Last month, I got my first look at AMD's OEM-only version of the Radeon RX 9050. This one here, the one with just 4 GB of VRAM. And I had to pay $1,500 Aussie to get it. And it was every bit as horrible as I expected. It was so bad, in fact, that I deemed it a scam, which caused a small amount of controversy. A vocal minority claimed clickbait, which has become the boy who cried wolf at this point. But I did make a strong case for why the 9050 is a scam, or at least I believe it's a scam. And thankfully, after polling our audience, it looks like the majority of you do agree.
Even so, I have to say I came away from that content quite disappointed with the PC hardware enthusiast community, and I'll get to that a bit later on in the video. But in short, I did find it disappointing how the community could get so upset about a product like the unlaunched 12 GB RTX 4080 only to seemingly not appear to give two shits about the 4 GB RX 9050, which in my opinion is an even bigger problem. This is a much worse product, and of course, it hasn't been unlaunched. Anyway, as I said, I'll talk more about that later in the video. For now, I want to show you just how bad the 4 GB RX 9050 really is.
Now, in my first video, I noted that I was at the time testing a best-case scenario by using the Ryzen 7 9800X3D and then 32 GB of high-speed DDR5 memory, which likely isn't going to be a real-world configuration for this particular GPU given that it is an extreme entry-level product. But at the time I wanted to create an apples-to-apples comparison with the entry-level GPUs that we already had data for. Today though we are retesting the 4 GB RX 9050 with a more real-world configuration. So that's going to be fun.
Today's video is brought to you by the Gigabyte MO32U24, a high-end, high-quality gaming monitor using the latest 32-inch 4K QD OLED panel. This brand-new panel for 2026 features upgraded performance compared to first-gen QD OLEDs, including higher brightness, DisplayHDR True Black 500 certification, and an advanced obsidian shield film that improves black levels and offers greater scratch resistance. This fast 240Hz monitor offers an elite gaming experience with beautiful HDR visuals and amazing motion clarity. Plus, it has one of the best feature sets I've seen on a Gigabyte monitor, including things like an OLED anti-flicker mode and a KVM switch. If you're after a QD OLED, the MO32U24 delivers outstanding value with a great price in many regions. I recommend it. So, to learn more, check the links in the description below.
Okay, so the real-world configuration that we've gone with is the one that I actually bought the RX 9050 4 GB in. Though, due to shipping costs, I brought in just the graphics card from the US because the entire PC was going to be hideously expensive. And I'd already spent, like I said, $1,500 just to get the card. So, I didn't want to spend any more money on shipping than I had to. But anyway, the PC that I bought this particular card in was the CyberPowerPC Gaming Master Gaming Desktop. I'm pretty sure that was the name. Um, not the best name in the world. Anyway, it was $850 US and it came with the Ryzen 5 5500 processor, just 8 GB of DDR4 memory and then a 500 GB SSD.
Now, I already have a Ryzen 5 5500 and an 8 GB stick of DDR4-3600 memory. So, I thought let's recreate the CyberPowerPC Gaming Master gaming desktop configuration and see how it performs. About that though, with just 8 GB of RAM, the system was nearly unusable. I could test some very basic games, but anything demanding was a complete nightmare. Either because it took forever to get into the game or because the game would crash while I was testing. There's a few issues here. Other than the obvious, that running a modern gaming system with just 8 GB of RAM isn't going to result in a fun time. But making matters worse was the fact that the discrete graphics card had just 4 GB of VRAM. And most modern games, even at 1080p with the lowest possible settings, consume more than 4 GB of VRAM. And as a result, the additional data will spill over into system memory. But without enough system memory, that data ends up in the page file on the local storage, in this case, the SSD. And by this point, performance is so poor, you're really no longer gaming.
But even getting to a point where you could try and game was extremely difficult with this configuration. Shader compilation took forever and even loading into the menus, again, it was a real challenge. In fact, it was so bad that I just gave up. Sorry, guys. I couldn't do it to myself. I just don't have that kind of time to waste. So, I dumped out the 8 GB stick and replaced it with a pair of 16 GB modules. A bit less real world, I know, but the Ryzen 5 5500 is slow enough. Speaking of which, there's another major issue with this configuration that I'm yet to touch on, and that is the CPU. Whereas most Zen 3 processors support PCI Express 4.0, the 5500 is limited to PCI Express 3.0, which halves the bandwidth across the PCI Express bus. So, although the full x16 interface is supported, bandwidth has been reduced from the 63 GB per second we'd see when using PCIe 5.0, which was enabled with the 9800X3D, down to just 16 GB with PCIe 3.0 using the Ryzen 5 5500, which I don't need to tell you that is a lot less than the 144 GB of bandwidth the 4 GB 9050 has to its local VRAM. So performance for the 4 GB RX 9050 relative to the 8 GB version using the Ryzen 5 5500 is likely going to be significantly worse than what we saw when using the 9800X3D.
Now, for all of this testing, I've done my best to keep VRAM usage as low as possible by benchmarking at 1080p with some of the lower quality presets that are available. And I should note that most of these configurations still did exceed 4 GB of VRAM as it is very hard to keep modern games below 4 GB. Even so, the RX 9050 will occasionally perform better than you might expect when exceeding the VRAM buffer. And there is a reason for this, and that being that we might not be exceeding the VRAM capacity by as much as you think, as some games simply reduce the amount of data that's offloaded to the VRAM, opting to cull textures and other details in the process, meaning they basically go missing, and this results in much worse image quality, and we will also take a quick look at that. Anyway, there is a lot to go over, so let's jump into the blue bar graphs.
Starting with Forza Horizon 6, we see that previously when using the 9800X3D, that the 8 GB 9050 was 47% faster than the OEM 4 GB version. So, a massive margin and an unsatisfactory difference between two products sold under the same name. But if you use a more modest CPU like the Ryzen 5 5500, things do get much worse. With the more limited PCI Express bandwidth, the 8 GB 9050 wound up delivering 98% more performance, almost twice the performance at 1080p using just the medium quality preset. And keep in mind, both GPUs are exactly the same as in the physical GPU configuration, though the 4 GB version has half the memory capacity and half the memory bandwidth. And as you can see, that completely cripples performance in this example, which makes me wonder how you could possibly defend AMD and the decision here. It's beyond me.
The Assassin's Creed Black Flag Resynced performance is also shockingly poor. Using the 9800X3D, the 8 GB 9050 was 46% faster, which is a ridiculous margin to begin with, but when we switched over to the Ryzen 5 5500, that margin ballooned out to almost 80%. And in fact, the 1% lows were 109% greater for the 8 GB model. Insane stuff and another example of how scummy the 4 GB OEM version is.
Next up, we have The Last of Us Part II Remastered. And it's more bad news. Previous testing with the 9800X3D found the 8 GB version of the RX 9050 to be on average 31% faster. So, a pretty big margin there, but it manages to get worse, much worse with the Ryzen 5 processor. This more realistic configuration saw the 8 GB version delivering 81% more performance, jumping up from just 54 FPS to 98 FPS.
And we're looking at yet another blowout in Battlefield 6. Previously, I was very unimpressed with the 38% performance margin seen when testing with the 9800X3D, but that pales in comparison to what we're seeing here with the Ryzen 5 5500, as the margin explodes to 102%. The 4 GB OEM model was only able to average 44 FPS, less than half the performance of the 8 GB version.
Things don't really get any better in Marvel Rivals either, using the medium quality settings at just 1080p. We see when testing with the 9800X3D previously that the 8 GB version was 44% faster, which again in my opinion is completely unacceptable, that kind of margin with two products that share the same name. But look, it gets much worse when you don't have PCI Express 5.0 and DDR5 memory helping to mask the deficiencies of the OEM model. Armed with just the Ryzen 5 5500, the margin grew to 71% in favor of the 8 GB model, going from a disgraceful 34 FPS to a much more playable and enjoyable 58 FPS. Oh, and we also saw a stupid increase for the 1% lows in Marvel Rivals at 1080p using modest quality settings. This isn't exactly a cutting-edge title with maxed-out settings.
Interestingly, performance in Crimson Desert does go largely unchanged for the Radeon GPUs. Despite using a much slower processor with the Ryzen 5 5500, both the 4 and 8 GB versions scaled much the same. Though, that did mean the 8 GB model was still 38% faster. Unfortunately though, the ARC overhead issue popped up in this test. Intel still needs to do some driver work for this title. Performance was more than halved with the 5500.
F1 25 using the medium preset at 1080p ran well enough on the 4 GB 9050 when paired with the 9800X3D. Though the 8 GB retail version was still 41% faster, which we felt was far too significant of a difference for two products that share the same name. But the situation gets far worse when using a processor like the Ryzen 5 5500, as this configuration saw the 8 GB model delivering 128% more performance, going from 76 FPS all the way up to 173 FPS. And yet the margins somehow are even more extreme when looking at the 1% lows here. The 8 GB model was 167% faster, jumping up from a pathetic 49 FPS to a smooth 131 FPS. Again, I can't get over the fact that AMD thought it was okay to give these products the same name. And worse still, there's no shortage of gamers that seem to be defending that decision.
Interestingly, Cyberpunk 2077 Phantom Liberty scaled very similarly for the Radeon GPUs using either the 9800X3D or the 5500. With the 9800X3D, the 8 GB version of the RX 9050 was 8% faster, and that margin grew to just 14% with the 5500. We are very CPU limited with the Ryzen 5 processor, so performance across the board was very poor.
Lego Batman: Legacy of the Dark Knight is the first example where the margin between the 4 and 8 GB versions of the RX 9050 were actually reduced with the slower Ryzen 5 processor. Using the 9800X3D, the 8 GB model was an insane 130% faster. And while the margin is still very massive using the 5500, it did come down to 81%. In short, the CPU bottleneck is limiting the 8 GB model more than the PCIe bandwidth is crippling the 4 GB version.
The results seen when testing with Pragmata were not favorable for the OEM model, to say the least. The fact that the 8 GB 9050 was 31% faster when paired with the 9800X3D was already a big problem, but that margin blew out to 80% with the Ryzen 5 5500. Worse still, whereas the 8 GB model will see budget gamers enjoying 1% lows of over 60fps when gaming at 1080p with the balance preset, the 4 GB model reduces performance here to just over 30fps.
Counter-Strike 2 isn't exactly the most demanding video game out there, but even here, the 8 GB 9050 is still a good bit faster than the 4 GB model, delivering around 25% greater performance seen when using either the 9800X3D or the 5500.
Another game that's not particularly demanding is Tom Clancy's Rainbow Six Siege, especially when using the high preset, though I have fixed the render resolution to 100%. So, 1080p. Despite that, the 4 GB 9050 gets blown out of the water here. Using the 9800X3D, the 8 GB version was 62% faster. And although that margin did come down with the 5500, the retail version was still 56% faster.
Moving on to 007 First Light. This title didn't play well with the 4 GB 9050, even when using the 9800X3D. Here it averaged just 38 FPS with 1% lows of 26 FPS, making the 8 GB model 53% faster when comparing the average frame rate and an insane 96% faster for the 1% lows. That's bad, but it only gets worse with the 5500. Here, the 8 GB model maintained similar performance to what was seen when using the 9800X3D. The same, however, cannot be said for the 4 GB model. Its performance plummeted as a result. The 8 GB 9050 is now 155% faster when compared to the average frame rate and 667% faster for the 1% lows. In other words, the 4 GB model was completely broken here.
The 8 GB RX 9050 and Intel Arc B570 didn't drop much performance when going from the 9800X3D to the 5500 in Resident Evil Requiem. The same, however, cannot be said about the 4 GB OEM version of the RX 9050. The 8 GB model was already 38% faster when using the 9800X3D, but that margin exploded to 71% when paired with the Ryzen 5 5500. Here, the 4 GB version couldn't even average 60fps, while the 8 GB model pushed over 80fps.
Cities Skylines 2 goes from bad to worse with the 4 GB 9050. And by that I mean performance when using the 9800X3D was quite bad to begin with, and the switch to the 5500 did it no favors, reducing the average frame rate by 35%. But the real story is much worse, as the game was unplayable here with much lower 1% lows, just 3 FPS, which translated into constant stuttering and hitching, often suffering long pauses. The 1% lows for the 8 GB version, they're not exactly amazing either, but the 367% increase to the 1% lows, it was noticeable.
Space Marine 2 shows pretty similar scaling between the 9800X3D and 5500 for the 9050 variants. The 8 GB model is 46 to 48% faster than the 4 GB version, which is obviously a substantial performance increase, but sadly for the OEM model, that isn't the full story. And as I noted towards the start of the video, what we're about to look at here does apply to a lot of the games tested.
Now, the full story isn't just massively degraded FPS performance for the 4 GB model, but also heavily compromised visual quality because it only has 4 GB of VRAM. Space Marine 2 looks awful on a 4 GB graphics card. There's loads of muddy, low-resolution textures and missing details. I often hear people say, 'Not enough VRAM? That's fine, just turn textures down and you're good to go.' In the case of the 4 GB 9050 though, that's not really true, as performance generally still sucks and you're missing half the memory bandwidth. But also, textures are the key ingredient for good graphics. Without them, you just have a game with a flat presentation that lacks any real detail, just as we're seeing here, for example. The difference between the 4 GB and 8 GB models is night and day. The 8 GB version looks so much better, it's not even funny.
Finally, we have Fortnite. And while the Ryzen 5 5500 severely limits performance here, knocking the 8 GB 9050 down to 202 FPS, the performance hit is far worse for the 4 GB model, which dropped down to just 157 FPS on average. And this does result in a noticeably worse experience in Fortnite. In the end, the 8 GB model went from 21% faster using the 9800X3D to 29% faster with the 5500.
Now, calculating the geo mean across the 17 games tested, we see just how bad the 4 GB 9050 really is. When paired with the 9800X3D, the 8 GB 9050 was an average 41% faster, which is certainly bad enough. But with a slower processor like the 5500, that margin extends to 68% with a 109% increase to the 1% lows. That is an insane difference. And despite using low to medium settings at just 1080p, we couldn't even average 60fps across the games tested with the 4 GB model. So, I think it's fair to say the 4 GB OEM-only version of the 9050 is a true abomination. And I certainly stand by my initial claims. This thing is a scam. Plain and simple. By the very definition of the word, it is clearly a scam.
Now, if all of that wasn't bad enough, before we wrap things up, I just wanted to show you how we can really crush the 4 GB 9050 further. Seems hard to believe, but we can do so while keeping the 8 GB model above 60fps. In some examples, comfortably above. F1 25, we can enable the ultra high preset but disable RT effects, and this increases the margin in the 8 GB model's favor from 128% with the medium settings to 162%.
The 8 GB 9050 was able to render 58 FPS on average using the ultra preset in Forza Horizon 6, and these settings reduced the 4 GB version to just 15fps. So whereas the 8 GB model was 47% faster when using the medium settings, it's 287% faster using ultra. We're also looking at a similar story in Resident Evil Requiem. Using the high preset, the margin extends out to 164% in favor of the 8 GB model, up from 71%. So clearly you have to be far more careful with what settings you use on the 4 GB OEM version.
Okay, so the Radeon RX 9050 4 GB was horrible when I paired it with the Ryzen 7 9800X3D. And as we've just seen, when using a slower processor that's limited to older memory technology and bus interfaces, the situation just gets worse. In many cases, much, much worse. This is a problem because a lot of people interested in buying the most entry-level discrete GPUs, they're not usually pairing them with the latest and greatest CPUs. In fact, they're likely not even on the latest platforms. And in the case of the 4 GB 9050, the pre-built PCs that it's typically sold in, well, they're very low-end PCs like the CyberPowerPC that I bought it in. Of course, the big issue with the 4 GB 9050 isn't that it's slow or that it's really, really slow on older systems using DDR4 memory and PCIe 3.0 or probably even PCIe 4.0.
Rather, the biggest issue is the name. AMD is selling two very different models with the exact same name, and they're attempting to hide this much worse version in pre-built systems. That sucks, and really shame on AMD. They're actively deceiving customers. They only promote the performance numbers of the 8 GB model, and the official product page makes no mention of the 4 GB version. It just states that memory configurations are available up to 8 GB. So, as I said, it's a scam.
But another big problem we're seeing is with the PC hardware enthusiast scene itself. People looking down on other people for buying low-end hardware, or in particular what we're seeing a lot of, especially over on sites like Reddit, but also we see more than we'd like to in the YouTube comment section, is people crapping on those who aren't PC hardware enthusiasts that end up buying hardware like the 4 GB 9050 because that's all they can afford or they think all RX 9050s are the same. And we're talking about people who are unwittingly victims of products like the 4 GB 9050. They're often just brushed off as idiots who should know better. Most people who do this are likely doing so in an effort to try and defend their favorite mega corporation. So, in this case, AMD, saying anything bad about AMD instantly upsets them. So, when you say a product like the RX 9050 shouldn't have two wildly different configurations sold under the same name with the much worse version being hidden in OEM PCs, their knee-jerk reaction is to say, 'Yeah, well, those people are just dumb.' It's obviously an unhelpful and unintelligent response. But to be fair, it's an indefensible act that AMD is committing here. So, coming up with a solid defense, it's going to be pretty difficult. And when a position is too difficult to defend, but you feel the need to out of some misplaced brand loyalty, emotional attacks where you call potential victims dumb is the obvious solution.
I guess on a bit of a side note, it is incredible to me how many people on sites like Reddit claim that we're out-of-touch YouTubers for crapping on a product like the 4 GB RX 9050, only to then call people who buy the 4 GB RX 9050 without realizing it's much slower than the 8 GB version dumb or that they need to do their research. Reddit experts aside though, I do think the PC hardware enthusiast community in general is selective with the products they care about. And look, I do get it. If you're typically buying in the range of an RTX 5080, you don't really care about a product like the RX 9050.
And speaking of GeForce, just in case any AMD fans out there think that we just turn a blind eye to Nvidia, we are well aware that Nvidia wrote the book on screwing over their most cash-strapped customers. In fact, at this point, I almost think that Nvidia views it as more of a sport. Jensen does believe scarcity is a good thing. After all, gamers were able to force Nvidia to walk back or unlaunch the infamous RTX 4080 12 GB back in 2022. And people were outraged that there were two versions of the same product. One with 16 GB and then a gimped 12 GB version with fewer cores and reduced memory bandwidth. And the backlash was severe and rightfully so. But AMD is effectively doing the same thing with the RX 9050. In fact, in many ways, this product is much worse. The 12 GB 4080 had 25% less memory and memory bandwidth along with 21% fewer cores, but it was also slated to cost 25% less. Look, still a garbage-tier anti-consumer product for sure, but with the RX 9050, you're getting 50% less memory and memory bandwidth for what is on average at least 25% less performance. And worse still, you don't even know the price. But it's certainly not going to be 25% less.
Although we can't prove what this thing costs, as AMD selling the 4 GB model as an OEM-only part directly to its partners, who won't be allowed to disclose the purchase price, it's extremely likely that the 4 GB model is at most 10% less expensive than the regular, or let's say retail, 8 GB model. And the PCs that this thing is popping up in certainly seem to suggest that this thing isn't much cheaper than the retail version. But regardless of all of that, the issue is that we don't actually know what the 4 GB 9050 costs.
So to recap, the missing memory on the 4 GB version of the RX 9050 is worse than what we saw from the unlaunched RTX 4080, as is the reduced memory bandwidth. And it's highly likely that the asking price is also worse. You're also starting from a far worse position with this product. The GPU itself is already very underpowered, and the retail 8 GB configuration really should be the absolute bare minimum in 2026. So having what is already a minimum specification, well, it's almost criminal. I really think as a community we need to send a stronger message to these companies that selling wildly different performing products under the same name is not okay. Whether it be a $1,000 product or a $100 product, it really shouldn't matter.
And that's pretty much all I've got to say about this. I again strongly believe this product is a scam. AMD hiding it away in OEM-only PC, well, pre-built PCs. Very bad, bad look and just takes advantage of a vulnerable section of the market. So yeah, I again, I think there should be more pushback against this sort of product. And I'm aware that, you know, it's not just AMD doing this. Nvidia does it. You know, Intel's done this sort of stuff in the past. It's, I'm not, it doesn't really matter. I hate it when someone's like, 'But Nvidia, it's all bad.' And, you know, we've pointed out in the past when other companies have done it, but we want to see it stopped. But we want to see the community get more upset about these sorts of products because yeah, we really need to signal to these companies that it's not a good thing to do.
Obviously, I get the AI demand and stuff. Gamers aren't really a priority right now. But, you know, the tide may turn on that one. And it's better to unite against products like this. I just think this is an indefensible act. Nobody should be defending this in any capacity. All the excuses or whatever interpretations of what's going on, it's just, it's not helpful. So yeah. Anyway, I am keen though to hear what you guys think about this product given the updated testing with the Ryzen 5 5500, which was the CPU that this card shipped with in that pre-built system.
Admittedly, that pre-built did only have 8 GB of DDR4, whereas I tested with 32 GB of low latency, which you could say is not real world, but the 8 GB was basically unusable. So, I had to increase the memory. That's all I had in the DDR4 flavor. So, yeah, you could argue that that's still like a better-case scenario, but I think the pitch is pretty clear, right? It wasn't good on the 9800X3D, and it's even worse with the 5500. And there are existing products like the Arc B570 that are better value and products that you should probably buy instead of this garbage. Or unfortunately ending up, no one's going to go out of their way to buy this product. It's more that people are going to unwittingly purchase this product thinking they're getting what is the advertised performance on say AMD's product page for the RX 9050, only to end up with this dodgy version that they don't really advertise and most people aren't aware exists.
Anyway, broken record at this point. You guys have seen all the testing. You've heard my thoughts on it. So now I'm keen to hear your thoughts. So I'll be sure to check out your comments and read over all of them. But that is going to do it for this video. If you'd like to support the channel more directly, allow us to buy garbage-tier products like this to show you how bad they are, we have the join button or Patreon links are in the video description. Check those out if you're interested. It also gives you access to our exclusive Discord server where we can talk to you directly about products like this. We have live streams for members, Q&A stuff, behind-the-scenes content, a lot of cool things there. So check it out if you're interested. If not, that's perfectly fine. And of course, I would like to thank you for watching this video. I'm your host, Steve. I'll see you again next time.