One of the things I’ve been working on is a way for blocking the dGPU vents for when more CPU cooling is needed and there dGPU is either not in use, or doesn’t need as much airflow. This is version 3 of my attempt at this, and it may be the final version. Their job is simple: block the dGPU vents when I want them to, and stay out of the way (but available) when I don’t. The blockers just slide on the guide track. This also allows full linear control for how much blockage is desired :D.
STL’s:
Note: I recommend printing in at least PETG for the added heat resistance.
supportBar.stl (684 Bytes)
gpuPlugV3.stl (3.0 KB)
gpuEndCapBeta.stl (6.9 KB)
For assembly, I super glued 2 of the support bars together lengthwise, and then glued them into the end caps. I may add adhesive under the support bars just to keep them from moving around too much, but it was designed specifically so that this would not be required. The whole system hooks into the cpu vent as a latching point and is strong enough to hold the weight of my entire FW 16 on that latch. To remove, just unlatch!
Admittedly… after overhauling my cooling solution, there’s really no need for this. I did it anyways.
4 Likes
Out of curiosity is there a reason you have what appears to be 3 small sections for each side rather than 1 large and one small? Was it just easier to print 3 of the same size?
Only reason is that 1 model was simpler than 2 in modeling. A single large one for the center sliders would work just as well. Good question 
1 Like
I’ve added a wedge airflow cover for the bottom vents now, check it out! I messed with many different versions, most of which used the CPU vent ports to hang the wedge from a clip like the dGPU vent blockers do… but this just proved to be kind of miserable in practice. So, I opted to cave and use an adhesive, in this case: Velcro. That approach was an immediate improvement, and looking at the Velcro material specs sheets suggests that the adhesive should be fine with the temperatures at play at the bottom surface of the laptop.
Was cutting out the holes for the bottom vents necessary? No.
Was it recommended? No.
Was it fun? Well, also no.
But does it just scratch an itch I had? YES.
wedge_beta4.stl (11.4 KB)
Here’s a pile of all the other prototypes I made before settling on the version above:
So, I did some rudimentary testing (played an hour of Overwatch) and noticed a couple things.
Temps: they didn’t really change. This is kind of expected though, as I wasn’t maxing out the fans before, so I’d guess the fans just continued to balance out at whatever point they like (for me, that’s around 65c on the GPU and 70c on the CPU).
Noise. On the Performance profile, there wasn’t much of a noticeable difference, at least not one that I can be confident in. But on the balanced profile, holy cow the impact was immediate. Playing Overwatch on balanced had the fans whisper quiet at what I would consider an idling noise level (assuming the fans aren’t off entirely). Only once I switched to Performance profile did the fans begin to sound like a gaming laptop’s fans. It’s possible even on performance they’re quieter, but I can’t say for sure. I don’t have strong data.
In theory, the power draw should also be lessened. Given that I’ve reduced the air resistance, the fans should need less power to move the same amount of air. I’m hoping that this translates into a reduction in the impact of the battery flipping issue under high load, but I’ll need more testing to validate that. Unfortunately, I don’t expect these power efficiency gains to affect battery life in practice, because my battery-operated workloads rarely turn the fans on to begin with anyways. Prior testing tells me though that before cutting out my bottom vents my fans drew 10W total (they’re rated for 12W). I’m not sure if that 2W delta is a result of my removal of the top vent cover or not.
Update. Using framework-control I was able to run a simple test to check if there’s any power savings from opening the bottom ports. It looks like removing that air-restriction dropped my fan’s max-power usage by about 2.25W. If the battery is being drained at a rate of 10W during heavy load, then this translates into an up-to +20% time-delay before the battery flipping issue occurs (in theory) and ever so slightly increases the window within which the battery draw is not needed to begin with. So… I’ll call that a win.
3 Likes
I love seeing your crazy FW16! At this point I don’t think I’d consider it a “laptop” 
, but the improvements and efficiency gains are interesting to see. It does show just how much we give up for a polished portable laptop. If I had the time, money, and means I’d probably try to water cool mine.
2 Likes
Thanks! It’s been a lot of fun, and yes, I’m definitely stretching the definition of laptop as far as I can. I’ve thought about water cooling, and (through work) even ended up visiting a data center immersion cooling vendor’s HQ and got the chance to ask them about immersion cooling a laptop. To no surprise, they said that’s a terrible idea… but to my surprise they said that if I did pull it off, they estimated about a 6% (IIRC) increase in battery life.
So far, I’ve decided that both aren’t worth it, even by my flexible standards. Liquid cooling could be worth it if it was like those other setups I’ve seen where it’s liquid cooling in addition to the air cooling. Actually, now that I think about it, the main downside to those was the extra space it needs and that often requires cutting into the chassis… which I’ve already done. Hmmmmmmmmmm.
3 Likes
Immersion cooling is a bit different than water cooling. I’m not entirely sure how you would do that with a laptop aside from just dunking a laptop in mineral oil or something. As for the water cooling I think it might be easier. I have seen LTT water cool a laptop before so it’s definitely possible. I think the main benefit would be pulling the heat away from the laptop. You could use larger, quieter, better cooling fans to dissipate the heat for a less noisy laptop in conjunction with a radiator obviously. Also, there is a potential for lower temps. The downside is it becomes significantly less portable. IIRC LTT use quick connects for the connections between the laptop and the radiator.
I do think the FW16 is much better suited for something like water cooling given you can remove the expansion bay. I think the best solution to go about it would be to have a whole new expansion bay shell for the GPU. Something much larger to accommodate the water cooling plates while still able to connect to the FW16.
1 Like
Especially with the good stuff like 3M novec and stuff having been banned for being really bad for the enviroment
Essentially you’d just need to get access to the heat-pipes and somehow mount some form of low-profile water-block and quick disconnects to it.
1 Like
I had the opportunity to test the impact of the dGPU vent blockers on fan RPM and thought I’d share.
Under a heavy 100% CPU load (compiling a gentoo liveiso in this case) my fan speeds (as reported by sensors) averaged ~4350 and ~4100 RPM for fans 1 and 2 with the vents open. I don’t know why they’re different, but oh well. When I then close the vents, those values dropped to a repeatable 4100 and 3900, which is about 6%.
Now, it’s worth mentioning that I have also removed nearly all the protective grill material from my fans, and so my device is operating very closely to the performance of an open-air bench test. I do not know how or if that impacts the relationship between closed/open vent RPMs, but it definitely impacts the actual RPM values themselves.
For now, I’m satisfied in knowing that there seems to be (on my machine) a 6% drop in fan speed when I use the vent covers, which also results in perceptibly quieter fans. Just thought I’d share, that’s all. 6% is not nearly as much as I would’ve guessed intuitively… but I imagine that speaks to the engineering that went into making the laptop by people far more knowledgeable about this than I.
1 Like
