AMD Radeon RX Vega gets as many stream processors as R9 Fury X

After leaving the performance segment on walkover AMD is about to launch the graphics architecture Vega, which will seriously take up the fight against Nvidia. The architectural details were presented at the beginning of the year, and on several occasions the company has stated that the launch will take place during the second quarter.

Contents hide

1 Specifications: – AMD Radeon RX Vega

2 Related posts:

Ahead of the launch of Vega, one is released Direct Rendering ManagerUpdate (DRM) for Linux, which adds architecture support. This means that several details about the specifications are strengthened and that Vega in many respects is reminiscent of AMD’s long-distance Radeon R9 Fury X with the graphics circuit Fiji.

Vega 10	Fiji
case CHIP_VEGA10: adev->gfx.config.max_shader_engines = 4; adev->gfx.config.max_tile_pipes = 8; adev->gfx.config.max_cu_per_sh = 16; adev->gfx.config.max_sh_per_se = 1; adev->gfx.config.max_backends_per_se = 4; adev->gfx.config.max_texture_channel_caches = 16; adev->gfx.config.max_gprs = 256; adev->gfx.config.max_gs_threads = 32; adev->gfx.config.max_hw_contexts = 8;	case CHIP_FIJI: adev->gfx.config.max_shader_engines = 4; adev->gfx.config.max_tile_pipes = 16; adev->gfx.config.max_cu_per_sh = 16; adev->gfx.config.max_sh_per_se = 1; adev->gfx.config.max_backends_per_se = 4; adev->gfx.config.max_texture_channel_caches = 16; adev->gfx.config.max_gprs = 256; adev->gfx.config.max_gs_threads = 32; adev->gfx.config.max_hw_contexts = 8;

The update states that Vega will have 4 shader motors, each of which is equipped with 16 calculation units. By all accounts, AMD retains the same structure as previous graphics architectures. The number of stream processors, texture units and raster units should thus remain at the same level as Fiji.

Read This Now: Nvidia Releases Geforce 495 Driver For Linux

Specifications: – AMD Radeon RX Vega

	RX Vega	RX 580	RX 570	R9 Fury X
Technical	14 nm	14 nm	14 nm	28 nm
Circuit	Vega 10	Polaris 20	Polaris 20	Fiji XT
Circuit surface	?	232 mm²	232 mm²	596 mm²
Transistors	?	5.7 billion	5.7 billion	8.9 billion
Architecture	NCU	GCN 4.0	GCN 4.0	GCN 3.0
Streamprocessorer	4 096 st.	2 304 st.	2 048 st.	4 096 st.
Texture units	256 st.	144 st.	128 st.	256 st.
Raster units	64 st.	32 st.	32 st.	64 st.
Bass frequency	?	1 257 MHz	1 168 MHz	–
Turbo frequency	>1 526 MHz	1 340 MHz	1 244 MHz	1 050 MHz
Computational power	12 500 GFLOPS	6 175 GFLOPS	5 095 GFLOPS	8 602 GFLOPS
Memory bus	2 048-bit	256-bit	256-bit	4 096-bit
Memory amount	8 GB HBM2	4/8 GB GDDR5	4/8 GB	4 GB HBM
Memory frequency	2 000 MHz	8 000 MHz	7 000 MHz	1 000 MHz
Memory bandwidth	512 GB/s	256 GB/s	224 GB/s	512 GB/s
Power supply	?	8-pin	6-pin	8+8-pin
TBP	?	185 W	150 W	275 W

Read This Now: Nvidia Geforce RTX 2080 Ti with two fans in reference design finds its way to the web

Although the Radeon RX Vega has similarities to the R9 Fury X, the new architecture Next-Generation Compute Unit (NCU) will offer better performance at a given clock frequency. At the same time, AMD has already announced the Radeon Instinct MI25 for machine learning and which will use the same graphics circuit.

Vega Final Presentation-page-023.jpg

According to AMD, the Radeon Instinct MI25 has a theoretical performance of 25 TFLOPS in half-precision (FP16) calculations or 12.5 TFLOPS in full-precision (FP32). To reach this level, a clock frequency of at least 1,526 MHz is required, significantly higher than the R9 Fury X and AMD’s Radeon RX 500 series graphics cards.

Guru3D’s forum also publishes performance results from 3DMark Time Spy of two upcoming graphics cards from AMD, which is believed to be Vega 10 and a scaled-down variant alternative the smaller circuit Vega 11. The graphics score in the first screenshot is as high as 10,158, which is about nine percent higher than a Geforce GTX 1080 Ti which in SweClocker’s tests lands at 9,302 points in the standard version.

Read This Now: Hackers circumvent Nvidia's limitation on virtual graphics hardware

The graphics score in the second screenshot is 5,721, which is 34 percent higher than the 4,265 points AMD Radeon RX 580 can handle. More interesting is a comparison with the Geforce GTX 1070, whose result of 5,786 points means that the differences are well within what can be considered a margin of error.

► Read all articles about #AMD Vega

The launch of the Vega architecture takes place during the second quarter of the year and according to AMD, the top model will be named Radeon RX Vega.