P2 - Cool, Hot, or Wot?
Peter Jakacki
Posts: 10,193
Yeah, the P2 is a very cool chip indeed but it does not run cool. Some say it runs hot. Wot then is hot?
Let's check it out with a P2D2 that I have mounted flipped with the chip facing down and a thermocouple taped with polyimide to the rear of the pcb smack bang center of the P2 hotspot. My bench P1 reports the ambient as 26.7'C with 55% RH in still air, no fans or A/C etc.
Now I will test the P2 using a new version of TAQOZ that has VGA and a keyboard connected as well as other devices. I am mainly using two cogs, one for TAQOZ and one for 640x480x8 VGA and the TAQOZ console is switched over to VGA and keyboard displaying text and bmp graphics etc. The 12MHz clock divider is set to 1 and I am multiplying from there.
So I wouldn't operate the P2 at 300MHz without cooling unless it was used indoors where the ambient is low but by the same token even a tiny bit of heatsinking with my finger will immediately cool it and even using a very small 25mm fan will prevent it from ever feeling hot to the touch even at 324MHz. When I get some new chips during the week I will check one out on the current P2D2 before deciding if I want to add extra layers. My opinion is that I don't need extra thermal layers, but if run at 300MHz or so I just need a tiny bit of air flow or place the back of the P2D2 against a metal case.
BTW - activating all the cogs doesn't really seem to make much difference.
Let's check it out with a P2D2 that I have mounted flipped with the chip facing down and a thermocouple taped with polyimide to the rear of the pcb smack bang center of the P2 hotspot. My bench P1 reports the ambient as 26.7'C with 55% RH in still air, no fans or A/C etc.
Now I will test the P2 using a new version of TAQOZ that has VGA and a keyboard connected as well as other devices. I am mainly using two cogs, one for TAQOZ and one for 640x480x8 VGA and the TAQOZ console is switched over to VGA and keyboard displaying text and bmp graphics etc. The 12MHz clock divider is set to 1 and I am multiplying from there.
>5 mins FINGER >2 mins NO COOLING TOUCH 25mm FAN NOTES 100MHZ 43'C 33'C 120MHZ 45'C 36'C slow to heat 180MHZ 54'C 47'C 39'C 240MHZ 59'C 50'C 40'C 300MHZ 63'C 52'C 43'C 324MHZ 66'C 55'C 44'C burns my finger
So I wouldn't operate the P2 at 300MHz without cooling unless it was used indoors where the ambient is low but by the same token even a tiny bit of heatsinking with my finger will immediately cool it and even using a very small 25mm fan will prevent it from ever feeling hot to the touch even at 324MHz. When I get some new chips during the week I will check one out on the current P2D2 before deciding if I want to add extra layers. My opinion is that I don't need extra thermal layers, but if run at 300MHz or so I just need a tiny bit of air flow or place the back of the P2D2 against a metal case.
BTW - activating all the cogs doesn't really seem to make much difference.
Comments
Almost all kits for the raspberrypis, now have heatsinks included, and their cases have fans now too.
...
Nuthin wrong with that. Heck, take advantage of the pi form factor/pinout/usbpower methods, etc... and use them with the p2, and boards.... and you have cases and addon-"hats", power supplies, and heatsink kits all ready to rebrand. ;D
This data makes it look like maybe a nice heat sink would be good enough.
With a fan if enclosed...
I think a fan would only be "required" if you are doing the extreme overclock or have it enclosed in a small case that doesn't have much room for normal convection.
I haven't had time to test this thoroughly yet. Still working on the VGA and streamer.
These tests were mainly to confirm "wot hot is" since you never actually measured it except with your finger
BTW, I start up with black text on a white background. It helps your monitor's autoset if you have a non-black background, especially at the edges.
Added digital, not replace.
True, but that is once-only.
Temperature does affect Analog operation, and it does affect reliability.
General industry claims of 10~15°C elevation halving expected life, and thermal cycling over full range reducing lifetimes by 8x, show that °C is the enemy of reliability.
? fuzzy ?
Can you expand ? Do both your working monitors show the same effect ? Is that over whole display area ?
Maybe the DAC speed degrades with temperature, which would give a loss of focus type effect.
There are already P2D2 with 2 layers, for 2 layer testing.
To me it makes more sense to next do a 4 layer P2D2, with any 'free' copper added. eg pcbway offers 1.5oz inner layers for no extra $
You can still use Chip's board for larger area and thicker copper comparisons.
I have a feeling solder mask is not a good thermal conductor...
Then, I'll see about silver pasting on a heat sink.
There are some figures here for it
https://www.intersil.com/content/dam/Intersil/whitepapers/power-module/carry-heat-away-from-power-module.pdf
While its a whole lot thermally conductive compared to the copper layer beneath it, 0.2 C per watt is only going to add ~ half a degree to the overall temp rise, and thats based on 1 square inch rather than the 2 or more we're talking about
There are also some solder-able small heatsinks that could suit P2, and for those bare copper would help.
I'm not sure which is best for simple radiation, exposed PCB, or black resist ?
I am only at 148.5MHz so there seems to be a heat problem here.
Strange, can you measure the actual temperature then. This is my clock header I use:
My temp probes are somewhere still in boxes after the last move. Insufficient time to unpack atm.
I'll post the latest code as soon as I complete the latest tidy-up.
-Phil
P2 is not dissipating kilowatts, but a 'small plastic case' is probably off the table, and not likely with the larger package anyway.
The simple trick is to move the air with a decent fan. The fan can be tiny, it just has to move the air.
You can still have a fully sealed, even IP68 stainless steel enclosure, so long as you circulate the air within. It really isn't a big problem (based on a few decades making products typically around 10~18W dissipation).
Remember we're running torture tests to push the limits. What current would a P1 running 8 cog torture test draw at 100 MHz? If someone has some P1 torture test code, we can test a FLIP side by side, MIPs for MIPs, with the thermal camera.
-Phil
On the fans thing, use a quality fan with a decent L10 rating, not a cheap CPU cooler designed to give you an upgrade every 3 years. From the other thread where we dealt with same:-
If you look at the specs on quality (eg) ebm papst fans you'll see many models that have L10 figures well over 10 years. If you go by IPC L10 figures, many approach 20 years, and L10 means 90% of them continue operating past that point.
I recently did an LED-string special effect (shooting star, aurora) for a school play. I used a FLiP module (one of Parallax's best products!) and had no concerns about heat, even in the small, plastic single-gang junction box I designed the driver PCB around. What I hadn't planned on was the 1/8th-brick 24V-5V DC-DC converter plugged into the same PCB getting hot, given its 90+% efficiency rating. But get hot it does, even when idling, even with a heatsink. So I had to drill 1" holes in the box, above and below the converter and cover them with screen wire to keep the bugs out. That was enough to keep things relatively cool. But there's no way that such a small box -- even a metal one -- could have radiated the heat generated therein.
-Phil
Presumably bigger than the 1" holes...
-Phil