DM8148 DDR3 1.35V

Stephen,

From the DM8148 DS section 6.2, Recommended Operating Conditions, The DDR3 mode has 1.43-1.58V.  So, 1.35V is not recommended.

BR,

Viet

Stephen,

I checked with designer and DDR IOs does support 1.35v (DDR3L) operation though the performance will be lower compared to a 1.5v (DDR3). Performance impact is expected to be 5-10% though DDR3L is yet to be characterized on DM8148 silicon.

Hope this help,

Viet

Hi Viet,

Are here any plans to characterize 1.35V DDR3 on DM8148?

Their 32 bit packge and power savings is attractive for minor performance estimated impact.

There are 4gb x32 devices sampling now.

Patrick

support or not support ?????

is there any response ???

Yes, it should work, however we have not characterized this interface with DDR3L yet. Our existing EVM and testing has been done with DDR3 only.

I have three boards with DDR3 and one with DDR3L. All boards with DDR3 are working and the bord with DDR3L are not working. At the board with DDR3L i could load u-boot at OCMC RAM, but could not run it at DDR3L. I still don't know if its a coincidence(mounting issue) itś really dont work with DDR3L.

best regards

Hi ,

We have validate DDR3L on DM385 device. I am not sure if it also works for DM8148.

Basically you need to lower down the frequency of DDR.

And ,  I am not sure  you have modify the DDR controllor registers which is related to DDR timing parameters, and , does the param of DDR3L different from DDR3 parts?       Use the xls below to caculate the register value and modify in uboot.

http://processors.wiki.ti.com/index.php/File:DM814x_DDR_Controller_Register_Configuration_spreadsheet_v1.0.zip

BR,

Eason

Eason Wang said:

I am not sure  you have modify the DDR controllor registers which is related to DDR timing parameters, and , does the param of DDR3L different from DDR3 parts?

Looking at some parts which officially support DDR3L, such as the DM38x and AM335x, their datasheets don't seem to make any distinction between DDR3 and DDR3L other than for the voltage range given in the electrical specifications and the TRMs doesn't even mention DDR3L.

So it seems that the differences are purely electrical, and I'd expect no configuration differences other than possibly requiring more relaxed timing.  I didn't see any mention of timing differences in the DM38x and AM335x datasheets though, but perhaps the RAM chips themselves do.

Guilherme Haas said:

I have three boards with DDR3 and one with DDR3L. All boards with DDR3 are working and the bord with DDR3L are not working.

If the issue is related purely due to the lower voltage of DDR3L, an interesting experiment may be to slightly lower the 3.3V (and maybe the 1.8V) supply rails (within the allowed margins) as well:  for example, if the "max 2.0V difference" limit between the 3.3V and 1.8V supplies accidently also exists between the 3.3V and DDR supplies on the DM814x, then 1.35V is getting dangerously close.

Lowering clock speeds during initial testing is of course a good idea as well.

(Curiously though, while the timing-sensitivity of DDR3 is often mentioned, it seems to be very tolerant our board: I seem to be able to use it at 533 MHz at OPP100 on a BCYE0 part, and still get very wide min/max margins from the byte-leveling program under those circumstances...)

Guilherme Haas said:

I still don't know if its a coincidence(mounting issue)

Connection/mounting problems can sometimes be easy to spot by interactively "playing" with the memory through the memory browser via JTAG.  For example, one of our prototype boards apparently sustained some kind of damage and accesses work fine for 3 of the RAM chips but sometimes end up at a wrong address for the other 5 chips.  Such patterns are of course a lot more informative than the generic crash you'd get from u-boot.