Intel Arc A380 review: Best for video, weak for gaming

The Intel Arc A380 has got to be one of the worst graphics card launches in history – not necessarily the hardware itself, but the commercial launch of the hardware. By all indications, Intel knew the drivers were broken when the hardware was ready for release earlier this year. Rather than take enough time to fix drivers before the retail launch, and with time running out as new AMD and Nvidia GPUs are on the horizon, Intel has decided to ship its Arc GPUs in first in China – probably not the kind of approach a company would take. take if the product was worthy of inclusion on our list of best graphics cards.

Several months later, after much negative publicity thanks to GPUs making their way to other shores, and with numerous driver updates, the Arc A380 has officially launched in the US with a starting price of $139 (opens in a new tab). The unique offer on Newegg has sold out and is currently out of stock, but this is likely more to do with limited supplies than high demand. Still, the A380 isn’t too bad, and we’re happy to see Team Blue join the dedicated GPU market for the first time in over 24 years. (And no, I’m not really counting the Intel DG1 from last yearbecause it only worked on specific motherboards.)

How does the Arc A380 compare to competing AMD and Nvidia GPUs, and what’s all the hype around AV1 hardware encoding acceleration? You can see where it lands in our GPU Benchmark Hierarchy, which if you want a spoiler is… not good. But let’s get to the details.

Arc Alchemist Architecture Summary

(Image credit: Intel)

We have provided extensive coverage on Intel Arc Alchemist Architecture, dating from about a year ago. When we first wrote this article, we were planning a late 2021 or early 2022 launch. not even a full version, at least not yet. The Arc A380 is just the first salvo, right at the bottom of the price and performance scale. We have seen many signs of the Arc A750 fasterwhich seems close to RTX-3060 performance based on Intel’s own benchmarks, and expected to launch next month. What about the even faster Arc A770 or the mid-level Arc A580 and other products? Only time will tell.

Arc Alchemist represents a departure from previous Intel graphics designs. There’s probably a lot of overlap in some elements, but Intel has changed the names of some of the core elements. Gone are the “execution units (EU)”, which are now called vector engines (VE). Each VE can compute eight FP32 operations per cycle, which is loosely translated to “GPU cores” or GPU shaders and is roughly equivalent to AMD and Nvidia shaders.

Intel packs 16 VEs into a single Xe-Core, which also includes other features. Each Xe-Core therefore has 128 shader cores and roughly translates to the equivalent of an AMD Compute Unit (CU) or an Nvidia Streaming Multiprocessor (SM). They’re essentially all SIMD (single instruction multiple data) designs, and like the competition, Arc Alchemist has upgraded the shaders to meet the full DirectX 12 Ultimate feature set.

(Image credit: Intel)

This naturally means ray tracing hardware is built into the design, and Intel has one ray tracing unit (RTU) per Xe-Core. The exact details of the ray tracing hardware aren’t entirely clear yet, though based on testing, each Intel RTU could decently match an Nvidia Ampere RT core.

Intel didn’t stop there. Along with VEs and RTUs and other typical graphics hardware, Intel has also added Matrix Engines, which it calls XMX Engines (Xe Matrix eXtensions). These are similar in principle to Nvidia’s Tensor Cores and are designed to analyze a lot of less precise data for machine learning and other uses. An XMX engine is 1024 bits wide and can process 64 FP16 operations or 128 INT8 operations per cycle, which gives Arc GPUs relatively large computing power.

Intel Arc A380 Specifications

With that brief architectural overview out of the way, here are the Arc A380 specs, compared to a few competing AMD and Nvidia GPUs. Although we provide theoretical performance here, remember that not all teraflops and teraops are created equal. We need real-world testing to see what kind of real-world performance the architecture can deliver.

GPU Specifications
Graphic card Bow A380 RX 6500 XT RX-6400 GTX 1650 Super GTX-1650
Architecture ACM-G11 Navi 24 Navi 24 TU116 TU117
process technology TSMC N6 TSMC N6 TSMC N6 TSMC 12FFN TSMC 12FFN
Transistors (Billions) 7.2 5.4 5.4 6.6 4.7
Die size (mm^2) 157 107 107 284 200
SM/CU/Xe-Cores 8 16 12 20 14
GPU Cores (Shaders) 1024 1024 768 1280 896
Tensor cores 128
Ray tracing ‘cores’ 8 16 12
Base clock (MHz) 2000 2310 1923 1530 1485
Boost Clock (MHz) 2450 2815 2321 1725 1665
VRAM Speed ​​(Gbps) 15.5 18 16 12 8
VRAM (GB) 6 4 4 4 4
VRAM bus width 96 64 64 128 128
ORP 32 32 32 48 32
TMU 64 64 48 80 56
TFLOPS FP32 (boost) 5 5.8 3.6 4.4 3
TFLOPS FP16 (MXM/Tensor if available) 40 11.6 7.2 8.8 6
Bandwidth (GBps) 186 144 128 192 128
Video encoding H.264, H.265, AV1, VP9 H.264, H.265 (Turing) H.264, H.265 (Volta)
TDP (watts) 75 107 53 100 75
Release date June 2022 Jan 2022 Jan 2022 November 2019 Apr 2019
Introductory price $139 $199 $159 $159 $149

On paper, Intel’s Arc A380 essentially competes with AMD RX 6500 XT and RX-6400Where Nvidia’s GTX 1650 Super and GTX-1650. Its price is slightly lower than the competition, especially when it comes to current online prices for new cards, with roughly similar features. There are, however, some important caveats to note.

Nvidia doesn’t have ray tracing hardware under the RTX 3050 (or RTX 2060). Similarly, none of the AMD or Nvidia GPUs in this segment support tensor hardware, giving Intel a potential advantage in deep learning and AI applications – we’ve included FP16 throughput for cores GPUs on AMD and Nvidia cards for reference, though not entirely apples to apples.

Intel is the only GPU company that currently has AV1 and VP9 hardware-accelerated video encoding. We expect AMD and Nvidia to add AV1 support to their upcoming RDNA 3 and Ada architectures, and possibly VP9 as well, but we don’t have official confirmation on how that will pan out. We’ll also look at performance and encoding quality later in this review, but note that the GTX 1650 uses Nvidia’s older NVENC hardware which offers lower quality output than the newer version of Turing (and Ampere).

The Arc A380 has a theoretical compute performance of 5.0 teraflops, putting it slightly behind the RX 6500 XT but ahead of everything else. It is also the only GPU in this price range to come with 6GB of GDDR6 memory, with a 96-bit memory interface. This gives the A380 more memory bandwidth than AMD but without Infinity Cache, and less memory bandwidth than Nvidia’s GPUs. Power consumption targets 75W, although overclocked cards can exceed this, just like with AMD and Nvidia GPUs.

(Image credit: Tom’s Hardware)

Ray-tracing capabilities are harder to pin down. To quickly recap, Nvidia’s Turing architecture on the RTX 20-series GPUs had full hardware ray tracing capabilities, and each RT core can perform four ray/triangle intersection calculations per cycle, along with support hardware for BVH (bounding volume hierarchy) traversal. Nvidia’s Ampere architecture added a second ray/triangle intersection unit to the RT cores, potentially doubling the throughput. In practice, Nvidia claims Ampère’s RT cores are typically 75% faster, as they can’t always fill all execution slots. AMD’s RDNA 2 architecture is similar to Turing in that it can perform four ray/triangle intersection calculations per ray accelerator. However, it uses GPU shaders for BVH traversal, which is slower and more memory intensive.

From what we can verify (and we’ve asked Intel for clarification), Intel’s RTUs are similar to Nvidia’s Turing RT cores in that they can perform four ray/triangle intersections per cycle. We’re still trying to figure out if they include BVH traversal hardware, or if they’re more like AMD’s ray accelerators and use GPU shaders for BVH traversal. It looks like they include some BVH material and might actually be pretty decent.

Still, with just eight RTUs, the A380 certainly won’t be a ray tracing powerhouse – Nvidia, for example, has 20 or more RT cores in its RTX lineup, or 16 if you include the mobile RTX 3050 in the list, Nvidia’s slowest RTX chip. . AMD, on the other hand, has as few as 12 ray accelerators in its RX 6000 series parts, and built-in RDNA 2 implementations like the Steam Deck can have as few as eight RAs, although RT performance naturally suffers a lot. – not that you need ray tracing, even four years after hardware first supported the feature.

Comments are closed.