PINE64
1.3GHz - Printable Version

+- PINE64 (https://forum.pine64.org)
+-- Forum: ROCK64 (https://forum.pine64.org/forumdisplay.php?fid=85)
+--- Forum: General Discussion on ROCK64 (https://forum.pine64.org/forumdisplay.php?fid=86)
+--- Thread: 1.3GHz (/showthread.php?tid=4984)

Pages: 1 2


1.3GHz - stuartiannaylor - 08-21-2017

I am presuming the Rock64 max clock speed is set in the DTS.

I just wondered if 1.3GHz had been chosen as a safe level without cooling as the Rock64 is supplied.
Did anyone do any testing with passive and active cooling at higher clock speeds or did 1.3GHz seem to be the limit.

I am presuming with the 1.5Ghz rating its a junction temperature cooling decision and wondered if a kernel/image might be supplied at higher clockspeeds for those who want to employ increased cooling solutions.


RE: 1.3GHz - MarkHaysHarris777 - 08-21-2017

(08-21-2017, 12:21 PM)stuartiannaylor Wrote: I am presuming the Rock64 max clock speed is set in the DTS.

I just wondered if 1.3GHz had been chosen as a safe level without cooling as the Rock64 is supplied.
Did anyone do any testing with passive and active cooling at higher clock speeds or did 1.3GHz seem to be the limit.

I am presuming with the 1.5Ghz rating its a junction temperature cooling decision and wondered if a kernel/image might be supplied at higher clockspeeds for those who want to employ increased cooling solutions.


Overclocking is not recommended;

however,  its your board.  You can do as you like with it-- but the images need to be provided for best engineering practices. As always with any good engineering setup, somebody will always want to experiment with overclocking even though its a dangerous practice that reduces the life of the processor and is prone to processing instabilities;  anyone is free to over-clock to their hearts content.

IMHO, intelligent operation of the system should always employ the proper ondemand scaling governor ( or other scaling governors as required ) so that the system is not always cranking at the "hottest" .  Over-clocking almost always introduces instability and an element of uncertainty;  as well diminishing returns.

Do as you will.


RE: 1.3GHz - blue - 08-21-2017

pretty sure there's a usr/local/sbin script where you can have all the freq fun you want.


RE: 1.3GHz - stuartiannaylor - 08-21-2017

(08-21-2017, 12:58 PM)MarkHaysHarris777 Wrote:
(08-21-2017, 12:21 PM)stuartiannaylor Wrote: I am presuming the Rock64 max clock speed is set in the DTS.

I just wondered if 1.3GHz had been chosen as a safe level without cooling as the Rock64 is supplied.
Did anyone do any testing with passive and active cooling at higher clock speeds or did 1.3GHz seem to be the limit.

I am presuming with the 1.5Ghz rating its a junction temperature cooling decision and wondered if a kernel/image might be supplied at higher clockspeeds for those who want to employ increased cooling solutions.


Overclocking is not recommended;

however,  its your board.  You can do as you like with it-- but the images need to be provided for best engineering practices. As always with any good engineering setup, somebody will always want to experiment with overclocking even though its a dangerous practice that reduces the life of the processor and is prone to processing instabilities;  anyone is free to over-clock to their hearts content.

IMHO, intelligent operation of the system should always employ the proper ondemand scaling governor ( or other scaling governors as required ) so that the system is not always cranking at the "hottest" .  Over-clocking almost always introduces instability and an element of uncertainty;  as well diminishing returns.

Do as you will.

The RK3328 is a 1.5Ghz device, so essentially we are not overclocking.
[font=Arial, 宋体,]"Quad-Core Cortex-A53,up to 1.5GHz" is how rockchip advertise the RK3328.
[/font]

Its why I was asking about the current chosen 1.3Ghz and was that to provide a safe platform with no cooling?
The Rock64 has been supplied under clocked and expect this can be tweaked is maybe the engineering practices employed where to allow the use of bare chips.

(08-21-2017, 01:25 PM)blue Wrote: pretty sure there's a usr/local/sbin script where you can have all the freq fun you want.

Not sure as I think the options are set in https://github.com/ayufan-rock64/linux-u-boot/blob/5504f96aad80b125d1d23271cf63bec2c22064fd/arch/arm/dts/rk3328.dtsi

Code:
cpu0_opp_table: opp_table0 {
compatible = "operating-points-v2";
opp-shared;

opp@408000000 {
opp-hz = /bits/ 64 <408000000>;
opp-microvolt = <950000>;
clock-latency-ns = <40000>;
opp-suspend;
};
opp@600000000 {
opp-hz = /bits/ 64 <600000000>;
opp-microvolt = <950000>;
clock-latency-ns = <40000>;
};
opp@816000000 {
opp-hz = /bits/ 64 <816000000>;
opp-microvolt = <1000000>;
clock-latency-ns = <40000>;
};
opp@1008000000 {
opp-hz = /bits/ 64 <1008000000>;
opp-microvolt = <1100000>;
clock-latency-ns = <40000>;
};
opp@1200000000 {
opp-hz = /bits/ 64 <1200000000>;
opp-microvolt = <1225000>;
clock-latency-ns = <40000>;
};
opp@1296000000 {
opp-hz = /bits/ 64 <1296000000>;
opp-microvolt = <1300000>;
clock-latency-ns = <40000>;
};
};
I think that sets the max that you could call from a script.
I just have a hunch that all the engineering has been on the assumption of passive cooled set-top boxes.
Maybe with a squeeze and active cooling you could ramp things up, but I think it would have to be added here.


RE: 1.3GHz - MarkHaysHarris777 - 08-21-2017

(08-21-2017, 01:36 PM)stuartiannaylor Wrote:
(08-21-2017, 12:58 PM)MarkHaysHarris777 Wrote:
(08-21-2017, 12:21 PM)stuartiannaylor Wrote: I am presuming the Rock64 max clock speed is set in the DTS.

I just wondered if 1.3GHz had been chosen as a safe level without cooling as the Rock64 is supplied.
Did anyone do any testing with passive and active cooling at higher clock speeds or did 1.3GHz seem to be the limit.

I am presuming with the 1.5Ghz rating its a junction temperature cooling decision and wondered if a kernel/image might be supplied at higher clockspeeds for those who want to employ increased cooling solutions.


Overclocking is not recommended;

however,  its your board.  You can do as you like with it-- but the images need to be provided for best engineering practices. As always with any good engineering setup, somebody will always want to experiment with overclocking even though its a dangerous practice that reduces the life of the processor and is prone to processing instabilities;  anyone is free to over-clock to their hearts content.

IMHO, intelligent operation of the system should always employ the proper ondemand scaling governor ( or other scaling governors as required ) so that the system is not always cranking at the "hottest" .  Over-clocking almost always introduces instability and an element of uncertainty;  as well diminishing returns.

Do as you will.

The RK3328 is a 1.5Ghz device, so essentially we are not overclocking.
"Quad-Core Cortex-A53,up to 1.5GHz" is how rockchip advertise the RK3328.   <snip>


Yes, you are over-clocking;

...  there is more involved than the cpu|gpu clock.  The clocks are all interrelated, and memory and memory speed is a key factor , particularly for stability.

We have a small baby issue in that these boards are coming out with a mix of memory chips -- some at the PC-1866 and some at the PC-1600.  Although some of the boards have the faster memory, the boards are all "spec'd" for PC-1600;   you don't want to be messing around with the clock(s)!   Do so as an experiment sure, but for reliable operation and stability, leave them alone.


RE: 1.3GHz - stuartiannaylor - 08-21-2017

(08-21-2017, 02:04 PM)MarkHaysHarris777 Wrote:
(08-21-2017, 01:36 PM)stuartiannaylor Wrote:
(08-21-2017, 12:58 PM)MarkHaysHarris777 Wrote:
(08-21-2017, 12:21 PM)stuartiannaylor Wrote: I am presuming the Rock64 max clock speed is set in the DTS.

I just wondered if 1.3GHz had been chosen as a safe level without cooling as the Rock64 is supplied.
Did anyone do any testing with passive and active cooling at higher clock speeds or did 1.3GHz seem to be the limit.

I am presuming with the 1.5Ghz rating its a junction temperature cooling decision and wondered if a kernel/image might be supplied at higher clockspeeds for those who want to employ increased cooling solutions.


Overclocking is not recommended;

however,  its your board.  You can do as you like with it-- but the images need to be provided for best engineering practices. As always with any good engineering setup, somebody will always want to experiment with overclocking even though its a dangerous practice that reduces the life of the processor and is prone to processing instabilities;  anyone is free to over-clock to their hearts content.

IMHO, intelligent operation of the system should always employ the proper ondemand scaling governor ( or other scaling governors as required ) so that the system is not always cranking at the "hottest" .  Over-clocking almost always introduces instability and an element of uncertainty;  as well diminishing returns.

Do as you will.

The RK3328 is a 1.5Ghz device, so essentially we are not overclocking.
"Quad-Core Cortex-A53,up to 1.5GHz" is how rockchip advertise the RK3328.   <snip>


Yes, you are over-clocking;

...  there is more involved than the cpu|gpu clock.  The clocks are all interrelated, and memory and memory speed is a key factor , particularly for stability.

We have a small baby issue in that these boards are coming out with a mix of memory chips -- some at the PC-1866 and some at the PC-1600.  Although some of the boards have the faster memory, the boards are all "spec'd" for PC-1600;   you don't want to be messing around with the clock(s)!   Do so as an experiment sure, but for reliable operation and stability, leave them alone.

The cpu frequency is separate from the DDR controller which is in the RK3328 and a bit confused about  "the boards are all "spec'd" for PC-1600" irrespective of problems of PC-1866 or PC-1600 the cpu frequency will make that no better or worse.

I was just interested if anyone had any knowledge and any interest in tinkering but actually if we don't go past the the published maximums then we are not overclocking.
Pointless argument though but maybe the wiki should be changed to Quad-core Cortex-A53 up to 1.3GHz CPU
http://wiki.pine64.org/index.php/ROCK64_Main_Page#CPU_Architecture

Irrespective of Marcus says I am thinking there might be a little room to 'over clock' if you have active cooling as there is a little bit of headroom until you hit the specified maxes then maybe you could push that a little with a slight 'overclock'

It seems the max voltage was becoming close currently set at 1.3 for to achieve 1.3GHz with a specified max of 1.35 with a supply of 1.48.
http://opensource.rock-chips.com/images/d/d7/Rockchip_RK3328_Datasheet_V1.1-20170309.pdf
Dunno just thought I would ask?

The pre-production ones apparently ran hot and just wondered what voltage and frequency they where set to?
Where they originally what its says in the RK datasheet Max CPU frequency of A53 1.4 GHz


RE: 1.3GHz - xalius - 08-21-2017

I think RK went back a bit from the original whitepaper which had 1.5Ghz in the feature list, their original Android images based on 3.10.x still have OPPs for 1.4 and 1.5Ghz, but their newer datasheets say 1.4Ghz max atm, maybe they had issues after ramping up production... on the 4.4.x kernels the maximum OPP I got to run stable with a heatsink was 1.4Ghz so far...


RE: 1.3GHz - stuartiannaylor - 08-21-2017

(08-21-2017, 03:46 PM)xalius Wrote: I think RK went back a bit from the original whitepaper which had 1.5Ghz in the feature list, their original Android images based on 3.10.x still have OPPs for 1.4 and 1.5Ghz, but their newer datasheets say 1.4Ghz max atm, maybe they had issues after ramping up production... on the 4.4.x kernels the maximum OPP I got to run stable with a heatsink was 1.4Ghz so far...

Yeah I noticed with peters video it looked very stable @ 1.3 with just a passive heatsink.
https://www.youtube.com/watch?v=umLnsBlOtmo

I was thinking 1.4 is prob just junction temp stability but its a 7% increase and just wondered if we could have an image with a disclaimer 'This image must be run with a heatsink' or something to that effect.
I noticed the 1.4Ghz max reference also which is looking like a max but maybe that is an assumption of passive cooling and 1.5 is achievable with subsequent cooling.
It would only be a 15% increase and an OC to the new reference, but hey maybe.
Did you notice the original voltage levels for 1.4 & 1.5?
Been trying to find them but got lost in git-hub Smile

I ordered some bits to make an I2C pwm 2 channel fan controller that should come on at full blast then ramp down under control.
Just tinkering, but interested in pushing the stable max, purely out of interest as intrigued to what you could do with this little board.


RE: 1.3GHz - Drsdroid - 08-21-2017

I was playing HydroThunder while scraping games in RetroArch and listening to music last night and everything was running fine, but the CPU climbed to 106... I wouldn't recommend an overclock. If anything, maybe an underclock to be safe lol


RE: 1.3GHz - stuartiannaylor - 08-21-2017

(08-21-2017, 05:05 PM)Drsdroid Wrote: I was playing HydroThunder while scraping games in RetroArch and listening to music last night and everything was running fine, but the CPU climbed to 106... I wouldn't recommend an overclock. If anything, maybe an underclock to be safe lol

Just out of interest did you have a heatsink or any active cooling or just bare?