It feels virtually redundant to name the brand new Nvidia RTX 3080 the quickest graphics card you’ll be able to purchase. I imply, these days, it completely is, and after the primary disclose that used to be just about inevitable. The preliminary efficiency figures from the fairway crew’s grand unveiling have been no shaggy dog story: this factor takes the RTX 2080 Ti outdoor and offers it a nice, sound kicking. It isn’t even shut.
The ol’ Turing GPU’s now whimpering within the nook, its vapour chambers leaking out over my ground as its enthusiasts fitfully spin their final.
Which is an attractive impressive consequence whilst you imagine the top-end Turing chip used to be the former holder of the ‘quickest graphics card’ name, with its daring specifications checklist and $1,200 sticky label value. The truth this $699 card could make any such efficiency step up, on the identical value because the RTX 2080 Tremendous it is changing, is fantastic.
It is without delay testomony to the brand new Nvidia Ampere structure, the paintings its engineers have installed, and perhaps to the extra aggressive panorama of 2020.
The Nvidia RTX 3080 is the flagship graphics card of this new Ampere era. A minimum of that is what Jen-Hsun and Co. is asking it, in spite of the RTX 3090, with a awesome specifications checklist, launching very, very quickly. However that is a $1,500 GPU, and then again a lot you may want the borderline offensive energy the RTX 3090 is promising, the RTX 3080 is the way more reasonable choice for the ones folks with out a Hollywood financial institution steadiness.
So, how has Nvidia controlled not to simply ship RTX 2080 Ti ranges of gaming efficiency, however fly previous it at an extraordinary fee of knots? It is a mix of sensible design alternatives, a complete bunch extra transistors, and a few severe engineering.
Nvidia RTX 3080 specifications
The important thing specifications, maximum particularly with regards to gaming, are the ones closely beefed up CUDA core numbers. We are seeing successfully two times the collection of cores in comparison to what we have been anticipating, and excess of the an identical GPU of the former Turing era. That is right down to the redesigned information paths Nvidia has fitted into the GA102 silicon on the middle of the RTX 3080.
The entire GA102 GPU, the similar very important chip which powers each the RTX 3080 and the impending RTX 3090, homes 10,752 CUDA cores. The cut-down model dropped into the RTX 3080, then again, sports activities eight,704. This is nonetheless an enormous build up over the paltry three,zero72 CUDA cores Nvidia launched the RTX 2080 Tremendous with simply final yr. Despite the fact that that kinda is dependent upon your definition of a ‘core’.
Those cores are arrayed throughout 68 streaming multiprocessors (SMs), this means that that you simply get a couple of extra of the devoted 2d gen RT Cores too. On account of the way in which the brand new structure is designed, then again, you do get fewer AI-driving Tensor Cores.
RTX 3080 specifications
GPU – GA102
Lithography – Samsung 8nm
Die dimension – 628.4mm2
Transistors – 28.three bn
CUDA cores – eight,704
SMs – 68
RT Cores – 68
Tensor Cores – 272
GPU Spice up clock – 1,710MHz
Reminiscence bus – 320-bit
Reminiscence capability – 10GB GDDR6X
Reminiscence velocity – 19Gbps
Reminiscence bandwidth – 760GB/s
TGP – 320W
The RTX 3080 GPU is constructed at the Samsung 8N design, which is nominally an 8nm manufacturing node, with a die dimension of 628.4mm2. That is smaller than the end Turing chip, which used to be 754mm2, however packs in any other 10 billion transistors at 28.three billion transistors. It’s despite the fact that a bigger slice of silicon than the TU104 which powered the an identical degree card of the final era, the RTX 2080.
A facet-by-side have a look at the Spice up clock velocity of the RTX 3080 when put next with the RTX 2080 Tremendous would possibly have you ever being concerned for the efficiency of the GPU because it rated slower, however in our checking out we have now clock speeds smartly north of one,800MHz as usual—at a an identical degree to the last-gen Turing card.
You do get quicker reminiscence too, with 10GB of GDDR6X reminiscence, working throughout an aggregated 320-bit reminiscence interface. That delivers an attractive impressive 760GB/s of bandwidth when put next with the sub-500GB/s of the usual GDDR6 of the Turing card.
However you’re indubitably going to wish some severe PSU efficiency to your new card, as a result of this Founders Version has a board energy ranking of 320W, and we have now observed this model of the RTX 3080 working smartly above that on moderate, with peaks as much as 368W in tough video games.
Nominally Nvidia is suggesting a 750W PSU at the least, however that isn’t truly taking into consideration the upper energy attracts of a few fashionable CPUs. In case you are rocking a Core i9 10900Okay, for instance, you can most probably want 850W a minimum of.
Fortunately the cutting edge cooling answer at the RTX 3080 Founders Version is indubitably as much as the duty. It is a large, dense mountain of vapor chambers and direct touch heatpipes, which is helping the cardboard care for the ones clock speeds means above the rated Spice up numbers.
The PCB itself is a customized design, one we are not seeing on third-party playing cards, a minimum of no longer but anyhow. It is a strangely compact board, with all of the important componentry squeezed right into a small area to permit for a cutout to allow the second one relax spinner to do its task.
The throughflow design of this secondary fan pulls air around the cooling block, giving further chip-chilling features along the direct fan sat atop the GPU itself. The speculation is that one vents out the again of the graphics card and the opposite out the again of the PC.
My simplest worry with this design is that chances are you’ll finally end up getting toasted air circulating round your CPU’s air-cooler in a standard mounting, which might impact processor temperatures. Nvidia maintains that it might haven’t any have an effect on in any respect, and if you are rocking an all-in-one liquid cooler it is indubitably no longer going to subject.
Nvidia RTX 3080 structure
The brand new SM design, and subsequently the core-count, is likely one of the key differentiators between the Turing and Ampere GPU architectures. When Nvidia talks about its GPU cores, it is referencing 32-bit floating level (FP32) devices. Maximum graphics workloads are according to floating level operations (on the release of Turing Nvidia claimed that for each 100 FP directions there have been 36 integer pipe directions), and so doubling the collection of to be had devices hugely hurries up maximum gaming duties.
In Turing Nvidia made FP32 and INT workloads concurrent, which means portions of the GPU were not twiddling their silicon thumbs looking ahead to one or the opposite form of process to be finished ahead of getting on with its personal factor. “Earlier than Turing [the GPU] used to be simplest in a position to truly factor one math instruction in step with clock,” explains Jonah Alben, senior VP of GPU engineering at Nvidia, “and we discovered that used to be keeping again efficiency in a large number of instances.”
Like Turing, there are nonetheless two information paths within the Ampere design, however whilst in the past they have been cut up down the road between integer and floating level operations—concurrent despite the fact that they have been—Ampere now permits FP32 workloads on both information trail.
“Now it is an FP32-plus-integer unit that has enabled us to double up our efficiency on the ones floating level heavy workloads,” Alben tells us. “So now we’ve got wonderful efficiency for that roughly workload, like denoising and different symbol processing workloads you can see in a sport engine.”
Nvidia’s streaming multiprocessors (SMs) are cut up into 4 walls, each and every with a devoted FP32 information trail, however in Ampere the second one trail gives each integer and FP32 operations. That implies there may be doubtlessly two times the collection of FP32 devices to be had, despite the fact that given there’ll nonetheless be instances the place that moment information trail needs to be working integer operations you are no longer at all times getting two times the floating level efficiency.
Nvidia has additionally beefed up the cache across the enhanced information paths as a result of, as Alben explains “feeding that, in fact is at all times a problem. The quicker those processors cross, the extra it in fact issues that they’ve a large number of bandwidth to be fed. And this used to be extra of a vital architectural alternate the place we principally replumbed the pipeline to our L1 cache method in order that now has double the bandwidth.”
With fashionable sport engines, then again, particularly transferring into this next-generation of gaming, it isn’t with regards to floating level or integer operations. With real-time ray tracing set to turn into ever extra prevalent in video games, due to the brand new Sony and Microsoft mini PCs launching this yr, Nvidia has taken the chance to replace the RT Cores first offered within the Turing RTX GPUs.
At its most elementary, the RT Core is a particular fastened serve as unit devoted to accelerating the bounding quantity hierarchy (BVH) set of rules on the middle of ray tracing workloads. It is a workload not like anything you do in a shader core, so it is very inefficient to take a look at and run them on a typical GPU. Simply have a look at how smartly the GTX 1080 offers with ray tracing for a touch…
There are two portions to this devoted ray tracing core, person who varieties the bounding bins, and any other which then strains triangle intersection charges for the other rays. However the first-gen RT Core is certainly not the completed product.
“We discovered that we had truly nice bounding field intersection charges,” says Alben, “however the triangle intersection charges on the finish of the tracing is usually a bottleneck.”
So now the RT is in a position to do extra in parallel; so for some rays it is going to be tracing the bounding bins whilst it is going to be calculating the triangle intersections for others.
“So we’ve got separate assets for those two devices, they may be able to run in parallel, and the triangle aspect has double the efficiency of the final time.”
This added parallelism, in addition to a brand new method to hint triangles via time and area to hurry up the calculation of movement blur results in ray traced scenes, manner there may be authentic acceleration of ray tracing efficiency at the complete.
The 0.33 piece of the Ampere GPU puzzle is the third gen Tensor Core. That is purely a slice of silicon devoted to deep studying and AI, however ahead of you scan right down to the rating and snappy summation on the finish, it truly is related to gaming. It is the a part of the GPU which is accountable for DLSS, or deep studying tremendous sampling, or ‘that little bit of magic that makes ray traced video games in fact playable’.
In a little bit of a switcheroo there are in fact fewer Tensor Cores in step with SM than with Turing. “We discovered that it used to be in fact higher to have a smaller collection of cores in step with SM,” Alben explains, “after which make each and every core a lot more tough than the former era.”
The fundamental efficiency of the Tensor Core has been doubled in Turing, however Nvidia may be the use of ‘Sparse Deep Studying’ to make the entire procedure extra environment friendly. It is principally some way of pruning away a host of maths that is not truly important for the general result, and that critically hurries up the efficiency of DLSS 2.zero. And DLSS 2.1 when it comes too.
Nvidia may be introducing DLSS 8K, which leverage a brand new 9x tremendous decision scaling issue with the purpose of bringing 60fps gaming inside of succeed in at 8K resolutions, even on ray tracing enabled video games. Rattling. However that isn’t precisely inside of succeed in of the RTX 3080 itself and is extra set out for the impending RTX 3090, the cardboard Nvidia is asking the “global’s first 8K HDR gaming GPU.”
However then there is the brand new reminiscence sub-system which is related to the RTX 3080. Running with Micron, and the DRAM business at massive, Nvidia has helped create a brand new upper velocity model of GDDR6, catchily titled, GDDR6X. The use of some snazzy new signalling inside of this new reminiscence sort GDDR6X is in a position to shift extra information round and do it at quicker charges. Nowadays’s data-hungry sport engines, particularly the ones keyed into ray tracing workloads, should thrive at the more moderen reminiscence tech.
It is sophisticated stuff, however all you truly want to know is that it is speedy, like 19Gbps speedy at the RTX 3080. That every one manner there is a massive quantity of reminiscence bandwidth to be had to the Ampere card, with 760GB/s at the RTX 3080 vs. 496GB/s at the RTX 2080 Tremendous.
What else is reasonably advanced is the RTX IO characteristic wrapped up within the new RTX 30-series design. Those new Nvidia Ampere GPUs toughen Microsoft’s DirectStorage API, one of the thrilling options coming to next-gen consoles. And which means that Nvidia is sort of taking the CPU out of the equation with regards to decompressing the large high-fidelity, high-resolution scenes that may fill out next-gen video games.
However we are not anticipating RTX IO enabled video games to reach till smartly into subsequent yr, so presently it is a little bit of a dormant characteristic of the RTX 3080. Although, as the contest is need to say, that should imply the Ampere structure ages like superb wine as the following era of gaming rolls alongside.
What could make a distinction these days, then again, is Nvidia Reflex. This can be a characteristic being baked into one of the most peak esports video games of these days, comparable to Fortnite, Apex Legends, and Valorant, and is aimed toward offering an SDK which permits devs to drop settings into their video games to cut back method latency whilst you hit GPU-intensive scenes.
Similarly to how G-Sync synchronises the GPU with the track, Reflex does the similar between CPU and GPU with the outcome being that there are by no means frames ready within the render queue. It additionally permits video games to pattern the person mouse click on on the very final moment to get the enter lag down up to conceivable.
Reflex must be enabled at a sport degree then again, nevertheless it additionally companions up with GSync. There is a new module going into the most recent 360Hz gaming shows from Asus, and its ilk, which incorporates the Reflex Latency Analyser, and that permits you to measure mouse, show, and entire method latency with out the desire for specialist apparatus. Neatly, with the exception of a dear new 360Hz panel anyhow.
That would possibly not reduce latency itself, however would possibly assist you to determine what the bottleneck is to your method. Or a minimum of come to the belief that bottleneck is, smartly, you.
Benchmarks and function
Nvidia RTX 3080 benchmarks and function
You will have to know the way that is going to head by means of now. The RTX 3080 is an absolute monster of a gaming GPU, and is in a position to ship body charges which make the $1,200 RTX 2080 Ti seem like a totally mid-range graphics card.
On the lowest 1080p decision ranges you’re simplest getting a modest efficiency bump over the Ti, however that is when you are speaking about body charges coming into the 200+ degree. At that time it is much less about graphics energy and extra about the remainder of the method. Losing in a bit extra FP32 cores ain’t going to do so much there.
However whilst you shift gears into 1440p and 4K resolutions, then again, particularly within the extra tough sport engines, that is when the RTX 3080’s efficiency chops truly come to the fore. When the going will get difficult I have been seeing body charges jumps of between 25 % and 43 % over the RTX 2080 Ti.
That is an enormous efficiency uplift over absolutely the top-end of no longer simply the former era of Nvidia graphics playing cards, however what used to be absolutely the quickest client GPU ever made. And whilst you imagine the RTX 3080 is not even essentially the most performant of the brand new Ampere era you’ll be able to see what the architectural adjustments have achieved for the gaming prowess of the brand new GeForce playing cards.
In step with Nvidia we mustn’t truly be speaking about its comparative efficiency with the Ti as a result of that isn’t truly the cardboard the RTX 3080 is changing. That is the RTX 2080 Tremendous, and right here it is even worse information for individuals who dropped the most efficient a part of a grand on a brand new GPU during the last 12 months. Coming in on the identical $699 sticky label value because the RTX 2080 Tremendous used to be introduced at, the RTX 3080 destroys it.
However the place Nvidia has been speaking about efficiency as much as two times that of the unique RTX 2080 this is simplest in a couple of very particular titles. In my checking out simplest Shadow of the Tomb Raider items any such state of affairs, however nonetheless doubling the 4K body charges of the Tremendous is massively spectacular.
That mentioned, the uplift remains to be vital over the former era around the board: my 1440p positive factors moderate out at 50 % quicker body charges and the 4K efficiency spice up is even upper at 65 % on moderate.
The sped up second-gen RT Cores are very a lot in play whilst you take a look at the relative ray traced efficiency of the Turing and Ampere playing cards. All you truly want to know is that you’ll be able to now get upper ray traced body charges than the former era may just be offering with out the characteristic became on. That highlights simply the place this new structure has moved the sport on in that regard.
Principally, it’s secure to head RTX on in-game now.
Ray tracing benchmarks
Ray tracing benchmarks
Ray tracing benchmarks
Ray tracing benchmarks
Ray tracing benchmarks
Ray tracing benchmarks
Ray tracing benchmarks
Nvidia has additionally been prepared to speak so much in regards to the efficiency in step with Watt positive factors the Samsung 8N procedure, and Ampere structure efficiencies, have enabled it to drop into the RTX 30-series. And truthfully I feel a large number of that pre-launch noise is as a result of those are nonetheless extremely thirsty graphics playing cards. The height Wattage I have were given out of this Founders Version card, with its cutting edge PCB and 12-pin PSU connector is 368W, smartly over the rated 320W TGP Nvidia has been touting.
On moderate, then again, I am seeing board energy of 324W, which sits extra in step with the specs, however remains to be a hell of a large number of PSU juice.
But it’s nonetheless a extra environment friendly GPU than the cardboard it is changing, and the end RTX 2080 Ti too. Taking into account Turing used to be already an attractive environment friendly structure in its personal proper that is an excellent feat.
In the true global, with the Ampere GPU left to its personal units, chowing down on as a lot energy as your PC may give it, you are simplest shopping at round 19 % upper efficiency in step with Watt when put next with the RTX 2080 Tremendous. That is at 4K, and the delta virtually vanishes decrease down the decision scale. At 1080p it is simply 2 %.
Thermals and gear
Thermals and gear
Thermals and gear
CPU – Intel Core i7 10700Okay
Motherboard – MSI MPG Z490 Gaming Carbon WiFi
RAM – Corsair Vengeance RGB Professional @ three,200MHz
CPU cooler – Corsair H100i RGB Professional XT
PSU – NZXT 850W
Chassis – DimasTech Mini V2
Nvidia has been boasting of a 1.9x perf/W spice up over Turing, then again, however you’ll be able to simplest see that during scenarios the place the former era is truly pushing its toughest. The fairway crew has proven when the RTX 2080 is maxing out at 55fps with all of the bells and RTX whistles enabled in Regulate, the RTX 3080 is sort of two times as environment friendly. However that is an outlier.
We now have capped each the RTX 2080 Tremendous and the RTX 3080 at 60fps at 1440p in quite a lot of titles after which the perf/W delta is once more so much smaller. This is till you get to ray traced benchmarks—at that time the redesigned RT Cores and further FP32 devices come into play and you notice a much more vital spice up in potency. In Shadow of the Tomb Raider and Metro Exodus, with ray tracing set to Extremely, you get 38 % and 46 % upper perf/W respectively.
And the way does that new chip-chiller carry out? Remarkably smartly. Have in mind the ones days of your reference card cooler maxing out at 90°C and that being superb? Neatly, not more. The RTX 3080 peaks at round 75°C in my checking out, and averages out at 68°C, in spite of that vast quantity of energy flooding throughout the method.
It isn’t loud both, even whilst you truly push the cardboard. Plus it has got the 0dB fan characteristic, this means that they do not even get started spinning till you peak the mid sixties in relation to temps.
We will have to look how smartly it offers with overclocking, then again, as a result of I am nonetheless ready on unlock drivers and the brand new OC gear to truly become familiar with that. At the moment it is simply no longer solid working a lot above its usual clock velocity.
Nvidia RTX 3080 verdict
Select your expletive, since the RTX 3080 Founders Version is ****ing nice. A ways from simply being the reference model of the ‘flagship’ Ampere graphics card, as a result of that cooler and redesigned PCB, it’s the final expression of the RTX 3080. And when you’re with any other model of this fine-ass GPU then I am afraid you are more likely to really feel a little bit exhausting achieved by means of.
And likelihood is that you will have to make do. The Founders Version playing cards don’t seem to be going to be produced in the similar volumes because the third-party variations from the likes of Asus, Palit, and Colourful, to call however the few we have now observed. This can be a complete new era of card, it is critically nice, and that suggests inventory will most probably vanish of the RTX 3080 FE virtually once it is introduced, and would possibly take some time to go back and degree out.
However I would most probably counsel you wait, as a result of I feel it is price it for any such handsome, impressively performant model of the Ampere GPU. It could no longer essentially be the good, shopping on the huge heat-sinks on one of the most AIB choices, however it is most probably the only I would be happiest having in my rig.
The extraordinary energy calls for are fascinating then again, and most probably a stick that the contest will need to use to overcome Ampere with. Although I’d guess any considerations are not going to sway the typical gamer clear of selecting Ampere for his or her tough new GPU. If you wish to be specifically inexperienced in regards to the energy you utilize gaming then you are no longer working a desktop PC with a high-end discrete graphics card.
The efficiency uplift you recover from the former era is very large, and virtually extraordinary. Perhaps we have now turn into used to extra iterative generational deltas, particularly taking into consideration the slight distinction between the GTX 1080 Ti and RTX 2080, for instance. And we are going to proceed beating the Ti drum, as it bears repeating—the reality the $699 RTX 3080 can completely destroy the $1,200 RTX 2080 Ti is reasonably staggering.
In fact that perhaps says so much in regards to the marketplace presently when put next with two years in the past. Again then there used to be not anything the contest may just produce to head up in opposition to even the GTX 1080 Ti, however with the AMD RX 6000 at the horizon on the finish of October, and prompt to be a good bit quicker than the ol’ RTX 2080 Ti, Nvidia could not have long gone for a minor efficiency replace this time round.
There may be doubtlessly extra pageant within the high-end GPU area once more, and with the RTX 30-series lineup introduced thus far it looks as if the fairway crew goes for some surprise and awe ways forward of the impending Radeon release. So, thank you AMD for pushing Nvidia, as a result of we are getting some nice playing cards being launched to loom over Large Navi.
Jen-Hsun and Co. has long gone large with the RTX 3080, and the outcome is a phenomenal gaming card that units a brand new benchmark for each high-end 4K gaming efficiency and for ray tracing. And truthfully that has us salivating over the possibility of each the $499 RTX 3070 and the ludicrous doable of the $1,500 RTX 3090.