Version 4.6.0 Beta 10 of MSI Afterburner Introduces OC Scanner for Pascal

[xpredator_13]

Should I leave the fan curve be controlled by the OC Scanner? It sets my fan curve to AUTO...if I click the manual curve option, it ramp ups my fan speed...shouldn't this be at 100% when scanning? I mean, thermals can affect the total boost to nvidia cards...

 

[Deleted member 50521]

Just follow the directions in the video with a minor difference:


After scanning IMMEDIATELY apply your curve and save it to a profile. Or your curve will be lost and you have to start all over again.

Also do not slide the power and temp limiter before scanning. That will create instability issues. Leave them as default and do the scanning.

As for fan curves you can do 100% fan speed before scanning, but then that means the stability is only good if you blast it at 100% fan all the time.

For me I just used a custom fan curve and start scanning afterwards.

 

[AlexUnwinder]

AB scanner implementation was not as good as X1.

You're mistaking. There is no "AB scanner implementation", there is no "X1 scanner implementation". So both you and those who declare absolutely opposite thing (there are a lot of people online claiming that AB scanner implementation is better for them) are mistaking, any RTX overclocking application with NVIDIA Scanner technology support is just a GUI frontend for NVIDIA's own stress-testing API. And in any case testing is done by NV components rather than by third party vendor technology.

 

[Ferrum Master]

BTW,

Has anyone found how to save complete OC profiles + OSD settings? Where are the config files? There is no import or export function.

 

[Vya Domus]

I mean, thermals can affect the total boost to nvidia cards...

This only tests for stability at certain frequency points, thermals and power limits only determine for how long those clocks can be kept.

 

[commission3r]

Does'nt work with windows 8.1 keeps crashing

 

[EarthDog]

Also do not slide the power and temp limiter before scanning. That will create instability issues. Leave them as default and do the scanning.

Always cranked it... that how you OC these cards...

Wonder if @AlexUnwinder can confirm this? I didnt experience any issues doing so.

 

[Ferrum Master]

Always cranked it... that how you OC these cards...

Absolutely, because not doing so the test looses it's basic idea.

 

[Deleted member 50521]

BTW,

Has anyone found how to save complete OC profiles + OSD settings? Where are the config files? There is no import or export function.

You gonna have to re-scan I guess

 

[EarthDog]

Hoping this isnt another lack of caffeine moment...lol...

...what does scanning have to do with saving profiles? I dont use the OSD...but wondering how scanning helps save profiles.

 

[Slizzo]

After you do the scan, you have to save the OC it finds to a profile, otherwise when the program relaunches it will have no memory of what you did.

You need to save the results to a new profile after doing the scan, then set it as the active profile so it uses it when your card clocks up.

 

[Mayclore]

I can open the editor window, but clicking the OC Scanner button does nothing.

Yeah, same issue. Nothing happens.

 

[Wavetrex]

Tried it, very conservative, you'd be better served doing it manually.

So, I just made 3-4 runs on my Zotac 1080 Amp! which is factory overclocked to +75 Mhz

Noticed same thing as you, this OC scanner is quite conservative, and it resulted in different attempts between +60 and +80 Mhz, putting the card at the same level as it starts from bios, and the curve barely touching 2000 Mhz.
However, doing manual overclock (120% TDP, linked) I can push it to around +125 without voltage tweaks, and +150 by adding a bit to max voltage (+10%), occasionally seeing 2100Mhz, with measurably higher scores in benchmarks.
I probably have a pretty good chip, if I would convert this card into water cooled via custom loop block to keep the temps under 60 degrees I could probably even reach 2150 Mhz without crashing. (Tried it on air and it wasn't stable, reaching 90+ degrees even with the fan at 100% and crashing after 20-40 seconds)

I think this scanner API was designed for "reference" cards which do not come already overclocked. For factory overclocked models it does ... well, nothing.

 

[Gasaraki]

Always cranked it... that how you OC these cards...

Wonder if @AlexUnwinder can confirm this? I didnt experience any issues doing so.

No. That's not how you use the scanner. You don't always CRANK it.

You leave the power limit at 100%, run scanner, apply settings, turn up power limits.

 

[EarthDog]

No. That's not how you use the scanner. You don't always CRANK it.

You leave the power limit at 100%, run scanner, apply settings, turn up power limits.

Why would I base the overclock off the 100% power limit? Wouldn't I want it to find the highest stable clock possible within the raised/max power limit? You wont get as much with leaving at default. Doesnt the algo check clocks, voltage, temps, and power limit? So the curve would be lower since chances are it hits the power limit.

 

[Shou Miko]

I think I will try that auto oc function for my GTX 1080 Ti Hybrid even I am just bump up the power limit and it does nicely

 

[Deleted member 67555]

So it took about 45 minutes for this to figure out what took me a week but with a better fan curve....it's 5mhz less than what I had.
No more thinking to OC a GFX for me... it's just clicky clicky now...

 

[Enterprise24]

The curve that OC scanner suggest is very aggressive on my card. It set 2139Mhz at only 1.062V !! I know that my card need 1.1V for that clock... Testing with Heaven and got black screen immediately.

 

[Shou Miko]

Am I the only one after this Afterburner update that has problems with the OSD (On Screen Display) via RivaTuner?

Mine keeps disappearing and if I reboot my Windows 10 it works again.

 

[Assimilator]

I must say, this is a damn useful and impressive feature - it honed right in on the exact maximum stable value that took me many long hours of tweaking to determine. Next NVIDIA card I buy, I'm probably gonna run this OC Scanner against it first thing, then never touch the clocks again.

(I don't understand why you have to change to a shitty MSI AB skin to get access to the feature, but oh well.)

I think this scanner API was designed for "reference" cards which do not come already overclocked. For factory overclocked models it does ... well, nothing.

Nope, I have one of these https://www.techpowerup.com/gpu-specs/msi-gtx-1070-armor-oc.b3654 and default boost is 1746, OC Scanner pushed it all the way to 1961.5 (with +100 vCore/126% power).

 

[AlexUnwinder]

(I don't understand why you have to change to a shitty MSI AB skin to get access to the feature, but oh well.)

You may access voltage/frequency editor with keyboard shortcut (Ctrl+F) from any pre-Pascal (V2 and V3) or third party skin having no dedicated buttons for OC scanner and VF editor in main GUI.

 

[Mussels]

You may access voltage/frequency editor with keyboard shortcut (Ctrl+F) from any pre-Pascal (V2 and V3) or third party skin having no dedicated buttons for OC scanner and VF editor in main GUI.

thanks, i couldnt find that menu

 

[D_Lo]

Hi,

I've been following this forum from time to time, and only joined today for my first post. Although technically minded I have limited overclocking experience, please bear with me, thanks.

Two Gigabyte RTX 2080 cards were tested to check and compare their health and performance:
VCard 1=Windforce 8Gb OC (Bios F4Beta) and VCard 2= Gaming 8Gb OC (Bios F8), with max Power of 120 and 122 %TDP after Bios upgrade.

There was no problem running AB OC scanner its included Test with the first card. On the other hand, for the second one (Gaming) the monitor screen went black for several seconds (1s to 10s), and usually, this happened just at the change over from Point 1 to 2, and 3 to 4 and the end of the scan. The Power Limit (and Temperature Limit) slider was tested between default and max out and also the voltage control and monitor locks were toggled and not to avail. I let the room temperature drop for the tests between 13.6-15.7C, GPU max temp =56-72C (+42 to 52 relative to RT), Gaming has a little better cooler than Windforce. The Fan was tested with Auto and custom curve and also with a set value of 80-100%.

For stress tests, I have also used 3DMark , FurMark, Heaven, ... None of these runs exposed instability for either of these cards though. The best average OC from MSI AB OC Scan tool were: Windforce = 73MHz (max GPU Temp = 68C, RT=13.6) and Gaming = 61MHz (Max GPU temp =62C, RT=15.6), repeats for gaming had more variability 53-61MHz, averaging in the 50s.

Is the screen going black a typical "instability" as quoted below?
Can these be a sign that the "Gaming" card having some underlying issues not present in the "Windforce" card? Is it power limited? (Did forget that the CPU had been set on Boost 4200 -> 4800 MHz when Benchmark a couple of days ago. Platinum 550W PSU, M.2 SSD, 3 120 PWM case fan, low profile Noctua CPU fan. Silverstone confirmed the PSU should be fine.)

Interestingly the hickups of the Gaming card never prevented it from failing the OC scan, and often the scan at point 4 was taking longer with no black screen until the end of run scan flicker.
Another note, I noticed that with constant fan setting, say 100%, for Point 4, the fan control seemed to have switched to Auto/Custom and then reverted back to 100% once the scan had ended. Anyone else noticed this?

The goal here is not to overclock my GPU but to test the health and performance of these cards. One was a new card but looked like an opened box (Windforce), the other was a like-new item (Gaming), which didn't come in the original box. This made me concerned and suspicious (especially for the Gaming OC card - hence these quick series of benchmarking- and stress-tests. I will keep one in the end and these quick tests were carried out to help decide.

Any thoughts and recommendations for testing these two cards?

Thank you for your inputs, it'll be greatly appreciated.
D

Just follow the directions in the video with a minor difference:


After scanning IMMEDIATELY apply your curve and save it to a profile. Or your curve will be lost and you have to start all over again.

Also do not slide the power and temp limiter before scanning. That will create instability issues. Leave them as default and do the scanning.

As for fan curves you can do 100% fan speed before scanning, but then that means the stability is only good if you blast it at 100% fan all the time.

For me I just used a custom fan curve and start scanning afterwards.

 

[D_Lo]

Update:

The Gigabyte RTX 2080 Gaming OC was further tested to look into possible insufficient PSU power. The very first change was to go in my MSI Z270 Carbon Pro Motherboard Bios to switch off the I7-7700K CPU Boost option from 4800MHz back to 4200MHz. The control fo the case fans were modified in stages as follow:
1) First, the three 120 PWN arctic case fans were set to manual values, and so was the trio fans on the GPU in an effort to minimize fluctuation in load and noise
2) Once the case ran sufficiently cool, on a second pass both an HDD and an Optical DD were removed.
3) Fan control was changed back Custom profile in the end, with at RT=17C, Idle GPU Temp=28C.

The trends looked less noisy but overall the average OC frequencies achieved were between 53 to 56MHz.
For the last OC scan, at the end of the pass for Point 4 (i.e just before results), GPU Temp max=54C (Custom profile GPU Fan=54%), GPU core clock 2040MHz. Voltage locks were back ON and Limit Power=100, Limite Temp=83C. Scan succeeded, average overclock is 53MHz, and a message about Dominant limiters: Voltage, No load.

The goal was leaner power usage on my HTPC/VR system, and it seems that it helped but the instabilities still occur. It is not clear cut my Silverston Strider Platinum 80 Plus 550W PSU is really insufficient. https://www.amazon.com/gp/product/B017WL5UIG/ref=oh_aui_search_asin_title?ie=UTF8&th=1
The few PSU calculators I tried indicated that even with peak load for all components, the requirement was expected to be between 480-510W - and that was before removing the two drives and there one PCI-e x1 TV tuner I can remove too.

Can this black screen instabilities a cause for concern (i.e. GPU health/performance)? Are they really indicative of PSU issue?

Was this GPU abused already (before I got) or is it just a case of the "Silicon lottery" that it performs poorly on these quick scans and stress tests in comparison with the other Graphic card, the RTX 2080 Windofrce 8Gb OC?

On a side note, when the GPU Trio Fans were set to a fixed value, the value did vary for point 3 and point. Usually, for point 3, it dropped below and it ramped back up and overshot for point 4 tests. ANyone else noticed, and is it expected?

I am learning a lot in the process and there are quite a few variables. Thank you for reading my posts today. Looking forward to some pointers and clarifications about Afterburner's OC Scan tool.

Thank you.
D

 

[Mussels]

The scanners arent magic, it could just be picking an OC thats too high for your card