it is not for less but we will not enter morbid), it seems to me a good benchmark since the staging of objects, ships, plotters, sound, graphic effects on screen is very high, and it is perfect to quantify the improvement in Fps which is what ultimately interests us. Obviously we have used the R9 390x of our team and a 4690K @ 4400Mhz under Windows 10.

WE RECOMMEND YOU We explain why AMD improves more than Nvidia when moving to DirectX 12

This is the bench of the section that encompasses the entire PC, mostly using Gpu.

And finally the CPU test, to see how the Api releases the performance of the processor.

As we see the improvement of Fps is drastic, and not only that but as we say, the general quality by putting more units on the screen, effects and other things. It is the only tangible proof that we have but it should be taken with tweezers since it may not turn out that way in all games, since each one has differences in its style, whether it is arcade, role-playing, shooter, etc., but if the improvements go this way We will be facing one of the biggest changes in recent years.

Now, we are going to see how the calls from the cpu to the gpu affect, based on the test that 3DMark brings, with the same graph and equipment.

Yes, as we can see, the number of calls that Dx12 can execute against Dx11 cannot be compared. But what are drawcalls ? As a simple explanation I will tell you that they are the total of "meshes" drawn after the batching process, and this is the process where the engine combines the rendering of different objects in a single drawcall to try to avoid overloading the cpu, and as we see in the graph, the difference is astonishing.

Well and now let's finally get out of technicalities and focus on what matters to us, the games.

What card do I buy?

If you still do not have any graphics card compatible with these characteristics or simply want to upgrade to something better, we recommend that you buy what we always recommend, quality / price. Not all the games that are going to come are Dx12 and most of what there are to date are Dx9 or 11, therefore it is convenient to have a fixed price line and start from that base.

There are always alternatives of less than € 200 such as the Gtx 950 or AMD R7 370, above these the GTX960 and AMD R9 380, and so on, always having something balanced and according to our team and needs. It is still very early in my opinion to buy thinking about the feature levels or if I need X or Y based on a game since there is still nothing clear or explained except in what we have concentrated on which is roughly the release of the cpu and the improvement of fps passing from one api to another to put it in the simplest way.

With this in mind, all the gpus in the current market and probably many of you have / have a base support for Dx12 which is the main feature, and therefore the one that interests us until we see future games.

And what games come to us?

Gear of War Ultimate

In the remainder of 2015, there will be few games that will support Directx12 at first, and the first of all (not counting Ashes since it is pre-beta) is Fable Legends, which will be released for PC and Xbox. One in October.

According to the sources that scamper the internet, the improvements from Dx11 to 12 are strong, giving Fps rates in Dx11 of 43fps while in Dx12 we go to 53Fps, an improvement more than worthy for a start where, in addition, the minimums improve in higher proportion than the average.

On the other hand, by the end of the year and specifically in December, the new Hitman will arrive.

Where we are really interested in the subject is 2016 where the list is lengthened, where the Ark Survival Evolved will receive a patch which is gaining adherents every week that passes, in February 2016 the new Deus Ex Manking Divided, Sea of Thieves, Star Citizen comes out, Gear of War Ultimate, Day Z, Arma 3 and although it is not yet known which game will be the first, DICE, creator of games like Battlefield, already has its Frostbite 3 engine running on Dx12, although I would bet that it will be Battlefront, given the expectation it is generating and being a multiplayer title, perhaps where all this paraphernalia could be used the most.

Anyway, I think I did not leave anything in the pipeline and now you are a little calmer, so we say goodbye here and soon you will have an article about Vulkan, the new Api from the parents of OpenGL and belonging to the Kronos group, which comes to Let's be the Dx12 “competition” as it has been these long years and how briefly Mantle was, the one we think started to make everyone nervous.

We leave you with some images of the titles mentioned behind and we said goodbye !.