dx12 AMD vs nVidia의 부풀려진 오해와 진실

WCCFTech의 독점기사입니다.

출처 분명히 밝히고 퍼가주시기 바랍니다.

대부분 일반 유저들의 인식대로 암드만 비동기방식을 지원하는게 아니다.

엔비디아의 맥스웰 2.0 아키텍쳐(9xx시리즈 이상)도 31개의 비동기 쉐이더를 지원한다.

AMD GCN 1.0 – 1 Graphics Queue + 16 Compute Queues ( 7900 series, 7800 series , 280 series, 270 series, 370 series)

AMD GCN 1.0 – 1 Graphics Queue + 8 Compute Queues ( 7700 series, 250 series )

AMD GCN 1.1 – 1 Graphics Queue + 64 Compute Queues ( R9 290, R9 390 series )

AMD GCN 1.1 – 1 Graphics Queue + 16 Compute Queues ( HD 7790, R7 260 series, R7 360 )

AMD GCN 1.2 – 1 Graphics Queue + 64 Compute Queues ( R9 285/380, R9 Fury series, R9 Nano )

Nvidia Kepler – 1 Graphics, Mixed Mode not Supported (32 Pure Compute)

Nvidia Maxwell 1.0 – 1 Graphics Queue, Mixed Mode not Supported (32 Pure Compute)

Nvidia Maxwell 2.0 – 1 Graphics Queue + 31 Compute Queues

The various DirectX 12 feature levels

A feature level will however, usually require a similarly named API to access its features in their entirety. So basically, all GPUs conforming to FL 11_0 through FL 12_1 can run the DirectX 12 API completely and fully. The much hyped about advantage that is the reduction of CPU overhead – everyone will get that (provided you fall in the FL 11_0 to 12_1 band). The thing is however, these new GPU had new hardware features, something that only the DirectX 12 API can finally access: so new standards had to be created: namely FL 12_0 and 12_1.

Graphic cards supporting the following feature levels can run DirectX 12. The qualifying requirement for the particular feature level itself is also given:

FL 11_0: Supports Resource Binding Tier 1 and Tiled Resources Tier 1
FL 12_0: Supports Resource Binding Tier 2, Tiled Resources Tier 2 and Typed UAV Tier 1
FL 12_1: Conservative Rasterization Tier 1 and Raster Order Views (ROV)

Now that we know what the definitions are, here is the complete specification table of all IHVs with released hardware (including the latest Skylake iGPU and GM200):

IHV Hardware Specification Comparison


Architecture	GCN 1.0	GCN 1.1	GCN 1.2	Fermi	Kepler	Maxwell 1.0	Maxwell 2.0	Haswell	Broadwell	Skylake
Resource Binding	3	3	3	1	2	2	2	1	1	2
Conservative Rasterization	0	0	0	0	0	0	2	0	0	0
Tiled Resources	1	2	2	1	2	2	3	1	1	2
Raster Order Views	No	No	No	No	No	No	Yes	Yes	Yes	Yes
Typed UAV Formats	1	2	2	0	0	1	1	0	0	1
Feature Level Specification	11_1	12_0	12_0	11_0 +Partial11_1 support	11_0 +Partial11_1 support	12_0	12_1	11_1	11_1	12_0

각 제조사의 아키텍쳐별 dx12 지원 현황

암드의 현재 모든 라인업은 비동기 쉐이더를 지원하지만 엔당은 맥스웰 2.0에 와서야 31개의 Queue를 지원하기 시작했다. 이것도 암드의 64개에 비해서는 적지만 Maxwell 2.0의 dx12성능은 암드와 동일하거나 조금 좋다.


Radeon HD 7730	Cape Verde LE	384	24	8	GCN 1.0	3	1	No	1	11_1	1/8
Radeon HD 7750	Cape Verde PRO	512	32	16	GCN 1.0	3	1	No	1	11_1	1/8
Radeon HD 7770	Cape Verde XT	640	40	16	GCN 1.0	3	1	No	1	11_1	1/8
Radeon HD 7790	Bonaire XT	896	56	26	GCN 1.1	3	2	No	2	12_0	1/16
Radeon HD 7850	Pitcairn PRO	1024	64	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon HD 7870	Pitcairn XT	1280	80	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon HD 7870 XT	Tahiti LE	1536	96	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon HD 7950	Tahiti Pro	1792	112	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon HD 7970	Tahiti XT	2048	128	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon HD 7990	Tahiti XT x2	2048 x2	128 x2	32 x2	GCN 1.0	3	1	No	1	11_1	1/16 x2
Radeon R7 240	Oland Pro	320	20	8	GCN 1.1	3	1	No	1	12_0	1/8
Radeon R7 250	Oland XT	384	24	8	GCN 1.1	3	1	No	1	12_0	1/8
Radeon R7 260	Bonaire PRO	768	48	16	GCN 1.1	3	2	No	2	12_0	1/16
Radeon R7 260X	Bonaire XT	896	56	16	GCN 1.1	3	2	No	2	12_0	1/16
Radeon R7 265	Curacao PRO	1024	64	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon R7 360	Tobago	768	48	16	GCN 1.1	3	2	No	2	12_0	1/16
Radeon R7 370	Trinidad Pro	1024	64	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon R9 270X	Curacao XT	1280	80	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon R9 280	Tahiti PRO	1792	112	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon R9 280X	Tahiti XT2/XTL	2048	128	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon R9 285	Tonga PRO	1792	112	32	GCN 1.2	3	2	No	2	12_0	1/64
Radeon R9 290	Hawaii PRO	2560	160	64	GCN 1.1	3	2	No	2	12_0	1/64
Radeon R9 290X	Hawaii XT	2816	176	64	GCN 1.1	3	2	No	2	12_0	1/64
Radeon R9 295X2	Vesuvius	2816 x2	176 x2	64 x2	GCN 1.1	3	2	No	2	12_0	1/64 x2
Radeon R9 370X	Trinidad XT	1280	80	32	GCN 1.0	3	1	No	1	11_1	1/16
Radeon R9 380	Antigua	1792	112	32	GCN 1.2	3	2	No	2	12_0	1/64
Radeon R9 390	Grenada Pro	2560	160	64	GCN 1.1	3	2	No	2	12_0	1/64
Radeon R9 390X	Grenada XT	2816	176	64	GCN 1.1	3	2	No	2	12_0	1/64
Radeon R9 Fury	Fiji PRO	3584	224	64	GCN 1.2	3	2	No	2	12_0	1/64
Radeon R9 Fury X	Fiji XT	4096	256	64	GCN 1.2	3	2	No	2	12_0	1/64
Radeon R9 Nano	Fiji XT	4096	256	64	GCN 1.2	3	2	No	2	12_0	1/64

AMD DX12 Graphic Cards List Notes:

AMD 는 31개의 모델이 DX12를 지원한다.
R9 390, 390x는 GCN1.1아키이고 R9 285, 380, Fury, Fury X, Nano는 GCN 1.2이다. (????)


Geforce GT 440	GF108	96	16	4	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTS 450	GF106	192	32	16	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 460 SE	GF104	288	48	32	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 460	GF104	336	56	24	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 465	GF100	352	44	32	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 470	GF100	448	56	40	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 480	GF100	480	60	48	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce 520	GF119	48	8	4	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 545 DDR3	GF116	144	24	16	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 550 Ti	GF116	192	32	24	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 560	GF114	336	56	32	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 560 Ti	GF114	384	64	32	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 560 Ti 448 Cores Limited Edition	GF110	448	56	40	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 570	GF110	480	60	40	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 580	GF110	512	64	48	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GTX 590	GF110 (x2)	1024	128	96	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
Geforce GT 610	GF119	48	8	4	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
Geforce GT 620	GF108	96	16	4	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
Geforce GT 640	GF116	144	24	24	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
GeForce GT 730 (128b DDR3)	GF108	96	16	4	Fermi	1	0	1	No	0	D3D12 Support Delayed.	N/A
Geforce GT 630	GK107	192	16	16	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
Geforce GT 630 Rev 2	GK208	384	16	8	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
Geforce GT 640 Rev 2	Gk208	384	32	8	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 650	GK107	384	32	16	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 650 Ti	GK106	768	64	16	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 650 Ti Boost	GK106	768	64	24	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 660	GK106	960	80	24	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 660 Ti	GK104	1344	112	24	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
Geforce GTX 670	GK104	1344	112	32	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 680	GK104	1536	128	32	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 690	GK104 (x2)	3072	256	64	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GT 720	GK208	192	16	8	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GT 730 (GDDR5)	GK208	384	32	8	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GT 740	GK107	384	32	16	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 760	GK104	1152	96	32	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 760 Ti	GK104	1344	112	32	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 770	GK104	1536	128	32	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 780	GK110	2304	192	48	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 780 Ti	GK110	2880	240	48	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX Titan	GK110	2688	224	48	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX Titan Black	GK110	2880	240	48	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX Titan Z	GK110 (x2)	5760	480	96	Kepler	2	0	2	No	0	11_0 +Partial11_1 support	1
GeForce GTX 750	GM107	512	32	16	Maxwell 1st Generation	2	0	2	No	1	12_0	1
GeForce GTX 750 Ti	GM107	640	40	16	Maxwell 1st Generation	2	0	2	No	1	12_0	1
GeForce GTX 950	GM206	768	48	32	Maxwell 2nd Generation	2	2	3	Yes	1	12_1	1/31
GeForce GTX 960	GM206	1024	64	32	Maxwell 2nd Generation	2	2	3	Yes	1	12_1	1/31
GeForce GTX 970	GM204	1664	104	56	Maxwell 2nd Generation	2	2	3	Yes	1	12_1	1/31
GeForce GTX 980	GM204	2048	128	64	Maxwell 2nd Generation	2	2	3	Yes	1	12_1	1/31
GeForce GTX 980 Ti	GM200	2816	176	96	Maxwell 2nd Generation	2	2	3	Yes	1	12_1	1/31
GeForce GTX Titan X	GM200	3072	192	96	Maxwell 2nd Generation	2	2	3	Yes	1	12_1	1/31

Nvidia Graphic Cards List Notes:

Nvidia has a total of 50 base models promised with DX12 API support but only 30 base models currently supporting DX12 API. (Nvidia has yet to make on its promise of DX12 support for Fermi cards.
엔당은 50개모델이 dx12를 지원할거라 약속했었으나 현재는 Fermi를 제외한 30개의 모델만 지원한다.

So summarizing, all hardware vendors fully and completely support the DirectX 12 API.

No hardware vendor can claim 100% support of all hardware features and the differences are usually negligible in nature. If one is deciding by features observable by the end user and gaming experience, the vote might fall in favour of Nvidia with its Feature Level 12_1 support which will allow advanced illumination visual effects in next generation games. That said, there are ways to simulate the effects without much of a performance hit for Radeons as well. If we are talking about performance increase (in terms of untapped potential, not maximum potential) then an argument can be made for AMD with its ASync advantage.

요약하자면 암드든 엔당이든 인텔이든 dx12를 다 지원하는게 맞고 사용자의 체감 성능 차이만 존재할 뿐이다.

현재 nVidia만 FL12_1표준을 지원하고 있다.

암드 조차도 FL12_0까지 밖에 지원하지 못하고 있는 상황이다.