DIRECTX 12 API OVERHEAD TEST

jaffusmax

Just thought I would share my test results. Some stonking good results!
I am using WINDOWS 10 Pro Insider Preview 10166

DX11 Multi-threaded draw calls per second
2 530 168

DX11 Single-threaded draw calls per second
1 339 413

DX12 draw calls per second
18 323 089

That is a 624.18% & 1267.99% respective performance boost!!!!!!

http://www.3dmark.com/3dm/7755326?


API Overhead Test 1.2
Add to compare
Valid result
NVIDIA GeForce GTX 970(1x) and Intel Core i7-4770K

DX11 Multi-threaded draw calls per second
2 530 168

DX11 Single-threaded draw calls per second
1 339 413

DX12 draw calls per second
18 323 089

Mantle draw calls per second
0

Show Result Details
Run details
Hide result
Delete result
View benchmark run

Name
Edit
Description
Edit
User
JAFFUSMAXIMUS

Graphics Card

Graphics Card
NVIDIA GeForce GTX 970
Vendor
EVGA Corporation
# of cards
1
SLI / CrossFire
Off
Memory
4,096 MB
Core clock
1,165 MHz
Memory bus clock
1,753 MHz
Driver name
NVIDIA GeForce GTX 970
Driver version
10.18.13.5330

Processor

Processor
Intel Core i7-4770K
Reported stock core clock
3,500 MHz
Maximum turbo core clock
3,790 MHz
Physical / logical processors
1 / 8
# of cores
4
Package
LGA1150
Manufacturing process
22 nm
TDP
84 W

General

Operating system
64-bit Windows 8.1 (6.3.10166)
Motherboard
Gigabyte Technology Co., Ltd. H87-HD3
Memory
8,192 MB
Module 1
4,096 MB Crucial Technology DDR3 @ 800 MHz
Module 2
4,096 MB Crucial Technology DDR3 @ 800 MHz
Hard drive model
120 GB Samsung SSD 840 EVO 120GB

jumbobreakfast

Sounds good, hopefully the games developers can make use of it.

Zillah

Can someone describe what is happening for those of us who are not computer engineers, please?

Spear

Zillah wrote: »

Can someone describe what is happening for those of us who are not computer engineers, please?

Previous DirectX versions are horribly bottlenecked in their ability to issue instructions. The AMD Mantle API was an attempt to bypass part of this issue. But Microsoft have attempted to sort this in DX 12, hence the massive leap in the amount calls DX12 can issue compared to the previous release.

Kiith

New software, makes things go quicker

jumbobreakfast

DirectX 9 was developed in the age of single core CPUs where just one core talked to the graphics card even when multicore CPUs came out.

DX11 allows multiple CPU cores to perform calculations that help generate the next frame but they all speak to the GPU through just one core. (hence the bottleneck). This was probably easier for games developers to adopt coming from DX9.

DX12 allows all cores to talk to the GPU simultaneously.

Games developers still need to write their game engines differently to take advantage of DX12 and the synthetic benchmarks that are posted above only show the top speed when using very simple instructions. It does show the potential though

Zillah

So this is the first time that the software has been designed from the ground up to actually take advantage of multiple cores?

Kind of disappointing that it took this long

jaffusmax

If you have a CPU with integrated graphics and a GPU apparently the system can use both to compliment each other with DirectX12. The whole system becomes much more efficient and powerful. In theory it works but lets hope game devs make the most use out of it!