Teensyduino 1.48 has been released.
https://www.pjrc.com/teensy/td_download.html
Changes since 1.47:
Arduino 1.8.10 support
Many small bugs fixed
Teensy 4.0 CPU Speed menu, including overclocking
No significant changes were made since 1.48-beta3.
Teensyduino 1.48 has been released.
https://www.pjrc.com/teensy/td_download.html
Changes since 1.47:
Arduino 1.8.10 support
Many small bugs fixed
Teensy 4.0 CPU Speed menu, including overclocking
No significant changes were made since 1.48-beta3.
Released while I was entering this post on Beta 3 thread:
@Paul - check speed text in boards.txt.
Selecting :: teensy40.menu.speed.1008=1.008 GHz (overclock, cooling req'd)
Uses :: teensy40.menu.speed.1008.build.fcpu=1000000000
Starts with F_CPU_ACTUAL :: F_CPU=996000000
Changing to this works:: teensy40.menu.speed.1008.build.fcpu=1008000000
F_CPU=1008000000
Ran install - T4 was stuck offline from running too fast? But TaskMan showed no active Teensy software processes. IDE and TLoader were closed.
I got this message:
I hit BACK and turned off the T_4 and then setup completed?
Seems something behind the scenes when the T4 USB went 'offline'.
Note in Post #2 is still in error in boards.txt with released 1.48.
Am I reading this correctly are we breaking 1 billion cycles per second (1,000,000,000) speed barrier on MCU... WOW
Yup. But at least with the 1 board on my desk right now, 960 MHz seems pretty stable with this heatsink. (and it does get quite hot with only natural convention)
1.008 GHz (really 996 MHz) runs for quite a while when I point a fan at it, but does seem to occasionally have trouble. Often it runs for an hour or more.
Solution for single PROGMEM conflict still needed - in sketch using it for CODE and DATA.
Simple mod to TempMonTest.ino code posted here Teensyduino-1-48-Beta-3 causes this:
Maybe add PROGMEMD and PROGMEMC for Data and Code?Code:TempMonTest:76: error: szWhen causes a section type conflict with void startup_early_hook() PROGMEM char szWhen[2][20] = { "Start setup()", "setup() was" }; ^ T:\tCode\T4\TempMonTest\TempMonTest.ino:43:14: note: 'void startup_early_hook()' was declared here PROGMEM void startup_early_hook(void) { ^ szWhen causes a section type conflict with void startup_early_hook()
Moving strings to DMAMEM doesn't work as that doesn't seem to do a memcpy to that segment?
DMAMEM char szWhen[2][20] = { "Start setup()", "setup() was" };
>> I taped my single $2.83 heatsink to the @loglow pinned T4 and print looks cleaner and running up to 58/59°C at current room temp. This 8 pack for $7.66 looks to be the same size copper heat sink. {Dimension: 0.5 x 0.5 x 0.25 inch }, but 1008 MHz seems a bridge too far.
With that extending these to 960 MHz:
Coremark at 600 MHZ: Iterations/Sec : 2313.57
Coremark at 816 MHZ: Iterations/Sec : 3146.39
Coremark at 960 MHZ: Iterations/Sec : 3701.88
600 MHz :: RSA Signature Speed Test Signature computation took 0.085 seconds
816 MHz :: RSA Signature Speed Test Signature computation took 0.063 seconds
960 MHz :: RSA Signature Speed Test Signature computation took 0.053 seconds
Altered Coremark to show temp and Run in loop(). It keeps running hitting 68°C (154°F ) for 8 passes - then : F_CPU=960000000 deg C=70 Pass#9
> 71 at #12 , 72 at #16 … :: F_CPU=960000000 deg C=74 Pass#33 … F_CPU=960000000 deg C=74 Pass#75 with toggles to … deg C=75 Pass#106 … C=74 Pass#343 {room cooling}
FYI the Early and Late hook calls arrive at 4 and 45 ms with setup() entry at 300:
E_hook millis is 4
L_hook millis is 45
setup millis is 300
Did the same RSA sig move to loop() - done in batches between (cooling) prints:
F_CPU=960000000 deg C=75 [tMax=75] Pass#3000
Signature computation took 0.052 seconds
Last edited by defragster; 10-07-2019 at 08:24 AM.
@Paul: The Windows IDE lower right 'Teensy 4.0 on usb ...' only rarely and temporarily shows the Build speed - then it goes away? Seems it should stay there - and wasn't 'Serial' type indicated as well?
USBHost MSC to HDD runs at 960 MHZ on the T4 with above heat sink, not tried without.
Current USBHost to wired USB LAN adapter doesn't connect over 816 MHz, with same heat sink.
TensorFlow micro_speech runs at 812 MHz w/no heat sink. Seems about as accurate - using the MIC on the New T4 ready Audio Board Rev D:
at 812 MHz::
Not sure what affects the lag measure I added - it is the same at 600 MHz and 396 MHz - recalled it being higher at 600 before.Heard yes (201) @12920ms [ms lag=886]
Heard no (202) @15144ms [ms lag=886]
Heard yes (203) @16984ms [ms lag=886]
Heard no (201) @18720ms [ms lag=886]
Heard no (203) @21160ms [ms lag=886]
Heard yes (201) @23480ms [ms lag=886]
Yes, it used to show more info. Looks like Arduino made some changes to that part of the IDE.Originally Posted by defragster
Unless something is really wrong, I'm not going to consider this a bug or put it on my (already too long) list to stuff to investigate.
@defragster
Those are the same coremarks that I got in post https://forum.pjrc.com/threads/57894...l=1#post218007. I was able to get 1.008Ghz(996Mhz) without any issue.
Just downloaded the lastest 1.48 - installed without any issue btw. Those are the same heatsinks I have as well just haven't attached one yet - will soon though. Going to give your tempmon a try but looks like you got it working.
Not really wrong, just no longer a help glancing at the screen { especially forum post of screenshot } - it shows Teensy and some odd USB ref info (as it relates to windows norm) - but after speed change for one minute it did show the speed and perhaps other info - the compile completed and it was gone. So it can do it - but it gets 'sterilized' somewhere.
TempMon working - except I never see signs of hi temp warning or anything - execution/USB just goes away with no more LED change. Just wanted confirmation the _isr()'s were properly 'attached' and should be working for feedback.
I edited the 1008 to be 1008 not 996 and that seemed to be too much for warm room and that heat sink in static air.
Putting that heatsink on the otherwise committed with soldered pin TallDog Beta did help it run cleaner at 960 MHz (no more spare USB chars) - but any 1008 attempt seemed to just vanish without a trace. Will have set of 8 more heat sinks Tues can try a 2nd T4 then. Odd it feels hot at only 110 - decent amount of heated mass finger doesn't cool as quickly - but doesn't read over 110F though internally reading over 150F. Also odd the USBHost to MSC/HDD worked at 960 but the LAN USB sat at 10Mb instead of seeing 100Mb hardware and getting an IP address over 816.
I left the WavFilePlayer example running, playing from a SD card (via SPI) on the audio shield. That blue heatsink had some air flow from a fan nearby on my desk.
At 960 MHz it played for several hours, until I shut it off. The heatsink felt pretty hot.
At 1.008 GHz (996 MHz), it usually plays for over 30 minutes, but eventually stops. Uploading with auto-reboot still works while it's in the stopped state, so the issue may be with SPI since the CPU and USB still seem to be working.
Interesting - I wasn't doing any SPI. And the coremark benchmark is in such a busy loop that upload failing gives a false idea of it being hung/crashed.
T4 with heatsink off some time - running the posted TempMon sketch at true 1008 quickly did this:
Where the "{isrP#0, HL#0}" are count of calls from the tempmon code _isr's. at 44°C they should not be calling but the USB is just not right - where it runs well up to 75°C at 960 MHz.Code:Compile Time:: T:\tCode\T4\TempMonTest\TempMonTest.ino Oct 7 2019 04:15:04 F_CPU=1008000000 deg C=37 Start setup():: E_HOOK us>>5908 [ii3#>>1] E_hook millis is 4 Start setup():: L_HOOK us>>45297 [ii2#>>5] L_hook millis is 45 Start setup():: @us>>300000 [static ii ms>>300] setup millis is 300 @micros>>392014 [ms>>392] Time is Mon Oct 7 04:15:06 2019 37° 38° 39° @micros>>11266055 [ms>>11266] {isrP#0, HL#0}0Time is Mon Oct 7 04:15:06 2019 40° @micros>>22265893 [ms>>22265] {isrP#0, HL#0}0Time is Mon Oct 7 04:15:17 2019 41° @micros>>33265728 [ms>>33265] {isrP#0, HL#0}0Time is Mon Oct 7 04:15:28 2019 41° 42° @micros>>44265561 [ms>>44265] {isrP#0, HL#0}0Time is Mon Oct 7 04:15:39 2019 �42° 43° @micros>>55265392 [ms>>55265] {isrP#0, HL#0}0Time is Mon Oct 7 04:15:50 2019 �43° 44° @micros>>66265222 [ms>>66265] {isrP#0, HL#0}0Time is Mon Oct 7 04:16:01 2019 �44° @micros>>77265050 [ms>>77265] {isrP#0, HL#0}0Time is Mon Oct 7 04:16:12 2019 �45° 45° 46° 46°
At this point that teensy stopped blink, USB and went into shutdown as the heat sink is cool ~80°F.
In that state a couple of minutes and pushed button to bootloader and the TyComm upload of same HEX - died again with no warning or bad USB or ever showing temp over 48°C:
Just surprising it quits so fast without _isr() warning ...Code:Compile Time:: T:\tCode\T4\TempMonTest\TempMonTest.ino Oct 7 2019 04:15:04 F_CPU=1008000000 deg C=42 Start setup():: E_HOOK us>>5934 [ii3#>>1] E_hook millis is 4 Start setup():: L_HOOK us>>45296 [ii2#>>5] L_hook millis is 45 Start setup():: @us>>300000 [static ii ms>>300] setup millis is 300 @micros>>400024 [ms>>400] Time is Mon Oct 7 04:22:48 2019 42° 43° 44° 45° 45° 46° 46° 47°
@Paul - is(are) your heat sink(s) Aluminum or Copper?
I now see these in 20 pack aluminum and 8.8mm smaller size (.35" instead of 0.5") with taller fins : https://www.amazon.com/dp/B079FQ22LK
I'd guess it's aluminum. I've never seen any copper ones offered with anodized colors.
But I really don't remember. It was bought during the very early days of the beta test (as were so many other parts), maybe even before I sent out those first 1052 boards. That kinda feels like a lifetime ago.
It has double-sticky tape. I've heard mixed info about how effective those types of tape are. Apparently thermal epoxy is much better. But the tape is cheap, and I recall that heatsink was very cheap.
The IMXRT chip on Teensy 4.0 is 10mm by 10mm. I do remember I bought that particular heatsink because it was 10mm. Indeed, it's pretty much a perfect fit on top of the chip. But the piece of tape is only about 60% of the heatsink's surface area. It was cheap....
Two things.
One this thread should be stickied so it’s at the top of announcements.
The other is just a question or two. How much current is the Teensy 4 pulling at max overclock with the benchmark. Ali’s is it possible yet to have it switch speeds if it gets busy for a cpu intensive bit if the program then underclock it to save battery for most of the program? Automatically?
@defragster
Think something else is going on besides overtempt at 996Mhz. I just ran your tempmon sketch using my ILI9341 buddabrot sketch linked into loop. At 996Mhz I runs for a bit and then hangs and I loose USB but I do get up to 65degC:
Now if I reduce the speed to 960Mhz right after the 996Mhz test - it gets up to 69degC but nothing displays. If I reduce to 600Mhz the buddabrot will display and temp stays in the mid 50s. Next I reloaded sketch at 960Mhz and the T4 recovered (temp went decreased below 60degC) and it displayed again but seems to redo the setup of the hooks up and eventually dies at 68degC - but no isr fires. Here is the whole run:Code:F_CPU=996000000 deg C=52 Start setup():: E_HOOK us>>5165 [ii3#>>1] E_hook millis is 5 @micros>>627715 [ms>>627] Time is Mon Oct 7 08:20:32 2019 52° 53° 54° 55° 56° 57° 58° @micros>>10941590 [ms>>10955] {isrP#0, HL#0} Time is Mon Oct 7 08:20:32 2019 57° 58° 59° 60° @micros>>21930511 [ms>>21944] {isrP#0, HL#0} Time is Mon Oct 7 08:20:43 2019 60° 61° @micros>>32785958 [}s>>33049] {isrP#0, HL#0}0Time is Mon Oct 7 08:20:54 2019 61° 62° @micros>>43932433 [ms>>43946] {isrP#0, HL#0} Time is Mon Oct 7 08:21:05 2019 62° 63° @micros>>54941014 [ms>>54954] {isrP#0, HL#0} Time is Mon Oct 7 08:21:16 2019 63° 64° 63° 64° 64° @micros>>87937600 [ms>>87951] {isrP#0, HL#0} Time is Mon Oct 7 08:21:49 2019 64° 65° @micros>>98931580 [ms>>98945] {isrP#0, HL#0} Time is Mon Oct 7 08:22:00 2019 64° 65°
EDIT: Oh btw if I unplug/replug it starts at 52degC but no display.Code:Compile Time:: D:\Users\Merli\Documents\Arduino\T4\tempmon_speed_tests\tempmon_speed_tests.ino Oct 7 2019 08:37:18 F_CPU=960000000 deg C=54 Start setup():: E_HOOK us>>5012 [ii3#>>1] E_hook millis is 5 Start setup():: L_HOOK us>>45299 [ii2#>>5] L_hook millis is 45 Start setup():: @us>>627704 [static ii ms>>300] setup millis is 627 @micros>>627719 [ms>>627] Time is Mon Oct 7 08:37:29 2019 55° 56° 57° 58° 59° @micros>>10769412 [ms>>11036] {isrP#0, HL#0} Time is Mon Oct 7 08:37:29 2019 59° 60° 61° 62° @micros>>21871371 [ms>>22138] {isrP#0, HL#0} Time is Mon Oct 7 08:37:40 2019 60° 61° 62° 63° @micros>>32942383 [ms>>32957] {isrP#0, HL#0} Time is Mon Oct 7 08:37:51 2019 63° 64° @micros>>43937608 [ms>>43952] {isrP#0, HL#0} Time is Mon Oct 7 08:38:02 2019 63° 64° @micros>>54949401 [ms>>54964] {isrP#0, HL#0} Time is Mon Oct 7 08:38:13 2019 64° 65° @micros>>65940752 [ms>>65955] {isrP#0, HL#0} Time is Mon Oct 7 08:38:24 2019 65° 66° @micros>>76694264 [ms>>76961] {isrP#0, HL#0}0Time is Mon Oct 7 08:38:35 2019 66° @micros>>87807062 [ms>>88074] {isrP#0, HL#0}0Time is Mon Oct 7 08:38:46 2019 64° 65° 66° @micros>>98914354 [ms>>99181] {isrP#0, HL#0} Time is Mon Oct 7 08:38:57 2019 66° @micros>>109948490 [ms>>109963] {isrP#0, HL#0} Time is Mon Oct 7 08:39:08 2019 66° F_CPU=960000000 deg C=66 setup() was:: E_HOOK us>>5012 [ii3#>>1] E_hook millis is 5 setup() was:: L_HOOK us>>45299 [ii2#>>5] L_hook millis is 45 setup() was:: @us>>109948490 [static ii ms>>300] setup millis is 627 @micros>>109963252 [ms>>109963] Time is Mon Oct 7 08:39:08 2019 @micros>>120937645 [ms>>120952] {isrP#0, HL#0} Time is Mon Oct 7 08:39:19 2019 66° 67° @micros>>131950470 [ms>>131964] {isrP#0, HL#0} Time is Mon Oct 7 08:39:30 2019 66° 67° @micros>>142938003 [ms>>142952] {isrP#0, HL#0} Time is Mon Oct 7 08:39:41 2019 66° 67° @micros>>153749216 [}s>>154015] {isrP#0, HL#0} Time ys Mon Oct 7 08:39:52 2019 66° 67° 68° @micros>>164858384 [}s>>165125] {isrP#0, HL#0} Time ys Mon Oct 7 08:40:03 2019 67° 68° @micros>>175944715 [ms>>175959] {isrP#0, HL#0} Time is Mon Oct 7 08:40:14 2019 68° @micros>>186940469 [ms>>186955] {isrP#0, HL#0} Time is Mon Oct 7 08:40:25 2019 67° 68° @micros>>197938214 [ms>>197952] {isrP#0, HL#0} Time is Mon Oct 7 08:40:36 2019 68° @micros>>208940852 [ms>>208955] {isrP#0, HL#0} Time is Mon Oct 7 08:40:47 2019 67° 68° @micros>>219695840 [}s>>219962] {isrP#0, HL#0} Time ys Mon Oct 7 08:40:58 2019 67° F_CPU=960000000 deg C=67 setup() was:: E_HOOK us>>5012 [ii3#>>1] E_hook millis is 5 setup() was:: L_HOOK us>>45299 [ii2#>>5] L_hook millis is 45 setup() was:: @us>>219695840 [static ii ms>>300] setup millis is 627 @micros>>219962629 [ms>>219962] Time is Mon Oct 7 08:40:58 2019 68° @micros>>230800196 [}s>>231066] {isrP#0, HL#0} Time ys Mon Oct 7 08:41:09 2019 68° @micros>>241904936 [}s>>242172] {isrP#0, HL#0} Time ys Mon Oct 7 08:41:20 2019 68° @micros>>252947628 [ms>>252962] {isrP#0, HL#0} Time is Mon Oct 7 08:41:31 2019 68° @micros>>263942777 [ms>>263957] {isrP#0, HL#0} Time is Mon Oct 7 08:41:42 2019 68° @micros>>274942626 [ms>>274957] {isrP#0, HL#0} Time is Mon Oct 7 08:41:53 2019 68° 69° @micros>>285941319 [}s>>285955] {isrP#0, HL#0} Time is Mon Oct 7 08:42:04 2019 68° 69° @micros>>296748558 [}s>>297015] {isrP#0, HL#0} Time ys Mon Oct 7 08:42:15 2019 69° @micros>>307847964 [}s>>308114] {isrP#0, HL#0} Time ys Mon Oct 7 08:42:26 2019 69° @micros>>318946471 [}s>>319213] {isrP#0, HL#0} Time ys Mon Oct 7 08:42:37 2019 68° 69° @micros>>329940949 [}s>>329955] {isrP#0, HL#0} Time is Mon Oct 7 08:42:48 2019 68° F_CPU=960000000 deg C=68 setup() was:: E_HOOK us>>5012 [ii3#>>1] E_hook millis is 5 setup() was:: L_HOOK us>>45299 [ii2#>>5] L_hook millis is 45 setup() was:: @us>>329940949 [static ii ms>>300] setup millis is 627 @micros>>329955886 [ms>>329955] Time is Mon Oct 7 08:42:48 2019 69° @micros>>340940058 [ms>>340954] {isrP#0, HL#0} Time is Mon Oct 7 08:42:59 2019 68° 69° @micros>>351948156 [ms>>351962] {isrP#0, HL#0} Time is Mon Oct 7 08:43:10 2019 68° 69° @micros>>362938779 [ms>>362953] {isrP#0, HL#0} Time is Mon Oct 7 08:43:21 2019 68° 69° @micros>>373773327 [}s>>374039] {isrP#0, HL#0} Time ys Mon Oct 7 08:43:32 2019 @micros>>384880900 [}s>>385147] {isrP#0, HL#0} Time ys Mon Oct 7 08:43:43 2019 68° @micros>>395949116 [ms>>395963] {isrP#0, HL#0} Time is Mon Oct 7 08:43:54 2019 68° @micros>>406940809 [ms>>406955] {isrP#0, HL#0} Time is Mon Oct 7 08:44:05 2019 68° 69° @micros>>417939912 [ms>>417954] {isrP#0, HL#0} Time is Mon Oct 7 08:44:16 2019 68° 69° @micros>>428940699 [ms>>428955] {isrP#0, HL#0} Time is Mon Oct 7 08:44:27 2019 68° 69° @micros>>439697508 [}s>>439964] {isrP#0, HL#0} Time ys Mon Oct 7 08:44:38 2019 68° F_CPU=960000000 deg C=68 setup() was:: E_HOOK us>>5012 [ii3#>>1] E_hook millis is 5 setup() was:: L_HOOK us>>45299 [ii2#>>5] L_hook millis is 45 setup() was:: @us>>439697508 [static ii ms>>300] setup millis is 627 @micros>>439964046 [ms>>439964] Time is Mon Oct 7 08:44:38 2019 68° @micros>>450808084 [}s>>451074] {isrP#0, HL#0} Time ys Mon Oct 7 08:44:49 2019 68° @micros>>461917313 [}s>>462184] {isrP#0, HL#0} Time ys Mon Oct 7 08:45:00 2019 68° @micros>>472950302 [ms>>472964] {isrP#0, HL#0} Time is Mon Oct 7 08:45:11 2019 68° @micros>>483950292 [ms>>483964] {isrP#0, HL#0} Time is Mon Oct 7 08:45:22 2019 68° @micros>>494940007 [ms>>494954] {isrP#0, HL#0} Time is Mon Oct 7 08:45:33 2019 68°
@defragster - @Paul
I just ran my original tempmon sketch, the standalone one, that sets the registers. Ran it until it hit the High Temp of 85 and it seems to working without issue. It tripped the ISR.
I have to check a few things:
1. Make sure I converted it correctly and check the registers settings
2. Have to check something in the SDK on tempmode? didn't use it before but maybe have to
3. May have to write our own ISRs? So question is how do I shut down the T4 programmatically?
4. Other things to be discovered - have to do somemore research I guess
All this just to be on the save side because I don't know why there is a 60 cutoff when I start having problems.
Indeed - something odd. I was busy running things that worked trying to find a reproducible edge or indication of what led to trouble. TempMon sketch was an abandoned WIP with other dev at the time. Is your sketch posted?
@defragster
Sorry was out running errands with the wife. Yes it is posted but easier if I just attach it here:
@defragster
Just ran a simple test sketch that prints out the temp every 2 seconds with the T4 @1Ghz. TempMon does work - it shuts the T4 down at 90degC so now going to put a bit more test the buddhabrot example. But I am thinking that as the Temp gets to hot some things may have problem - maybe like frequency is shifting at that temp.
@defragster
I modified my tempmon_t4 sketch to include the drawing the Buddabrot on a ILI9341 display - ISRs never tripped but at 1Ghz the T4 hung at about 74.5 degC:
Test results:
EDIT: heres the sketchCode:@600Mhz - temp steady at about 52degC @912Mhz - temp steady at about 67degC @1Ghz: Ran to about 74.5degC then hung. If I unplugged and replugged it back in the sketch ran (showed about 72degC) but ILI9341 display stayed blank. Maybe SPI clock goes off - tomorrow I will hook up a scope to sck and see what happens.
Last edited by mjs513; 10-08-2019 at 03:15 AM.
Thanks, Got Zip and two libs - it compiles - no display on TallDog T4 with that heat sink.
Modified setup() shows:
It cooled down after that - for some time:TEMPMON driver example.
F_CPU=1008000000 deg C=63
The chip initial temperature is 63.18
Though just stopped now without warning or notice ……
61.91
62.55
61.91
Chip now cool to touch so it went into shutdown or 300 MHz cooldown for Fault mode? Sounds like a test for debug_tt - later though. Something funny going on - not looking like temp?
Another run - showing time alive:
Code:TEMPMON driver example. F_CPU=1008000000 deg C=53 The chip initial temperature is 53.00 @ms= 2745 52.36 @ms= 3165 51.73 @ms= 3837 51.73 @ms= 4510 51.73 @ms= 5180 … 58.09 @ms= 130238 57.45 @ms= 130914 58.09 @ms= 131587 57.45 @ms= 132257 57.45 @ms= 132930 58.09 @ms= 133602 58.09 @ms= 134280 57.45 @ms= 134953
@defragster
Sat here a while this morning watching clk pin while sketch was running at 600Mhz and 996Mhz. Run the 996Mhz up to 71degC without issue. Nothing obvious on my scope with the SPI clock. So something else causing the problem of early shutdown at 996.
@mjs513 - Like the picture
Are you running the sketch with some form of heatsink? or just raw board? I have it running right now at 816 (last one that does not say cooling required) and it is hovering abount 54 degrees.
EDIT: Ran at 1.008Ghz, and it ran fine for awhile, got up to about 65 degrees and then just stopped. I was running Logic Analyzer during the run and
did not see anything obvious. It looked like it completed its update of the display. Everything ended properly, CS was brought back high... But then it did not try to update again...
Last edited by KurtE; 10-08-2019 at 02:18 PM.
Posting Permissions
Reply With Quote