Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
12 years ago

Cyclone 3 vs Cyclone 3 LS Power Dissipation

I'm currently working on estimating moving from a Cyclone 3 (EP3C120F484I7) to a 3 LS (EP3CLS200F780I7). I'm trying to get an estimated change in power requirements. I migrated my design without the pin settings to the 3 LS to try and get an estimate (invalid pin placements). I expected to see a number simliar to the 3 if not a little higher, yet what I'm finding is a 10% decrease in power at worst-case settings with matching junction temperatures of 85°C. This comparison was done through Quartus's Powerplay Analyzer (11.1 Build 259 01/25/2012 SP 2 SJ Full Version).

Device

ep3c120f484i7

ep3cls200f780i7

Core Static Thermal Power Dissipation

566.77 mW

452.87 mW

Spoke with a Rep. and their idea is the 3LS possibly uses 60nm while the 3 uses 65nm, which would account for the drop. If not they said it would just be within the tools margin of error (I'm not using a VCD for toggle rates as a valid gate level isn't possible atm). In that case I understand the % error that can occur, yet checking with the EPE spreadsheet with no resources used I see the following static power:

EP3C120F484 - 119mW

EP3CLS200F780 - 149mW

EP3CLS200F484 - 149mW

Is the estimated power loss I'm seeing due to the 65/60nm difference (is that true?), are there any other possibilities of what is dropping my static power?

7 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Actually I just checked with the following parameters and got more matching results with the EPE

    Power Characteristics: Maximum

    Junction Temp: 85°C

    EP3C120F484 - 617mW

    EP3CLS200F780 - 571mW

    EP3CLS200F484 - 571mW
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Did you recompile the design when you switched to the LS part? You say you have invalid pin assignments, could it be possible that Quartus optimized away a part of your design for this reason?

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Yea, when I did the initial recompile on the LS part it failed due to the invalid pin assignments. Just to see if my design would compile I removed my pin QSF file. I was able to recompile and all the I/O was not mapped but have the default 2.5v/8mA values.

    The compile report still accounts for the same# of pins as on the 3 and comparing the log files I'm not seeing any optimized parts of the design. Comparing LE usage the LS compile used 0.2% more which I'm disgarding as an issue as there are MegaCore components and I would'nt expect a 1:1 copy across the families (I'm in the process of getting a gate level verification of the port).

    I'm not necissarly looking to pick through the design to see if a block on the 3 ran at higher power than the LS, more so trying to understand what's going on in general that the LS runs at a lower static power in worst case situations than the 3 (atleast according to the static power of the EPE and PowerPlay). Is it just physical design advancements?
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Thanks for the document, that helped.

    --- Quote Start ---

    the static power being higher is reassuring (that there isn't a bug/omission)

    --- Quote End ---

    Are you saying the Cyclone 3's static power being higher is reassuring?

    I'm trying to see why the at 25°C and normal conditions the 3 is lower than the LS, but at 85°c and max conditions the LS is lower than the 3. If that's normal then thats great!
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    no, i'm saying the 200 kLE Cyclone III LS static power being larger than the 120 kLE Cyclone III makes sense

    i don't know why the overall power is lower
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    The Cyclone 3 was originally 65 nm, but went through a die shrink to 60 nm after the Cyclone IV came out.

    (Remember seeing a notice about it, and it).

    That being said I'm sure the tools still use the 65 nm Power numbers, since there are both 65 and 60 nm parts out there.

    It sounds like the Cyclone 3 LS has always been 60 nm. So it's probably shows some better power numbers. (At one point I think they said it was marketed as the "Low Power" family members for Cyclone III, but that has since changed.

    Pete