I’ve recently tested 120 MHz and would like to use it permanently if the MCU is happy long term.
Cheers
In short, the upper temperature for stability is the actual silicon temperature. .. but we measure that at the package. Overclocking increases the core temperature and that heat must migrate from the silicon hot-spot to the packaging to the outside (ambient, heatsink, circuit board, etc.) The heat transfer is not immediate, it takes time. Cheap uC’s are different from the $$$ microprocessors used in PC’s that have a dedicated heatsink interface.
Overview: https://www.mclpcb.com/ic-packaging-information/
General rule-of-thumb: if you must ask, do not do it
Ray
ST take great care in preventing you setting inappropriate clocks in their utilities.
Having said that, I would guess that heat comes from clock speed but also from peripheral output drivers. So if you don’t take large GPIO currents, you can probably afford to clock the core faster.
Likewise a benign environment is safer than extremes.
You will still be limited by access speed to Flash memory.
Life is a lot easier if you you just follow the rules. Buy an F4 or F7 or better. They are designed to cache the Flash to service faster cores. The F446 is 180MHz.
I happily run an AVR Xmega at 64MHz.
The real mystery is why punters are wedded to a first generation Cortex like the F103.
Especially when there are far better chips. The F103 is 11 years old (I think)
David.
[david.prentice – Sun Sep 09, 2018 3:33 pm] –
The real mystery is why punters are wedded to a first generation Cortex like the F103.
Especially when there are far better chips. The F103 is 11 years old (I think)
punters are a cheap lot
[Rick Kimball – Sun Sep 09, 2018 3:51 pm] –
…
punters are a cheap lot
+1
PS: and many of us cheap rascals are off in $5 rPi land or ESP32 stuff… under $5.
So, yes, if one understands the issues of “real time” processing vs “OS latencies” then life in rPi land can be most rewarding:
Raspberry RGB LED binky written in C:
https://www.hackster.io/rayburne/inexpe … ng-updated
The ESP32 does have FreeRTOS integrated, but I have not found a significant latency issue with the dual-core uC. The benefit of integrated Bluetooth and WiFi in the rPi Zero-W and ESP32 greatly simplifies connectivity at the expense of being somewhat power-hungry… nothing comes for free.
Ray
[Rick Kimball – Sun Sep 09, 2018 3:51 pm] –[david.prentice – Sun Sep 09, 2018 3:33 pm] –
The real mystery is why punters are wedded to a first generation Cortex like the F103.
Especially when there are far better chips. The F103 is 11 years old (I think)punters are a cheap lot
STM still sell millions of STM8s every year, so I’d say that companies are ultra cheap.
It makes commercial sense to use hardware which only has adequate speed etc to support the product, as that’s likely to be the cheapest option.
Using boards like the ESP32 or RPI Zero , makes sense for home projects , where development time and and simplicity are more important, but has big down sides for commercial products , as the additional complexities of these devices , leads to greater risk when supporting the product.
[RogerClark – Sun Sep 09, 2018 9:32 pm] –
…
STM still sell millions of STM8s every year, so I’d say that companies are ultra cheap.
It makes commercial sense to use hardware which only has adequate speed etc to support the product, as that’s likely to be the cheapest option.Using boards like the ESP32 or RPI Zero , makes sense for home projects , where development time and and simplicity are more important, but has big down sides for commercial products , as the additional complexities of these devices , leads to greater risk when supporting the product.
I believe an analysis will show that your premise is based on uC models that are still in full production. As the life cycle begins to wind down, some production managers will secure larger than needed inventories, reducing supplies in the supply chain and running up costs. Like thermal runaway, other production managers will do the same and costs go up in response.
Other design managers may know of a V2 or V3 of a product that will need additional internal hw and therefore management may elect to design using a uC with more capability and use firmware to add the new capabilities without changing hardware design. They will pay a wee bit more on the front end but make up the difference in larger volume discounts and the savings in PC boards, tooling, and assembly.
Ray
[david.prentice – Sun Sep 09, 2018 3:33 pm] –
ST take great care in preventing you setting inappropriate clocks in their utilities.
Thanks.
The real question is if 120mHz is actually ‘overclocking’ and if 72MHz is actually a conservative speed.
[Jye – Mon Sep 10, 2018 1:00 am] –
…
The real question is if 120mHz is actually ‘overclocking’ and if 72MHz is actually a conservative speed.
No, that is a defective thought. The published specs are:
ARM®32-bit Cortex®-M3 CPU Core
72 MHz maximum frequency,1.25 DMIPS/MHz (Dhrystone 2.1) performance at 0 wait state memory access
Single-cycle multiplication and hardware division
We all understand that many members have clocked the STM32F103C8 at 120MHz, but that is “over clocking” the published specification. There is a thread about this: viewtopic.php?t=432&start=20 But whimsical wishing does not change the laws of physics.
Ray
[Jye – Mon Sep 10, 2018 1:00 am] –[david.prentice – Sun Sep 09, 2018 3:33 pm] –
ST take great care in preventing you setting inappropriate clocks in their utilities.Thanks.
The real question is if 120mHz is actually ‘overclocking’ and if 72MHz is actually a conservative speed.
The 72MHz clock speed is the design speed of the device, to meet the environmental conditions spec and the Mean Time Between Failure.
It’s not possible to know whether sustained overclockimg will cause any problems.
However I agree with Ray etc… If the longevity and reliability of the device is an issue to you, then don’t run it outside of its design spec envelope.
Use a faster device.
i doubt it would be quite possible to keep usb working in the normal manner running off other higher mhz e.g. 90-120 mhz
in addition, running at higher mhz normally means needing a higher voltage in the mcu core which means higher heat dissipation.
so if one is willing to go the distance say to use a heat sink to keep it operating at the lower temperatures it may after all be possible run it at the higher mhz in the usual context. my guess is 72mhz is pretty much a most stable speed considering no additional heat sink etc
But I acknowledge the responses to just use 72, and I’m not sure the end application needs to use 120.
