Framework Laptop 16

“Thanks for the clarification. If the cases fans are not arranged correctly or if the physical gpu is placed in such a way that its own fans are getting chocked then yes, a desktop gpu can throttle. But if you have a gpu freely standing and in the open air (Like many of these Egpu set ups do) then it won't throttle. So, no laptop company makes a x16 connection because that is completely up to either Intel or AMD as they are the ones making the chips and the chips can only support a specific amount of Pcie lanes. Whatever Pcie lanes you have is based on what they give you and it is up to the laptop maker to distribute them accordingly. A lot of laptop gpus (So mobile gpus) are x8, not x16 for example. This means that if you have a mainboard with x20 lanes (Like the 7840hs iirc) that means you have can have one Mobile gpu connection (8 lanes), 2 M.2 slots (Combined 8 lanes), and then the final 4 lanes are split between the other parts of the mainboard (Like the wifi card). Because of this we have never seen a free floating x16 lane before. Many of the people who use oculink 4i for example (The most common form of oculink) either do so by using a dedicated oculink port in the device or by using an M.2 to Oculink adapter (So theya re sacrificing a storage slot for the connection). Now, major companies have tried in the past to get into the EGPU market, this was most prevalent when the Thunderbolt 3 connector just came out and a few years after it. The good thing about Thunderbolt (Or just USB-C in general) is that the controllers that limit its bandwidth are also the things that make it hotswapable (Unlike OCulink), so you can just plug it in without hesitation. The issue is that TB3 only had a Pcie 3.0 x 4 connection, so 32 Gbps (practically 22 to 24 gbps when you factor in the controller). This limited connection is what threw it away from the normal consumer market and into a niche area. Having that low of a bandwidth just leads to massive stuttering and really low average fps (Something you can still kind of see with Thunderbolt 5). The final thing that is worth mentioning is the fact that mobile CPUs used to not be nearly as good as they currently are. Back then you could realistically hook up a mobile cpu to a desktop card and find that it would be bottlenecking the GPU. It also doesn't help that the manufacturer could have limited your access to adjusting the CPUs settings as well, so you probably couldn't adjust things so that it could work. Now, egpus did make a resurgence in China thanks to Crypto miners funny enough. This is actually how Oculink (A connection that was originally meant to be used in large data centers) ended up finding its way into commercial usage. This is also why you see most of the oculink stuff come from china too. These days its mostly a thing you'll see in handheld PCs or desktop minipcs. In fact, the majority of the EGPU market and products are mostly tailored to both handheld and mini pc users. Now, ”