Gadgetory

well it's been a while since we've had a serious fuckup moment on this channel and we just had one right now and we had a great dress rehearsal of this video because I just got done recording it and went to turn off my wireless mic and it was already off because the batteries died mid video I totally pulled a Paul's hardware did what it's so small you really think so what does the wife thing well she says it gets the job done - ugh that smell the fractal design No - OH - it's not about how big it is but all about how you use it what you see next to me here is a lineup of graphics cards that I've been collecting because we all know I collect graphics cards like old kids to the 50s collected baseball cards or maybe kids at the 80s - because I collected baseball cards and I've obviously moved on to bigger and better collecting but I do like to play around with them for the sake of science and today we are going to talk about Moore's law we're going to try to apply it to these graphics cards and see just how far we've come in the last 5 years with the nvidia side of graphics card starting at the GTX 580 now the reason why we're doing Nvidia and not AMD is simply because I'm having a hard time sourcing the cards I need for AMD because most of them seem to have died in self obliterating balls of fire over time because you know AMD graphics cards are pretty damn hot and a lot of them have seen premature death due to heat consumption and you know spontaneous combustion I can get my hands on a 59 70 69 70 7970 and what was a - 90 - 90 ex has already had those and a 390 ex because I don't have a reference 390x I only have the one water-cooled unit if I can get my hands on those cards I will do the same thing I'm doing today for AMD and then we'll do a collective video where we talk about the generational improvements in performance between both sides and see you know kind of who has really been progressing farther over time and we'll just let those numbers do the now this isn't an AMD versus nvidia video don't look at it that way this is just the cards that i happen to have on hand now this is my GTX 580 this is mine I've actually had this card now for quite a while I actually bought this card new back in November of 2010 when they launched and this has got a lot of nostalgia with it I actually sold it to a friend who upgraded his system and when he did I was like dude can I get that 580 back and he he actually gave me back the 580 so that I could do this video awesome this card was overclocked and water-cooled when I had it clearly we put it back in stock reference format we also have a GTX 680 this was sent over by Jakob a TV ga thank you sir this is one of his personal cards my 680 s I sold my 680 that I had to my brother-in-law so I don't have access to that anymore gtx 780 this is actually coconut monkeys card thank you coconut monkey for let me borrow your card he's currently buy it borrowing my geforce gtx 980ti g1 gaming from gigabyte and when I told him I was done doing the benchmarking he was like oh do I have to give back the gigabyte card he referred to it as the gt-r of graphics cards I told me I had to give back soon but he can borrow it for a little while longer this is his card and then I do have my personal gtx 980 classified kingpin actually it is a classified classified kingpin card but yes this is not a reference card however I did get a reference bios from EVGA that will slap into this card and turn this card into a reference card where it has no benefits of being the classified of a kingpin or any of that it will not boost any farther it will not overclock any farther it will not even increase the power target or any of that stuff when I plug this in and just let it go it will do the exact same behavior as a reference 980 so this is not going to be indicative of the classified kingpins performance keep that in mind and then just for you know laughs I went ahead put the 980 or the 989 and 980ti we did a tight necks in there now obviously there's a couple of cards missing from this lineup where I'm missing the original Titan we are missing the Titan black we are missing the 980ti and the 780ti now the reason why I didn't include those as those cards are pretty much variations of the same architecture that was released incrementally throughout the year rather than being a generational improvement to the lineup what does that mean it means there was no 680 Ti and there was no 580 ti so I decided in order to keep it as apples-to-apples as I can we are comparing these bare-bones 80 series lineup and then comparing to the tight Nexus because that's the single fastest GPU Nvidia makes right now with all of its CUDA cores unlocked and I figured it'd be fun just to see how that kind of McMaster maybe in the future when we get more TI models because maybe they started the TI model of the 80 series with the 780 maybe then we'll do this video again in another couple of years so we can do 5 or 6 cards and see how the generational gap of the TI cards have actually improved so now that we've rambled on quite a bit and I don't think there's anything else to mention here other than again all of these cards are reference design with the exception of the kingpin on there oh yeah I think that's pretty much all there is to do now I just want to talk about the methodology here real quick the reason why I did all of this was I wanted to see if Moore's law even came close to being true for graphics cards and the reason why I mentioned that is because Moore's law is something I've always been interested in especially being a part of technology and loving computers and transistors and integrated circuits and all that sort of stuff Moore's law Gordon Moore was actually a co-founder of something that is in my eye Gordon Moore put this in my eye what the hell no he didn't really do that don't don't get mad at him Gordon Moore is a core co-founder core founder yeah I guess he kind of is a core founder too he is the cool I cannot say that he is the co-founder of Intel and he had an observation over the years and especially on paper that the amount of transistors in a dense integrated circuit had been doubling every two years that's a pretty aggressive curve and you consider if they've made that observation back in the 70s where we are now in the you know 2015 we have come a long way with billions of transistors in a single unit now obviously these GPUs are not going to hold true of being twice as fast every two years or maybe it will GPU performance I don't know if that's going to integrate directly to the comparison of cpu performance which is what they were more or less talking about then because they didn't have graphics units then like we do now so that's why we're doing this video how far have we come let's go ahead and find out now obviously I only use firestrike for this test and the reason why I did that is it wouldn't have been fair to take the older cards and put them against newer titles because remember as the GPU performance and the GPU technologies increase and evolve so does the api's and so do the developers utilizing api's available for current hardware in most cases I mean do I need to even mention Witcher 3 and hare work some is an example so it would not have been fair to take games like Far Cry 4 or Witcher 3 or Crysis 3 and put them against these cards because the GTX 580 was the original Crysis days where it was like can it run Crysis once again recording J you're wrong it wasn't crisis it was crisis to the original Crysis can it run Crysis was kind of during the 8800 GT x and 8800 ultra days which again I owned one of those and yeah it didn't do a great job at running crisis but the 580 was around during the crisis too era just one to correct myself once again in post processing because recording J is a dumb ass and editing J is a little bit smarter I don't even know if it would have loaded with Crysis 3 so what I did instead was I took the fire strike GPU benchmark and then we ran it in normal extreme and ultra and as you saw the performances on there were pretty consistent when it came to the amount of increased a little bit more than that going from 1440p 580 to 680 at 35 but then 28 29 and 23 where we had less of an increase going from 980 to Titan X at 1440p so as you can see 980 is looking like a really good sweet spot for 1440p gaming especially compared to a card like the Titan X which is twice the price now if we look at 4k obviously 50% and then a 65% jump and then 30 and 25 where as you can see as cards got newer the focus turned towards higher resolution because the 580 if you guys were not into PC parts when this thing was new the 1080p was looked at like 4k now where 1080p was that resolution like oh man can it run 1080p now everyone's like 1080p who cares my phone has more pixels than 1080p but now 4k is obviously the thing and now we're hearing about 5k and obviously there was eight Caleb's being talked about for a little while there as well it's kind of getting insane we're pixel rate is increasing faster than GPU speed is that's why hopefully Pascal is going to be quite the improvement now I wanted to say that with the pretty consistent numbers jumping you know improvement wise from generation generation could we predict what the next graphics card speed was going to be like from Nvidia I would have said yes if we were still using standard GPUs surrounded by the RAM chips that we're seeing today however that's all going to change obviously with HBM especially with AV HBM gen 2 which is what Pascal is supposed to be now a lot of people have really spread some misinformation about Pascal is supposed to be 10 times faster than Titan X oh my god the 4k can you imagine that's not what Shen what Jensen said he said ten times the compute power so Pascal he was talking about being a professional environment powerhouse for video editing and graphics design Direct compute and stuff like that not gaming the jury is still out on what's going to happen with Pascal and gaming we're at the wait till at least next year to find that one out I'm looking forward to it but for shits and giggles though I also included skunk works in this lineup because like I said this was my exact graphics card right here and I was curious as to well you know I had this car in this card in 2010 how far has my gaming experience increased since then well here are those numbers and as you can see it's pretty freaking I mean it's it it pushed everything down so far you can't even read the numbers on the chart for the lower-end GPUs anymore that it's so squished but we were 9 times 3 faster 9.3 times faster than a single 580 at 1080p 12 times faster than a single 580 at 1440p and at 4k 34.6 times faster remember this isn't % this is times faster that's just for fun holy crap man no no skunkworks getting kind of old though it's probably time to upgrade well I hope this video was fun for you guys and like I said I will go back and do this for AMD if I can get my hands on the cards that I need I might have to reach out to some of the vendors and see if they have any old like RMA units sitting around or something that could at least get me through the test but otherwise I hope you guys had fun with this sake of science type of video science I like science and stuff if you guys enjoyed this video hit that like button if you thought that this was stupid then by all means feel free to hit that dislike button believe it or not and actually helps as well follow on social media if you guys want to chitchat about random stuff and Technology and all that otherwise I will see you in the next video and thanks for watching this little experiment of mine let me know if you want me to do more stuff like this if you want me to do more direct comparisons let me know what and get ready to do kind of an all in one water cooling shootout to where we take a few different water coolers and see which is best I like doing comparative videos shootouts are fun it's like watching people fight I love watching people fight like UFC is freaking awesome so is Ronda Rousey Ronda is very arousing if you know what I mean yeah but she got me in a leglock I'd probably squeal like a pig in a good way