Haha, I am glad all this resulted in one more member of the community ;).
2 Likes
A little bit of devil’s advocate here:
Depends on what you do with it. If you’re running workloads that are already close to the limit of the available performance, then no, a 5090 will not run and run and run. (E.g, if you want to run the biggest possible AI models of the current time, 5090 might be ok for 1 year, then obsolete).
Likewise, the current fw16 specs might be outdated for playing whatever the next AAA game will be in just a couple years, sure, but it’s also not obsolete at all for the next probably ten years if you use it for office tasks and light gaming 
.
Also, a precondition to using it for years and years and years is that it doesn’t break in a majorly unfixable way, which is more likely with integrated CPU + GPU on single mobo designs.
(Edit: I’m also not an expert on GPU manufacturing but I think there’s also the argument that they can’t just make everything a 5090. I believe there’s tiers / different quality yields which get turned into different products along the stack)
2 Likes
At least in case of the desktop the 5090 is already the worse bins, the really good samples of the chip go into the rtx pro cards. The mobile 5090 uses (probably efficiency focused) bins of the same gpu as the desktop 5080/5070ti.
3 Likes
Hi, yes, some of the points certainly make sense, and mind you, there is also the price to consider ;). But I maintain the overall argument that a 5090 (even at 100W) will allow you to “buy and forget” anything for many years to come. Meanwhile, spending 780 Euros on 8GB of VRAM at the end of 2025 will be truly sad in 2027 for gamers and other demanding use cases. And just to note, I am bringing these things forward only regarding sustainability. High-end components simply do have an advantage here and should be considered from this angle.
PS: and I did assume that people that purchase dedicated graphics would like to do more demanding tasks (beyond office for example). If you are only into light tasks in any way, then I think the onboard graphics go a long way already - no real need to buy a RTX 5070. The current onboard graphics literally start Cyberpunk (which is insane) and also let you play Baldur’s Gate fairly well. Great tech that they included there.
3 Likes
I agree with you. I think having higher end hardware will keep it in use for longer. For my desktop I’m still running an Intel i7 6850K with a GTX 1080. For nearly all of my use cases this still functions very well. The GTX 1080 is now roughly equivalent with the newest low tier hardware (RTX 5050). Given that the RTX 5050 is low tier hardware it’s expected to perform for a year or two before needing a replacement. If I had gotten a low tier GTX card rather than a high-end card, I would have needed to replace it a lot sooner. Even worse is if I had done that and replaced it with another low tier card I likely would have needed to replace it again.
@Kyuhae It does depend on what you do with it, but in that example, you’re working on something that is bleeding edge and will probably require bleeding edge hardware to go with it. You wouldn’t want to get an RTX 5070 if an RTX 5090 is required. You’d need to replace it even sooner assuming it even worked at all for that situation. RTX 5090 may obsolete in a year, but that would mean the RTX 5070 would likely be obsolete the moment you bought it. Either way the higher end hardware will last longer.
1 Like
This seems very wrong to me. I have a PC with a GTX 750 Ti in it and if you mainly play MMOs, Indies, and JRPGS, that actually still kind of works okay more than 11 years later.
Low tier hardware will play low tier (in terms of tech reqs) games for a long time. The things that are starting to cause problems for Maxwell/Pascal cards are missing features like DLSS and ray tracing that are missing equally from my 750Ti and my Titan Xp 
Not recommending anyone go out and get a 750Ti in 2025 (2GB of VRAM 
), but I think people that are mainly interested in the latest life sims will be well served by a 5050 just as far into the future as AAA obsessed pixel peepers are by a 5090…
3 Likes
This is true to some extent, but the point I’m getting at is that the low tier bar (as with mid and high tier) tends to move up over time. I’m aware that isn’t entirely the case in all situations (like old games not getting updates or new indie games with lower spec requirements), but generally with newer games there’s a newer performance requirement. Also, performance requirements may chang over time for games that are more of a live service/subscription such as MMOs. An example of this is FFXIV dropping support for DirectX 9. If you had old enough hardware that didn’t support the newer version of DirectX you would be required to upgrade or stop playing.
That being said if you’re going to keep playing a game from 2010 with hardware from 2010 then it will be fine until the hardware can no longer function. If you plan on buying newer games, then the likelihood of needing newer hardware increases. Higher end hardware will meet more of those requirements for longer than lower tier hardware.
Sure, but DX10 was introduced in low tier and high tier GPUs in a single generation, so higher tier hardware wouldn’t actually help.
My point is basically that feature set evolution is generational, but horsepower requirements vary more by genre/market segment than by year, which your example appears to agree with.
2 Likes
That’s a great point. I suppose the big difference between hardware tier is the performance/quality you prefer. In most games your performance and visual quality will remain the same within the same game. Although I mentioned feature set which is typically generational, I still stand by that the minimum spec requirements bar tends to rise over time, sometimes due to feature set, but not always.
On the CPU side of things IPC (instructions per a clock) tends to move the bar up and having higher end hardware there can make up for the IPC improvements of newer CPUs (think quantity vs quality). I guess this would also apply to GPUs where the higher number of cores in the higher tier would be more likely to keep up with efficiency improvements of newer GPUs.
With any given hardware your settings will likely decrease over time with newer games/software. Older games would run at high or ultra, and newer games medium to low till they couldn’t run a new game. If your hardware is already running at or near low, there isn’t much headroom as that bar rises. Keep in mind that’s generally for any newer game or software and that not all new games or software may require more performance.
Exactly my point and I would agree fully.
1 Like
This is still a nice setup :). Ah yes, the Pascal cards… I truly respect people that are still on Pascal. 
Just to say, my motivation for my comments actually stems from my own experience from Desktop cards too. I have a RTX3090 for many years already and it is just too good to require any replacement yet. And this is such good news, also considering how hopelessly overpriced and also flawed the following generations were, in particular the 50s series (unfortunately).
And this is also exemplified by ebay prices by the way. Ridiculously, my RTX 3090 costs the SAME on ebay as the price that I bought it for (used and fully refurbished) 3 years ago - no loss of value. Crazy. Meanwhile, I have seen a lot of RTX 3080s that are stone cold dead on the sidewalk and cost next to nothing - because they don’t have 24 GB of VRAM. With a 3090, you can still go on for some time. People notice this and so yet another argument would be pricing.
A Framework 5070 GPU module will be worth very, very little in the future (sadly). If it was a 5090, you would lose less money in the long term potentially because it could seriously be resold or used as a powerful EGPU.
High-end flagships save you money in the future, and are more sustainable. That’s at least my own experience.
PS: and no, not everything should be a “5090”, most people don’t want to spend that kind of money or have such a large graphics module. But my main point is that “higher-end” hardware (e.g. 16GB of VRAM) is more sustainable, and can be resold for a longer period of time. It would be a useful consideration for Frameworks sustainability concept I think.
3 Likes
I could order a FW16 or I could order a Thinkpad P1 Gen 6
FW16 Specs:
Ryzen 7 AI 350 8 cores
RX 5070 8GB 100W
No ram
No storage
No OS
240W PSU ($150)
1 year warranty
$3158 + ram + storage + OS
Thinkpad P1 Gen 6 Specs:
Intel i9-13900H 8 + 4 cores
RTX 4090 16GB 80W
64GB ram
1TB storage
4 years of warranty
$3325.50
FW16 Pros:
Has upgradable mainboard
Has upgradable GPU
Easier to replace input components
Can run full power off of USB-C
Second nvme slot but not full length
Thinkpad P1 Gen 6 Pros:
Much smaller chassis
Higher performance GPU even at a lower TGP
GPU has more VRAM so will be viable longer than a 5070
Has better speakers
Has better screen (Better IPS in this case but 60hz oled is also available)
Has 3 years additional warranty for a comparable price
On sale Legion Pro are even better value and have very good build quality. I wouldn’t touch bargain gaming laptops however. Don’t get me wrong the FW16 is very cool but from someone who does not use their laptop as a desktop replacement but as a laptop and has traveled with a MBP16 for work FW16 is a no go. It’s too expensive for what it delivers for a desktop replacement anyway. It doesn’t deliver much performance for being so large. It doesn’t compete in many areas. It does however compete in upgradability and serviceability but not without sacrifices. For the price I can have a well equipped desktop and a FW13. Which is what I have already but I don’t like being tied to my desk so I tend to use my Legion Pro laptop with 5800H + RTX 3070. It’s pretty big and heavy but the FW16 is even larger.
Seems like it really depends on the country as well. In Germany I would have a hard time to find something with these specs for a comparable price to the framework.
And I would not want to use one of those 1500€ gaming laptops as well: bad screens with 300 nits, 100% plastic chassis, short runtime on battery are just some of the reasons.
In my case that Thinkpad is $2200 and the FW16 with your specs is $2324, which looks to me like a Framework win. I haven’t used a Thinkpad P1, but I have used a Dell Precision 7670, and I did not like it. It was heavy, thick, ran hot, got bad battery life, I didn’t like using a barrel charger, didn’t like the keyboard… it really sucked. The 13900 Intel is 2-3 generations behind the “AI” Ryzens. That Thinkpad’s on clearance now, on the Lenovo website. It’d be fairer to compare it to the original Framework 16 I think.
I can’t speak to the Lenovos, but the Dell was smaller technically but handled like it was larger, because so much of the Framework’s thickness owes to its rubber feet.
(I found dimensions.)
(… Doh! Editing the dimensions to match parawizard’s comment below.)
| Laptop | Height (mm) | Width (mm) | Depth (mm) | Weight (kg) |
|---|---|---|---|---|
| Dell Precision 7670 | 24.7 | 356 | 258 | 2.7 |
| Lenovo P1 6th Gen | 17.5 | 359.5 | 253.8 | 1.9 |
| Framework 16 | 20.95 | 356.58 | 290.2 | 2.4 |
| MBP 16 (2019) | 16.2 | 357.9 | 245.9 | 2.0 |
Either way though, I think the machines are cost-competitive. It’s more down to preference. That was my point. If you were willing to use one of those cheap gaming computers, and just wanted the most frames/dollar, the Framework 16 wouldn’t make any sense.
The cheapest I can get a new Framework 16 is a DIY Ryzen 7 7840HS without RAM or storage, with a 7700s for $1688. Compare that to, on Amazon, “gaming laptop” gets me one of,
- MSI i9-14900HX w/5070, 32 GB of RAM, 1 TB SSD for $1500
- Asus i7-14650HX w/5060, 16 GB of RAM, 1 TB SSD for $1275
- Asus Core Ultra 9 275HX w/5070 Ti, 32 GB of RAM, 1 TB SSD for $2160
There’s probably a quality / personal preference and values argument, but there isn’t a video game frames/dollar comparison to be made really. That’s what I was trying to get at.
(… is that Thinkpad as good a deal on a mobile 4090 as I think it is? That’s impressive, if the thermals and power are good.)
2 Likes
My goodness, that’s a tough one, that exact Lenovo notebook is the one I have been looking for too as a potential replacement.
Just a couple of things to consider:
- the Lenovo can sustain 105w with a bios update (look for reddit threads where it is explained, for example). Lenovo has, after 5 generations, f i n a l l y managed to sustain the high-end GPU, and it is even beyond the 80w advertised. So much more performance to gain there than 80w.
- the lenovo is virtually indestructible - you can keep it for 10 years. To be considered in the equation.
- Framework has OCuLink 4i compatibility (search for the thread here in the forum) - making it an almost unique and true desktop replacement.
Very difficult choice. Two of the best notebooks out there. Razer Blade 16 would complete the top 3 I would say. It depends on your use case of course. Personally, I would go with the Lenovo if you can find it for a good price. But the Framework is more of an adventure, it has a lot to offer in the future. No clear winner here, they are both too good.
The DIY FW16 I spec’d doesn’t have ram or storage. It also only competes with the almost 3 generation old P1 Gen 6 in CPU performance and cannot compete in GPU performance. It also has less warranty and is 30% larger.
Your dimensions are for the FW16 without DGPU
With Expansion Bay Shell:
Height: 17.95mm
Width: 356.58mm
Depth: 270.00mm
Weight: 2.10kg
With Graphics Module:
Height: 20.95mm
Width: 356.58mm
Depth: 290.20mm
Weight: 2.40kg
The 4090 is definitely a lot more GPU. The P1 Gen 7 unfortunately topped out at a 4070 which is a large step backwards. The FW16 competes better with it than the Gen 6. The Gen 8 should be announced this month. Obviously it will be sometime before any deals come up on it. Business laptops typically are not a good deal without discounts. Lenovo typically has large discounts on their product line.
Not interested in OcuLink at all here. A small mitx desktop is not much more expensive than running an external gpu. I’ve done the external gpu thing in the past and I am ok with not doing it ever again. The FW16 is certainly in desktop replacement territory size wise but unfortunately does require an egpu to compete with any desktop graphics.
I fixed my dimensions, thanks.
FWIW, just to make the decision harder, the 7940HS (presumably 7840HS too) have much lower power consumption under load than the 13900H (which makes sense given the process node).
I’ve got a pretty strong anti-Lenovo bias, but I also really like Framework. I don’t know where I’d come down if I had to choose, if Lenovo hadn’t pulled that crap with the spyware. (thought it was more recent than that… I’m getting old.) Anyways, my thought is, once a company bites, you have to put it down. It’s developed a taste for human flesh.
1 Like
The Framework 16 lost in wifi battery testing to the P1 Gen 6 despite this in Notebookcheck’s reviews. It may win in idle but don’t have the numbers for that.
Personally I don’t trust any manufacturer to install Windows or Linux for me including Framework.
To be fair though FW16 is in a larger size class entirely. It is not quite as big as some of the P series but it is closest to the P16s in thickness but in height and depth it surpases all p series laptops by at least an extra inch.
Just want to point out that when it comes to mobile GPU benchmarks/comparisons, you have to be really careful with exactly what parameters that chip is running under. That techpowerup article is a great example, as it quotes the 5070 Mobile’s TDP as 50W, base clock of 907MHz, and a boost clock of 1,425 MHz.
Meanwhile, framework’s soon to be released 5070 dGPU for the 16 has a TGP of 100W (yes, I know TGP != TDP), base of 2 GHz and a boost of 2.4GHz. Glancing at various benchmarks puts the 5070 about 20% faster than the 7700S and 1080ti.
3 Likes