Categories
hardware

Nvidia GeForce RTX 3080: Everything We Know

The GeForce RTX 3080 is approaching September 17. Powered via the Ampere structure, it is probably the most important generational soar in efficiency now we have ever observed from Nvidia. During the revealing of the GeForce RTX 3080, Jensen Huang mentioned it will be “two times as rapid” because the RTX 2080 … and on the identical $699 value. If that is even shut to true, it’s going to be one of the very best graphics playing cards round as soon as it arrives, and will have to eclipse each and every different card in our GPU hierarchy. It’s no longer on my own both, as it’s going to be joined via the GeForce RTX 3090 and GeForce RTX 3070. But at part the cost of the 3090, that is the extra good excessive efficiency GPU. Here’s the whole lot we all know concerning the GeForce RTX 3080, together with specs, value, options, unlock date, and extra.

We’ve mentioned the finer main points of Ampere somewhere else, and right here our focal point goes to be at the RTX 3080. It’s the primary GPU made the usage of Samsung’s 8nm (8N) procedure generation, while the larger Nvidia A100 makes use of TSMC’s 7N tech. How a lot of a distinction does that make, if any? It’s tough to mention, however obviously Nvidia used to be ready to construct a big an impressive new GPU.

Nvidia GeForce RTX 3080 At A Glance: 

  • 10GB GDDR6X at 19 Gbps
  • 8704 CUDA cores and 29.8 TFLOPS
  • 1.9X extra environment friendly than Turing
  • Release Date: September 17, 2020
  • Price: Starting at $699
Nvidia GeForce RTX 3080 Specifications
GPUGA103
Graphics CardGeForce RTX 3080
Process (nm)Samsung 8N
Transistors (billion)28
Die Size (mm^2)?
SMs68
CUDA Cores8704
RT Cores68
Tensor Cores272
Boost Clock (MHz)1710
VRAM Speed (Gbps)19
VRAM (GB)10
Bus Width320
ROPs80
TMUs544
GFLOPS FP3229768
RT TFLOPS58
Tensor TFLOPS (FP16)238
Bandwidth (GB/s)760
TBP (watts)320
Launch DateSeptember 17, 2020
Launch Price$699

After months of leaks and pretend specs, we in the end have the entire monty, and wow is it stunning. 8704 CUDA cores. What!? That’s an enormous leap in shader counts, and it sounds as if Nvidia has elected to double the choice of cores in line with SM. We’ll to find out extra within the coming weeks, however the result’s greater than double the computational efficiency of the former halo card, the RTX 2080 Ti.

Also spectacular is that even if there are fewer Tensor cores, each and every one is four instances as potent because the 2d gen Tensor cores in Turing, yielding a internet doubling in Tensor efficiency as smartly. There’s extra.

The 2d technology RT cores in Ampere are 1.7 instances extra robust than the first gen RT cores in Turing. Nvidia says the RTX 2080 Ti may do 34 TFLOPS value of RT calculations, and the RTX 3080 can do 58 TFLOPS value. That’s going to be vital, because it looks as if the approaching release video games for a better technology PlayStation 5 and Xbox Series X consoles are going to be pushing extra RT results than earlier video games. Think extra alongside the strains of Control than Battlefield V, in different phrases.

Even reminiscence bandwidth sees a considerable spice up, going from 616 GBps to 760 GBps. We have no idea what different tweaks Nvidia has made to the reminiscence subsystem, however we think it makes higher use of the to be had bandwidth, this means that much more than a 23% development. And that is in opposition to the 2080 Ti; as opposed to the RTX 2080, the RTX 3080 has 70% extra bandwidth.

Nvidia GeForce RTX 3080 reference card (Image credit score: Nvidia)

Meet the Nvidia GeForce RTX 3080

To the wonder of no one, given the entire leaks, Nvidia’s reference fashion RTX 3080 sports activities a radically made over cooler design. It’s nonetheless a dual-fan configuration, however the ‘rear’ fan (relative to the IO ports) blows throughout the cooling fins whilst the ‘entrance’ fan exhausts out the again of the PC. We have no idea of Nvidia nonetheless plans to name its personal card Founders Edition fashions, and we will see a host of customized designs from Nvidia’s add-in board (AIB) companions, however the reference RTX 3080 seems not anything like the former technology Nvidia playing cards.

Perhaps that is for the most productive. While I just like the glance of the present RTX 20-series Founders Edition line of playing cards, I’ve spotted that the backplates at the GPUs can get extraordinarily scorching whilst gaming—particularly at the RTX 2080 Ti and 2080 Super. The new design will have to have much less turbulence over the radiator fins, that means much less noise and optimistically decrease temperatures. That’s vital because the RTX 3080 will recreation a 320W TDP. That’s the easiest TDP to date for a single-GPU card (although the RTX 3090 bumps as much as 350W).

Given the transfer to Samsung’s 8nm lithography, the leap in energy is slightly unexpected. However, Nvidia says efficiency is incessantly capped via the volume of cooling and gear a design can deplete, and it obviously sought after to move giant with the RTX 3080. It will likely be fascinating to peer how the customized AIB designs fare when in comparison to Nvidia’s new cooler, which is one of the issues we are taking a look ahead to checking out.

(Image credit score: Nvidia)

Nvidia GeForce RTX 3080 Features

Nvidia hasn’t formally coated all of the options within the GeForce RTX 3080, however we do know numerous different further items of data. We have no idea die dimension for instance, however we do know that it comprises 28 billion transistors. Considering the GA100 has 54 billion transistors in an 826mm sq. bundle, the GA103 is most definitely with reference to part that dimension (450mm sq., give or take).

Nvidia’s information middle ambitions are on a unique trail than the patron playing cards, which is most definitely a excellent factor. Consumers do not usually want FP64 computing, and HBM2 with its silicon interposer requirement can build up prices considerably. Get rid of the ones pieces and upload in ray tracing and GDDR6X reminiscence, and you’ve got a much more sexy gaming GPU.

Let’s additionally briefly speak about the TSMC N7 vs. Samsung 8N side. TSMC’s procedure is considered being awesome to Samsung’s, however there may be additionally extra call for for N7, together with from Nvidia’s personal A100, plus AMD’s Zen 2 and Navi 1x chips, and most definitely Navi 2x as smartly. And Intel is also doing a little chips on N7 as smartly. So, if Samsung’s 8N is not too a long way off of TSMC’s N7, it could actually is sensible for Nvidia to department out to different choices.

There are some new options that we anticipated, like HDMI 2.1 enhance. That’s provide, which permits the RTX 3080 to power as much as 8K120 HDR by means of DSC. On the opposite hand, DisplayPort continues to be caught at 1.4a. That’s sufficient for 8K60 with DSC, however taking into consideration DisplayPort 2.0 has been outlined for over a 12 months now, we idea it will make an look. The internet result’s that the place the RTX 2080 Founders Edition incorporated three DP1.4 ports, one HDMI 2.0 port, and a USB-C VirtualLink port, the RTX 3080 will it sounds as if send with one HDMI 2.1 port and three DisplayPort 1.4a ports (no phrase on VirtualLink).

(Image credit score: Nvidia)

GeForce RTX 3080 Performance

Given the rise in shader cores, plus the quicker RT cores and Tensor cores, efficiency expectancies are top. At one level, Jensen mentioned the RTX 3080 could be greater than two times as rapid because the RTX 2080. That’s nearly without a doubt simplest going to be at 4K, since the RTX 3080 goes to run into CPU bottlenecks at decrease resolutions. If you are nonetheless gaming on a 1080p track, purchasing a 3080 most definitely will have to be on the most sensible of your listing.

Wait, what about 360Hz 1080p shows? Sure, the ones could possibly use the quicker GPU, with the exception of maximum video games are nonetheless going to be CPU restricted to framerates a long way under that mark. It’s most commonly going to be esports video games like CSGO and LoL that push past 360 fps. But possibly with the entire ray tracing results cranked up, 1080p local rendering will nonetheless be GPU restricted. It’s going to rely so much at the particular person sport design reasonably than any common metric.

On the opposite finish of the spectrum, there may be the 8K enhance we have been simply discussing. Native 8K rendering is not going to be common any time quickly. I’d additionally query whether or not it is actually wanted for PC gaming on a track, however let’s no longer cross there. 8K with DLSS then again is not a lot of a stretch.

We already be expecting the RTX 3080 to be moderately slightly sooner than the RTX 2080 Ti, and the RTX 2080 Ti a minimum of comes somewhat with reference to dealing with 4K at 60 fps or extra in lots of video games. Boost efficiency via 30-50% (or extra) after which run DLSS at the consequence, and all at once 8K gaming is not so far-fetched. It’s no longer 8K rendering within the true sense of the phrase, but when the end result seems higher? Sure, opt for it. Not that you are more likely to see the variation if you are sitting various ft away out of your track — or if you are getting outdated like me and not have superior eyesight.

Going again to the CPU bottleneck, that is additionally going to turn out to be a significant factor for the RTX 3080. We’ve already observed CPU bottlenecks in lots of video games with the RTX 2080 Ti, to the purpose the place you incessantly want to run at 1440p or 4K prior to you turn out to be GPU restricted. (See: Microsoft Flight Simulator and Project CARS 3.) Considering the theoretical efficiency of just about 30 TFLOPS is greater than double the RTX 2080 Ti, there is a excellent likelihood many video games will stay CPU restricted even at 1440p.

We additionally checked out Core i9-9900Ok vs. Ryzen 9 3900X gaming efficiency with a big suite of GPUs just lately, in preparation for the next-generation GPUs. Even although Intel nonetheless lacks a desktop PCIe Gen4 answer, that mustn’t subject a lot, and CPU efficiency goes to be an element. We’ll be doing our very best to check the brand new GPUs on each testbeds (and improve to Zen 3 and Comet Lake as smartly), however if you are eyeing the RTX 3080 you will have to most definitely have a minimum of a Core i7-9700Ok or higher CPU first.

Frankly, on paper the RTX 3080 seems even sooner than what Jensen claimed. 29.8 TFLOPS of FP32 shader efficiency, 238 TFLOPS of FP16 Tensor core efficiency, and 58 TFLOPS of ray tracing efficiency? Only the very last thing is not simply greater than double an RTX 2080 Ti. Perhaps reminiscence bandwidth will restrict exact efficiency, or possibly the structure cannot maintain the ones height values. More most probably, although, it’ll be CPUs preserving again the GeForce RTX 3080.

(Image credit score: COLORFUL)

Will There Be an RTX 3080 Ti or RTX 3080 Super?

So a long way, Nvidia has simplest discussed ‘vanilla’ naming at the three RTX 30-series GPUs. Rumors of RTX 3080 Ti have been commonplace a number of months in the past, and there may be nonetheless room for one thing in between the RTX 3080 and RTX 3090. Nvidia has a $7,00 card and a $1500 card; why no longer a middle-ground $1000 fashion?

In the non permanent, it sort of feels not likely Nvidia should fill that hole. On paper the 3090 is simplest 20% sooner than the 3080, so there may be no longer numerous wiggle room for an element that is 10% sooner/slower than the present playing cards. More possible is that we will see a refresh of the RTX 30-series in a couple of 12 months (or six to nine months?), the place Nvidia bumps efficiency and assists in keeping the similar value, very similar to the 20-series Super fashions.

The identical applies to the distance between the RTX 3070 and RTX 3080. There’s certainly Nvidia may attempt to fill it, however there is no actual want to accomplish that at the moment. Waiting some time for the preliminary hype to die down after which popping out with ‘new’ moderately sooner fashions is what we typically see.

But TSMC is getting with reference to manufacturing on it is N5 procedure, to not point out N7P and N6, so there would possibly no longer also be time to do a refresh prior to Nvidia strikes to a brand new node with even higher efficiency. The 16nm/12nm node caught round for moderately some time, and 28nm lasted even longer prior to it. Now there are a host of competing choices, so who is aware of what is going to occur with new GPUs subsequent 12 months.

(Image credit score: Nvidia)

GeForce RTX 3080: The Bottom Line

The final main GPU release from Nvidia used to be two years in the past. In the period in-between, there were a slew of top of the range, midrange, and funds choices, plus product refreshes. That does not cover the truth that it is been some time since there used to be a brand new king of the hill. Of route the GeForce RTX 3080 is not the brand new king; that will be the RTX 3090. It’s the penultimate RTX 30-series card in the intervening time, however it is nonetheless a large step ahead for gaming efficiency.

We’re desperate to in truth get our palms at the {hardware} (and drivers) and start checking out. We have no idea when our overview will cross reside, however both on or simply prior to September 17 is a protected guess. How a lot will your CPU subject? What answer will have to you be operating prior to the RTX 3080 is sensible? Does PCIe Gen4 subject in any respect? These are all excellent questions that we will do our very best to respond to.

The excellent information is that our previous issues about pricing have grew to become out to be a minimum of partly improper. The RTX 3080 is retaining Nvidia’s RTX 2080 Super beginning value of $699. Will you be capable to purchase a card at release for that value? Probably no longer, however give it a month or two and issues will have to fortify.

The actual query now’s simply how aggressive AMD’s Big Navi goes to be? Maybe the backlash in opposition to the Turing costs at release is accountable, however now we have additionally heard rumblings that Big Navi goes to be moderately excellent. Probably no longer 30 TFLOPS excellent, however the entire identical, Nvidia most probably would not be doing a 30 TFLOPS gaming card for $700 if Big Navi used to be simplest going to compete with the former technology RTX 20-series GPUs. Or in all probability Nvidia is all for avid gamers moving to the next-gen consoles?

Either method, we will know extra in only a couple brief weeks. Stay tuned for our complete overview, and in case you’ve been saving your pennies and preserving directly to a prior technology graphics card, it may well be time to position it to leisure and leap at the ray tracing bandwagon. (Yes, I’m taking a look at you Cyberpunk 2077!)