https://www.dropbox.com/s/rpnbqhv17lcytdu/Captura%20de%20pantalla%202019-05-08%20a%20las%2020.39.05.png?dl=0Hello PAtrike I leave you a capture with URL format to not occupy space in the post.
It is a small problem with the native OC, I already realized a while ago but now they are asking me a lot in the telegram group and I decided to notify them.
As you will see in the capture in a rig of 6 RTX 2080, all buy at the same time in the same batch, same model, same bios. I mean 6 exactly the same cards.
How can it be that applying a native OC profile, some have the highest core, and others the lowest core ?, and even the voltages change but that does not worry me much, the voltage does not touch it.
But if it is very rare, that if I apply an OC, some cards remain in 1600 and others exceed 1750 of the clock, when all supposedly the same OC is applied to it. As a result it happens that many times the miner fails and restarts, it is not serious but annoying and the only way is to lower the OC of the whole rig, at the cost of losing power to have stability.
I did the test with Aftherburner and being all the same, they were all at the same core speed. That does not happen with the native OC, which seems to apply some cards if not others, and I do not understand.
The great advantage of the native OC is its ease, but I already made it clear that the work with fans does not work, and you will see the amount of complaints that you will have in summer, I only use Aftherburner for the control of fans.
Check the programming of the native Oc. Because I can understand that if I have a rig with several brands, that's what happens. But having the same model bought at the same time in the same batch, with the same bios, in the same machine. Because some rise more the core and others do not?
What I do not want is to use the MSI server again for the OC.
Given that everything else is the same in terms of mining algorithm and software, it would be interesting to see if there are any differences in the GPU clocking settings when Awesome Miner reads them from the nVidia driver.
Can you open the GPU clocking dialog in Awesome Miner and click the View GPU Details button to generate a full report of the GPU settings. Please do this once after you set the clocking with MSI Afterburner and once after you set the clocking with Native Overclocking. Any differences in these two reports?
I do the tests you want, but I can not understand it with the automatic translation. If you give me a capture of that option, I'll understand. But I do not get to understand it well, sorry
I will first do a few more tests myself. I will get back to you if I need more information. Thanks for your questions about this.
I think I need some guidance from you or someone else with MSI Afterburner skills here related to how to lock the Core clock based on the behavior described above.
What Awesome Miner and most nVidia clocking tools do is to set the Core Clock Boost, which is a way to increase or decrease the Core Clock frequency relative a base line. It's an offset like -50MHz or +75MHz. The nVidia GPU boost feature will then use this as input when setting the Core Clock speed, and this speed can change any second based on many factors.
I've tested with nVidia P104 and nVidia 2080. No matter how I set the Core Clock setting in MSI Afterburner, it still behaves like when I'm clocking via Awesome Miner with a Core Clock speed that isn't constant during mining. If I use algorithms like Neoscrypt, I see Core clock variations all the time. If I use memory heavy algorithms like Ethereum or Phi2, the Core clock is however much more stable. Same behavior no matter if I change the Core Clock via Awesome Miner or via MSI Afterburner.
Did I miss something in MSI Afterburner to get another behavior? Are there any other clocking tools like nVidiaInspector that also would behave like MSI Afterburner or is MSI Afterburner the exception here?
I know that the core can vary a bit, that is normal, it is not exact all the time.
That's not the problem. The problem that using 6 equal cards, some always have the higher core than others. They all oscillate a bit inside your watch. But it is not the same to oscillate between 1600-1650 all the time, and that other similar cards are oscillating between 1720 and 1770. Always the same cards.
I mean, I know that the core can oscillate a bit, I see it in the cards, but some are placed in a higher core and oscillate there, and others are in a lower core oscillating there. When the OC is the same for all and they are exactly the same card purchased in the same lot.
If you look again at my capture that you leave, you will see that for example the memory frequency is exact on all the cards. But you will see cards in 1600 core and oscillate there, and others that do not fall below 1720 oscillating there.
I would understand that, if they were cards of different brands, because I myself have rigs with multi brand cards, and there I can not comment. I comment the case when all the cards are the same, and accepting their oscillation of Core, some for whatever reason, they are placed higher and others below, a difference of +100 in core I do not see it right. In Aftherburner I saw the oscillation too, but they were all in a range of core closest to each other.
In any case, I only report what happens to me and to other people in my telegram group. If it can be fixed well, if not, then we will continue like this, I will not remove the native OC because I like it a lot, except for the control of FAN, which is why I use Aftherburner.
If this can be improved then better. If this is so and you can not do anything, then I'm satisfied. I love native OC, I only intend to help with my observations and I can not always guess what I say.
Also thinking a bit, I do not know to what extent the chipset of the motherboard can influence in this sense, in its internal handling of the PCI.
Anyway, if you Patrike, after investigating it, believe that everything is fine, because it is already there, I accept it, it will be like that.
In general there can be variations on a chip level between GPU's even if it's the same model and brand. This goes for both GPU's and CPU's and is also the reason why some GPU's/CPU's can clock higher or get lower temperature than others - even if it's the exact same model. In overclocking competitions they hand pick their CPU's for a reason. For GPU's you can actually see the "ASIC Quality" value in the TechPowerUp GPU-Z tool, on the Advanced tab (although I don't know how accurate this number is).
Anyway, thanks for all feedback. Although there could be something Afterburner do different than Awesome Miner here, I don't think we will be able to find out what that would be. In general it also looks like the clocking set by Awesome Miner works fine.
Also thanks to @moppidoo and @joseph32 for all good information and feedback.