480 Memory Strap Timings

TruthchanterTruthchanter Member Posts: 549 ✭✭✭
edited September 2016 in Mining
For the 470, I have had great success putting the 1500mhz timing into the 1625, 1750, and 2000. I did hear someone used the 1625 or 1750 for the 2000 so im curious how that works compared to 1500..

For the 480, it seems less clear. Boysie said he thought he did the 1625 timing into the 2000. i checked the Polaris Bios Editor with the default bios compared to his custom bioses but it made no sense. Maybe he used a hex editor instead.

What is optimum for the 480 2000 timing strap? I'm currently using @Heliox 's original 480 modded bios and not sure it has the best strap timing for 2000 compared to the newer ones he and @boysie did.

Comments

  • TruthchanterTruthchanter Member Posts: 549 ✭✭✭
    edited September 2016
    Here's my 480 results right now (still may optimize mem strap timing, undervolts, and mem oc further):

    975-980w @ wall 6x480 whole system. 172-175mhs 22dcri 1920sia (29-29.2 per card) 1095/890mv 2125/890mv heliox original rom.

    What is confusing me with these 480 roms is that boysies and heliox roms use different 2000 mem strap values:

    Heliox: 777000000000000022CC1C00106A6D4DD0571016B90D060C006AE70014051420FA8900A0030000001B11333DC0303A17

    Boysie:
    555000000000000022CC1C00CE616C47D0570F15B48C250B006AE7000B031420FA8900A003000000190F2F39B22D3517

    Would I get higher mhs with boysies?

    EDIT: I edited heliox original 480 mod (which i was using for the stats at the beginning of this post) and put the 2000 strap value i listed for boysie in it. Wierdly i'm getting like 17-19s mhs at 2200 mem this way but if i put the mem to like 2080 or 2125, mhs goes back up to normal range around ~29 So there is some wierd stuff going on and i think if im getting the same mhs with this new strap value at 2125 compared with heliox's at 2200. there may be a better strap value to use with 2200. I will continue experimenting, hopefully at least one of you guys will chime in with something good.
    Post edited by Truthchanter on
  • boysieboysie Member Posts: 591 ✭✭✭
    you are basically doing what I did, whack it in and see what you get...

    for me the lower the speed the less power that's needed to do the same thing.

    Boysie
  • TruthchanterTruthchanter Member Posts: 549 ✭✭✭
    edited September 2016
    One wierd thing I just noticed is i'm getting higher sia AND slightly higher eth with dcri 30 compared to dcri 22. thats with this custom mix of your and heliox's mods on a single 480

    my eth mhs significantly goes down if i lower my dcri even further to 14
  • boysieboysie Member Posts: 591 ✭✭✭
    hehe, if you map it out even with stock bios that happens in various parts. Its weird I know but I guess if you are trying to do 2 things at once then the best point will be where they interfere with each other the less.
  • bctopicsbctopics Member Posts: 333 ✭✭

    Here's my 480 results right now (still may optimize mem strap timing, undervolts, and mem oc further):

    975-980w @ wall 6x480 whole system. 172-175mhs 22dcri 1920sia (29-29.2 per card) 1095/890mv 2125/890mv heliox original rom.

    What is confusing me with these 480 roms is that boysies and heliox roms use different 2000 mem strap values:

    Heliox: 777000000000000022CC1C00106A6D4DD0571016B90D060C006AE70014051420FA8900A0030000001B11333DC0303A17

    Boysie:
    555000000000000022CC1C00CE616C47D0570F15B48C250B006AE7000B031420FA8900A003000000190F2F39B22D3517

    Would I get higher mhs with boysies?

    EDIT: I edited heliox original 480 mod (which i was using for the stats at the beginning of this post) and put the 2000 strap value i listed for boysie in it. Wierdly i'm getting like 17-19s mhs at 2200 mem this way but if i put the mem to like 2080 or 2125, mhs goes back up to normal range around ~29 So there is some wierd stuff going on and i think if im getting the same mhs with this new strap value at 2125 compared with heliox's at 2200. there may be a better strap value to use with 2200. I will continue experimenting, hopefully at least one of you guys will chime in with something good.

    heloix and boysie edit there rom's differently.
  • TruthchanterTruthchanter Member Posts: 549 ✭✭✭
    bctopics said:

    Here's my 480 results right now (still may optimize mem strap timing, undervolts, and mem oc further):

    975-980w @ wall 6x480 whole system. 172-175mhs 22dcri 1920sia (29-29.2 per card) 1095/890mv 2125/890mv heliox original rom.

    What is confusing me with these 480 roms is that boysies and heliox roms use different 2000 mem strap values:

    Heliox: 777000000000000022CC1C00106A6D4DD0571016B90D060C006AE70014051420FA8900A0030000001B11333DC0303A17

    Boysie:
    555000000000000022CC1C00CE616C47D0570F15B48C250B006AE7000B031420FA8900A003000000190F2F39B22D3517

    Would I get higher mhs with boysies?

    EDIT: I edited heliox original 480 mod (which i was using for the stats at the beginning of this post) and put the 2000 strap value i listed for boysie in it. Wierdly i'm getting like 17-19s mhs at 2200 mem this way but if i put the mem to like 2080 or 2125, mhs goes back up to normal range around ~29 So there is some wierd stuff going on and i think if im getting the same mhs with this new strap value at 2125 compared with heliox's at 2200. there may be a better strap value to use with 2200. I will continue experimenting, hopefully at least one of you guys will chime in with something good.

    heloix and boysie edit there rom's differently.
    Yes, one of my main points is that Heliox's at 2200 and boysies at 2125 yield about the same mhs... (so you would think that boysies at 2200 would be the highest but it is way lower there than it should be/bugs out)...

    therefore Heliox's may not be optimal at 2200 since boysies yields the same at 2125 (talking about memory timings here)
  • LogicaluserLogicaluser Member Posts: 214 ✭✭
    edited September 2016
    It's always a balancing act between clockspeed and timing tightness.
    Similar ETH performance can be had using a wide variety of combinations that could very well be +15% apart in terms of clocks.

    Boysie seems to be using tighter timings for that strap range, far too tight to allow for +2150mhz, but as you say they achieve similar results despite the lower clockspeeds.
  • YogiYogi Member Posts: 147 ✭✭
    The tighter the timings you use the less clock speed you need to get the same work done.
    But also core and mem clocks have to work synchrinously as well to a point.

    What I really mean by this is there is essentially a optimal point for all of the following (for each card).

    1. Core Speed
    2. Mem Speed
    3. Mem Timings
    4. Core Volatge
    5. Mem Volatage

    There are some other factors as well but then you really start looking at minor performance increases. I think the main factors are as above.

    By having more agressive timings on the same Core / Mem speed you get more throughput with the card.
    So you can start adjusting Core / Mem voltage then and lower clock / mem speeds to get lower throughput based on increased timings but lower power.

    Essentially though it's always a balancing act between these 5 main things.
    Balancing them well in a generic enough way to be viable for 99.99% of cards is the hard bit :)

  • cscheatcscheat Member Posts: 147 ✭✭
    if heliox & boysie have the same mh, i prefer to use the one with lower power draw.
Sign In or Register to comment.