The tale of the cases


Grench said:
If we're going to turn this case thread into another SoC dream thread, may as well throw this in.

I have no idea if this would fit the thermal envelope of the Pyra or if the 21mm X 21mm physical size is too large.
https://www.renesas.com/en-us/solutions/automotive/products/rcar-h3.html
But - it has a fun specsheet and there is a dev board for it.
https://www.renesas.com/en-us/solutions/automotive/adas/solution-kits/r-car-starter-kit.html
Note the big heatsink & fan. If used in a handheld fashion, this monster would have to be seriously throttled back.
Automotive grade so it should have availability for an extended time.
Linux is listed first in Supported OSs.

Probably too expensive, not available to the project, too power hungry, etc... But - it has a fun spec sheet.
Click to expand...
it's not a soc dream post. It's a planning for the future by preparing it right now post. That's it. That's what it was about because since the thing is doing extreme thermal throttling right now and since it's made to go to 1.5ghz it should be able to do 1.5ghz at less than 90*C. So there's little to no reason to not do the dual heatsinks. It'll reduce the overall temp of the soc and will also thus make even the socs that would normally have to be junked due to running too hot be useable.
 
133794m3r said:
it's not a soc dream post. It's a planning for the future by preparing it right now post. That's it. That's what it was about because since the thing is doing extreme thermal throttling right now and since it's made to go to 1.5ghz it should be able to do 1.5ghz at less than 90*C. So there's little to no reason to not do the dual heatsinks. It'll reduce the overall temp of the soc and will also thus make even the socs that would normally have to be junked due to running too hot be useable.
Click to expand...
You seem to be under the impression that CPU & SoC should be able to run balls out continuously. Good luck with that. Everything post the Intel Pentium Pro throttles.

Today's thermal solutions are about letting the SoC throttle less OR finding an acceptable clock rate for continuous.

I'm good with it if it can run 2 cores @250 MHz continuous no throttle inside a handheld computer with passive cooling.
 
I guess I don't understand what is going on in here. Why do we need a better cooling than we need now to prepare for a future CPU board? Wouldn't it make more sense to deal with that when we get there since a heatsink or other cooling solution isn't something permanent?
 
Grench said:
You seem to be under the impression that CPU & SoC should be able to run balls out continuously. Good luck with that. Everything post the Intel Pentium Pro throttles.
Click to expand...
Yup, my new work laptop logs stuff like "CPU0: Package temperature above threshold, cpu clock throttled" whenever I compile things or do other really resource intensive stuff. The CPU is Intel Core i7-7500U for reference. I think two metrics should be provided: peak performance and continuous performance. Currently it's typical to only provide the former.
 
133794m3r said:
since it's made to go to 1.5ghz it should be able to do 1.5ghz at less than 90*C.
Click to expand...
Sadly that seems to be a thing that will never happen, you'd best temper your expectations.
And I'm not saying in just this product, I'm saying that that's the direction the market has gone, seems like this is how CPUs are being made and marketed: built for bursts but normal operation throttled to maintain heat control, but marketing will always list the highest number. It's unfortunately confusing for someone who grew up as processors got more and more powerful, especially if you lived through Intel's near-sighted "more Mhz is literally all that matters!" phase, but I've come to realize it's actually quite the reasonable tradeoff.
If you don't like it then you should be complaining to Intel and ARM, ED has no control over what they say a CPU is "made to go to".
 
Grench said:
Everything post the Intel Pentium Pro throttles.
Click to expand...

I'm quite impressed that the Conroe era 2,4GHz Pentium Dual Core chip in my inherited desktop build machine hasn't been seen to throttle yet under load. It'll drop the CPU freq as soon as you take the load off, presumably to bring the core temp down quicker, but I've never seen it do that while stuff is still running. I think it's running the stock CPU fan and one case fan, and I think I'm using the stock linux sheduler cpu governor.

It's completely irrelevant to this thread of course, since that box could easily hold around 100 pyra in I reckon, but as an aside desktop CPUs can get quite a lot faster than the Pentium Pro and not throttle.
 
levi said:
I'm quite impressed that the Conroe era 2,4GHz Pentium Dual Core chip in my inherited desktop build machine hasn't been seen to throttle yet under load. It'll drop the CPU freq as soon as you take the load off, presumably to bring the core temp down quicker, but I've never seen it do that while stuff is still running. I think it's running the stock CPU fan and one case fan, and I think I'm using the stock linux sheduler cpu governor.

It's completely irrelevant to this thread of course, since that box could easily hold around 100 pyra in I reckon, but as an aside desktop CPUs can get quite a lot faster than the Pentium Pro and not throttle.
Click to expand...

With an appropriately sized active cooling solution with variable ramping fans and no overclocking, nearly any SoC can be prevented from throttling.

In theory, you could stop or remove the fan to test the CPU's throttling capabilities, but I don't suggest this for any computer you actually want to use ever again. If it fails to throttle, it could very well melt the system board.
 
Well of course if there's infinite space for a cooler, eventually you can make any CPU not throttle, with the main limits being current draw (because the faster you overclock, the higher voltage must be upped for stability).

Kind of like how if we returned to hard drives platters the size of these, using modern technological methods...

IBM_350_RAMAC.jpg


...we'd have no problem being in the petabytes of storage on a home PC in all likelihood.

(For the curious: That stored 3.75 MB, spinning at 1200 RPM, and is 50 platters with 100 writing surfaces overall.)
 
Last edited:
If they still cost the same as the 350RAMAC did then not many houses would actually have one. I'm personally content that I can get a hard disc these days in the same form factor as I did in 1994 with 25000 times the capacity and it cost me about a tenth of the cost.

Grench said:
With an appropriately sized active cooling solution with variable ramping fans and no overclocking, nearly any SoC can be prevented from throttling.

In theory, you could stop or remove the fan to test the CPU's throttling capabilities, but I don't suggest this for any computer you actually want to use ever again. If it fails to throttle, it could very well melt the system board.
Click to expand...

I did that briefly while monitoring the temperature one time. It did ramp up from there quite quickly enough for me to quickly get the fan going again.

I guess with your comment about the Pentium Pro you mean that was the last time that x86 CPUs IIRC were not shipped with active cooling. It did have a big metal heatsink (as did devices dating back to late 386s at least). I'm not sure whether the early Pentium2 chips that came in those weird vertical plastic cases had active cooling inside those boxes or not. Traditionally ARM chips up until about 2010 came bare, but these days more and more of them are needing cooling solutions to run fast for any appreciable length of time, from passive radiating heatsinks to active cooling. I agree that moving from passive to active cooling is a significant change in the technlogy, but from another perspective is all part of the same journey, and the distinction didn't occur to me when I read your post. Personally I consider the move from bare chips to all that involving paste, clamps and relatively big chunks of metal a bigger step in the evolution of chip cooling, but maybe that's just me.
 
levi said:
If they still cost the same as the 350RAMAC did then not many houses would actually have one. I'm personally content that I can get a hard disc these days in the same form factor as I did in 1994 with 25000 times the capacity and it cost me about a tenth of the cost.



I did that briefly while monitoring the temperature one time. It did ramp up from there quite quickly enough for me to quickly get the fan going again.

I guess with your comment about the Pentium Pro you mean that was the last time that x86 CPUs IIRC were not shipped with active cooling. It did have a big metal heatsink (as did devices dating back to late 386s at least). I'm not sure whether the early Pentium2 chips that came in those weird vertical plastic cases had active cooling inside those boxes or not. Traditionally ARM chips up until about 2010 came bare, but these days more and more of them are needing cooling solutions to run fast for any appreciable length of time, from passive radiating heatsinks to active cooling. I agree that moving from passive to active cooling is a significant change in the technlogy, but from another perspective is all part of the same journey, and the distinction didn't occur to me when I read your post. Personally I consider the move from bare chips to all that involving paste, clamps and relatively big chunks of metal a bigger step in the evolution of chip cooling, but maybe that's just me.
Click to expand...

No - the Pentium Pro was generally expected to be under a heat sink with a fan. That is not the same thing as throttling. Keep in mind, we're talking about 20 years ago, which is a lifetime in CPU advancement.

http://www.tomshardware.com/answers/id-2841505/intel-cpu-thermal-protect.html

I used the Pentium Pro as an example of CPUs that did not have the ability to 'throttle down' when under load to prevent overheating. I don't think they even had temperature sensors in the CPU back then. I/we used to just strap a heatsink & fan to it and hope that the solution supplied enough cooling to keep the thing running - no real way to even measure how hot the thing ran. They didn't even report temperatures back to the BIOS.
 
Yes, that's true. I've only learned how to monitor temperatures in my OSes in the past year or two. Before that I'd have to guess by the temperature of the air coming out of the vents, and the fact the machine had or hadn't crashed yet. Provided you yank the power after the first crash the silicon's usually intact once it's cooled down enough to be usable again.
 
ED talked about doing 2 heatsinks on the system not just a single one. Plus he talked about adding a fan to the system if he had to. A very small one. My post was about doing the dual heatsinks and then keeping the space for the fan in the mould and it could still fit. That's what I'm talking about. You might as well go with the plan that you already had.

CPUs do throttle but not if they're cooled properly. On x86 you won't burst anymore upon reaching too high temps. My laptop doesn't go below the 2.4ghz that the standard clocks are. Plus you're talking about future socs as if they won't be even more insane in terms of thermal runaway. Doing what you suggest will require the system to be completely redone for each SOC. Unless ED lied in his original plans for the pyra it's supposed to just replace the cpu board for future versions. And if you're doing that, you can't after the fact add heatsinks to the case and get back space you don't have.
 
@133794m3r your laptop must not have very good battery life.

Code: 
$ cpufreq-info
cpufrequtils 008: cpufreq-info (C) Dominik Brodowski 2004-2009
Report errors and bugs to cpufreq@vger.kernel.org, please.
analyzing CPU 0:
  driver: intel_pstate
  CPUs which run at the same hardware frequency: 0
  CPUs which need to have their frequency coordinated by software: 0
  maximum transition latency: 0.97 ms.
  hardware limits: 400 MHz - 2.40 GHz
  available cpufreq governors: performance, powersave
  current policy: frequency should be within 400 MHz and 2.40 GHz.
                  The governor "powersave" may decide which speed to use
                  within this range.
  current CPU frequency is 892 MHz.
analyzing CPU 1:
  driver: intel_pstate
  CPUs which run at the same hardware frequency: 1
  CPUs which need to have their frequency coordinated by software: 1
  maximum transition latency: 0.97 ms.
  hardware limits: 400 MHz - 2.40 GHz
  available cpufreq governors: performance, powersave
  current policy: frequency should be within 400 MHz and 2.40 GHz.
                  The governor "powersave" may decide which speed to use
                  within this range.
  current CPU frequency is 854 MHz.
analyzing CPU 2:
  driver: intel_pstate
  CPUs which run at the same hardware frequency: 2
  CPUs which need to have their frequency coordinated by software: 2
  maximum transition latency: 0.97 ms.
  hardware limits: 400 MHz - 2.40 GHz
  available cpufreq governors: performance, powersave
  current policy: frequency should be within 400 MHz and 2.40 GHz.
                  The governor "powersave" may decide which speed to use
                  within this range.
  current CPU frequency is 632 MHz.
analyzing CPU 3:
  driver: intel_pstate
  CPUs which run at the same hardware frequency: 3
  CPUs which need to have their frequency coordinated by software: 3
  maximum transition latency: 0.97 ms.
  hardware limits: 400 MHz - 2.40 GHz
  available cpufreq governors: performance, powersave
  current policy: frequency should be within 400 MHz and 2.40 GHz.
                  The governor "powersave" may decide which speed to use
                  within this range.
  current CPU frequency is 772 MHz.

admittedly i'm only doing some light web browsing...
 
levi said:
If they still cost the same as the 350RAMAC did then not many houses would actually have one. I'm personally content that I can get a hard disc these days in the same form factor as I did in 1994 with 25000 times the capacity and it cost me about a tenth of the cost.
Click to expand...
Considering that in the 1960s this was Space-Age technology and nowadays it's old hat, pretty sure it wouldn't cost nearly that much. It would still cost some, though.

The closest modern analogue would be something you're probably quite familiar with if you're well-versed with building PCs - a NAS box.
 
You must log in or register to reply here.

Similar threads

EvilDragon
From Greece with love.
3 4 5
Replies
92
Views
78K
matzesu
matzesu
EvilDragon
Finalizing the PCBs
11 12 13
Replies
258
Views
131K
ible
ible
EvilDragon
Let's move the moulds!
8 9 10
Replies
197
Views
101K
Djhg2000
D
EvilDragon
The following is based on a true story.
5 6 7
Replies
130
Views
61K
EricB
E
EvilDragon
Wrapping everything up!
19 20 21
Replies
404
Views
303K
Lightkey
L
Back
Top