Xex360With Pascal they put the same desktop GPUs on laptops, just reduced clocks, now they sell an overpriced 3080ti that is basically a slow 3070ti, for 5000€ or so I expect to see the full GPU at least the 3080.

Xex360With Pascal they put the same desktop GPUs on laptops, just reduced clocks, now they sell an overpriced 3080ti that is basically a slow 3070ti, for 5000€ or so I expect to see the full GPU at least the 3080.

maxflyThere aren't many options for ANY kind of wcing in a mass produced laptop. There are simply far too many variables at play not to mention the problems associated with trying to physically fit something that might actually work into such a tight space.
Once you get the idea of a custom wced loop out of your head and look at it from such a small physical area. Your not left with much else than what we see here.
Unless you think your customer might be willing to carry around a 240rad and res/pump combo along with their laptop (not likely). It could work but then you've got to design some kind of ultra flat cpu and gpu blocks that wont turn your laptop into something military grade thick. Again the space constraints.
With all the complaints we see with people complaining about their gaming laptops overheating...I think this is a good example of people thinking out of the box for once.

The excellent cooling system of the XMG NEO 17 is one of the most striking features of the laptop, which is designed for uncompromising high performance. The interconnected system with five heat pipes and four heat sinks and air outlets allows the processor and graphics card to be pushed to their ultimate limits, while the two 11 mm fans are characterised by an unobtrusive, low-frequency sound characteristic.

Xex360I'm curious though, is AMD also not supporting this, I had a 6800xt and during gameplay seemed to hover around 230w, with lower voltages and clocks it would be a better solution to the hopelessly slow 3080ti mobile.

bugNot to mention "liquid cooling" is not mentioned on the official website anyway, it's just an unfortunate addition on TPU.

ValantarDesktop PC watercooling: here's my $200, 1kg pure copper GPU waterblock with precision machined microchannels and highly optimized flow paths, alongside my $150 CPU block that's nearly as complex.

Laptop water cooling: yeah, so there's this tiny, kinda flat copper pipe here, you see, that runs kind of across the GPU and GPU in a little loop, on top of the other heatpipes, and it's got water in it.

lexluthermiesterActually, physics says it should work. You can joke, but I'm betting it's effective. That said...

bug175W is about the limit of what it can do. Whereas on a desktop, you can "easily" manage 100W on top of that. Of course, it's still physics. Airflow, more material and more room to work with have to amount to something.

lexluthermiesterActually, physics says it should work. You can joke, but I'm betting it's effective. That said...

ValantarI think this is the same OEM solution that LTT has covered a few times, and apparently it works kind of decently

ValantarWay ahead of you there:

The fact that it works doesn't make the difference between this and desktop custom water cooling any less funny.

AssimilatorIt's so funny how obviously shoehorned in there that one pathetic pipe is. It doesn't even cover the maximum surface area it could, FFS! Just goes to show that fools and their money...

ValantarPascal is also pretty much the only series in which this has been true. And it makes perfect sense - names are only arbitrary signifiers of relative performance between tiers, and laptops are not desktops, so there's no reason to expect them to be identical, especially given the drastically different power and thermal envelopes. And, once again, you're not getting a GA102 in a mobile form factor - it's too large, needs too much space for VRAM, needs too wide a VRM to reasonably fit in even a large gaming laptop, even if they underclocked it for efficiency. It would be a super expensive, bespoke product tier that would only exist in DTR laptops - which barely sell at all. It would lose money for both Nvidia and laptop makers. As @bug said, you put way too much weight on model numbers. Higher number=better, but mobile and desktop are entirely separate lists, and tiers are not cross-comparable.

Xex360Before we had an "m" after the name to differentiate between laptop and desktop GPUs, with Pascal (and later Turing) nVidia removed it because they put the same die on laptops.
The rest of your arguments fall apart as the 3080ti laptops are insanely expensive, and Turing laptop GPU had huge dies like the 2080 super which had 200w unlike the fake 3080ti mobile with 175w.
You and @bug are naïve enough to fall to nVidia marketing, the new naming scheme is a lie to mislead people (alongside the awful power nonsense).

ValantarWell, it's good to know that we are dealing with someone with deep engineering knowledge on a level rivalling at least Nvidia's.

/s

We never saw a mobile TU102 implementation, so... yeah. Still topped out at the second largest die, still topped out at 256-bit memory. Same as now. Turing didn't have quite the same massive transient power spikes of Ampere either, which might go some way towards explaining its marginally higher power limits. But OEMs are also free to configure the 3080 Ti mobile at 175W+ if they want to - that this hasn't happened likely says more about diminishing returns for performance than anything else. Also, do you really need that 'm'? It's a laptop gpu. It sits on a laptop motherboard. It is not a desktop GPU - it can't be. It's obvious Nvidia doesn't want to highlight the difference between the two, but... does it matter? Yes, it's a branding and marketing exercise that removes a tiny sliver of clarity, but the only point at which I find it even remotely problematic is if it were actually confusing to buyers - and I don't believe there are enough people cross-shopping laptops and desktops for that to be much of a problem.

Xex360You are completely missing my point, nVidia shouldn't use misleading names that's all, they used to have a reasonable naming scheme.

lexluthermiesterNo, they saw right through your point. You are missing some context which Valantar was trying to help you understand.

Xex360What context? The only thing I read was excuses/speculatios to justify nVidia's misleading marketing. None of my arguments were addressed.

bugThere are no arguments to address. The correlation between laptop and desktop part names does not exist anymore. Simple as that.

Xex360What context? The only thing I read was excuses/speculatios to justify nVidia's misleading marketing. None of my arguments were addressed.

Xex360Didn't know that, probably nVidia doesn't want competition for their pathetic laptop GPUs, with Pascal they managed to put the same GPUs on laptops (better with the 1070 which had more cuda cores).

Xex360With Pascal they put the same desktop GPUs on laptops, just reduced clocks

ValantarPascal is also pretty much the only series in which this has been true. And it makes perfect sense - names are only arbitrary signifiers of relative performance between tiers, and laptops are not desktops, so there's no reason to expect them to be identical, especially given the drastically different power and thermal envelopes.

Xex360Before we had an "m" after the name to differentiate between laptop and desktop GPUs,

Xex360with Pascal (and later Turing) nVidia removed it because they put the same die on laptops.

Xex360The rest of your arguments fall apart as the 3080ti laptops are insanely expensive

Xex360and Turing laptop GPU had huge dies like the 2080 super which had 200w unlike the fake 3080ti mobile with 175w.

ValantarYour arguments weren't addressed? Have you even had multiple? Let's see:


These are essentially the same thing being said, just from slightly different angles. "Laptop GPUs used to be closer to desktop GPUs, or named differently." Which was, among other places, addressed in this post:

And once again:

This is just naming. Names are arbitrary. You can think that one naming convention is better or worse than another, but that's just an opinion.

No, they removed it because they wanted to remove the stigma of "m GPUs are crap" which had - deservedly - cemented itself over the years.

For the response to that, again, see above.

Moving on:

This isn't an argument, at least not one that relates to the topic at hand whatsoever. Does one segment being more expensive somehow make desktop and laptop GPUs inherently comparable? What? Does the fact that a large excavator costs more than a family sedan suddenly render them somehow comparable? Different things are different things. Mobile GPUs are not desktop GPUs.

And you have already been told that this is not true. The size difference between mobile Turing and Ampere is marginal - they're both huge. Mobile Turing topped out at TU104. I was actually wrong before, as Ampere doesn't top out at GA104, but has the GA103S as its largest mobile variant. These are very comparable in terms of die size - 545mm2 and 496mm2 respectively. Yes, GA103S is marginally smaller, but not to a degree that matters. You have also persitently failed to address the actually space-consuming components of a laptop GPU, which matter far more than die size: the board space needed for ancillary circuitry, mainly VRMs and VRAM. Ampere boosts much more aggressively than Turing and as a consequence has significant power excursions/current spikes, and thus need a beefier VRM for the same wattage. They also top out at the same 256-bit VRAM bus, with 8 VRAM dice - they're at the same level.

What does this tell us? Well, we can use our eyes and look at the boards of desktop GPUs and compare them to the boards of high end gaming laptops. Two things are clear: high end GPUs have huge and/or jam-packed boards with a lot of Z-height for VRM components; laptops are always jam-packed, do not have the luxury for lots of Z-height (outside of a few very thick laptops), and are thus more limited in what you can pack into them. Which is why there's no TU102 or GA102 laptop variants. There just isn't room, outside of the tiny, hyper-expensive niche of DTR laptops - and developing a specific SKU for that would be a massive money sink for everyone involved.

You're also blatantly ignoring the massive power consumption increase from Turing to Ampere and its attendant consequences when looking at mobile SKUs in your insistence on comparing them to desktop cards. The 2080 Super mobile has a 100W lower TDP than its desktop counterpart (configurable up to a 50W lower TDP), at 150W (up to 200W) vs. 250W. So, that's a reasonable gap, right? But how does that change when the 3080 Ti has a full 100W TDP increase over the 2080 Super? That delta grows - you can't just choose to increase laptop TDPs by 100W - physics doesn't work that way. We've also seen a fundamental change in how laptop GPUs are segmented from Nvidia, with the Max-Q labeling being abandoned, as they've realized that it's better to allow each OEM to configure the GPU to what their chassis can actually cool (with a lower bound for acceptable performance per model) than forcing fixed configurations that might not fit a given thermal envelope. This has the disadvantage of less clarity for consumers, as you can't look for the Max-Q label for low power or its absence for high performance, but also has the advantage of providing a wider, more diverse set of options for consumers as laptops can be designed at any point in between those two extremes.

There is also of course the fact that with Turing, 80-tier desktop GPUs were 04-series (with the exception of the Ti, which had no mobile counterpart), while with Ampere, they are 02. This also changes things, no? Again: are you expecting GA102 on mobile? That is a significantly larger die - 628 mm2 - again, and crucially, one equipped with 12 memory channels. Would you want a mobile GA102 cut down to 8 channels? Or are you just selectively ignoring the impossibility of implementing this much VRAM in any reasonably sized gaming laptop?

And, once again, remember the market segments and how they are evolving. Gaming laptops are shrinking. Drastically - the 17" segment today is dominated by thin-bezeled laptops around the size of 15" laptops just a few years ago. Larger than 17" is essentially nothing. And they're getting thinner and more portable too. While you could argue that the growth of SFF means desktops are also shrinking, it's still a world apart in terms of cooling capability and what can actually be fit within the confines of the chassis. A GA102 laptop GPU isn't feasible in the types of laptops that sell in any type of quantity today. Period.

This isn't "falling for Nvidia's marketing". It is taking into account the actual relevant realities surrounding these two separate product segments, and accepting that despite having similar names, names are ultimately arbitrary and for GPUs only designate relative performance within their same generation and brand (with a lot of flexibility). You are insisting that the names should be non-arbitrary, i.e. that there must be a material commonality between a mobile 3080 Ti and a desktop 3080 Ti, beyond both of them being at or near the top of their respective product stacks - extreme performance, highly priced GPUs in their respective markets. This insistence of yours is misguided and misunderstood, and from your arguments seems to be based on some weak reasoning and unrealistic desires. You cannot compare a 3080 Ti laptop to a 3080 Ti desktop and expect similar performance - they'll both be crazy fast, but still wildly different. That was true with Turing and Pascal too. And with previous generations.

Xex360Again you fail miserably to address my points or even understand them. All the points you raised I'm aware of them but are irrelevant.
As I said earlier we can agree to disagree.

Xex360Again you fail miserably to address my points or even understand them. All the points you raised I'm aware of them but are irrelevant.
As I said earlier we can agree to disagree.

Valantar... so, perhaps try to expand on your arguments, if they are so incomprehensible to us? You said you have brought up arguments that haven't been addressed; I quoted the majority of your postings in this thread and showed how it had been responded to. What are we missing, beyond your vague and general thing of disagreeing that desktop and laptop GPUs should be considered separately due to being distinct and separate types of products?