Assembling a computer from 40,000 v.4

-Sergey Viktorovich-
Belomorkanal
insel260
Eh, the current confuses me with one thing .... This one ... Forgive once again SV for giving work ..
http://users.playground.ru/862121/

SUSUL1
First of all, I didn't quite grasp the connection between multithreading and DX12.
Secondly, as you know, MS offers a free upgrade from Win7 to Win10. Now they earn a little differently.
Thirdly, MS does not compete with anyone in this case, and it makes no sense for anyone to cut fingers.

Belomorkanal
dobro4insky
Gentlemen, will you break off a lot if there is a photo? And then I like that, too, I remember being in such situations more than once. I casually mentioned my config and all sorts of comrades immediately attacked me to prove to me that my words are untenable, and I myself do not deserve a single drop of trust. True, after proofs, all these comrades look extremely untenable.

PS Dear SV, as far as I know, the iron forum does not moderate anymore, do not remember it in vain for nothing.

DartMaul
"Gentlemen, will you break off a lot if there is a photo?"
Pfft. As it is generally do not care.

DartMaul
I will break off very much ....)))
That's just .....
Spoiler

Belomorkanal
For what, then, such a hysteria?

dobro4insky
Not ready to read, to be honest. We've all been sitting here for a long time, and so we know what to expect from whom. So I read the posts of only individuals, don't blame me.

dobro4insky
Any bot. Eh bots got divorced right now, probably soon as in Fallout everyone will start paranoia, everyone will see bots in every user with zaregan in the late 15's and early 16's.

DartMaul
What kind of hysteria?
And I hope that I also get on the list of those individuals. If yes, then very nice ... not a drop of sarcasm.

DartMaul Do
you think he will post something ...? And please explain in Russian what it means
Spoiler
You have roccat, as far as we know ...

DartMaul
PS Dear SV, as far as I know, the iron forum does not moderate anymore, do not take it in vain for nothing.
Sorry, I didn't know.

DartMaul Casually
mentioned his config

But this Persian obviously did it casually. He came with the explicit intention of showing the members of the forum what they are all plebeians, and he is PEPKA EMPEROR.

dobro4insky
The mouse has a horizontal scroll wheel. If you bind the side scrolling to it (which, by the way, is written below as a comment to the option), you can set the step 1 of the scroll. For example, in Excel, if 3 is set, it will scroll by 3 columns to the side.

I don’t know who invested what, it doesn’t interest me much.

Belomorkanal
If I had not read, I would not have answered.
Spartak_13
Does it bother you so much?

DartMaul
there is no direct connection between multithreading and DX12, but the new API is declared as the most optimized for this business (I did not understand a lot from the presentation, but the manager drew beautiful graphics and focused on multithreading), especially for games. As for the competition from Microsoft. then he competes with himself, there is still not a small proportion of those. who sits on XP. and Windows 7 is still installed on 60% of computers. And just a few days ago, the generous shop was shut down, from now on the upgrade to Win10 will cost $ 50 full. and in order to sell a new version of Windows, it must destroy the old Windows. We saw very well how marketers and game manufacturers work when XP spread rot.
Literally two years will pass and we will see only DX12 in the requirements for games

SUSUL1
but the new API is declared as the most optimized for this case.
As I understand it, you do not quite understand what the graphics API is responsible for. There is multithreading in processor calculations (parallelization between cores) and there is graphic multithreading, the same CF / SLI. Separately, I would like to note that the ability to use GPUs from different manufacturers was not done in order to make AMD and Nvidia friends, but in order to use both discrete and integrated graphics (which is now in most laptops and PCs). Hence the optimization for multi-GPU.
The great thing about DX12 is that it provides a simpler and more flexible GPU environment. In fact, it takes over some of the functions that game drivers and engines currently perform. I would venture to suggest that the presentation you saw most likely views DX12 as a software development environment rather than a customer product.

from now on, the upgrade to Win10 will cost 50 full $. I
went to the MS website specially - I don't see anything like that there. Until now, they offer to upgrade for free (thank God, they don’t offer me, I have an enterprise Windows)
But even so, $ 50 for an upgrade is just prettier than $ 200 for a new Win10.

You don't have to tell me how marketing works, I have a MSU diploma on this topic)
As for MS's interest in the transition of users to the new axis, the reason is obvious.
1) Win10 is a cross platform axis. MS are trying to create a complete software environment for all devices from PCs to smartphones. This is essentially Apple's answer.
2) Win10 is tailored for the use of various MS services, mainly cloud ones. This is the collection of analytics and additional binding of the user to the company's products. If you have a PC on win10 and say a laptop or tablet, the likelihood that you will buy a smartphone on win10 multiplies.
3) It is much easier to support one axis than several. And it is extremely difficult to switch developers' attention to the new axis when most Windows users are using older versions.

DartMaul
I 'm so worried that I can't eat.
He slightly distorted the famous phrase.

DartMaul
Switching developers' attention is as easy as shelling pears , a suitcase of money brought to the main office instantly attracts their attention. And you don't need to mention everyone, it's enough to fund Catlofield4 or Ass Ass in Cread8 and the fans will voluntarily, for free and with foam at the mouth, steam up how cool it is to sit on Windows10 when everyone else is playing 40FPS or should buy ASUS AresII =)
I think that all the new chips DX12 could be implemented in Win7, but programmers want to eat well =) So Windows should be sold ...
PS
Where did you find Windows for $ 200 ?!
RUB 7900 there is a home ... and where is it still sold at the old price
or do you fundamentally need a pro version?

SUSUL1
200 $ on MS Win10 Pro Box
11500 on softline.

MS does not fund the development of the games you mentioned, and fans will eat what they are offered.
If you think that DX12 compared to DX11 gives some stunning increase in fps, you are greatly mistaken.
In general, this myth, which has grown by leaps and bounds, that DX12 is capable of magically drastically affecting the performance in games, is nothing more than an empty hope multiplied by a lack of understanding of what the new API is. DX12 does not increase performance. Moreover, in terms of the operation of the video card itself, it is no different from DX11. And if a card with the new API will do exactly the same as with the old one, why would it suddenly become faster?

http://www.amd.com/ru-ru/innovations/software-technologies/directx12
Here is a link where it is popular for users, it is described what DX12 gives. Unlike most of my links, this one is in Russian!
I want to immediately comment on the content on the link in order to correctly place accents:
1) In the first picture, you can see an explanation of what you did not understand in the presentation from MS. DX12, unlike the 11th, is multi-threaded. That is, Direct3D (DX command operator) will now use multi-core, which will certainly increase its performance. As a result, the delay in the message of the command to the video card will become less.
Here I want to immediately mention a number of nuances: the fact that the engine starts rendering multithreaded does not make the algorithm shorter, but rather the opposite. If you play a processor-dependent game in which you have 90% + CPU utilization, you will not feel any increase. If you remember, I at one time repeatedly said that percent is always a bottleneck in the system, no matter how fast it is. This is another example. People who understand this, know how to tune the system, bought a normal percentage and were able to overclock it correctly, have largely exhausted the optimization potential in this area. The only ones to whom it really will give an increase are people with 3-4 vidyahs.
2) As for asynchronous shaders and other crap: do you know how hypertrading works in Intel? The meaning is exactly the same. Use the time while the kernel is waiting for data to be loaded to perform other tasks. Probably someone will be useful, but taking into account 12MB of cache in the process, 4-channel overclocked low-latency memory, PCI-Ev3 x16 for each card and 350 + GB / s memory bandwidth of the cards themselves, this is unlikely to help me.
3) As for draw call: it's just a drawing command. And it does not mean at all that the video card, even with the current limit for this operation, is capable of performing more of them.

To summarize: if you are using some AMD 8-core, without overclocking, and you have a video card of at least 980Ti, or better 2, I congratulate you - DX12 will save you (although of course it will not add common sense to you, which by the way is very regrettable) ...
If you have 5960 and 4 980Ti, then the increase will probably be noticeable as well. But people with such assemblies usually do not play at 1080p, which means that most of the power of the views is fed to rendering ultra-high resolutions, at which the load on the graphics increases multiple times, but in general it does not affect the percentage.
Of course, the promise to reduce delays is very captivating, which seems to be supposed to cure the so infuriating spikes in games. But here there is one unpleasant but: to paraphrase the well-known saying - for each cunning API there is a programmer with shitty code. And given that 2015 has truly become the year of non-working games, we will all lag, regardless of the power of the computers and DX versions.

As for draw call: it's just a draw command. And it does not mean at all that the video card, even with the current limit for this operation, is capable of performing more of them.
in general, it means if there will be a greater supercapitalism of disconnected computations or more actions performed at the level of video card subroutines without accessing the CPU, this again corresponds to multiparasitism. otherwise, there are many shader processors in modern video cards and they are capable of performing many complex mathematical transformations. plus the memory becomes very fast and it becomes possible to use it more as a cache. however, most likely everything will be divided by zero since almost all advanced shaders are associated with raster processors.
There will be no spoilers and gains in the first place because of the sequence of processing by the engine, for example, a particular example - grass. as a result, the engine is forced to calculate all overlaps and all operations using highly linear algorithms. and your own grass smokes are amorphous materials with complex lighting effects, the only place where it should be drafted. without this, the gain is either 0 or negative with an error, because the api compiler becomes more complicated and longer.
the second special case is actually an increase in the number of identical objects of different variations bypassing the processor. to wait a long time for the first, then all sorts of 980s will most likely turn into a pumpkin and the grass in 3D fog with rays will make it dravukaly kala and this is not necessary for consoles, because consoles do not experience problems with dravukals in the ratio of processor and shader performance.
the final reconciliation is a fierce polygonal compaction. a lot of polygon = a lot of drafts and the computer turns into feces. but for console reasons, 300k polygon models are unrealizable.
the main thing is that what could increase the productivity of the peki is actually unrealizable and the potential is not revealed.
As for the SLI and MGPU, I just suppose on the contrary that their efficiency will fall due to the peculiarities of compiling task packages for them. or maybe such a disgrace is going on that some developer can even fix such a compilation technically impracticable because of the GPU procedural generation, and it will be similar to the current state of ogl in terms of support for MGPU. in order for GPU 2 to find out what is happening on another GPU, it needs to send data that is not predictable back to the CPU, and this breaks the whole point of such actions and reduces the bandwidth or such algorithms are tedious where the GPU knows exactly what to do and it is known that two reverse unpredictable actions do not overlap.
in the first couples, a significant plus in performance will be received only by the radeon red-line cross-fines, because they say that the cross works more often using the 1hpu = 1 frame method, and this ineffectively uses the processor time and results in micro delays. so 2-4 card systems should probably run 20% faster at once only due to this
for this, due to the described reasons, an ordinary household baker with an i7 and a game video card grieves and twists or something like a 100% mmm increase in FPS deposits should not be expected. mk will not make a significant difference for him. and when it will be necessary in any case to think about Windows with dx12 and a new pitch. I even question that pitch now suffers from dravkal problems because I don't quite understand what exactly is meant by them. because even my pitch is capable of squeezing out a monstrous amount of highly polygonal objects with current textures while reducing the load on the GPU, of course, the GPU is always a weak point.
the only thing that is interesting is some sort of memory optimization. but here I can't say anything about pitch

DartMaul
Isn't the algorithm for calculating visible objects changed in DH12?
Advertising benchmarks show exactly the increase in FPS, they say, in DX12, a simplified calculation of objects that do not fall into the frame is implemented, which allows you to increase the detail and place more objects.
As in this benchmark

dobro4insky
Well, of course I don’t believe it. Do I look like a naive simpleton?

SUSUL1
Dear, if you want to continue your conversation with me, please stick your ear out of the armored train, otherwise I kind of wrote a wall, especially for you, with explanations, and you are the same to me again.
DX12 does not contain any changes in the video algorithm. The changes apply only to the part of the engine that is rendered on the process. If the vidyaha (hi) were limited by the productivity of the processor, then there will be an increase, if not, then no.

The test used the most simple objects (in fact, one-color cubes) that the vidyakha can draw in almost unlimited quantities, because the resource intensity of such operations tends to zero. This was done on purpose so that it was the percent that became the bottleneck of the algorithm. In real games, this is far from always the case, because there are 1) uberHD textures with a resolution of 100500x100500 2) models with 100500 polygons (especially if you still use tessellation), anti-aliasing (sometimes using the principle of multisampling), shadows, anti-aliasing of shadows, and a lot what else that guys usually turn off because they don't pull, and then they complain about not poor graphene.

PS I will soon cross you off the list of read authors, if you don’t stop writing nonsense.

SUSUL1
dx12 does not change the algorithm for calculating visible objects, there are
such technologies and all kinds of streaming technologies have probably been from the time of sruzis 2, and all the advanced engines have long shown detail only in the visible part of the screen, starting from the eclipse and ending with any sorting and cleaning of objects covered by other objects.
and in the benchmark, rather what I wrote about pitch - procedural GPU generation of a pair of houses by the power of a video card.