25 December 2025

Analyse This: The Steam Machine's Specs are bad, no matter what Valve might say...

 
Mana from heaven or a gift from the devil?
 
Much has been made over Valve's reveal of the Steam Machine. With opinions varying from gaming saviour to irrelevant devil. Okay, maybe things aren't quite that polarising - I'm trying to make the intro a bit spicier! However, it's clear there are two main emerging narratives surrounding the soon-to-be-released Steam Machine - one positive, one negative. 
 
Personally, I lie on the negative side of the equation for a couple of specific reasons with orbit around the fact that it's just not good enough
 
So, let's get into why I feel that way...  
 
 

Playing to the Crowd...


Many people (Valve included) have pointed out that the Steam Machine has specs which are better than 70% of all PCs on Steam.

Aside from the fact that the Steam Hardware Survey is fundamentally flawed in its approach from a statistical perspective, basing your decisions and arguments on this average is also logically flawed from several points of view:
  • The current average is the worst place to put your product.
  • The markets the Steam Machine will be placed onto at the price point it is at are not going to move the needle in a positive direction.

Let's address the first point.
 
While I have my own reasons, I'll defer to a graphics engineer's opinion to begin with:
 
Here's the issue: there are many more low-end computers in the world than there are high-end. The average will be skewed towards the low-end and those systems will be impacted by the second point:

The markets where the lowest-end hardware will be present will be overly represented by very old hardware combinations. Not to beat around the bush - there are reasons why people in certain parts of the world are more likely to sport hardware that may be a decade old, or more*.
 
The issue, then is that the Steam Machine is not going to those countries. 100% a guarantee. At this point, Valve sells directly to 30 countries, with another 4 countries through third party sellers. That doesn't include most of Asia, South America, Africa, the Middle East. Just check out this survey from last year regarding the top 20 countries on Steam over a two week period...
 
Sure, some of those countries which are excluded have extremely low-income and maybe could not afford  a Steam Machine but the argument could be made that those same countries also potentially lack a PC gaming culture and also internet infrastructure which would support Steam content dissemination... 

*Sometimes, it's purely economic - low salaries, high tariffs or import taxes on computer technology that is not manufactured in the specific country. However, sometimes you just can't easily get access to the technology. For example, I've wanted an Optane drive for years and even tried to purchase one when I spotted deals. However, each and every time i was stopped by my banks - they wouldn't let me purchase from the USA. I couldn't get around the restriction because the amount was relatively large and a suspicious purchase and by the time I could get to speak to a customer service agent, the deals were either finished or stock was gone, etc.

There's a well-known statistical meme where if you have two groups with different averages and you selectively re-classify one or more of them and place them in the other group the average of both groups can increase.
 
This can come about when the result for the population which is shifted to the other group has values below the average in the group they are part of but above the average of the group they move into. This can be (and has been) used by entities* to cook the books on various initiatives for which positive outcomes are "required" or "expected" and may be necessary to justify ongoing funding and/or employment. The thing is, you can do the opposite and lower both averages.
*Usually political in nature... It's usually one of the reasons why new metrics or ways of measuring metrics are introduced. Cynical, I know.

Coming back to the second point, the people who are going to be purchasing the Steam Machine are primarily going to be in relatively affluent countries that have ease of access to high-performing computer hardware.
 
The Steam Machine has a good chance to lower the overall specifications of the "average" system on the Steam Hardware Survey by working its way into homes which already have a high-end desktop system as an entertainment system addition. It won't be likely to raise the floor of the low-spec gamers on that survey.
 
So, the comparison is meaningless. It's a pointless statistic... 
 
 

  

Misaligned Priorities... 

 
Now, we get onto the specifications themselves.
 
The Zen 4 APU* is limited to 30 W, so will be around the performance of a part like the Ryzen 5 7545U 28 W APU and, tied to this relationship is the fact that there will not be 6 "full" cores but instead 2x full cores and 4x compact cores**. So, a laptop CPU with laptop TDP. However, that's mostly fine. We're not going to be being held back by this relatively recent CPU architecture which can perform really well even at 30 W.
*Minus the integrated GPU...

**Compact cores, otherwise known as Zen "c" cores, are lower power and also operate at lower frequencies than the full cores but are otherwise feature equivalent. 

The system memory is 16GB - which is fine. According to my own trending and predictions, 16GB will be the most required for the next couple of years, and second most required after that. It's likely that the RAMpocalypse will result in developers keeping 16GB as the most required option for a few more years than I was initially predicting...
 
The Steam Machine also has two models - one with 512 GB storage and the other with 2 TB storage. The former is woefully inadequate. It's bad.... with the sizes of games over the last 10 years, you're looking at being able to install only a few - maybe less than a dozen. If you go back further or play indie games, you're golden.
 
 
512 GB! Really?! Come on, Valve...

 
The GPU, on the other hand is a disaster.
 
Not only is the suped-up version of the RX 7600M in the Steam Machine a pretty weak GPU - cut down from the RX 7600 and reduced to 110 W TDP - but the technology is more than two years old, at this point. We're already on the much more effective, efficient and "fixed" RDNA4 in current desktop parts and RDNA5 is likely to enter the market at either late 2026 or mid-2027. 
 
RDNA3 is a stop-gap architecture which failed to deliver on the promises that AMD themselves made in the reveal. It's not a good architecture for ray tracing and isn't very performant when compared to the Nvidia counterparts and RDNA4. For N33 (the lower-end GPU chip for RDNA3) the performance uplift over N23 (RDNA2) was around 3-5% for the same core frequency. 
 
Therefore, based on my own estimation, the chosen specs for the Steam Machine GPU will be around the performance of an RX 6600, maybe very slightly faster. (And this was before I was aware of the RX 7600M entry in the TechPowerUp GPU database!)
 

This watered-down mobile GPU also has the misfortune to be paired with only 8 GB VRAM. This is unfortunate because Valve themselves are positioning this as a 4K console - something which it will struggle with, even when using upscaling...
 
You only need to look at the majority of large tech outlets and you will see story after story after video of how 8 GB of VRAM is not enough in 2025 unless you're playing on low settings and at 1080p and I have to agree with them when it comes to the AAA titles that get released every year. For older games and indie games, it's going to be fine. However, my trending is predicting that the average recommended VRAM quantity will be greater than 8 GB in 2026. 
 
So, just when the Steam Machine is launching, it's already backward-facing in both GPU technology as well as capability. 
 
 
The RX 7600M is slightly below the RX 6600, so the increased power and slight up-clock will put it around the latter's performance, maybe 1-2% faster... via TechPowerUp

 

How to Fix the Steam Machine... 

 
In my opinion, the minimum storage size should be 1 TB. So, the two models should be 1 TB and 2 TB. It's likely that they made the decision they did because of SSD and RAM prices. The 512 GB model is likely to be more reasonably priced whereas the 2 TB model is probably going to be over the cost of a simple storage upgrade as it would have existed over the course of 2024 and the first 7 months of 2025...
 
So, this is likely a forced issue and likely, if the situation improves, Valve will address this in the future. The GPU issue is not so easy to fix.
 
Now, I could just wax lyrical and say that Valve should add more VRAM (insert comments about RAM shortage and vast overpricing here) and a more powerful GPU. However, the problem here is that AMD doesn't have one!
 
They have zero RDNA4 mobile GPUs on the market*, and from the 7000 series lineup, the next powerful GPU is the 7800M which is a chiplet-based design (very bad) with 12 GB of VRAM (good) with 180 W TDP (sort of bad) that just would not fit in the TDP, nor the budget of the Steam Machine. 
*And to the best of my knowledge, they have not even announced any!
Their other options are iGPUs which, unfortunately, top-out at 12 CU** - 43% of the quantity in the RX 7600M or, if you prefer, 16 CU less. These iGPUs are good for very light gaming (mostly 2D or simplistic 3D from many years ago) but not for the types of games that players may wish to experience on their TV.
 
The only products which have a significant amount of graphics processing power on their APUs are the already existent consoles (PS5, XSS/XSX) - none of which may be utilised by Valve.
**There does exist the Ryzen AI 9 HX 370 with 16 CU of RDNA 3.5 but, let's face it, Valve can't afford that chip - it's in high demand and ends up in €960+ devices...
There are the rumoured/inevitable RDNA4 mobile parts which will likely be announced at CES in January 2026 but those are set to release after Valve's intended launch window. Added to this stumbling block is the rumour that the RX 9060M will feature only 8 GB of VRAM, leaving it without any real advantage at 4K resolution - something which you really don't want to try, even on the desktop GPU with 16 GB. It's just not powerful enough for it.
 
So, in reality, the way to "fix" the Steam Machine is to not launch it in 2026...
 
Not really a very good take away but, as I've tried to point out, here, it's just not a very good system to be releasing now...
 
 

Conclusion...

  
So, will I get a Steam Machine? I'm damn tempted to! But I could also just roll my own with the much superior hardware I already have on hand. Will it work as smoothly? Probably not, but I'd get a lot better performance when it does work!
 
Is the Steam Machine bad? Objectively, I think yes. 
 
It's likely going to be a poorly-priced and positioned product in the market in 2026. However, consumers are fed-up with Microsoft - both for Windows OS and for Xbox. There is a market for secondary systems which play undemanding games on the TV and also streaming within the home from a more powerful system running Steam on Windows.
 
Could Valve have made a better system? In this time and place? Most certainly not.
 
They are constrained by the lack of advancement in the low-end of the GPU technology stack as much as the rest of us are - and that's a damn shame!
 
The Steam Machine v1 will, therefore, be a shadow of what it could have been. A good CPU, decent platform but crippled GPU that won't even support the FSR4 feature set.

No comments: