The Intel Arc A310: top tier display adapter gaming

The Intel Arc A310 is not intended to be a card for gaming, but I’ve checked, and they can’t legally stop you.

Welcome to Esk Computers, I’m Scott and today on the desk of Esk, I think you know where this is going.

I’d picked this up for what I think was sensible use case in a home server, to sit alongside a more powerful card I would use for gaming. It would be used for its well regarded and pretty up to date transcoding engine to stream media, as it has hardware support for newer formats like VC1 that the 1080Ti would have to pass back to the Xeons to deal with.

A cunning plan, hampered only by the weird dual Xeon motherboard not booting when two graphics cards are installed, at least as far as I’ve been able to tell after relentless prodding. I did briefly try this in my computer running alongside my 7800XT for a similar reason, as my video editing package Davinci Resolve might benefit slightly from using that Intel engine to handle bouncing around in the timeline. It probably did, to be fair, but not by enough to justify keeping it installed and powered up continually, doing nothing ninety odd percent of the time.

As such, I can’t really think of a good reason to keep this around, so it will soon be featuring on an eBay near you. Before it departs Esk Gardens, we might as well give it a proper poke. As mentioned, it’s not a gaming card, really. It was announced with little fanfare at the tail end of 2022 as the most entriest level of Intel’s Alchemist line-up, and is much more suited to light CAD work than Triple A gaming. Indeed, it’s maybe even closer to display adapter territory than that.

This is the SPARKLE A310 ECO model, which is notable in that it’s using a reduced 50W TBP over the usual 75W, they claim without any affecting any function. This allows it to have this compact, single slot, low profile version, which I thought might be a useful option in the ultracompact builds I’d looked at a while ago.

It still crams in one HDMI port and two Mini DisplayPorts, and it’s DirectX 12 Ultimate compatible so in theory every modern game should launch on it. Playing them’s another matter, with only 4GB of GDDR6, and six of Intel’s Xe graphics cores. The A770 had 32 cores, so lots of cutting down going on. It should run at about 2000 MHz, although the datasheet lists 1750 MHz as the base clock. Techpowerup reckons this should be roughly GTX1050 levels of performance, so Intel are really catching up with entry level performance of a decade ago.

It might not be a 100% comparable test set up, but a while back we looked at the NVidia T1000, a card with a similar form factor, on a Xeon 2980 v4 system. That should have been powerful enough to max it out, so lets use that as a comparison point. It’s about a year older, with my model having 8GB of GDDR6, and somewhat more shading units but no raytracing hardware compared to the A310, but, let’s be realistic, we’re not expecting that to be on the table.

We’ve also just looked at getting the most out of the older P1000 via an extremely practical water cooling setup, and with that being a 4GB card it’s maybe a fairer comparison for the more demanding tests we’ll throw at it. As with the P1000, we’ll be testing on a platform using my 9800X3D processor on a B650E-F motherboard with 64GB of 6400MHz DDR5.

I’ve got a slightly cut down subset of the usual test subjects, some at reduced settings or resolutions to get something slightly playable, hopefully, so lets get cracking and see what it can do.

The cooler might be tiny, but it’s coping with the low power draw in Furmark. After this 15 minute hairy donut rendering party it’s levelling out at 68 degrees C. The sound of the titchy fan at full tilt is not loud, exactly, it’s more… annoying.

Taking a look at the venerable Fire Strike first, and the A310 is scoring 6698 points. The Nvidia T1000 we looked at scores 18% higher with 8164 points, and the P1000 was 24% slower at 5120 points.

It’s a bit closer in Time Spy, with the A310 getting 3165 points, 9% slower that the T1000’s 3486, and 45% faster than the P1000’s 1752.

Superposition at 1080p High Settings returns 3216 points on the A310, a good 29% slower than the T1000’s 4531. The P1000 scored 2925, so 9% slower than the A310.

The Shadow of the Tomb Raider benchmark at 1080p High sees the A310 get a slightly disappointing 38 average fps, 28% lower than the T1000’s 53 average fps, but 26% faster than the P1000’s 28 average fps.

Counter-Strike 2 with the FPS Heaven Benchmark at 1080p Very High Settings sees us get only 47 fps. Not only is that a massive 45% slower than the T1000’s 85 fps, it’s 11% slower than the P1000’s 52 average fps.

Dropping to competitive settings, and once again it’s a weird set of results from Counter Strike. The A310 gets a massive performance jump to 176 average fps, 23% faster than the T1000’s 143 fps, and 23% faster that the P1000’s 135 fps. Weird. I mean, good, obviously, but weird.

Unsurprisingly this isn’t a card well suited for Cyberpunking. In decreasing order of stupidity, at 1080p with the super-demanding Raytracting Overdrive setting, this gets 5 average fps, and with the merely highly-demanding Ray Tracing Ultra settings it gets over double that, although 11 fps is nothing to be excited about.

It’s almost getting to respectability at 1080p Ultra settings with 26 average fps. On these settings, the T1000 gets 31 fps, that’s 16% faster, and the P1000’s getting 15 fps, 42% slower.

Dropping down to 720p Low Settings gets a reasonably playable 66 average fps from the A310, although that’s still 20% slower than the T1000’s 83 fps. The P1000 slots neatly between them with 41 fps, 38% slower than the A310.

Marvel Rivals at 1080p Native resolution, Medium settings is a demanding test, and the A310’s only able to get to 14 average fps. That’s 50% of the T1000’s 28 average fps, and exactly the same 14 fps that the P1000 could get to.

Trying to get a bit closer to playability by dropping to Low settings and Ultra Performance XESS upscaling, we get 42 fps, at the cost of it looking like a fourth generation photocopy of itself.

Sticking with Unreal Engine 5 and the Black Myth Wukong Benchmark Tool, where at 1080p High, no RT or FSR, we get 9 fps, with it understandably complaining about low VRAM. That’s technically 40% slower than the T1000’s 15 fps, and 22% faster that the P1000’s 7 fps but let’s not pretend any of these results are viable.

Forza Horizon 5’s benchmark at 1080p Ultra settings gave a respectable 27 average fps on the A310, and dropping to Medium settings bumps us up to 60 average fps. At these settings it’s only 10% slower than the T1000’s 67 fps, and 35% faster than the P1000’s 39 average fps.

Just for fun, let’s throw a some newer games at it and see if any of them stick. The A310 does support all the latest APIs and so let’s see if there’s any mileage in that.

Doom: The Dark Ages of course wouldn’t even launch on non-RT capable cards, but this meets that requirement, albiet without having the horsepower to do much after it fires up. By dropping all the way to Low settings and throwing Ultra Performance XESS upscaling at it, we only get a cinematic average framerate of 24 fps, with 1% lows of 18 and 0.1% lows of 16 fps, with less than cinematic visual quality.

It’s actually somehow borderline playable in an N64 sort of way, but it would clearly be much happier running 2016’s Doom at it’s recommended settings, which are mostly Medium, where it gets 48 average fps and 1% lows of 34 fps with only the 0.1% lows of 8 fps letting the side down.

Silent Hill f gets a surprisingly decent showing, albeit a not very pretty one, at 1080p Low settings, scoring 36 average fps, with 1% lows of 22 and 0.1% lows of 10 fps.

Alan Wake 2’s not so happy though, and complains from the off about wanting at least another 2 gigs of VRAM, and that’s reflected in the 1080p Low settings average framerate of 14 fps, with 1% lows of 11 an 0.1% lows of 7 fps. It’s a bit more playable with ultra performance FSR enabled, getting 25 average fps, with 1 % lows of 20 and 0.1% lows of 9 fps, but it looks extremely dreadful. This isn’t the horror it ought to be evoking.

Death Stranding 2 gives us another playable with asterisks kind of result at 1080p Low settings, with ultra performance XeSS enabled it again takes a visual hit, but feels pretty stable to play at 35 average fps, with 1% and 0.1% lows of 25 fps.

Resident Evil Requiem 1080p Lowest upscaling off gets 49 average fps, with 1% lows of 36 and 0.1% lows of 26 fps in ths admittedly not too demanding early section, and doesn’t require quite the same visual compromises to get there.

Forza Horizon 6’s benchmark at 1080p Low, with no raytracing or upscaling gets a pretty decent 53 fps. Bombing around Tokyo gives me an average framerate of 56 fps, with 1% lows of 46 and 0.1% lows of 44 fps. That’s… pretty good, actually. Overall, sure, it’s not the optimum way to play newer games, but it’s coping with them better than I expected as long as you meet it halfway and accept radical cuts to the graphical fidelity to cram it into 4GB of VRAM.

So what on earth was the point of all this? Well, not everything has to have a point, you killjoy. Just let me live my life. Well, okay, let’s see if this makes any sense for low end, low profile gaming.

Averaged out the A310 is about 20-odd percent slower than the T1000, and 20-odd percent faster than the P1000, so as long as it’s going into a system with Resizeable BAR support it’s a contender for a more budget, single slot, half height build. That’s a niche of a niche, and looking at the value proposition it seems to be a tough sell. The A310 cost me £85, with the P1000 coming in at £45 and the T1000 costing me £100. I’d either be taking that 20% hit in performance and going with the practially half price P1000, or spending the little bit extra for the better performing, higher VRAM T1000.

Much faster low profile, single slot cards are of course available, with gaming focused like certain versions of the 3050, or workstation cards like the A1000, but they will be maybe three times the price of the A310, so a different market segment entirely.

While this does have the theoretical advantage of up to date API and driver support, meaning it can launch the latest games, it’s not an amazingly practical one, as it’s not fast enough to play most of them without more compromises than I’d like to make. Maybe this is more helpful with CAD packages or other workstation tasks, but that’s outside of my ever-limited knowledge base, so unless you really need that Intel video encoding juice, the A310 doesn’t seem to be worth the squeeze.

Of course, value and small form factors are rarely part of the same sentence so maybe this is asking the wrong question, but at any rate the answer is that while it’s not utterly unviable for a lower cost, lower size build, it’s not my first choice, and your money is much better spent on full sized components which will give much better performance at the same price point.

I think before flogging this on I will pop it into that little compact box we initally tested the P and T1000 in, and see if it’s as badly affected by the lack of resizeable BAR support as we expect.

Until then, if you have any questions or want further details please leave a comment down below, and if you enjoyed this videotronic missive then consider subscribing. Until next time, take care of yourself, and each other.