Forum
  • Search
  • Member List
  • Calendar
Hello There, Guest! Login Register — Login with Facebook

Turbo overclock
Thank you for your donation

Post Reply 
 
Thread Rating:
  • 0 Votes - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Threaded Mode | Linear Mode
Turbo overclock
16th Mar, 2014, 01:00 PM
Post: #6
mk01 Offline
Registered
Posts: 2,485
Joined: Mar 2013
Reputation: 209
RE: Turbo overclock
@min

emmc_pll_core=0 is good in /boot/config.txt,
xbian-package-firmware 1.4.13-1
and
latest kernel package 1.3-6.10

should restore the internal workings back as it was for months before RPI's Octobers updates to kernel & firmware.

for myself it is "true" as I can run again my top 950,450,500,2 . (I was never able to run turbo). If it doesn't work for you now, ... unfortunately I don't have other solution to try as above - or downgrade xbian-package-firmware below 1.4.7 and remove emmc_pll_core from config.txt. this would make your XBian the same as two/three months ago when 1.4.8 or higher of FW was posted to stable and we got reports about overclock affected.

this is one dimension of the story. second one is different approach on how XBian and Raspbian handles the clocks.

Raspbian turbo (1000,500,500) means you get 1000,500 (cpu/core) if system is topped to max going down to base (700,250) . Normally this sounds logically. But going to details it introduces quite significant lags & latencies for clk transitions. There we speak of microseconds what is quite significant impact. I will demonstrate:

Raspbian official test on Turbo calculate 78497 prime numbers in 3m29s. my Xbian with 950,450,500 in 0:03:20.846995. For user interactive apps & actions the steppings are even with worse impact.

running 1000/500 has even bad implications on clock syncs between busses. if busses are not running the same or whole number multiplication of base clock, the faster clock needs to put wait cycles. imagine two busses running 100 and 150mhz. they can speak to each other only each third cycle for 100mhz and each second for 150mhz bus.

and compare this to 100 / 160 setup. it will take 4 waste cycles on 160mhz bus to speak to 100mhz bus. and if those busses are transferring data between cpu / memory / mmc ? 100/150 will be faster than 100/160 with one cycle out of four. so 25% faster.

I don't remember now exactly the ratios for 1000/500 (they can be found on RPI forums) but I was always wondering why this setup. Probably the real-world use is running it at 700/250 anyhow so the drawbacks are not so visible Wink

I know that this info will not make your RPI be running Turbo again. And yes, I know that running core at 250 and 500(or mine 450) is hell a difference ! RPI actually start working Smile


You can keep us in loop in case you will manage some other testing.

Please read rules and do a search before you post! . FAQs . How to post log file? . Looking for answers? Please start here
Find all posts by this user
Quote this message in a reply
« Next Oldest | Next Newest »
Post Reply 


Messages In This Thread
Turbo overclock - min - 2nd Mar, 2014, 03:44 AM
RE: Turbo overclock - IriDium - 2nd Mar, 2014, 05:10 AM
Re: Turbo overclock - min - 2nd Mar, 2014, 07:09 AM
RE: Turbo overclock - CurlyMo - 2nd Mar, 2014, 09:42 AM
RE: Turbo overclock - min - 2nd Mar, 2014, 10:23 AM
RE: Turbo overclock - mk01 - 16th Mar, 2014 01:00 PM
RE: Turbo overclock - min - 29th Mar, 2014, 04:57 AM
RE: Turbo overclock - mk01 - 3rd Apr, 2014, 02:48 PM

  • View a Printable Version
  • Send this Thread to a Friend
  • Subscribe to this thread
Forum Jump:

Current time: 10th Jun, 2025, 07:19 PM Powered By MyBB, © 2002-2025 MyBB Group.