Tuesday, June 30th 2026
Steam Machine Crippled by Single-Channel RAM
Shortly after Valve launched the long-awaited Steam Machine, reviews were quick to point out its middling performance and high price, with the gaming console scoring in between the AMD Radeon RX 6600 and 7600 in gaming tests. As it turns out, one of the compromises Valve made in order to launch the Steam Machine at a slightly lower price was to use single-channel DDR5 and not dual-channel DDR5. Curiously, even though you might expect the Steam Machine to suffer from a GPU bottleneck, given the GPU specs and limited VRAM, recent tests by Gamers Nexus on YouTube show that the mini PC is capable of up to 19.4% more performance in CPU-intensive workloads.
The YouTube channel's testing team put the Steam Machine through a battery of CPU and GPU tests, and, as expected, the biggest differences showed up in CPU-intensive workloads, with GPU tests demonstrating a peak performance difference of 8.7% in Baldur's Gate 3 at 1080p in Ultra settings. That's not to say there weren't improvements to gaming workloads. Especially when targeting high frame rates in games, as would be the case with e-sports games, for instance, differences of up to 15% could be observed. These substantial differences were observed by Gamers Nexus when testing Baldur's Gate 3 at 1080p and low settings, where a 15.3% performance improvement was exhibited, while Outer Worlds 2 at 1080p and low settings also showed a 14.7% performance improvement going from single-channel to dual-channel SODIMMs. The other major outlier was Resident Evil 4 at 1080p and "prioritize performance" mode, which showed a 10% improvement in frame rate.The full video follows:
Steve Burke of Gamers Nexus also notes that, during the testing, the Steam Machine exhibited instability at boot with certain DDR5 SODIMM sticks, and the team had to track down a memory stick that exactly matched Valve's OEM SODIMM, which may prove challenging to end users. Valve currently does not list the DDR5 speed frequency compatibility for the Steam Machine, and certain memory timings and frequency combinations may prove problematic. That said, the Steam Machine only ships with DDR5-5600 with poor timings, so there may be more performance gains from installing faster RAM, if indeed any tuning can be done.
Source:
Gamers Nexus on YouTube
The YouTube channel's testing team put the Steam Machine through a battery of CPU and GPU tests, and, as expected, the biggest differences showed up in CPU-intensive workloads, with GPU tests demonstrating a peak performance difference of 8.7% in Baldur's Gate 3 at 1080p in Ultra settings. That's not to say there weren't improvements to gaming workloads. Especially when targeting high frame rates in games, as would be the case with e-sports games, for instance, differences of up to 15% could be observed. These substantial differences were observed by Gamers Nexus when testing Baldur's Gate 3 at 1080p and low settings, where a 15.3% performance improvement was exhibited, while Outer Worlds 2 at 1080p and low settings also showed a 14.7% performance improvement going from single-channel to dual-channel SODIMMs. The other major outlier was Resident Evil 4 at 1080p and "prioritize performance" mode, which showed a 10% improvement in frame rate.The full video follows:
When it comes to GPU benchmarks, Baldur's Gate 3 was the only outlier, with most games netting between 0.6% and 3.6% performance uplift when played at high or ultra graphics settings at 1080p. The most important takeaway here is that productivity workloads will see the biggest performance uplifts. That was highlighted by the 19.4% performance improvement in the 7Zip file compression benchmarking, but it may also be applicable to things like video encoding and decoding, but swapping out the RAM to dual-channel does not seem to affect rendering performance in applications like Reddit.
Steve Burke of Gamers Nexus also notes that, during the testing, the Steam Machine exhibited instability at boot with certain DDR5 SODIMM sticks, and the team had to track down a memory stick that exactly matched Valve's OEM SODIMM, which may prove challenging to end users. Valve currently does not list the DDR5 speed frequency compatibility for the Steam Machine, and certain memory timings and frequency combinations may prove problematic. That said, the Steam Machine only ships with DDR5-5600 with poor timings, so there may be more performance gains from installing faster RAM, if indeed any tuning can be done.

110 Comments on Steam Machine Crippled by Single-Channel RAM
There is performance lost, but it's in the region of 5-20% than 35%.
If talking all brand new parts then I certainly wish you luck in you endeavor of making a 600$ DIY Machine.I mean... I guess it's technically an APU, but the 740M's ginormous FOUR (4) CUs has it closer to a CPU with integrated graphics than anything else lol, the 760M jumps directly to double that (no merit there, it's really not a big jump) which is already a tad more gaming capable, though I still wouldn't expect anything more than heavily limited resolutions and graphics options (it's basically a Z2 handheld chip)
There was only one known case (and that was quite some time ago!) of a "false positive" involving an ACEMAGIC model—the specific model escapes me now—and Microsoft Defender Antivirus. At the time, several influencers exploited this to generate clicks. And yes, none of them subsequently apologized for their behavior. Where I live, this sort of thing is called defamation of character and is a punishable offense under Section 186 of the Criminal Code (StGB). Of course, that doesn't really help the manufacturer from the Far East much.
I fail to understand why the words "memory" and "IMHO" in a single sentence confuse you. Perhaps this clarifies things: "Memory" applies because the event is in the past—correct. "IMHO" is also appropriate because people like @Roph—and not for the first time—continue to claim otherwise, even though they really ought to know better by now. Whether he was successfully indoctrinated by various YouTube influencers, or simply has a personal issue with ACEMAGIC, remains unclear.
Regards,
Fred.
Regards,
Fred.
I had a single channel DDR3 system, let me tell you, at the performance you'd expect from the rest of the components, you'd still get a stutterfest in game if anything on the background was doing processing.
The system we talk about now is an APU based system, so for sure the RAM impact is going to be significant, this has been the gist of APUs since the dawn of time: RAM speeds are of major impact, and single/dual channel equally so. At high loads a single channel RAM system will just choke, its as simple as that.
'Why is 15% detrimental'... lol. The SM can hopefully do 60 FPS in your optimal environment... now that's capped to 51 FPS. Mate.... I get it, you like the Machine, but you're not being fair to yourself. Its hilarious and sad all at the same time. The copium knows no bounds. If there is any time critical workload, it is realtime processing in games.Ah yes, all of the fancy reasons why a 4090 or 5090 is also commanding 2-3k on the market.
Except that GPU is about 20x as fast.
What game can I play on a bespoke fan and high density board/daughterboard? Just curious :rolleyes:
--EDIT: I stand corrected, the video I/O is also one of those daughterboards
custom cooler? really? are we acting like it's not the same under the shrouds now? is there any physical differences in the 5080 FE & 5090 FE coolers?
you know damn well that they got basically one size fits all for a whole generation then they just put a different shroud on it with sometimes the same fans they put on the previous generationyeah? and? you can put an OS on it? didn't know a 5090 could have a controller hooked up to it, what's the *storage* on it?you think you're all smart with that remark but one can buy three Machines and still have money to spare for controllers AND screens before price matching a 5090
there's more fun to be had with a Machine that there will ever be in a GPU that requires a mortgage AND is an actual fire hazard
Just that is enough and speaks volumes here. FTR, I would never pay 2-3k for a GPU either. There are limits to insanity.Nope. Can a Steam Machine play Crysis? :roll: (I reckon it might actually)
Of course its a shitty comparison mate. But we are talking about a >1000 dollar device here and... the fan and daughterboard became a talking point to somehow justify that number. Like... what the actual F !You really don't need to spend over 1k to have that level of fun and play the exact same content on the exact same controller, you know? Which, by the by, also does NOT come with controller or screen for that serious money.
32 CU on graphics performance of 3060-4060
32 GB DDR Quad channel @8000 Mt cost MBO+APU+ cooler+32GB Quad channel : 1032 EU
PSU: Enermax Revolution 3 ERV 850G ATX 3.1 full modular 70 EU
Case: Deepcool CH160 Plus 44 EU
M2 1TB : Lexar NV790 172 EU
A total of 1317 EU with no Steam controller, Keyboard or Mouse.
Edit: I didn't added the fan as some will go for Noctua A12 or a cheaper P12
Roughly 100 Watt TDP.
Much better performance on 1080 P and longevity than SM
imo, its sad to lose performance on an already limited box but I understand it.
I also have no need for it, my 6L lan rig is about the same size and more powerful with 5600x and rtx4060. so... (shrugs)
but for those that want an appliance for the TV it has its uses.
I am mostly happy that it is spawning lots of knockoffs running SteamOS, that is a massive long term win for the ecosystem and frankly might be why Valve overpriced the box a bit.
I don't think it is a good deal, but its also not the worst in the current climate, yes you can build the same cheaper, but it is fairly bespoke and has engineering time to expense.
This dynamic is also proven by the above results, where you can see performance deficits on every single game that wants a bit more than bottom tier CPU performance. Now consider running a few things in background as well instead of a clean bench rig. % deficit go up - the RAM is a limiting factor.
Also, you won't have another Steam Machine next to yours running dual channel RAM to remind you of what you're missing, neither are you gonna run spreadsheets to calculate the difference. In a plain gaming experience, a few percent doesn't add anything to write home about. Note that 2.3% in Black Myth Wukong, which means a whopping 1 frame per second in real life.
It's easy to cry about the raw percentage, but one should understand what the data actually mean for the gaming experience.