635ml (21.5oz) Ryzen 2400G Mini-STX Build

[chinevo]

It's a bad idea to try cool 2200G\2400G with these coolers. The point is that there are throttling point of GPU at ~62C.

Six months ago, I tried Alseye IO-SH15-65 with 2400G and I had no throttling issue at 80C. Maybe it depends of the batch.

p.s. info about 3600G is fake. AMD doens't even start producing of these chips. I bet on 4\8 and 6\12 chips (of course there will be low-end athlons with 2\4) and I pretty doubt they hit 5.0 ghz ^^

I also think this is a fake. Moreover, I think that for this level of the GPU it will be absolutely enough to have a 6/12 chip with 4.0-4.3 GHz. The only thing that worries me is the number of computing units in the GPU. I believe that it will be increased to 20 as a result of the transition to 7nm. Any way, we have to wait Q3.

 

[dondan]

Why not just get an Dynatron A18?

 

[chinevo]

Why not just get an Dynatron A18?

There is a strict height limit - 12mm.
I can use Dynatron A18 but only without a fan, and I think, this is not a good idea.

 

[boingk]

I think I've got the Dynatron T71 on the way to play around with, should be interesting. Of note is that the stock blower cooler is rated at 9.3cfm - garbage airflow, especially for the (rated) 45dB noise output. I'll be looking at a slim-profile fan to partically cover the heatsink and blow air along the fins.

Throttling doesn't occur on my 2400G, and it gets up to about 75'C.

- boingk

 

[Valantar]

Could you send some links?

No problem The only "exotic" part I mentioned is the vapor chambers. There are some available through AliExpress:
https://www.aliexpress.com/wholesal...id=SB_20190401010636&SearchText=vapor+chamber
Most are sized for LED lamps and other weird form factors, but it should be possible to use some of them on a CPU as long as it doesn't obstruct the cooler mounting holes. AM4 mounting holes are spaced 54mm apart, so a 50mm vapor chamber should be pretty much perfect. If you can't find that, just find one that fits inside of the 90x54mm pattern of the mounting holes, e.g. a 65x80mm one like this.

After that, just find a suitable heatsink that's large enough to drill mounting holes through (ideally as large as you can fit on your board), drill the required holes, and mount with screws down into the standard backplate (I think the screws are M5, but don't quote me on that). I'd probably add some appropriately sized spacers (cut plastic tubing or similar) between the motherboard and heatsink for the screws to pass through to avoid over-tightening and crushing the vapor chamber.

I found this recommended as good, easy-to-use thermal adhesive, but given its rather abysmal thermal conductivity, you might be better off just stacking everything and using a thin layer of high-quality paste between both the CPU and vapor chamber and vapor chamber and heatsink. Just be sure to spread it thin between the vc and heatsink, as there won't be much pressure to spread it out there. Ideally these would be soldered together, but I don't think that's feasible without industrial equipment.

As for aluminium heatsinks, they're a dime a dozen (no, not literally) on Ebay and AliExpress - just look around until you find one that fits your measurements, fin direction/spacing/pitch, etc.. They can be ordered in custom sizes (at least in one dimension), from what I've seen. Copper would probably be better, but I wouldn't trust that to not corrode when put next to an aluminium vapor chamber.

As for fans, I'd probably go for a high-end gaming laptop fan, as large as can fit - current generations tend to be relatively quiet while delivering good airflow (for a radial fan, that is). Can be difficult to find without a part number, but searching Ebay for [laptop model]+[year]+replacement fan ought to work. I'd mount the fan off to the side of the heatsink with a 3D-printed casing (dependent on clearance, of course), and find a heatsink with relatively dense (ideally skived) fins - the one strong point of radial fans is static pressure, after all, which means that they're able to maintain their modest airflow even through large and dense heatsinks.

If I could find a top-down image of the A300 motherboard, I'd make a mock-up of how I'd put this together, but sadly I can't find one ...

 

[chinevo]

@Valantar
thanks a lot, I will research this topic.

PS: I watched der8auer's videos about 2400G delidding and AMD Ryzen direct die cooling. Decided to go "direct die" in my build. This not only helps make CPU/GPU cooler but also gives extra millimeters.

 

[Valantar]

@Valantar
thanks a lot, I will research this topic.

PS: I watched der8auer's videos about 2400G delidding and AMD Ryzen direct die cooling. Decided to go "direct die" in my build. This not only helps make CPU/GPU cooler but also gives extra millimeters.

Sounds good! That makes not over-tightening the cooler even more important, though, as you're concentrating the mounting force onto an even smaller area (and you now have the risk of cracking the die as well). I have no idea how resistant to crushing a vapor chamber like that is, but I don't expect it to be very durable (conventional vapor chamber designs have internal "support columns", but it's still a thin sandwich of metal and empty space/vapor). Might be a good idea to have some sort of CPU shim made (wow, this brings back some early 2000s memories!) to alleviate the risk of cracking the die and give a bit more support to the vapor chamber, but it'll need to be very precisely made to not lift the cooler off the die, of course.

 

[boingk]

You're going to go direct die, with a compromised air cooler, to save 3mm of height? Thats 67cc of volume over a 15x15cm STX area.

Dedicate? Sure. Efficient? Debatable.

Looking forward to whatever solution you go with, some creative thinking going here for sure!

 

[chinevo]

You're going to go direct die, with a compromised air cooler, to save 3mm of height? Thats 67cc of volume over a 15x15cm STX area.

Dedicate? Sure. Efficient? Debatable.

Looking forward to whatever solution you go with, some creative thinking going here for sure!

Yes, you are right about efficiency. But this is an uncompromising build. The main goal is to make it as small as possible, I don’t give a damn about practicality Cooler from 10 years old GPU, direct die, motherboard modifications, all of this stuff inspire me

 

[Valantar]

I made a rough mockup of a cooler layout using a 65x80mm vapor chamber, as big a heatsink as could fit, and two (random) laptop cooler fans (ended up using Sunon Maglev fans from the Asus ROG G771 - they're only 9mm thick, so they ought to fit well). Vapor chamber size is to scale. Heatsink size is around 144x93mm.

You'd have to remove the fans from their housings and rig up a custom mount and cover for the heatsink to ensure proper airflow, but a layout like this ought to be as good as you'll get for this kind of size. If I were actually making this, I'd probably want to have a heatsink that went underneath the fans, cut/machine the fins off in the relevant area, and mount the fans directly to the aluminimum base of the heatsink, then 3D print or otherwise rig up a surround for the fans to direct the airflow where it has to go.

Of course this needs to clear the on-board USB ports, m.2 slots and everything else, but it could work.

Edit: didn't realize until now that I've blocked two of the mounting holes with the fans. Oops. Could possibly still work with countersunk screws, but only if the fans are separable from the heatsink itself.

 

[chinevo]

@Valantar wow, exciting solution. But, as I understood after some research, the vapor chamber is excellent for delivering heat over a considerable distance, but for a small setup, it does not bring additional efficiency. Therefore, I decided to do without a vapor chamber in at least the first version. The chosen solution is a cooler from the Zotac 9800 GT and a direct die.

 

[Valantar]

@Valantar wow, exciting solution. But, as I understood after some research, the vapor chamber is excellent for delivering heat over a considerable distance, but for a small setup, it does not bring additional efficiency. Therefore, I decided to do without a vapor chamber in at least the first version. The chosen solution is a cooler from the Zotac 9800 GT and a direct die.

The Zotac cooler is probably the easiest solution, but your reasoning is a bit off. A vapor chamber is essentially just a very large, flat heatpipe. "Considerable distance" is as such an exaggeration - even tiny coolers like the Noctua L9a/i use heatpipes to great effect, after all. The point is to spread as much thermal energy to as large a surface area as possible, and using just massive metal is quite inefficient at this, especially if the base of your heatsink is on the thin side (thinner heatsink base = smaller cross sectional area = less efficient thermal transfer). This is why good heatsinks, even huge tower coolers, have their heatpipes spread through the fin stack rather than grouped together - the closer a surface is to an efficient thermal energy transfer matetial it is, the more energy will be transferred there, and the more energy will be dissipated when air flows over said surface.

Put into more practical terms, the difference between a pure metal heatsink over a heat source and one with a vapor chamber (or heatpipes) between the heat source and fin stack is that the former will only really cool well in a small-ish circle around the heat source (APU in this case) with the rest of the heatsink never really getting warm, while the one with the vapor chamber will be efficiently dissipating thermal energy across a far larger area, meaning that it can dissipate far more heat. The distances involved really don't need to be large for this to have an effect - there's a reason these are used in server heatsinks (or Nvidia's previous Founder's cards). Those coolers are less than 100x100mm after all. No great distances involved.

 

[chinevo]

@Valantar Thanks for the explanations, I totally agree. I think I'll try this approach in a few months.

 

[Vlad502]

I just edited post #34

 

[chinevo]

I just edited post #34

Thanks a lot!

Today I received my DeskMini A300 and made some measurements.

Here they are:

I just need to stick to the previous plan.
- cut heatsink
- remove VGA
- delid CPU

To be continued...

 

[chinevo]

I made some temperature tests. I hate all stress test software, that’s why I chose 10 min play Metro Exodus. Open stand, 1440x900, low graphic settings (40-60fps).

1. Stock cooler from Asrock (with DeskMini)
Noise: low
Temps: 55-57C (CPU & GPU)

2. Stock cooler from AMD (AMD Wraith Stealth)
Noise: low
Temp: 55-57C (CPU & GPU)

3. Zalman CNPS2X
Noise: medium
Temp: 62-64C (CPU & GPU)

4. Cooler from 8600 GTS
Noise: high
Temp: 72-74C (CPU & GPU)

5. Cooler from 9800 GT
Could not install without modification.

 

[Windfall]

I made some temperature tests. I hate all stress test software, that’s why I chose 10 min play Metro Exodus. Open stand, 1440x900, low graphic settings (40-60fps).

1. Stock cooler from Asrock (with DeskMini)
Noise: low
Temps: 55-57C (CPU & GPU)

2. Stock cooler from AMD (AMD Wraith Stealth)
Noise: low
Temp: 55-57C (CPU & GPU)

3. Zalman CNPS2X
Noise: medium
Temp: 62-64C (CPU & GPU)

4. Cooler from 8600 GTS
Noise: high
Temp: 72-74C (CPU & GPU)

5. Cooler from 9800 GT
Could not install without modification.

So the 8600GS cooler mounted without mods?

Get a Noctua fan, and dissect it.

 

[chinevo]

So the 8600GS cooler mounted without mods?

Yes, but of course I could not mount it, I just put cooler on CPU.

 

[Windfall]

Yes, but of course I could not mount it, I just put cooler on CPU.

Temps would be better if you could mount it, correct.

 

[Vlad502]

Today I received my DeskMini A300 and made some measurements.

Here they are:

About direct die cooling, look this video:
Socket plastic is a bit higher than CPU