Question - can the GPU in this kind of enclosure be upgraded in the future?

I'd rather just buy a laptop with an RTX 2070 already inside.

LFaWolfQuestion - can the GPU in this kind of enclosure be upgraded in the future?

Possibly but the usual issues may arise.

Keep in mind that physically the enclosure is very short with respect to how long some PCIe video cards can get these days. So whatever video card one would choose to replace the existing one with must meet the physical size limitations.

Also the power delivery system might have some limitations too so that would have to be looked into.

Empty 3rd party eGPU cases (bring your own GPU) tend to be physically bigger. However, paying ~$350 USD for an empty box is something many take issue with.

I was a early adopter of the alienware egpu solution, I wouldnt buy an enclosure again.

eGPUs get slapped around with such a massive performance penalty ~15%, and even high-end laptops suffer from power limitations of laptop parts, that you're better off buying a nice thin non-gaming laptop for work, save the money, and then having a budget gaming rig that's dedicated.

It's like a motorcycle side car. Interesting idea, but even the beater cars are better lol.

Gigabyte's Gaming Box series is one of very few eGPU solutions I would consider. Why? Size and portability. They're small and compact, have an internal PSU that's still replaceable (and not proprietary, even if it's an unusual standard), fit standard ITX-length GPUs, and are thus eminently portable while being both upgradeable and repairable. The only downside seems to be the noisy 40mm fans they use, but people seem to be having success upgrading them to far more quiet Noctuas. It's just too bad these boxes are so expensive, otherwise I would seriously consider saving up for one.

Still, this upgrade is... lackluster. Performance parity with the previous top end model (GTX 1080), yet I'm willing to bet it'll be more expensive. All for some features that you can't use yet/perhaps at all. No thanks.

How about a universal egpu chassis that can support multiple high bandwidth standards, leave it up to the buyer to put a card in

DarkswordI'd rather just buy a laptop with an RTX 2070 already inside.

Laptops that contain an RTX2070 don't meet the definition of a "mobile device" for most people.

eGPU is also pretty cost-effective for ultrabook owners (assuming they don't need fast GPU on the road).

notbLaptops that contain an RTX2070 don't meet the definition of a "mobile device" for most people.

eGPU is also pretty cost-effective for ultrabook owners (assuming they don't need fast GPU on the road).

Not to mention how much slower Laptop models are compared to desktop variants.
15-40% is not uncommon, as you also say they aren't really a mobile device anymore.
so you buy a RTX2070 laptop and you get RTX2060 performance ~

Imho, if I got a ryzen 4000 7nm chip with better gpu performance to play starcraft 2 for instance on the move.
AND! a connection between a gpu box and a laptop that has no real penalty like the x4 connection current boxes has I'd be rather sold.
But that's not ever gonna be the situation it seems soo.. desktop it is.

eidairaman1How about a universal egpu chassis that can support multiple high bandwidth standards, leave it up to the buyer to put a card in

Multiple high bandwidth standards? Are there viable, non-proprietary alternatives to Thunderbolt? Or are you talking about closed-off, entirely proprietary standards like Alienware's Graphics Amplifier port? And besides, wouldn't that require a PLX chip to switch between the output ports? Last I checked, even the simplest PLX chips are >$100. While I wish there were open alternatives to TB, at least Intel is moving it in that direction.

Other than that, there are plenty of BYO GPU boxes out there. Sadly they're all huge, hulking monstrosities often bigger than a well-built ITX gaming case - and still crazy expensive.

ImsochoboImho, if I got a ryzen 4000 7nm chip with better gpu performance to play starcraft 2 for instance on the move.
AND! a connection between a gpu box and a laptop that has no real penalty like the x4 connection current boxes has I'd be rather sold.
But that's not ever gonna be the situation it seems soo.. desktop it is.

So you're saying you don't think mobile Ryzen 7nm will have better GPU performance than today's thin-and-lights? That seems pessimistic. Or are you saying it won't be better than a 2070/1080? If so, you're right about that - we're not getting 180W dGPU performance in a 15-35W APU in quite a while yet. 7nm isn't that good.

I'm hopeful about the interconnect, though - AMD is bringing PCIe 4.0 this year, and it'd be weird if Intel didn't follow suit, including an upgrade to Thunderbolt. PCIe 3.0 8x equivalent is plenty for any current GPU, which is what we'll get with TB4 if Intel keeps up what they've been doing so far. Hopefully also fully backwards and forwards compatible, which seems more likely given the adoption of USB-C for TB3 and the general growth of TB availability in both hosts and peripherals.

Im hopeful, but this won't be until next year, at the earliest. I wouldn't be surprised if Intel kept TB4 Intel-exclusive in the beginning despite their stated goals of making it open and free, though.

ValantarOther than that, there are plenty of BYO GPU boxes out there. Sadly they're all huge, hulking monstrosities often bigger than a well-built ITX gaming case - and still crazy expensive.

But does comparing to ITX setups make any sense? The whole point of eGPU is to have a single system that can do 2 things, not 2 systems that you're forced to move between. :-)

Until we get a proper standard, eGPU plug&play sets seem like the best choice. And even they have big performance variance with different laptop brands. BYO GPU boxes are far from stable at this point and often need a lot of software tinkering to work.

So you're saying you don't think mobile Ryzen 7nm will have better GPU performance than today's thin-and-lights? That seems pessimistic.

Well, you can get an ultrabook with 1060 today and 1050 is quite common, so that's a tough call. No integrated GPU comes even close.

notbBut does comparing to ITX setups make any sense? The whole point of eGPU is to have a single system that can do 2 things, not 2 systems that you're forced to move between. :)

If what you're looking for is that specific, no, it doesn't make sense. But then you're only looking at one factor - ease of use through having a single do-it-all system. The next question ought to be what trade-offs are made to achieve this. First, you're paying a $3-500 premium for the eGPU chassis. For $500, you can build a pretty decent ITX gaming PC except for the GPU, which will likely outperform your $1000 ultrabook. Sure, managing two systems is a bit more of a hassle, but you'll gain performance and flexibility (two PCs can be useful, after all), and it might be a UX boon to have a dedicated gaming system, not cluttered with work-related stuff or other mess.

notbUntil we get a proper standard, eGPU plug&play sets seem like the best choice. And even they have big performance variance with different laptop brands. BYO GPU boxes are far from stable at this point and often need a lot of software tinkering to work.

I don't think we're going to get any more of a "proper" standard than TB3/Xconnect. Intel, AMD and Nvidia have all agreed to the standard, which means all relevant players are on board, and there are no technical solutions that can compete with TB3 at this point, nor announced for the near future. Still, as you say, it's an immature ecosystem, and plug-and-play PCIe over a third-party protocol is bound to be problematic, especially when implemented across a wide selection of vendors both for the host PC, the interconnect, and the GPU (or PCIe device), even if there's only one vendor for the chips themselves. Performance variance across laptop brands sounds weird, though. Is this with everything else equal, i.e. PCIe x4 uplinks, same CPU TDP (no cTDP-up or similar tweaks) and no thermal throttling? If so, that's weird indeed.

notbWell, you can get an ultrabook with 1060 today and 1050 is quite common, so that's a tough call. No integrated GPU comes even close.

You should look up what "ultrabook" means. Even if Intel has abandoned the term, it's cemented itself as the catch-all term for ultraportables or thin-and-lights. At the very best, you could include thin-and-light 15" laptops like the XPS 15 within that definition (though that is stretching it quite a bit), which max out at 1050Ti or 1050Ti MaxQ performance. MX150 25W is the most common ultrabook GPU, which is barely faster than AMD's current iGPUs. I doubt we'll reach GTX 1050 Ti levels of performance with a standard iGPU (memory bandwidth will make that very difficult), but if AMD comes up with something HBM-equipped, it'd be entirely possible. Their MCM strategy would help make this come true, at least. The lightest 1060-equipped laptops you can get are >4lbs/1.8kg and 15.6", and that's no ultrabook, even if they fulfill Intel's requirements from 2011.

ValantarIf what you're looking for is that specific, no, it doesn't make sense.

What I'm talking about is the prevailing way people buy and use PCs today. We own notebooks. It may not be true for "we" on this forum, but it is true for "we" on this planet.

But then you're only looking at one factor - ease of use through having a single do-it-all system.

Well, I'm surely not taking all factors into consideration. :)
But I feel quite safe talking about what majority does.

And do you think about all factors of having a desktop (even a small one)?
You may think it's just a tiny ITX case. But what else do you need? Well... a desk for starters, which takes huge amount of space and is always problematic to place properly. And then everything else: keyboard, monitor. I could go on and on, but it's not a place, clearly.

The next question ought to be what trade-offs are made to achieve this. First, you're paying a $3-500 premium for the eGPU chassis. For $500, you can build a pretty decent ITX gaming PC except for the GPU, which will likely outperform your $1000 ultrabook. Sure, managing two systems is a bit more of a hassle, but you'll gain performance and flexibility (two PCs can be useful, after all), and it might be a UX boon to have a dedicated gaming system, not cluttered with work-related stuff or other mess.

2 PCs can be useful for what exactly? And once you find an example, does it really apply to many people? :)

Why would a a typical eGPU buyer want a dedicated gaming system? How much are casual gamers spending on this hobby? 2h a week? 4-6h if they have a lot of free time? A dedicated PC makes no sense.
And on the other side of the market: if you're an avid gamer, playing 4h a day, it's unlikely you'll ever consider an eGPU anyway. But it's nice to have a choice, right?

But most importantly, eGPUs are not made for gaming only. They're used for work as well, so it makes a lot of sense to use them with your ultrabook "cluttered with work-related stuff or other mess". :)

notbWhat I'm talking about is the prevailing way people buy and use PCs today. We own notebooks. It may not be true for "we" on this forum, but it is true for "we" on this planet.

Oh, absolutely, a lot of people are even abandoning laptops for tablets and smartphones (though those tend to be people who don't actually need to do any work on them).

notbWell, I'm surely not taking all factors into consideration. :)
But I feel quite safe talking about what majority does.

Put that way, the majority buys gaming consoles. They cost the same as or less than an eGPU box without the GPU, deliver decent-to-good framerates at 1080p-ish resolutions, and are eminently plug-and-play. As such, the step to an eGPU requires someone willing to go the extra mile for something other than a console gaming setup. I'm also counting into this that there are few mainstream games that can benefit from a GPU upgrade over an iGPU that don't also exist on consoles.

notbAnd do you think about all factors of having a desktop (even a small one)?
You may think it's just a tiny ITX case. But what else do you need? Well... a desk for starters, which takes huge amount of space and is always problematic to place properly. And then everything else: keyboard, monitor. I could go on and on, but it's not a place, clearly.

Sure, eGPUs allow you to (re)use the display and inputs (well, not touchpad, not really) from your laptops, but with a serious performance deficit on top of the already noticeable TB3 bottleneck. This is why eGPUs are generally recommended in conjunction with a monitor, and most have USB connections for proper peripherals too. While this adds cost, it also improves the experience, but you're right that eGPUs have the upper hand in terms of base requirements here. OTOH, an SFF PC can live under/next to a TV without issues, removing the requirement for a desk.

notb2 PCs can be useful for what exactly? And once you find an example, does it really apply to many people? :)

Well, the first example is easy: more than one person can use them at the same time. Helpful if you have kids or anyone else in the household who might not have their own PC available. Second, you can use your laptop for game capture or streaming (unless you want spotless 4k capture, of course). Third, an at-home desktop can be repurposed for all manner of uses, like HTPC use, home server use, and so on. While this is getting advanced, HTPCs are reasonably popular.

notbWhy would a a typical eGPU buyer want a dedicated gaming system? How much are casual gamers spending on this hobby? 2h a week? 4-6h if they have a lot of free time? A dedicated PC makes no sense.
And on the other side of the market: if you're an avid gamer, playing 4h a day, it's unlikely you'll ever consider an eGPU anyway. But it's nice to have a choice, right?

This is where your logic fails the most visibly: if you're spending 2h a week, why on earth would you spend $6-700 on an eGPU? A $300 console would be a far better solution. eGPUs aren't for people dabbling in on-and-off casual gaming - they're too expensive for that.

I agree that eGPUs don't make much sense for dedicated gamers - their (current) niche seems to be people who already have a gaming setup at home but are looking for something more portable/flexible/different for secondary use. That's where my interest comes in, as I would like (but can't afford/am not willing to spend $6-700 on) a portable eGPU to allow some light-to-medium gaming on my laptop when travelling - or ideally even when I'm lying on the couch and the TV is taken. This is very much a luxury and not a necessity, and I don't think anyone is buying an eGPU for that reason, as there are far better alternatives if your priority is easy access to decent-quality gaming.

notbBut most importantly, eGPUs are not made for gaming only. They're used for work as well, so it makes a lot of sense to use them with your ultrabook "cluttered with work-related stuff or other mess". :)

Sure, if your work benefits from GPU acceleration beyond what an iGPU can do for you. That doesn't apply for the majority of people (most people don't do extensive media editing, CAD, or other compute-heavy work), but sure, for some. But then again, those people are unlikely to work from an ultrabook, at least as their only work PC. They're more likely to have an MBP 15", XPS 15, or similar GPU-equipped laptop, which means they're 50-150% further in in terms of initial investment for the laptop, too. I would imagine that to lessen the willingness to invest even more for increased GPU performance - if you've already spent $2500 on your laptop, another $6-700 is of course a smaller fraction of your total investment, but the total price suddenly rises to more than 3x a basic ultrabook, and you're partially voiding the usefulness of ~50% of the price you paid for your laptop.

In short: eGPUs in their current state (and likely for the foreseeable future, given prices and market conditions) aren't for people looking for a small boost in gaming performance for their one and only laptop - even though I'd imagine that to be a great market opportunity, there are no products even close to fitting that niche due to excessive prices. eGPUs are luxury products for those with special use cases.

ValantarMultiple high bandwidth standards? Are there viable, non-proprietary alternatives to Thunderbolt? Or are you talking about closed-off, entirely proprietary standards like Alienware's Graphics Amplifier port? And besides, wouldn't that require a PLX chip to switch between the output ports? Last I checked, even the simplest PLX chips are >$100. While I wish there were open alternatives to TB, at least Intel is moving it in that direction.

Other than that, there are plenty of BYO GPU boxes out there. Sadly they're all huge, hulking monstrosities often bigger than a well-built ITX gaming case - and still crazy expensive.

So you're saying you don't think mobile Ryzen 7nm will have better GPU performance than today's thin-and-lights? That seems pessimistic. Or are you saying it won't be better than a 2070/1080? If so, you're right about that - we're not getting 180W dGPU performance in a 15-35W APU in quite a while yet. 7nm isn't that good.

I'm hopeful about the interconnect, though - AMD is bringing PCIe 4.0 this year, and it'd be weird if Intel didn't follow suit, including an upgrade to Thunderbolt. PCIe 3.0 8x equivalent is plenty for any current GPU, which is what we'll get with TB4 if Intel keeps up what they've been doing so far. Hopefully also fully backwards and forwards compatible, which seems more likely given the adoption of USB-C for TB3 and the general growth of TB availability in both hosts and peripherals.

Im hopeful, but this won't be until next year, at the earliest. I wouldn't be surprised if Intel kept TB4 Intel-exclusive in the beginning despite their stated goals of making it open and free, though.

I meant that if it comes and it's likely.
Interconnects, yes they have bandwidth.
TB has latency, and a lot of it thus causing frametime issues!

ImsochoboI meant that if it comes and it's likely.
Interconnects, yes they have bandwidth.
TB has latency, and a lot of it thus causing frametime issues!

Yeah, latency is definitely an issue, but not one that's likely to be solved without creating a PCIe-only port (which would then see very limited aoption. Signal translation takes time, after all, no matter how well-designed the hardware doing the job is.

The conundrum with eGPUs is that they're too expensive to deliver low-to-mid-range performance (which is where it would shine), but lack the performance and have too many bottlenecks to deliver anything near high-end performance. I don't see the latter getting fixed any time soon, so for me the obvious solution (and where there ought to be the biggest market anyhow) is bringing down prices of entry-level and mid-range solutions. The hype around the (known scam) exklim eX core stands as tentative confirmation that a portable, GTX 1050-level eGPU at a reasonable price is something that a lot of people would be interested in. The closest to that on the market currently is Sonnet's RX 560 Breakaway Puck, which is quite expensive at $400 (and here in Norway even worse, thanks to VAT, a small market and a weak currency it's NOK 5500 or around $650 :o ). If it was $200-250, I'd expect it to sell like hotcakes.