Is there interest for an updated WattTool?

2

Comments

  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    Willdo, thanks for the links and info @suthek.
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    edited August 2017
    Uhh yeah. Those "beta" drivers gave me 8MH per card down from 26-30MH per card on my Windows 10 rig lol. Thanks AMD. (Yes I uninstalled the previous drivers using DDU beforehand; Claymore immediately crashes with those drivers on my Windows 7 machine)

    Extra thanks (NOT) to AMD for changing the interface, method names (X2...really?) and structs in the new driver. What the fuck ever happened to backwards compatibility? Oh wait...it's AMD, nevermind.

    After applying some very half-ass hacks (courtesy of AMD not having documentation for this anywhere) I was able to successfully get a working tool on the newest drivers. Try it out:

    http://elsonsmith.com/amd-gpu-tool/

    OR

    https://mega.nz/#!sNokxaiZ!Z8vdgDaQBZX1gUdlfgXYBSLKpEH-uBZUMCtiAlu-ixI
    Post edited by ecs87 on
  • sutheksuthek Member Posts: 324 ✭✭
    edited August 2017
    Prior to using your tool, new driver gives me 85mh/s (3 cards with stock clocks.)
    after using your tool, I'm at 12mh/s
    if I click the reset button in your tool... jumps back up to 85mh/s (As I reset each card I can see that card revert to fast clocks.)

    I'm thinking it's messing something up with clock speeds? Not really sure?

    Can you possibly release a version that only touches mem and clock voltage? not clock or mem speed?


    *EDIT*
    GOT YOUR TOOL WORKING!!!

    Steps followed:
    Installed latest beta driver (like you did)
    Used your tool (like you did) *kept open*
    started claymore. (like you did)
    **Make sure no overclocks or voltage settings are in your claymore start file.**
    got a horrible hashrate (like you did.)
    Next,
    Refreshed your tool's device list. *everything looks fine... has correct settings...Shrug a bit*
    Click apply on each card within your tool.

    Next, you'll see 30+ MH/s per card depending on your overclocks. (RX 580 in my case)

    I then opened Afterburner to tweak fan settings.

    Now I'm stable with RX 580 cards on latest beta driver.
    Cards running cold...... 900mvddc and 950cvddc

    Summary:
    Dual Mining SIA and ETH
    Before:
    86MH/s ETH
    1650MH/s SIA

    After:
    91.2MH/s ETH
    1730MH/s SIA

    This is awesome. I wonder if I'm one of the first people to be running RX 580 stable on new drivers?



    Post edited by suthek on
  • mjmeansmjmeans Member Posts: 65
    Where is the source for ecs87NextGearOD.exe?
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    edited August 2017
    You're right. My tool has a huge bug. It wasn't setting the performance levels back to ENABLED after hitting the SET button. Only after I fixed that did I realize that the voltage setting isn't working...

    EDIT: It seems that AMD's own official tool can't even set voltages...wtf? I figured afterburner can since they're tapping into the I2C settings IIRC.

    @mjmeans I still have to clean up the code for that section before putting it on GitHub. I doubt it'll happen before the entire switch over to the single C++ .exe is done (and that will be done after this entire new-driver/new-dll fiasco is resolved).
    Post edited by ecs87 on
  • sutheksuthek Member Posts: 324 ✭✭
    Status update:
    Still stable at 31.2mh after 7 hours.
    Although not quite perfect, your tool seems to "work".

    Although I can't see voltage, my temperatures are the exact same as when I was running an older driver with claymore pushing cvddc and mvddc...

    Pretty sure it's working?
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    I'm 99% sure voltage settings aren't working with the beta driver nor with 17.7.2. I checked with HWInfo64. Others can feel free to chime in here too
  • sutheksuthek Member Posts: 324 ✭✭
    ecs87 said:

    I'm 99% sure voltage settings aren't working with the beta driver nor with 17.7.2. I checked with HWInfo64. Others can feel free to chime in here too

    I'll open gpuz tonight and see what power usage I'm getting. (Hopefully gpuz shows something?)
    I know amd wattman showed correct 900/950 settings when using your tool.

    I would think if voltage settings weren't working, I'd be cooking my cards... But the fans are on auto and temps sitting around 60 - 70 degrees.
  • mjmeansmjmeans Member Posts: 65
    It doesn't matter if it is clean. The there are potentially dozens or hundreds of coders that could help the project and submit PR's with the changes you need.
  • sutheksuthek Member Posts: 324 ✭✭
    mjmeans said:

    It doesn't matter if it is clean. The there are potentially dozens or hundreds of coders that could help the project and submit PR's with the changes you need.

    I don't see the difference between ecs87 providing source code... and claymore providing source code for his miners? If he doesn't want to provide the code... it is what it is... buyer beware and hopefully ecs87 is a nice guy.
  • mjmeansmjmeans Member Posts: 65
    suthek said:

    mjmeans said:

    It doesn't matter if it is clean. The there are potentially dozens or hundreds of coders that could help the project and submit PR's with the changes you need.

    I don't see the difference between ecs87 providing source code... and claymore providing source code for his miners? If he doesn't want to provide the code... it is what it is... buyer beware and hopefully ecs87 is a nice guy.
    Claymore has closed source to protect dev fee and the author has openly declared it. There is no claim by OP that this is or will be a for profit project.

    OP said he would post the source as soon as he cleans it up. The point to posting source is: 1) the entire internet can help clean it up faster (possibly significantly faster); and, 2) if OP disappears like WattTool originator has done, the project doesn't die.

  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    edited August 2017
    I think I found the issue...it's the voltage numbers in the BIOS file now taking priority over what the program sends to it over AMD's ADL (dumb). If I set everything to say...850mv in the BIOS (other than the 300mhz power level) then tune using my program, I'm able to control voltage "normally".

    Additionally, I tried disabling all powerplay levels other than one of them which I modified to 900mv @ 1140mhz (and 900mv @ 2000mhz memory) and it worked without modifying the BIOS... Not too sure what else AMD has screwed up in the new drivers, but I believe that the issue lies within what I mentioned above, or with changes in PP levels.


    EDIT: I guess it wasn't the BIOS mv numbers although it did do something odd. It does change the voltage if you flash the BIOS -> reboot -> reset GPU OC settings -> tune...but if you reboot after tuning, then the voltage goes back to ~1000mv at max power level if you have it set for under 1000. Sigh.

    EDIT2: Voltage numbers seem to correlate with the clock speed. Very odd behavior happening:
    • if core voltage and mem voltage > core clock then it uses the highest voltage number
    • else if core clock > core voltage and mem voltage then it uses a mv number similar to the core clock
    The only viable solution I can see to these drivers are to only have one power level enabled, which means you will be running at a certain core clock, memory clock and voltage 24/7. This keeps the drivers from messing with the voltage altogether (and uhh...I guess this entire fiasco gives a new meaning to a DEDICATED mining rig):



    However, if you ever idle your desktop or rig, it's now using a lot more power. FWIW, I'd just wait and cross my fingers for 17.7.3. Hopefully they won't screw anything else up and we will have an updated ADL SDK.

    I've updated links on my site for users running 17.7.2 (and other beta drivers floating around out there) and for users on 17.7.1 and lower.
    Post edited by ecs87 on
  • smpsmp Member Posts: 3
    edited August 2017
    Any chance you will remove the annoying disclaimer? Or at least make it less obtrusive (like only show it once at first program run)?
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    smp said:

    Any chance you will remove the annoying disclaimer? Or at least make it less obtrusive (like only show it once at first program run)?

    I can look into making a checkmark. I just don't want anyone attempting to sue me because they burnt down their apartment "with my tool".
  • smpsmp Member Posts: 3
    It would also be nice if there was no need to manually refresh the device list. Can you make it do this on program run (like in WattTool)?
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    Sure thing, your requests have been merged into the program: https://mega.nz/#!AAJwxCiK!lPLMmPjS5B0J-SyOGZaIG2RWVB8jEO-ix6jNH3ToONw (or you can always download it from http://elsonsmith.com/amd-gpu-tool, use the non-17.7.2 version...I didn't make the updates to the beta (17.7.2) version of my tool because I'm unsure about keeping it up; the entire situation between the buggy drivers and the changed API make me think it's wiser to wait for 17.7.3 for a final update to the program).
  • smpsmp Member Posts: 3
  • sutheksuthek Member Posts: 324 ✭✭
    ecs87 said:

    wait for 17.7.3 for a final update to the program).

    I'm not convinced the next final release will have the DAG fix in it.

    I suspect this beta is a FORK from the regular drivers made only for miners.... as is.

  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    I don't think it's only a DAG issue but rather an OpenCL issue. By fixing it in their release drivers they not only appeal to miners but other facilities as well.

    But you are 100% right, for a company that didn't notify any devs that they changed the library interface with their GPUs...it wouldn't surprise me at all if they totally leave this out of their future drivers because of the bugs (which I'm sure is caused by the new ADL version rather than the OpenCL patches).
  • shooter40swshooter40sw Member Posts: 47
    @ecs87 Hi I plan to test the tool tonight, for an unmoded XFX RX480 non reference model, I got the 17.7.1 drivers installed I tried the new beta for blockchain but the volts only stay fixed for a few minutes and then spikes up the volts and temps, I did not really see any increase in the hash performance, before testing again what are the suggested settings for your wattool and do you suggest to test the beta drivers again? thanks.
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    Everyone's settings will be different because of the silicon lottery. I've personally reverted the drivers on my rigs from the beta back to 17.7.1. The extra MH was not worth the other issues it created.
  • shooter40swshooter40sw Member Posts: 47
    ecs87 said:

    Everyone's settings will be different because of the silicon lottery. I've personally reverted the drivers on my rigs from the beta back to 17.7.1. The extra MH was not worth the other issues it created.

    Thanks for the response, another question. Using your tool will improve hash or stability over the watman?, do the settings have to match somewhat the pics that you have posted? Thanks
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    This tool does the exact same thing as WattMan (plus a few more things) but on a less gfx-intensive scale. I didn't like how slow and clunky WattMan was so I made this.

    Your settings for each card will most likely not reflect my settings other than probably the clock speeds at higher power levels.
  • npittasnpittas Member Posts: 26
    In AB beta 15 you can change the core voltage by moving the slider of core/memory clock before you apply the core voltage change. This might be usefull to how or why this driver is not allowing you to change the wattage with your tool. Have you tried doing the same? I am away and cannot test by myself. Also there might be something in the new API that allows for voltage change to be applied only when you also change core clock
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    I've changed both clock and voltages somewhat successfully. The sequence doesn't matter. After a reboot GPU 0 will default back to ~1000mv (the rest of them keep their settings). Haven't looked more into it.

    MSI AB changes voltage through the I2C. Mine does not (it would be far too much work for me personally at this point to dig through enough documents to understand all there is to about the I2C).

    As far as I can tell, there are no new voltage interfaces exposed in the new adl dll. I've looked at it through a disassembler.
  • tinrooftinroof Member Posts: 26
    Hi guys, could you please put the downloads somewhere else like github? I'm not able to access the link provided. Thanks.
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    tinroof said:

    Hi guys, could you please put the downloads somewhere else like github? I'm not able to access the link provided. Thanks.

    Sorry about that. I just uploaded the zip here: https://github.com/ecs87/AMD-GPU-Tool/blob/master/src/NextGearAMDGPUToolOld.zip
  • tinrooftinroof Member Posts: 26
    ecs87 said:

    tinroof said:

    Hi guys, could you please put the downloads somewhere else like github? I'm not able to access the link provided. Thanks.

    Sorry about that. I just uploaded the zip here: https://github.com/ecs87/AMD-GPU-Tool/blob/master/src/NextGearAMDGPUToolOld.zip
    Thanks a lot
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    Did some more testing on the beta drivers (the newest one returns an amdocl64.dll error via Claymore...figures) and figured out some more things.

    Voltage settings do work...as long as you use a single power level...and as long as you don't have output from that card to something else. Yes, as soon as a monitor is hooked up, I see a voltage increase from 900mv to the "default" 1000mv. Unplug the monitor and 1000mv persists. Reboot the rig, start mining, and it's back at 900mv...until I plug the monitor back in. Stupid :#

    Also, on one run, my ASUS STRIX RX480 8GB's fans were non-functional. Meaning, the temps climbed rapidly with zero fan movement. Now, I was able to reboot and the fans magically worked the next run but yeah...for those of you using this "beta" driver, please be cautious.
  • ecs87ecs87 Dekalb, ILMember Posts: 339 ✭✭✭
    AMD responds to me...I didn't expect that.

    Dear Elson,

    Your service request : SR #{ticketno:[8200766582]} has been reviewed and updated.

    Response and Service Request History:

    Thank you for the email.

    I understand you are noticing GPU voltage defaults to 1000mv on your computer.

    You are observing this with your own overclocking tool developed using ADL SDK.

    In order to properly debug this issue, I would require some additional details about your system configuration.
    Details of the graphic card you are using?
    From when you started experiencing this issue? Did you make any hardware or software changes during this time, e.g. have you updated your AMD display driver or Windows update?

    Which driver version worked for you without any issues?

    Screenshot of the voltage you are observing? If possible provide a small video clip of the issue you are facing.

    Please provide a report of your system details by running the following Windows Utility so that I can have more information regarding your computer hardware and software (OS and Drivers info) and provide appropriate troubleshooting steps.
    Please follow the below steps to get the Dxdiag report of your computer.

    Go to Start > Type Run and type 'dxdiag' (without quotes please) in the 'Open' field and press OK button
    Wait for 'green' progress bar to finish loading and then press the 'Save All Information...'’ button
    Name the file with some useful info, like your name and the date, and save it where you can find it, like the Desktop, by pressing the 'Save' button
    Press the 'Exit' button to quit this application.

    Please attach the txt file in your response, so that I can assist you further.

    Thank you for contacting AMD.

    In order to update this service request, please respond, leaving the service request reference intact.

    Best regards,

    Santosh

    AMD Global Customer Care


    Hopefully this is enough information to get them started:

    Hi again,

    Thank you so much for responding, as I feel that this is a huge issue plaguing various users (even if they're not aware of it).

    Yes, I am observing these issues using my own tool developed with the ADL SDK, but your OFFICIAL program "WattMan" also exhibits the EXACT same behavior; so I guess what I'm trying to say is that my own tool is a moot point here since I've also tried using WattMan to rule out any possible issues with my program.

    The 3 Polaris cards that I'm using are: a ASUS STRIX RX480 8GB with Samsung memory, a Sapphire NITRO RX480 4GB with Hynix memory, and a Sapphire Nitro RX480 8GB with Micron memory.

    The issues started exactly when your development team modified the drivers to support the OverdriveX2 classes (I think it was with the 17.7.2 driver, as the 17.7.1 drivers don't use the OverdriveX2 classes in the ADL portion of your dll files when observed through IDA as far as I can tell).

    The only changes I made to affect voltage was updating my driver from 17.7.1 to 17.7.2. This effectively broke nearly everything.

    If I revert back to 17.7.1, everything works perfectly fine (including voltage control).

    Video #1: On 17.7.1 drivers and everything is working perfectly fine.
    https://drive.google.com/open?id=0B1UEeGfz47dkTGJoMlhDNzVJOW8

    Video #2: Installed 17.8.2 drivers and everything other than VOLTAGE works.
    https://drive.google.com/open?id=0B1UEeGfz47dkVERoeDFVanl5b00

    Video #3: Workaround to get voltage to work across any card that doesn't have a display connected to it. Possible driver bug(s) exposed.
    https://drive.google.com/open?id=0B1UEeGfz47dkOTY5NVB6T2NuWVE

    I've also attached dxdiag logs for both 17.7.1 (Elson8282017-1771.txt) and 17.8.2 (Elson8282017-1782.txt).

    Thanks again!

Sign In or Register to comment.