Flagship Tuneup History + Study

Flagship Tuneups are our star product, but there's quite a bit of history behind them!

The very beginning

I've been obsessed with making things run cooler for YEARS, even before I started The IT Club. I remember delidding old AMD CPUs with an X-acto knife (only to discover that some of them used solder TIM and I had just destroyed the chip), and obsessively reading Brad's Hacks's article on fixing the thermal situation of a Matebook X Pro. The first device whose heatsink I ever tried modifying was in 2019 so, an Asus Chromebox with a Coffee Lake i7 U-series that I flashed coreboot to in order run to run Windows (quad-core U-series chips were still in the vogue then, and they still work great today!). Ran great, but the fan was quite whiney. I was a broke high-school student and only had 3000+-grit sandpaper on me, so this first attempt was like four hours of attempting to mitigate the horrible surface flatness with sandpaper REALLY not designed for material removal.

When I started The Berkeley IT Club, the procedure we now call Flagship Tuneups was still just a thing I did to my own devices, like one of the many other niche computer enthusiasts on the interwebs. The process was still very rough around the edges, and I would need to iron out numerous issues before rolling this out as a truly hardware-agnostic service. 

The predecessor

I've always been a fan of bigger laptops. 15/16-inch devices built so big and tough where if you drop it, your first thought is whether the floor is okay. A lot of these laptops have quite cavernous internals, so there'd be plenty of room for thermal improvements. Most commonly, I'd add flat heatpipes along with the occasional small aluminum heatsink you'd find on a Raspberry Pi. Below is a P15v Gen 3 I used to own, where I added heatpipes on top of the factory heatsink with thermally-conductive epoxy. 

However, this process came with several downsides:

• Highly device-specific: I wouldn't be able to use a single heatpipe layout for more than a handful of devices.

• Material and shipping costs: Heatpipes can be surprisingly expensive, and the costs would add up over time with heatpipes that either don't fit or get damaged at some point.

• Long lead times: Most heatpipes take upwards of two weeks to ship, making them difficult to integrate into a commercial procedure (if I realized a heatpipe didn't fit and needed a different SKU, it'd take ages to ship! What would I say to a client who's already here?)

• Long procedure times: I needed to take extensive measurements and perform fit checks to find a combination of heatpipes that could fit, and combined with the hardening + curing times of thermally conductive epoxy (plus additional time for it to finish outgassing), this made the procedure a whole-day thing. People typically need their computers all day, so...

• Occupational hazards: Epoxy is dangerous! With enough exposure, people become allergic to it. It also smells terrible and emits harmful vapors when it's curing, and it leaves a mess if you're not careful.

• Inventory issues: Because of the long lead times and sheer number of heatpipe geometries and dimensions, I would need to carry scores, if not hundreds, of heatpipes on me if I wanted to launch this service commercially. This wouldn't even guarantee I'd have the heatpipe I'd need (what if I missed ordering one specific SKU?), and it would lock up tons of money that I could better use elsewhere.

• Airflow issues: In certain devices like Macbooks that intake air through side-facing vents and have it flow through the inside of the device, stacking heatpipes can seriously impede airflow and conversely make heat transfer less efficient. In devices with bottom-facing vents (see bottom-right image), having heatpipes too close to the vent can clog the vent and prevent air from adequately reaching components that might need it, like exposed VRMs.

• Ya can't bend a flat heatpipe: Laptops are the device category with the most thermal issues, and their low amount of internal space means you almost always need flat heatpipes. However, you can't bend a flat heatpipe outside of a factory, because it will bend and crease along itself, caving in its internal vapor cavity and rendering it useless. (Flat, bent heatpipes are commercially produced by bending a round heatpipe and then flattening it, which is something I am unable to do without extremely expensive tooling.)

• Limited device applicability: Sure, adding heatpipes works in a tank of a laptop with ample internal space. But many thinner laptops simply don't have space inside for heatpipes, so this process isn't applicable to every single device. And how would I add heatpipes to, say, the heatsink of a blower-style GPU? 

While adding heatpipes was a fun game of real-life Tetris for me, its fundamental limitations prevent it from being easily used in a commercially-available procedure. 

The new issues

Eventually, I came back to surface lapping. In theory, it's much better than adding heatpipes. It requires no additional space inside a device, doesn't increase weight, and addresses the problem of heat transfer closer to the source. However, there were a few issues that I noticed during field trials on both my own computers (I have a LOT of computers and am EXTREMELY obsessed with making each one run ):

• Creep and pressure distribution on thinner contact plates. Many laptops, especially thin-and-lights, use a contact plate made of a thin, stamped copper sheet. Because of manufacturing tolerances and the poor bending characteristics of a thin sheet, the surface flatness of a stamped contact plate is suboptimal out of the box. Combined with its low thickness (<1mm), I often found that lapping it flat risked causing the contact plate to warp over time due to creep, resulting in noticeably worse pressure distributions (initially it'd show high pressure in the center, but several months later that pressure would shift to the edges of the die).
  I fixed this by minimizing the use of low-grit, coarse sandpaper and shifting some of the material removal onto finer-grit sandpaper. For instance, grits 200-800 would wear away roughly 80% of the total thickness removed during the procedure, and grits 1,000-10,000 would wear away the last 20% of the thickness while also polishing the surface. I also exclude lower grits, like 80-180 grit, from being used to lap stamped contact plates because of their tendency to leave deeper scratches which require more material removal to buff out. In the end, these two techniques result in the same flat, mirrorlike surface ideal for heat transfer while consistently leaving enough thickness to make creep and other time-dependent deformation a nonissue. 

• VRM throttling on systems where power delivery was cooled primarily by direct airflow over the components instead of through a heatsink. Increasing the rate of heat transfer from the CPU/GPU could conversely increase VRM temperatures in systems that spin the fan slower in response to the increased efficiency of heat transfer from the processor.
  To fix this, I used several approaches:

  • In devices with metal chassis components (e.g. higher-end laptops with an aluminum bottom shell), I bridged VRM components (MOSFETs and chokes) to the chassis with high-end thermal pads. From my testing, the thickness of the thermal pads combined with the relatively low heat output of VRM circuitry meant that skin temperatures generally remained acceptable.

  • Burn-in testing. Even if I wasn't able to directly improve VRM cooling, I always perform burn-in testing under worst-case scenarios, and carefully monitor VRM temperatures through software like HWinfo to ensure that the VRMs remain at safe temperatures.

• TIM selection, not just for long-term reliability but also with flexible supply and low MOQs. Consumer-grade thermal paste is sold in relatively small volumes and generally prioritizes performance over long-term reliability (pump-out is a serious concern among nearly all thermal paste I tested).
  Fortunately, the popularization of Honeywell PTM7950 in the computer-enthusiast community instantly solved that issue. Not only does it have performance comparable to freshly-applied Thermal Grizzly Kryonaut (one of the highest-end consumer thermal pastes), it's practically immune to pump-out under even the worst conditions from my testing, and it can be ordered affordably and quickly in any quantity from many suppliers, ensuring that I'll never run out.

• Lapping surface flatness: While float glass is quite flat by nature, I found that it deformed somewhat under the pressure applied during the lapping process, resulting in a slightly concave heatsink surface. Furthermore, the heatpipes immediately below the contact surface are often uneven, resulting in further deviations from flatness.
  I resolved this by adhering a piece of float glass on the opposite side of the surface undergoing lapping, as well as using thicker glass and metal blocks (the steel block in particular is machined down to 1/10,000" flatness) on the lapped surface. This results in a noticeably flatter surface and better pressure distribution.

• Occupational hazards: Lapping, by nature, produces large amounts of copper dust, which is somewhat less-than-delightful to get into your lungs.
  To get around that, I use wet sanding, which traps copper particles in water and also washes away freshly-generated copper dust from the lapping surface, keeping the sandpaper from fouling while keeping my lungs clean. Furthermore, the copper-water slurry is easy to wipe off the work surface once lapping is complete.

The results

Because I took the time to resolve all these issues before launching Flagship Tuneups, I've seen excellent field reliability. It's been one year since I launched Flagship Tuneups, and in that time, I've performed it to over 40 diverse computers, including thin-and-lights, workstation laptops, desktops, consoles, workstations, and even a few servers. Not a single device has come back to me for servicing due to a problem caused by the Flagship Tuneup!

I'm very proud of our star product, and I actually redesigned the logo around that fact. We offer a service that no one else does, at a competitive price, with results unmatched by any other procedure. It was years in the making, and it's gone spectacularly thus far.