MSI AB / RTSS development news thread

Well, that’s really newbie oriented single click feature like NV were promoting it, so it is hard to do something wrong and no detailed guide is needed. You just open voltage/frequency curve editor exactly like you did it with Pascal GPUs before and click new "OC Scanner" button there (available only on RTX series cards). Additional MSI OC Scanner application will be started, then you click either "Test" or "Scan" button in the scanner's window and wait for result a few minutes. "Test" will take about 5 minutes and it will verify stability of your current overclocking settings (so you may manually overclock the system just like before) and return result as system stability confidence level (0% = unstable, 100% = stable). "Scan" will find optimal overclocking settings during approximately 20 minutes and you'll see new overclocked V/F curve in the end of scanning process. So you just apply overclocked curve in AB like any manual V/F curve modification, and that's all.

 

Alberto:
https://www.guru3d.com/articles_pages/msi_geforce_rtx_2080_ti_duke_review,29.html

I will likely follow up soon with a complete guide on how this works. But that should get you going.

 

It is starting being really boring, but exactly the same youtube channel giantmonkey101 tried to continue the hype and decided to play drama on exactly the same old story again, 3rd or 4th time in a row. As a result, I need again to deal with new wave of AMD fanboys commenting adding their own lie to it and making additional myths around it like "the issues is not just Vulkan/Async games, RTSS has been hammering performance across a wide selection of titles and API's over the last six months … their developers and the countless techtube reviewers who think everything is normal do deserve a bit of flak for never noticing.". I don't know how to comment it again, and honestly I'm starting to doubt that the person is searching for any technical comments because the previous comment I left under each of his videos about "performance bug" were simply ignored and story simply repeated over and over again.



On more positive side: we were working together with NV during last few days and analyzed possible ways of improving our current OC Scanner implementation for RTX series graphics cards. Their ideas were flying around improved visualization of overclocking scanning process. Previously we displayed just operating V/F point during scanning, now there will be full dynamically refreshing V/F curve so you'll be able to see how exactly the scanner is tuning the curve in real time. So probably there will be minor update of 4.6.0 soon.

 

I cannot give you exact ETA, sorry. It will be published in this thread when it is done.

 

Heh, it is rather hard to ignore and to NOT waste time on it because videos like this are constantly getting insane amount of misinformative or hate comments like this one. That's depressing as hell to read such things. But someone behind giantmonkey101 channel can enjoy his day of fame again.

 

.\ProfileTemplates\Global -> [OSD] -> PerformanceProfiler

 

NVIDIA seems to be satisfied with improvements in OC scanner, which we’ve implemented per their request after releasing a beta for RTX cards. So we’re releasing new slightly updated build of 4.6.0 beta 9. Nothing major inside, just small but useful GUI improvements making the process of overclocking scanning more convenient and informative for users. Changes are mostly covering OC scanner on new RTX series cards only, there is no need to upgrade from the previous 4.6.0 beta if you use any other GPUs. Full changes list comparing to 4.6.0 beta 9 build 13319 includes:

- Voltage/frequency curve editor is now updated on new GPU selection or closed if you select a GPU with no Voltage/Frequency curve control support, so you don’t need to reopen the editor when tuning the curve for multiple GPUs
- Improved hardware control shared memory interface:
- Now hardware control shared memory is immediately updated on selecting new GPU from GPU quick selection panel in GUI. It is no longer necessary to apply the settings in order to make new GPU selection visible to external applications connected to hardware control shared memory. So you may switch between multiple GPUs and test them from external overclocking scanner without closing and reopening it for each new GPU
- New hardware control shared memory interface command allows MSI Afterburner to refresh Voltage/Frequency curve editor to refresh the curve in real time per external application request
- Updated OC scanner is now refreshing Voltage/Frequency curve in real time during scanning process, so you may see how the scanner is locking voltage points and tuning each point of the curve while searching for the maximum stable GPU clock

New build is available here:

https://www.guru3d.com/files-details/msi-afterburner-beta-download.html

On a side note, after series of unpleasant discussions about RTSS with the next wave of “educated” AMD fanboys in different places during the last days, I decided to disable OSD support for Vulkan applications using present from compute queue on AMD cards (Doom and Wolfenstein 2). For experienced users understanding that performance penalty is expected in this case I’ll leave an ability to enable it via profiles.

 

> -Only as Info, because I know that you don't support un-released Windows Version. The aktuel Skip-Ahead Build has problems with Visual C++ 2008 an because of that Afterburner will not run. Perhaps MS start dropping support for the 2008 Version...

The chances that MS start dropping VC 2008 runtimes are close to zero. They still support VC 6.0 (1998 compiler) and older versions.

> -For the OC-Scanner, it take 4 Voltage points. Will it be possible to set this point Manually? Or Perhaps manual set the max Voltage for easier Under-Volt.

No, it is not customizable in NVIDIA Scanner implementation. MSI OC scanner's GUI already covering 99% of NVIDIA Scanner's API customization. You cannot alter anything else related to OC scanning process. The only customizable thing that I left hardcoded is a stress-testing time for "Test" mode. It is currently hardcoded in our scanner to take 5 minutes, but Scanner API is allowing to set custom testing time. But I'm not sure if it worth allowing end users to alter it in GUI.

 

> Not so shiny as Precision X1, but at least everything works like a charm. (Precision was an is a big mess)

We do not play marketing games with ambitious "completely new codebase" and "faster than ever" statements. Just an old polished codebase that works as fast as before.

> But you Answer let me ask an other Question; in Testing Mode, are only the 4 Points tested or the Stability in General?

Of course no, it is probing different parts of whole curve. And you can always look at V/F curve editor while stress-testing to see which point exactly is being tested.

> Doesn't the Scanner keep the VRam in consideration?

Scanner's algorithms are proprietary and not disclosed by NVIDIA. But the scanner API does take memory clock offset as one of its input values