Intel Arc A770

The Intel Arc A770 is a mid-range desktop GPU and is manufactured using the 6 nm process at TSMC. The graphics processor has 4,096 shader units and 8 or 16 GB GDDR6 VRAM with a 256-bit interface (16 Gbps, 512 GB/s).

Features

As the top model based on the Alchemist architecture, Intel offers the Arc A770, the top model in the series. As a completely new development, Intel wants to put pressure on AMD and Nvidia with these graphics cards. Features such as ray tracing are supported and the Intel Arc A770 comes with 32 RT cores. Intel is also implementing 512 Tensor Cores, which we can already find in the current graphics cards from Nvidia. With support for AV1, the graphics card is particularly interesting for content creators. Only the Ada Lovelace models from Nvidia and the RDNA3 GPUs from AMD are currently capable of hardware-side encoding and decoding with the new codec.

Performance

The performance of the Intel Arc A770 is comparable with an Nvidia GeForce RTX 4060 or an AMD Radeon RX 7600 comparable. The Intel model copes much better with newer games. With ray tracing enabled, the Intel Arc A770 can even outperform its two competitors. Based on the performance shown, the Intel Arc A770 is ideally suited for 1080p gaming in 2023.

Power consumption

The manufacturer specifies the power consumption of the Intel Arc A770 at 225 watts (TGP). This is slightly more than a Nvidia GeForce RTX 4060 or the AMD Radeon RX 7600.

Arc Series Arc A770 4096 @ 2.1 - 2.4 GHz 256 Bit @ 16000 MHz Arc A750 3584 @ 2.05 - 2.4 GHz 256 Bit @ 16000 MHz Arc A580 3072 @ 1.7 - 2 GHz 256 Bit @ 16000 MHz A770M 512 @ 1.65 GHz 256 Bit @ 16000 MHz A730M 384 @ 1.1 GHz 192 Bit @ 14000 MHz A570M 16 @ 0.9 - 1.3 GHz 128 Bit @ 14000 MHz A550M 16 @ 0.9 - 2.01 GHz 128 Bit @ 14000 MHz A530M 12 @ 0.9 - 1.3 GHz 128 Bit @ 14000 MHz A370M 8 @ 1.55 GHz 64 Bit @ 14000 MHz Arc A380 1024 @ 2 GHz 96 Bit @ 15500 MHz A310 6 @ 2 GHz 64 Bit @ 15500 MHz A350M 6 @ 1.15 GHz 64 Bit @ 14000 MHz
Codename	DG2-512
Architecture	Alchemist
Pipelines	4096 - unified
TMUs	256
ROPs	128
Raytracing Cores	32
Tensor / AI Cores	512
Core Speed	2100 - 2400 (Boost) MHz
Theoretical Performance	39.32 TFLOPS FP16, 19.66 TFLOPS FP32
Cache	L2: 16 MB
Memory Speed	16000 effective = 2000 MHz
Memory Bus Width	256 Bit
Memory Type	GDDR6
Max. Amount of Memory	16 GB
Shared Memory	no
Memory Bandwidth	512 GB/s
API	DirectX 12 Ultimate (12_2), Shader 6.6, OpenGL 4.6, OpenCL 3.0, Vulkan 1.3
Power Consumption	225 Watt
Transistor Count	21.7 Billion
Die Size	406 mm²
technology	6 nm
PCIe	4.0 x16
Displays	4 Displays (max.), HDMI 2.1, DisplayPort 2.0
Date of Announcement	12.10.2022

Benchmarks