Reliable hardware, yet performance metrics overall with significantly less potent CPUs.
The Intel Arc B580 graphics card left a good impression on us just a few months ago, because of its genuinely solid performance with a plentiful amount of VRAM at a mainstream price. However, we hadn’t gotten around to reviewing the more affordable B570 model—until now. This delay isn’t particularly bad, as it provides us the opportunity to test this GPU in real-world scenarios as it’s meant to be used, gaining insights on availability and conducting additional examinations on the driver overhead issues highlighted by tech reviews post-launch.
We will dive into that shortly, but for now, let’s address a simple question: what differentiates this supposedly $220 B570 card from the ostensibly $250 B580? The answer isn’t complicated—and this makes the B570 potentially the more intriguing choice of the two.
Specifically, we see a drop from 20 Xe cores and RT units to 18, XMX AI engines decrease from 160 to 144, and core clock speeds reduce by approximately 6%. The primary change is within the memory subsystem, which moves from 12GB of GDDR6 on a 192-bit bus to 10GB on a 160-bit bus, meaning the B570 only boasts 83% of the bandwidth offered to the full-fledged B580. Power consumption also significantly declines, from a rated 190W to just 150W for the B570.
These reductions in power consumption clearly seem proportional to the reductions elsewhere, and testing supports that assumption. Without revealing too much about the results in the subsequent sections, the B570 achieves around 89% of the B580’s performance while using 88% of its power in games like Alan Wake 2, with similar 87% performance for 88% of the power in Senua’s Saga: Hellblade 2.
Of course, these findings don’t exist in isolation, and it’s insightful to compare how the Arc cards stack against their nearest AMD and Nvidia counterparts regarding power efficiency. For instance, the Nvidia RTX 4060 outperforms the B580 in Alan Wake 2 while consuming only 80% of its power. Meanwhile, the AMD RX 7600 stands as the fastest option at 107% of the B580’s average frame rate, consuming 6% more power to achieve that.
If we assess these Alan Wake 2 outcomes in pure efficiency terms, measured in joules per frame, the RTX 4060 emerges as the clear leader at 3.19, followed closely by three other cards tightly grouped: AMD at 3.98, B580 at 4.03, and B570 at 4.06. This shifts when evaluating RT performance in Avatar: Frontiers of Pandora, where Nvidia again leads at 2.2 joules per frame against 2.94 and 3.00 joules for the B580 and B570, with AMD trailing at 4.29, making it the least power-efficient.
| Arc B580 | Arc B570 | |
|---|---|---|
| Xe Cores | 20 | 18 |
| Render Slices | 5 | 5 |
| RT Units | 20 | 18 |
| XMX AI Engines | 160 | 144 |
| Graphics Clock | 2670MHz | 2500MHz |
| Memory | 12GB | 10GB |
| Memory Interface | 192-bit | 160-bit |
| Memory Bandwidth | 456GB/s | 380GB/s |
| Peak TOPs | 233 | 203 |
| Total Board Power | 190W | 150W |
Certainly, power efficiency is just one aspect of a superior graphics card, and it’s essential to look at our raw performance results in greater detail—as we will on the following pages.
Before we proceed, it’s important to highlight the specific setup we are using for most of our performance testing—utilizing our standard high-end GPU benchmarking rig. This system primarily revolves around what is still the fastest gaming CPU, the AMD Ryzen 7 9800X3D, to shift the burden to the graphics card as much as possible. We also have 32GB of Corsair DDR5-6000 CL30 memory, a premium Asus ROG Crosshair X870E Hero motherboard, and a 1000W Corsair PSU.
We will also examine performance using more realistically matched components—particularly lower-end CPUs that tend to experience a greater performance hit due to Intel’s notably higher driver overhead—which we will explore on page eight.
Having stated that, let’s delve into the raw performance benchmarks to get started.