[U-Boot] DRAM clock frequency issue

[Greg Maitz]

Hi Chris. Thanks for your response. There was a typo earlier, the
correct U-boot version is 1.1.4. The current RAM consists of two
physical chips of 64 MB each, it is the Winbond W9751G6JB (DDR2).
1T/2T is not supported as I understand. I'll examine the board layout
to see how it is affecting the DDR.

On Sun, Jun 3, 2018 at 9:46 PM, Chris Packham <judge.packham at gmail.com> wrote:
>
>
> On Fri, 1 Jun 2018, 10:42 PM Greg Maitz, <ghh19622 at gmail.com> wrote:
>>
>> Hi - I have a target Atheros 9344 board on which I have loaded U-boot
>> version 1.23.2. On the reference board, the DRAM clock frequency has
>> been set to 400 MHz and it works well.
>
>
> Sounds like you've got a vendor bootloader. 1.23.2 isn't a version number
> used by u-boot. You might get more informed support from wherever you
> obtained this.
>
>> But when the same clock
>> frequency is used on my target board , mtest (memory test) fails.
>> However, when I reduced the clock frequency to 300 MHz, the mtest
>> passes and it appears to be stable.
>> The reference and target boards are alike in almost every respect.
>
>
> Double check this assertion. Even if things appear similar on a schematic
> the board layout is significant when it comes to ddr.
>
>> The
>> DRAM specifications are the same but the manufacturer is different.
>>
>>  I tried setting the DRAM timing parameters on my target DRAM to be
>> aligned with the numbers in the data sheet. When I load the Linux
>> kernel after this, there are a variety of kernel panics seen if I
>> leave it on overnight. They are OOPS messages such as "Unable to
>> handle kernel paging request at virtual address", I think they may be
>> related to the DRAM clock frequency setting.
>
>
> A plug here for memtester http://pyropus.ca/software/memtester/ it's been
> quite handy for excercising ram. Unfortunately it won't fix the problem, but
> it will let you know when you have.
>
>> I'd like to eventually use 400 MHz clock for the DRAM clock frequency
>> on my target board. If anyone has faced a similar scenario earlier,
>> kindly share your tips to debug the problem.
>
>
> I've used enabled 2T timing (assuming your ddr controller supports it) for
> similar issues in the past. If it works it's usually an indication of some
> hardware issue but it may get you through to the next board spin.