Gadgetory

everyone I'm joined by Tom Peterson distinguished engineer at Nvidia quite a prestigious title it seems prestigious yes thank you prestigious so Tom just got off stage here at a press event talking about overclocking for the new GPUs including the 20 80 and 20 80 TI yes well we did we just launched our touring products and we had our editor stay here it's great to see you again and I'm really excited because we we introduced a bunch of new things including something called the Nvidia scanner and the Nvidia scanner actually makes it possible to do one-click overclocking so a lot of new stuff before that this video is brought to you by the NZXT E Series power supplies the E Series PS use our high end power supplies with real time digital voltage and temperature monitoring per rail wattage measurements and data logging functionality for power usage the e series PSU is also come with a 10 year warranty all the way down to the 500 watt unit and they run fully modular and with silent fan operating modes learn more at the link below so we've we've heard this a lot one-click overclocking I've been hearing this for a long time to a we take something like that yeah yeah so one-click overclocking most the time in recent years it's gotten a lot better I'll say so we've done some man vs. machine things we're all overclocked we'll let it overclock it's pretty close yeah but I mean sell me on it why why should I trust it to handle it this time versus last time what's different fair enough question well I can tell you what's different this time is Nvidia is doing it so we've developed all the logic it's divided into two main chunks one is called the test program the other one is the test load and the test program kind of knows about how does our voltage frequency curve look so it knows exactly which voltage points to test to make things go really really quick and then since we can control the voltage exactly because we're Nvidia you know we can control that stuff there's no variability caused by games or load on the load side it's a mathematical test that we develop to exercise our GPU precisely so we can modulate that load based on the power or the temperature and effectively think of it as if you could get Jonah to like overclock your GPU for you and write giant programs to do it this is what he would do right and this is it's if you watch the thing it steps at one voltage runs a bunch of tests steps at another voltage runs a bunch of tests and it delivers a high quality curve in about 20 minutes so and this just to be clear here there is an OSI scanner program or there a couple of them at least previous generation how is this different from those well what's different is this doesn't crash it doesn't it doesn't like TDR it doesn't it doesn't fail because we're looking at a data comparison and that data comparison we can detect errors in far before we see any other like visual or Windows corruptions um the other thing is of course we know how our VF curve looks so we know which points to let go after so we're actually focusing on five different voltage points and interpolating between those voltage points to generate the kind of final curve so it's like all of the knowledge that we have in our chip team and our board team and our architecture team has been compressed into this test structure and our test load and then you know I have to bring up voltage of course I think on the in your presentation I think we're seeing something like what did stop at 2130 and 130 and one point zero something volts one point zero six something nice is it one point zero yes yes and that's why I have the question voltage I guess the community decision on this is that it's been locked is the the phrasing I'll use so tell me why I'm wrong ah well you're not wrong voltage is in fact controlled and restricted but I think you're the reason the community is unhappy I believe is that they feel like the voltages nefariously locked right that that Nvidia is locking this for poor reasons but the truth is we know exactly how our silicon behaves and we set our voltage limit like the peak voltage that we run at is set to be reliable so that we can run at that voltage for many many years five years is typical now as you increase the voltage beyond that we're going to start having fallout meaning that some chips are just going to die so it's our our kind of obligation to help gamers not blow up their cars everything else on that that UI for overclocked is non-destructive it doesn't void warranties we don't have like Oh tracking if you over voltage or not so what we're kind of trying to say is we want to make sure the people can enjoy overclocking they can experiment with it but they don't blow up expensive graphics cards voltage is the one thing that will blow up the card that is accurate you can play with the power you can play with the temperature you can play with the offset all of that is non-destructive but the voltage will not just perhaps blow it up right away but it'll also for sure degrade the performance of the chip over time so as kind of a promise to our end users its overclocking is free and it's safe and if we allow them to slide a voltage slider where you know suddenly they're going a to X the voltage we're gonna get a lot of surprised and unhappy gamers I'm gonna give you one more opportunity to tell me why I'm wrong before we move on to Foose 4.0 ok so the next one why is the voltage slider from 0 to 107 offset and Pascal or I would assume touring cards why is that not a placebo slider whoo well it does what it does I mean what that percentage means is how much of the available voltage Headroom is unlocked so it doesn't necessarily say what voltage are you gonna run at it just says what are the new voltage levels that are available so as an example normally you're not running at your peak voltage you're normally running lower because of temperature or power so just sliding that slider up doesn't change anything unless you actually hit those voltage levels so I think a little piece sometimes people are a little bit disappointed about how many levels that actually unlocks because it's really only gonna be a few ticks because we're going from five-year reliability to 1-year reliability and that translates into I think roughly 3 more voltage steps on most of our cards so it's not placebo it's just you know I want to be responsive but I just don't want people blowing up cards so it's like yeah yeah but I think we talked earlier and I know you would like a way to say I'm an expert I would like to play with my card so give me a ways that I can maybe go on a website get a cookie or maybe you know put a special USB key and I can make it harder but just don't lock me out entirely that's what you're saying so I'm gonna go look at that but right now obviously we don't have a change but it's it's not something that we're opposed to it's just you know we've got to find a way to do that without compromising quality and if any of you I guess on that topic have a specific idea relating to that let Tom know below actually let you know and then you know you can tell me yeah but let me know but you're gonna make me sift through the comment well yeah no yeah yes yeah you can tell Tom is a is high up in the company when he is delegating that to me you got it yeah so last question for you a quick overview which is asking a lot but boots 4.0 what's changed versus 3.0 what are give me the primary features that have changed okay well there's two things big one is that we have no longer had a card temperature limit so on boost 3.0 you set a temperature limit and that means that pretty much the car it's gonna stay below that limit and if the temperature ever rises above that limit then we cap all the way down to base o'clock and we stay at base clock let the fam ramp that's our default behavior you can slide the temperature slider up to kind of avoid that if you want to allow a higher temperature higher fan on boost 4.0 what we've done is actually created a plateau in that transition so it doesn't go from uncap down to base it actually goes from uncapped down to boost and it'll stay a boost for a while before the temperature goes a little bit too high and then it'll cap down to base so if you think about that it kind of says boost for dot o is no longer like this cliff it's more of a plateau and we're gonna allow temperature and power to rise up to a secondary limit where users can you know say I'd rather I'd rather trade off lowering clock for lower temperature now since that's such a user centric decision we've also enabled that curve in a new API so the boost for API allows end-users to sort of dial in their own boost algorithm and that will you know you can imagine somebody says I just want a hyper if I don't care about loudness or I want sort of a smoother transition between acoustics and performance and that's really what blues for is all about very cool so there's your quick overview of that otherwise we'll have articles online by the time this video goes live so check the link in the description though for more information tom thank you very much great to see you again brother for joining me and we'll see you all next time