Gadgetory

hey once we just got out of the nvidia event for the twenty seventy twenty eighty and twenty atti launch which includes official specifications and pricing from nvidia and video could have handled the event a lot better overall as we'll talk about momentarily but for now we do have the core specification z' that we can go through and some release dates and prices before that this video is brought to you by the EVGA CLC 280 liquid cooler people ask me how I keep cool during the summer with all this hair well I've tried a lot of different products and if you do exactly what I need many of them cause tangles or worse EVGA CLC 280 helps keep my core temperatures low they're in hot benchmarking sessions the CLC 280 is price competitive and focuses on performance for value offering a 280 liquid cooler at an affordable price get yours at the link in the description below hair mounting kit sold separately the r-tx 20 70 80 and 20 80 TI were all officially announced today the announcement could have been a lot better there were many things that were really not the best with the way NVIDIA handled this event but perhaps that will be discussed in a different video just to clarify something here on the pricing so there's two sets of pricing that Nvidia release there's one on their website which is for six hundred eight hundred and twelve hundred dollars and then there's one that was released at the event for five hundred seven hundred and a thousand dollars and that one had a different slide and we think it either coincides with the reference pricing versus Effie pricing so there might be two disparate SKUs or it could just be a miscommunication we're really not sure at this time other than that though we do have some prices from gigabyte which seems to coincide more closely with the higher prices so a gigabyte range of 790 to 834 the 2080 and 1170 to a bit more than that for the 2080 Ti so we have some partner card prices already it'll be linked in the article in the description below if you want more and September 20th for the availability of the 2088 2080 TI and the 2080 with a 20 at 70 TBD no really stay on that just yet the cooler is the most interesting next thing other than hard specs and that's because it looks like nvidia is going for a dual axial cooler much like most of their partners do or have done in the past in the renders if their representative of the real product they have 13 blades across two fans and they are positioned over the vrm and over the GPO core as you would expect so this is particularly interesting because moving to a dual axial cooler is going to be more similar to what the partners do especially for the two fan cards which means that I don't know we could see some of the partners getting pushed either out of the market or towards three fan coolers and that's probably why a lot of the leaks coolers you saw were three fan because Nvidia is doing this so very interesting very interesting and curious implications for the partners where I'll be honest I'm a bit concerned that there might be some Nvidia kind of pushing the partners around or out of the market at some of the price points where they've existed in the past but it all depends on how good that dual axial cooler is that Nvidia made the leaked PCB shots looked like 70 on power stages if that's accurate for the 20 80 TI then that's a pretty damn good BRM to what we also know is that the specs for the cards remain largely the same as what we saw leading up to launch so just to go over each card individually we have the 2070 up first this one has 23:04 cuda cores CUDA cores are just FP use so an FPU is not really a true core and a seems closer to that they can't do a lot of things that cores can do they can compute but they can't fetch and put stuff into registers from registers can't really work with the cache quite the same way as a core can so you get the idea but regardless 23:04 FP use on the 27 t it's got a 17 10 megahertz boost clock and this is for the founders Edition I need to note that - so there are two specs for the Nvidia cards today there's the founders Edition and the reference spec and the reference I believe is what the partners will be starting from the founders is a bit higher so it's kind of going back to the confusion that Nvidia created in Austin in 2016 when we had to make a special video clarifying what the heck founders edition means because at the time it actually wasn't special all it meant was reference except renamed but now it looks like it is going to be a bit special in that it's got a bit of a higher clock then the reference back so Fe might actually go towards what people mistakenly thought it was in 2016 that's kind of what it looks like right now base clock 1410 for both reference and Fe 14 gigabits per second memory speed 8 gigabytes of gddr5 86 bit memory interface for the 2074 48 gigabytes per second memory bandwidth and then all of the other stuff is as you would expect also these cards have NV link support which is pretty interesting too so we'll talk about that more perhaps in a secondary video but SLI as we know it is going away and NV link is what's going to be replacing it for the next card let's talk about the 2080 I'll just go into order here 2080 29:44 FP use or CUDA cores as they call it 1800 megahertz boost clock 1515 base and the reference non Fe version is 1710 for the boost clock that's the only difference the memory configuration is 8 gigabytes of gddr5 14 gigahertz for a second memory speed 256-bit memory interface width and 448 gigabyte per second memory bandwidth and then the 20 atti 4352 FP use for that one pretty big and 1635 boost clock or 1545 for the reference back as opposed to the efi spec 1350 for the base clock same memory except 11 gigabytes 352 bit boss so 16 16 gigabytes per second and that leads us back to a few other notes from the presentation or things that we kind of thought and put together while we were watching it first one and B link for SLI $80 for the bridge if you want it and NVIDIA frankly hasn't done a great job of convincing us that we need SLI or that it's worth buying but we'll see if that changes NV links quite a bit different it's possible that SLI could now come back into play with NV link if the developers support it that's the big question but either way it's a difference and it's interesting the cooler we kind of talked about a little unsure of how that's going to affect partners and partner relations but Nvidia is in a dominant position in the market and they don't really need to worry that much about much of anything the next part in the presentation routinely Jenson Wan stated that the it would take 10 10 a TTI's to keep up with 110 Turk or one note on this that does not mean that one GPU from a 1080i class card one 1080 TI is going to be that different a 10x change versus one tenser core that's not what that means so when they're talking about tensor core speeds and being 10 times the speed of as 1080i what is really being discussed is the deep learning capabilities machine learning ai processing these things are a bit different than what we deal with in graphics so 110 a TTI does not equal one-tenth of a touring GPU or even a tensor cores what they're really saying in fact what is happening is that 1 tensor core can do 10 times the work of a 1080i in specific applications like deep learning and machine learning so that I just thought was important to note because probably a lot of people might not catch that part image processing is a big one by the way like if you're processing billions of images or something like that next one so there was a lot of time spent talking about shadow shadow occlusion ray tracing recaps of the event from March all the same information tureens been in development for 10 years according to the CEO on stage and it's the second largest ever made behind the v100 for the quadric class high-end cars that were revealed previously the SM is brand new it can do independent fpn enter a integer operations to split color and address processing how that affects gaming or if it affects gaming TBD the next thing of note here so another item in the presentation 2x speed of a 1080 Ti but only with 4k DLS s mode enabled so one of the slides that show it's a 2x speed of a 1080i that may well certainly be true and probably is but the specific instance that was true was with 4k resolution with a new special mode enabled and this is similar to what we saw with the 1080i where yeah and 4k it was pretty substantially advantaged over the 1080 so not saying that it's misleading it really wasn't they had it on the screen it was pretty obvious but I did want to make sure that everyone in the audience caught what was on the screen which is that the 2x speed is under specific circumstances next one the phone number that was on the screen that was part of the joke was our TX 2080 so that name was revealed pretty early on the DG x4 Voltas so DG axe had four volts is on it processing that starwars image that was shown a long time ago half a year ago and the frame time for that was 55 milliseconds one taurine card was 45 milliseconds they wanted to point that out pascal was 308 and had an 8x claims difference by nvidia and also a taurine frame assembly so if we talk about the pipeline for touring frames with tracing one of these like Star Wars type images its raytrace integer 32 shading FP 32 shading and DNN processing it is semi asynchronous so FP 32 shading happens during the entire frame we have a block diagram of it and ray tracing and integer 32 happens simultaneously with FP 32 this is important as we saw with a Titan V asynchronous compute is taking a big step forward for NVIDIA in this generation so we talked about it then we confirmed it with Nvidia at the time asynchronous compute is if you're kind of looking past all the ray tracing stopping at just the hard performance numbers async compute is going to be where it's at for this car to this generation assuming it's integrated properly with Vulcan or whatever the case may be DNA and generates pixels that haven't been finished yet that's kind of interesting so new information with AI will be generated by the card so that it can finish frame is faster and leave some pixels unfinished and then fill them in synthetically also there's a new weighted number of tera r-tx ops they're calling it which is a waiting for FP 32 and teraflops per gig array so they went with Tara RTX ops and one note was they said 78 tera r-tx offs for the twenty Bedi I would assume Ti or basically touring card versus 12 tera r-tx ops it's a mouthful for Titan ax did not specify which Titan ax there are three or four of them so I could have used that clarification there and I think that pretty much covers it so big things here to talk about partner impact how does a dual fan axial card impact what the partners make this generation I think we're gonna see a lot of three fan cards as a result and probably depending on the price floor for them it's going to be harder to compete with nvidia at the absolute low end another potential point of impact pascal and AMD neither of which have the capabilities for processing things like rate racing that the touring cards that do so as rate racing becomes an actual thing in real time if it does in this generation there's a bit of a concern where if you have it enabled it's probably gonna kill the performance on both pascal and andy so it's not just an anti andy move its anti everything that's not touring which is fine that's progress and that's how technology moves it's just that keep in mind just like with all the other high-end effects with any game that's come out whether it's an AMD effect or an nvidia effect you'll have to disable it to improve the performance on older generation cards not that new but something with noting ray-tracing is the next kind of skeptical point where this requires developer integration RT x is an nvidia thane it's an sdk that will allow developers to more or less plug-and-play real time ray tracing but even still even with it being three programs it takes a while for developers to make that commitment games take years to develop and they probably haven't had access to the SDK for a significant period of time it's been months for sure but not on the scale of years as far as we're aware so it could be a while before real time rate racing games other than what was shown on stage the couple of them there were really take market share and that means that the rate racing stuff might not be super relevant for this generation or at least this launch but perhaps the future ones next big question what happens with the 2050 class card do the tensor cores on that have enough oomph to really actually do real-time ray tracing in any meaningful way so lots of stuff to think about with this launch but that's it for this one subscribe for more well definitely have more videos on this as we process the information go to storlock gamers access net to pick up one of our shirts mod mats or other merchandise and go to patreon.com/scishow sexist tell us that directly thank you for watching I'll see you all next time