If it looks too good to be true, it’s probably a load of cobblers, and that’s almost certainly the case with this ‘leak’ about an Nvidia GeForce RTX 5090 SE I saw being reposted by several usually-reliable people on X today. The first sign that something is up is that this new GPU apparently has 32GB of GDDR7 VRAM, but remarkably has an MSRP of $1,500, right in the middle of a memory pricing crisis.
Sorry, but no. The standard RTX 5090 might have an ostensible MSRP of $1,999, but anyone who’s tried to buy one in the real world right now knows you’ll have to pay at least 50% more for it. And that’s not just because of the demand for a high-end GPU – it’s mainly down to the cost of memory right now. If Nvidia is launching a new 32GB graphics card, I can guarantee it won’t cost $1,500.
Missed the bus
The next howler is that this 32GB of memory is apparently attached to a 384-bit wide bus, which is a technical impossibility with GDDR7 chip capacities. With that bus, Nvidia could either fit a GPU with 24GB, using 2GB chips, or 36GB using 3GB chips. There’s no option for a 32GB configuration with a 384-bit bus using this architecture.
What’s the source of this supposed leak? It comes from a site called GameGPU.com, in a story that doesn’t cite a single source. Instead, it just says that Nvidia is ‘rumored to be releasing’ this new card. There’s no link to a site where this information came from, nor even any reference to the site talking to people familiar with the matter. That’s all there is.
Who is Maxim Boldson?
Alarm bells also started ringing when I checked the author of the story. His name is Maxim Boldson, the apparent ‘chief editor’ of the site. There’s no email address to contact him, nor a single word on his biography page – just a picture of Thor from God of War Ragnarök.
Maxim Boldson appears to churn out 11 stories a day, and while I can’t say this for sure, it looks to me as though it’s generated with AI. It all looks a bit like the famously fictional ‘Shooter Jennings’ and ‘Brian Merrygold’, the AI writers who suddenly appeared on Videogamer.com, trashing that site’s reputation and getting it delisted from Google.
There are other clues in the long-winded phrases surrounding the details of that 5090 SE story, which bulk out the story but say very little. That $1,500 price, for example, “will help balance the cost of ready-to-use high-performance gaming systems.” Why is “ready-to-use” in there, and what does it mean? Pre-built? Wouldn’t a lower-cost RTX 5090 appeal to DIY builders too?
If Maxim Boldson isn’t a fictional AI writer, then he’s a real journalist who makes up uncited stories about graphics cards that can’t technically exist. Let’s face it, neither description is ideal.
Big specs mix ups
There are other oddities around the site – for example, looking at the Nvidia section shows a timeline of Nvidia graphics cards, but without a single mention of the famous GeForce 256. Nvidia apparently went straight from the “Riva GeForce” lineup to GeForce2, but Riva GeForce didn’t exist. I picked the page dedicated to Nvidia’s RTX 5070 Ti at random to see if it had any other silliness, and it didn’t disappoint.
“The RTX 5070 Ti is not a cut-down version of the 5080,” it says, despite earlier correctly identifying that it is indeed based on the same GB203 chip used in the 5080. “The RT and Tensor cores are more powerful, optimized for DLSS 4 and advanced geometry,” it says, which would be true if it were being compared to an Ada GPU, but there’s no mention of earlier GPUs in this paragraph – it’s talking about the 5070 Ti and 5070 here, which have the same core architecture.
There’s a reference to the 5070 Ti having 64MB of L2 cache, when it actually has 48MB (the 5080 has 64MB). This page also bizarrely says that, on the 5070 Ti, “Frame Generation works twice as stable as on 4070 Ti,” whatever that means, and that the 256-bit bus “provides huge bandwidth without the need for a super-high VRAM volume,” as if capacity has anything to do with bandwidth. This is just one page I picked at random – there are almost certainly loads more.
We often report on leaks and rumours on Club386, but we will always cite our sources when we do so, and we won’t make up any details, and we never use AI in any of our copy. That seems like the very minimum you should expect from a tech site, but these clearly aren’t normal times any more.
