Intel Arc B580 Content Creation Review

Introduction

Over two years after the launch of its first dedicated discrete GPU, Intel has returned with the second generation of discrete Arc graphics: the Intel Arc B580 “Battlemage”. Although this is a solidly entry-level graphics card, we are excited to see if Intel can deliver on its promises of stable drivers and good performance uplifts after its impressive first try.

Curiously, this generation, Intel is leading with what will likely be the low end of the product stack. With Alchemist, Intel launched the A770 and A750 GPUs first and then only released the low-end A580 a year later. We can, for purposes of actual products, ignore the A380. This time, it is beginning with the cut-down B580 and will presumably launch what would be B770-class cards in the next year; hopefully, we will receive an announcement about their future plans soon.

Intel is also taking a slightly different approach with its launch strategy this time, specifically around third-party models. Although there will be some availability of an Intel “Limited Edition” card (as pictured above), most video cards are expected to be sold through AIB partners, the number of which has expanded to include Acer, AsRock, Gunnir, Maxon, Sparkle, and Onix. Although lacking most of the “prestigious” OEMs like ASUS, MSI, and Gigabyte, we recently saw AMD take a similar path on its way to growing market share.

The B580 contains the second-generation Xe2 architecture, which we have previously seen only in Intel’s Lunar Lake laptop CPUs. The second-gen Xe-core promises major efficiency improvements and offers SIMD16 execution (1st-gen Xe cores supported only SIMD8). It also means new, second-gen Ray Tracing Units (RTUs) and XMX engines (specialized matrix engines, similar in purpose to NVIDIA’s Tensor cores).

The B580 also supports dual media engines, each with an encoder and decoder, for up to two 8K 10-bit workloads. In terms of codec support, Intel’s media engine has acceleration for HEVC 8, 10, and 12-bit (decode only) 4:2:0, 4:2:2, and 4:2:0, as well as AV1, among other codecs. The 10-bit 4:2:2 is of particular note, as currently, only Intel iGPUs and Alchemist cards support this common variant of H.265. In our testing, we found that it currently enjoys the same acceleration support in Premiere Pro as the Alchemist cards (for more info, see our pages on decode support in Premiere Pro and DaVinci Resolve), and believe it will be the same in DaVinci Resolve, although testing is ongoing.

Below, we have listed the most relevant GPU specifications from Intel and NVIDIA. For more information, visit Intel Ark, NVIDIA’s 40-series GeForce page, or AMD’s Radeon RX Page.

GPU Model	MSRP	VRAM	Shader Units	“Game” Clock	VRAM Bandwidth	TDP	Release Date
NVIDIA GeForce RTX 4070	$600	12 GB	5888	2480 MHz	504.2 GB/s	200 W	April 2023
NVIDIA GeForce 4060 Ti	$400	8 GB	4352	2540 MHz	288 GB/s	160 W	May 2023
Intel Arc A770 LE	$350	16 GB	4096	2100 MHz	560 GB/s	225 W	Oct. 2022
AMD Radeon RX 7600 XT	$330	16 GB	2048	2470 MHz	288 GB/s	190 W	May 2023
NVIDIA GeForce 4060	$300	8 GB	3072	2460 MHz	272 GB/s	115 W	June 2023
AMD Radeon RX 7600	$270	8 GB	2048	2250 MHz	288 GB/s	165 W	May 2023
Intel Arc B580	$250	12 GB	2560	2670 MHz	456 GB/s	190 W	Dec. 2024
Intel Arc A750	$250	8 GB	3548	2050 MHz	512 BG/s	225 W	Oct. 2022
Intel Arc A580	$180	8 GB	3072	1700 MHz	512 GB/s	185 W	Oct. 2023

Although we have only heard rumors so far of future Battlemage GPUs, it seems likely that Intel is imagining a shifting of the stack with this generation. The B580 carries an elevated price tag over the last-gen A580, instead matching the A750 at $250 MSRP. Interestingly, Intel has reduced the number of Xe Cores (which contain the shader units and ray tracing units) from both the A750 (28) and A580 (24) down to 20 for the B580. However, the clock speed of the cores has increased by nearly 1 GHz over the A580 and 600 MHz over the A750, so this, alongside the new Xe2 core microarchitecture, may allow for a solid performance gain regardless.

Intel also continues to offer good VRAM capacity and bandwidth for the money, with the B580 increasing VRAM to 12 GB from the A750 and A580’s 8 GB. This is a needed step as more applications become VRAM hungry, and previously, the first step above 8 GB occurred at $330 with the Radeon 7600 XT. On the NVIDIA side, this only occurs with the RTX 4070, giving you a mere 12 GB of VRAM for $600. Unlike NVIDIA, they also decided against slashing VRAM bandwidth, which is important in some workflows.

Ultimately, a GPU is the sum of its parts, so while looking at any individual major specification can be helpful to understand how a GPU may perform, it won’t ever tell the whole picture. Other factors like cache, bus clocks, core architecture, and drivers all play a part in determining performance. Additionally, bear in mind that shader units cannot be directly compared across generations, let alone across manufacturers. Nonetheless, it is good to see Intel continuing to push for a strong memory subsystem on even their budget parts, and hopefully, the Xe2 cores make up for the lower Xe core count.

Test Setup

For our GPU testing this time around, we have shifted to an AMD Ryzen 9 9950X-based platform rather than our typical Threadripper platform. The 9950X has fantastic all-around performance in most of our workflows and should let the video cards be the primary limiting factor where there is the possibility of a GPU bottleneck. However, this does mean that it is more difficult to compare these results directly to past GPU test results—though we have also overhauled many of our benchmarks since our last GPU roundup. For testing, we used the latest available GPU drivers and tested everything on the “balanced” Windows power profile. Resizeable BAR and “Above 4G Decoding” were enabled for every GPU as well.

For this review, we tested all of the low-end to mid-range cards we had on hand. Unfortunately, we do not have any AMD GPUs below the 7900XTX, nor the Intel A580. However, the nearest price competitor from the last gen is the A750, which makes that a more natural comparison. We included the A770 as well to see if Intel could achieve generational gains matching their previous flagship, as well as the lowest-end NVIDIA 40-series card in the 4060. Though notably more expensive than the $250 price point of the B580, we also included the 4060 Ti as we believed the B580 might be able to match it in some workflows, and the 4070 as it is the first NVIDIA GPU to have more than 8 GB of VRAM.

In terms of applications, Intel GPUs prove to be somewhat challenging for our traditional video card testing suite. They are supported, and we tested them in, the Adobe suite and DaVinci Resolve, as well as Unreal Engine and Topaz, but they are not supported in most of the rendering benchmarks that make up a good portion of our reviews. Specifically, they are not supported in Redshift (Cinebench) nor V-Ray, and we were also unable to properly test them in Blender despite nominal support.

Photography: Adobe Lightroom Classic

Lightroom is not typically an application that is associated with GPUs, and as such, we don’t typically include it in our GPU reviews. However, since Lightroom Classic 11, GPU acceleration has been enabled by default for Export on GPUs with sufficient VRAM. Due to this, we thought it would be interesting to include our Lightroom Classic benchmark as a result and see how much difference the quality of GPU makes. Since there is only one area in our benchmark where we can see an impact based on the GPU, we also pulled out that individual sub-result.

The B580 performs acceptably in Lightroom, with performance 24% above that of the RTX 4060 Ti in the Export test and 13% behind the A770. It also achieves large gains over the similarly priced last-gen A750. Ultimately, though, the effect on the overall score (chart #2) is fairly small, with most GPUs scoring within the margin of error.

Video Editing: Adobe Premiere Pro

In Premiere Pro, we found the Arc B580 to do relatively poorly, falling behind the A750 and only surpassing the RTX 4060. We’ll discuss why as we get into the subtests, but we certainly did not expect this and believe that there are either driver or application issues causing performance below what the card is actually capable of.

In our LongGOP tests (chart #2), we see the first issue with the B580: bad hardware acceleration. Particularly in the processing tests, the new Intel card is slower than expected, with the HEVC 4:2:2 tests being slower to run with hardware acceleration than in software. Although the B580 is still faster than the 4060 and 4060 Ti, we suspect that this is a bug and can be fixed with a driver update. It also means that the dGPU is slower for processing LongGOP codecs than an intel iGPU.

With RAW codecs (chart #3), NVIDIA has historically been dominant, and we see that trend continue. However, the B580 fails to make any progress in this area over the A750, instead falling solidly within the margin of error of the last-gen part.

In the GPU effects portion of the benchmark (chart #4), the B580 does very well, matching both the A770 and RTX 4060 Ti. Given that the B580 costs less than an RTX 4060 and has more VRAM than the RTX4060 Ti, this makes it a great budget GPU for these tasks in Premiere Pro. However, the poor performance in the rest of the application may hold it back (at least until we get software/driver updates).

Motion Graphics: Adobe After Effects

Much like Lightroom Classic, Adobe After Effects is relatively light on the GPU compared to most of the other applications we tested. We will be adding more GPU testing to our After Effects benchmark in the future, but for now, it makes up a relatively small portion of the overall score. Due to this, we’ve pulled out the GPU subscore alongside the Overall.

In the GPU portion of the benchmark, the B580 performs great, matching the A770 and falling behind only the 4070 (a card that costs more than twice as much). The lead over the rest of the cards, all of which perform similarly, is relatively small, but this is also the cheapest card tested. Impressive results from Intel! However, it is once again worthwhile to look at the overall impact of these results. In our current benchmark, the GPU performance has essentially a negligible impact on the overall score, with variance in other tests essentially washing it out.

Video Editing / Motion Graphics: DaVinci Resolve Studio

In DaVinci Resolve, the B580 makes respectable gains over the A750, with an overall score just slightly behind the A770. However, rather than consistent gains across the board, we saw some slightly odd behavior in Resolve in terms of which subtests saw performance improvements and which saw regressions. There also appears to be an issue in the benchmark where the B580 maxes out available VRAM (despite being a 12 GB card) and crashes during some combinations of our tests. We, again, suspect this is a driver bug.

In LongGOP workloads (chart #2), the B580 has the best performance of any video card we tested. This is largely due to the strong media engines on the B580. However, in the processing tests (essentially, decode), the B580 was still slower than a Quick Sync 200S iGPU. Overall, the performance of the media engines of all of the Arc cards in DaVinci Resolve has been somewhat confusing, and we suspect that application updates may be needed to fully extract the performance from the discrete GPUs.

NVIDIA continues to offer the best performance for working with RAW codecs (chart #3), and the B580 ends up 16% slower than the RTX 4060. This positions it only slightly ahead of the A750, but overall, you should probably stick with NVIDIA for RAW codecs, even on the low end.

Moving on to GPU effects (chart #4), the B580 is the second most performant card we tested, falling behind only the $600 RTX 4070, which is not a fair comparison. Although only slightly ahead of the A770, this places it 30% ahead of the 4060 and 18% ahead of the A750. It also offers more VRAM than either of the 60-class NVIDIA cards, making it a great deal for GPU effects in Resolve. Similarly, the B750 does well in Fusion tests (chart #5), beating the A750 by 16% and the A770 by 7%. However, it is still slower than all of the NVIDIA cards we tested, so it may not be the best budget card for heavy Fusion work. NVIDIA is also dominant with the AI workloads we test (chart #6), as the B580 ends up right in between the A770 and A750, with even the 4060 offering 43% more performance.

Game Dev / Virtual Production: Unreal Engine

For our Unreal Engine GPU benchmark, we test performance in 3 different scenes as 1080, 1440, and 4K resolutions. Two of those scenes use hardware acceleration, while one does not. Due to this, we can look at the results across a number of factors to see if there is anything interesting. For this review, we pulled out the overall score and then the results at 1080P, 1440P, and 4K.

Starting with the Overall geomean FPS, we find that the B580 is directly on par with the RTX 4060. This immediately makes the B580 a great value as it is less expensive and has 50% more VRAM. Impressively, we did not need to separate out RT and Rasterized performance, as although enabling RT comes with a performance penalty, the relative change in performance from enabling RT is the same between Intel and NVIDIA with the B580 and 4060. Enabling hardware-accelerated ray tracing on this class of card may be mostly academic, but NVIDIA does not hold a performance advantage when doing so.

Moving on to the results at various resolutions, we see that the 4060 and B580 are always almost identical. However, the 4060 generally performs better at 1080P, while the B580 is superior at 4K. For 1440P, they are essentially the same. In some ways, this may be the worst result for Intel, as a card of this caliber is not designed for 4K, but it is interesting to see how the architectures fare at differing resolutions.

Topaz Video AI

Topaz Video AI is a common application used to enhance video, typically through upscaling the pixel count or interpolating frames to increase the fps (or create a slow-mo effect). We use the built-in benchmark from Topaz and calculate a series of subscores and an overall score from the results. Historically, Intel has struggled to compete in this benchmark, especially with the interpolation tests. However, one of its complexities is that performance depends both on drivers and the underlying AI software toolkit (in the case of Intel, OpenVINO).

In this test, the B580 only outperforms the A750 but still falls behind the RTX 4060 and A770. For AI workloads, we’d probably recommend that users stick with NVIDIA, but due to the ongoing development of toolkits like OpenVINO, this could change in the future.

How good is the Intel Arc B580 for Content Creation?

Across the board, the Intel Arc B580 represents a solid generational improvement for Intel over their Alchemist line. Coming in at a mere $250, the B580 matches or exceeds the NVIDIA GeForce RTX 4060 ($300) in most workflows and occasionally punches up against the NVIDIA GeForce RTX 4060 Ti ($400). We found that the RT performance was particularly impressive, but that apparent bugs with the media engine caused performance penalties in areas that otherwise should have been a strength.

For video editing and motion graphics, the B580 ended up being something of a mixed bag. In LongGOP codecs, it was either the best option or the worst, depending on the specific application and flavor of H.264/HEVC. We expect that, to some extent, this is a bug, but until it is fixed, it is difficult to blanket recommend the B580 for use in this type of workflow. With RAW codecs, we still recommend an NVIDIA GPU. However, the B580 performed great in GPU effects in Premiere Pro, DaVinci Resolve, and After Effects, falling behind only the much more expensive RTX 4070.

In our Unreal Engine benchmark, the B580 was 31% faster than the last-gen A750, and matched the 4060 across the spectrum of resolutions and Ray Tracing. Given the higher VRAM capacity of the B580, it may make for a solid Unreal Engine GPU on the low end. For professionals, it may end up being too low end, though. These results also make us excited to see how it performs in gaming workloads, though you’ll have to check out other reviewers for that.

Overall, the B580 is a terrific budget GPU with plenty of VRAM and solid performance. We haven’t experienced any issues with the drivers in terms of bad installations, OS crashes, or setting wonkiness as we did with the initial Arc Alchemist launch—Intel seems to have ironed out many of the generic driver issues over the last few years. However, we have seen performance issues, particularly with hardware-accelerated media, that we believe to be driver-related. Hopefully, we see a patch in the near future to address those issues.

---

If you need a powerful workstation to tackle the applications we’ve tested, the Puget Systems workstations on our solutions page are tailored to excel in various software packages. If you prefer to take a more hands-on approach, our custom configuration page helps you to configure a workstation that matches your exact needs. Otherwise, if you would like more guidance in configuring a workstation that aligns with your unique workflow, our knowledgeable technology consultants are here to lend their expertise.