Last night NVIDIA did their live stream and Jensen delivered the news that a lot of us tech enthusiasts have been waiting for and that is RTX 3000. now they've announced three different products and that is the RTX 3090 the RTX 3080 and the RTX 3070 but what we have seen already is that another YouTube channel called digital foundry has uploaded gameplay footage with percentage differences between an RTX 3080 and an RTX 2080. so in today's blog i'm going to be doing some educated guesswork with the numbers that we already see in that we're going to grab those percentage numbers convert them back to fps and then guess the performance of the RTX 3070 and also the 3090 and even compare that to the RTX 2080 Ti to see the value propositions of these new cards in a number that's easy to digest.
Let's get on with it straight away looking at the numbers in the cuda core account from NVIDIA's own website we can see that this is essentially cuda core is NVIDIA's name for shader and AMD call it stream processor what they're doing here is something that you can't really compare directly with the RTX 2000 series or truing versus the new generation ampere because in one of the illustrations and one of the videos that they showed they had the shader like literally splitting into two kind of doing some matrix agent smith duplication kind of thing here and so what they're doing with the cuda core performance and also the cuda core count is actually a lot different than the previous generation so basically if you got out your good old calculator and you said oh wow the 3070 has 5888 cuda cores and the 2080Ti only has 4352 and then you said okay it's going to be 35 faster you would actually be making a wrong guess at least in my opinion and reason being is because what we can do however is look at the fps that's already out there on that 3080 versus that 2080 comparison that we saw over digital foundries basically it wasn't an fps comparison but again we'll throw it back in let's say the 2080 is a 100 fps baseline and what they showed was the 3080 was performing 80 faster so we can convert those numbers now into fps so we've got 180 fps versus 100 fps on the 2080 and so from here what we can do is actually then grab those cuda cores on the same generation and start to derive some numbers this is a fair practice for guesstimating numbers you just can't compare them from different generations because there's changes in the architecture so what we're going to see here i feel is that a 2080 ti is roughly 25 percent faster than the 2080 and so we put that in 125 fps and then we see that the rtx 3090 is going to be about 20 percent faster than the 3080 so we put that in around 216 fps now the 3070 actually has a lot lower cuda core count than the 3080 as well as the 3090 those being 8704 and 10496 cuda cores respectively there's also 28 billion transistors on both the higher core count gpu's versus the 17 billion on the 3070 and the 18.6 billionon the 2080 ti and so what this means is the RTX 3080 is actually going to be acute down version of the RTX 3090 rather than the 3070 being cut down version of the 3080 that's actually going to be a different die and size altogether and then basically taking the RTX 3070 cuda core count comparing it to the 3080.
We're going to have around 68 of that shading performance available on the 3070. however one big difference here is that they're using gddr6 memory on both the 3080 and 3090 so that's going to have more through put as well as a bigger memory bus compared to the 3070 whereas the rtx 3070 is gonna have gddr6 memory the same as the 2080 ti albeit it is going to be clocked at 16 gigabits per second versus 14 gigabits per second on that 2080 ti so significantly higher memory speeds and drastically higher cuda core counts is going to make the 3080 and also the 3090 in a different league compared to that of the 3070 and subsequently the 2080 ti which if i had to guesstimate on the 3070 now you can maybe expect around 120 fps versus 125 on the 2080 ti so although NVIDIA is saying that you're going to get pretty much the same performance as the 2080 ti i believe it'll be a little bit lower on release and then of course with driver optimizations over time you will see the 3070 beat out the 2080 ti and so with those guesstimated numbers out there we can look at the value proposition of the cars themselves where the RTX 3070 is going to come in with a 500 price point releasing in October and i'm going to guess that is early October because they're going to want to get these cards out to market as soon as possible whereas opposed to the 3080 that's coming out on September 17th and then the RTX 3090 is about a week later so the 38 is actually going to hit the market first and i believe it's going to be one of the best value cards in this lineup coming in with performance of a value proposition at least versus the 2080ti of what my numbers would guess here is 186 percent roughly 44 more fps and roughly 42 percent better valuable costing 70 of the original thousand dollar 2080 ti it's going to deliver much better fps it's also going to be much more power efficient and also deliver a better price point then look at the RTX 3070 that will be roughly just under 200 of the value of an RTX 2080 ti so that will edge out the 3080 but in the realm of getting the max fps possible that's why i'm leaning towards the RTX 3080 actually being one of the best buyers for enthusiast tech heads out there with this current lineup though lastly the big fish in the room the RTX3090 coming in with roughly 20 more performance than the RTX 3080 they're costing over double the price isnot going to be a value proposition it's also going to be designed for someone who wants to do say more 8k gaming with the 24 gigabyte vram buffer so it's going to have significantly more vram than the 3080 coming in with 10gigabytes of that stuff versus the 24 gigabytes and anotherthing about the 3092 is it's got the ability at least from nvidia's websiteto be able to be put into nvlink so if you want to do some serious 8k gaming with max ultra details and of course you've got an endless supply of money within that wallet of yours then NVIDIA is certainly going to be giving you that option so when we put it into a nutshell RTX3090 is not really going to be a value play but it's going to be for someone who wants the best experience possible on PC and of course they have a lot of money then we've got the value plays they're going to be the rtx 3080 and 3070 and also the upcoming 3060 i'm guessing is going to be bringing really good value to the table too.
So those numbers aside that's only half this video where we're looking at the raw price performance figures and i'm just having a bit of fun guesstimating the value and the numbers that are going to come out of these cards what NVIDIA showed in their presentation was a lot of different things in terms of not just they're moving to eight nano meters so you're going to get better performance per watt or better efficiency out of these cards but they're also showing things like improved ray tracing performance more tensor cores and also the shaders which they're saying are going to be more efficient too but on top of that they were working on dlss 2.0 which is essentially their super sampling tech which is going to be giving a better image but also giving more fps because it's doing it while super sampling up from an inferior image previously to fill in and basically let the GPU do the prediction on the image quality so at least that's my interpretation of it the next thing they announced on top of the GPU launch was one that was really cool for me personally and this was the reflex technology and that is that they hope to reduce the input lag by essentially working with the game developers to offer a new mode that's going to offer overall better input latency so what we can see with this graph that they showed here was that they had the average system input latency and at the end of the graph we could see that that line had moved drastically to the left but at the start of the graph the line had only moved a little bit to the right so basically what they're saying is they're going to be looking uh working with the game developers is my guess again in that they're going to be making sure that the slowest frames or the stuttering or micro stuttering is eliminated from the game by drawing the gpu frames as fast as possible another good thing is you don't have to own an rtx 3000 card to get this technology they're going tobe introducing it released from the graphs with gpus like the gtx 10 series and rtx 2000 series as well though another thing they did announce was with a new series of monitors they've got these 360 hertz monitors coming out but they're going to beading an input latency or system latency module into the monitors themselves so you monitor things like your pc latency your mouse latency or your total system input latency including the monitors by using this module sothat's going to bring for some really cool comparisons where I've been doing these in the past with 1000 fps camera i'm going to be definitely testing that against the 1000 fps camera and just seeing how cool this module really is where a lot of mouse reviewers out there especially my friend rocket jump ninja could really use a technology like this when they'redoing their mice reviews just to get you guys exact mouse input latency figures but speaking of reducing input output they did talk about rtx io which essentially looks to get really fast gen 4 ssds and load up things like textures and map pop-ins just directly from the ssd to the GPU where as opposed to beforehand it would have to go all the way through the CPU pipeline and then out to the GPU so NVIDIA claims this will bring more fps as well as reduced input latency on top of reducing input latencies by working with the game developers and their ap is so performance figures latency figures we're done talking about that but they also announced rtx broadcasting their software suite which is essentially going to help streamers without green screens so even like i'm using this backdrop here iwon't need a green screen and i could put myself in the middle of new York bringing a live stream minus the ambient noise well actually even if you had the ambient noise i think they got the rtx voice thing that could cancel out the ambient noise so you could just say hey i'm actually in new York and maybe people would believe you or they wouldn't believe you but this is essentially using your gpu just like rtx voice did to process the background and eliminate it and just keep the subject in focus how well that's actually going to work just in say a setup like mine i will test for you guys when we get the gpus as well as testing everything like this input latency numbers the performance when i get these cards in and i start testing them when the review embargoes drop i'll just give you guys the numbers straight and of course if you've got special requests on the testing that you want to see then do drop acomment in the comment section below and also let us know in the comment section below is this the gpu launch you've been waiting for which are the cards are you looking at getting or are you going to wait for say the rtx 3060 and possibly an rtx3050 to come out before you get your next lineup of GPU in that rig of yours though before we get on out here there's one more thing to talk about and that is the coolers and the power connectors that they've announced now they've listed on their website the 3090 is going to have 350 watt power requirement and with that power supply recommendations of 750 watts just to be on the safe side and that continues for the 3080 with its power requirement of 320 watts then down to the 3070 which is 220 watts and recommended power supply is 650 watt though the 12 pin power connector this was seen in the video footage so that will be a thing though i'm guessing NVIDIA is just going to throw in an adapter so if you've got a currently good high powered power supply you're going to have to worry about this and i believe this is going to be the way for quite some time to come though we are seeing custom aib models being shown out there in the wild and a lot of them are just still going with the eight pin connectors and in the case of a Zeus i can see a three eight pin requirement there so the 3090 is going to need a bit of power but at the same time it's going to be a very powerful GPU so naturally with that extra power consumption you're going to have to need four extra power connectors or in the case of NVIDIA introducing a new 12 pin power connector though the cooling design itself on the reference model is something to i guess look at and wait and see and see how it performs because the science and at least from what they show with the illustrations does make a lot of sense where you're going to have one part of the GPU venting out like a traditional reference card but then you're going to have the fresh air in a good airflow case by the way and i'm guessing in video saying well since you're buying these higher end cards you're most likely going to have a higher end case with good airflow so this is designed for a higher airflow case scenario where the back of the GPU is then going to flow air out over the CPU so would this make hotter CPU temps or is this designed with aio water coolers in mind i'll give you guys the numbers, in the meantime a lot of the aib partners are just going with the traditional just fans blowing air onto the GPU seems to work and it has worked for quite sometime it's still going to work absolutely fine and of course back to that age-old saying if it ain't broke don't fix it so speaking of not fixing things this question of the day we don't really need to fix it.
Comments
Post a Comment