Nvidia GeForce RTX 3080 Review

Nvidia claims the new GeForce RTX xxx graphics series provides a behemothic leap in raw graphics functioning, based on the company's latest Ampere compages and manufactured on Samsung's 8nm process. On paper, the GeForce RTX 3080 certainly looks mighty, packing an insane 28 billion transistors -- a 50% increase over the RTX 2080 Ti, the quondam earth'south fastest gaming GPU.

Today we tin finally show you how the GeForce RTX 3080 performs and you tin of course wait a boatload of criterion data in this review. There's plenty of architectural information we could go over and we could also explore the Founders Edition card's design, only both of those things hands warrant a dedicated article. Right now it'southward articulate about of you volition want the benchmark data first, and so we'll cover the hardware nuts and so jump into the results.

With Ampere, a big change has been made to the streaming multiprocessor (SM) and its 32-scrap floating point or FP32 throughput when compared to GeForce 20 Turing GPUs. Nvidia doubled the throughput by designing a new datapath for FP32 and INT32 operations which results in all iv partitions combined executing 128 FP32 operations per clock.

Nvidia says this helps improve gaming performance and while truthful, its effectiveness depends on a few things which we will explore in this review. Here'south a side-by-side specification comparison between the RTX 3080, 2080 and 2080 Ti.

	GeForce RTX 3080	GeForce RTX 2080	GeForce RTX 2080 Ti
Price (MSRP)	$700		$1200
Release Engagement	September 2020	September 2018
Process	Samsung 8N	TSMC12FFN
Transistors (billion)	28	13.6	18.half-dozen
Die Size (mm2)	628.4	545	754
Core Config	8704 / 272 / 96	2944 / 184 / 64	4352 / 272 / 88
Core Clock Frequency	1440 / 1710 MHz	1515 / 1710 MHz	1350 / 1545 MHz
Memory Chapters	10 GB	viii GB	11 GB
Memory Speed	nineteen Gbps	14 Gbps
Retentiveness Blazon	GDDR6X	GDDR6
Motorcoach Type / Bandwidth	320-bit / 760 GB/due south	256-bit / 448 GB/s	352-bit / 616 GB/southward
TGP	320 watt	215 watt	250 watt

When compared to the 2080 Ti we're looking at the same number of SM units with the RTX 3080, but thanks to the 2x FP32 redesign, Nvidia is claiming a 100% increase in CUDA cores. That doesn't mean the 3080 will exist 100% faster than the 2080 Ti, simply it volition offer a reasonable performance uplift. Point is, you can't brand a meaningful CUDA cadre count comparison between unlike generations.

Comparison the memory configuration is a more straightforward affair. The RTX 2080 Ti has an xi GB memory buffer, while the RTX 3080 offers ten GB. This is nonetheless a 25% increment over the vanilla RTX 2080, but it's surprising since yous'd wait a larger buffer for a card targeting 4K gaming.

The memory bandwidth has been greatly increased thanks to the use of xix Gbps GDDR6X retention on a 320-bit wide passenger vehicle, resulting in a peak bandwidth of 760 GB/s, a 23% increase from the 2080 Ti and seventy% from the 2080.

It's also worth noting the increment in TGP (Full Graphics Ability) from 215 watts for the RTX 2080 to 320 watts for the RTX 3080, a massive 49% increase and even a 28% increment from the 2080 Ti. It will be interesting to run into how well the RTX 3080 Founders Edition graphics card deals with all that power in one case it's converted to heat energy.

But plenty about all that, it's fourth dimension to get into the blue bar graphs. For testing we're using our Ryzen 9 3950X test arrangement, featuring four sticks of low latency DDR4-3200 memory operating in a dual-rank configuration. We too accept some Core i9-10900K and PCI Express 3.0 vs four.0 preliminary data that we'll await at, so permit'southward get to it.

Benchmarks

Starting with Expiry Stranding, we're looking at a xx% performance increase over the 2080 Ti at 1440p and a 41% increment over the RTX 2080. Not bad at all, only non the 80% we suspect many of you lot were expecting to see. We should notation the margins were identical on our Core i9-10900K test system.

Moving to 4K does improve margins for the RTX 3080 and now information technology's 29% faster than the 2080 Ti and 62% faster than the vanilla 2080, so that's starting to get upward there. As well 107 fps on average at 4K is a very impressive result, the game looked spectacular and every bit you might imagine, was very smooth.

Unfortunately the performance gains in Microsoft Flying Simulator 2020 at 1440p are very weak and this is largely due to our choice to use the 3950X, the game just doesn't utilise Ryzen processors very well correct now and is in desperate need of depression-level API support.

Here we're looking at a measly ix% increase over the 2080 Ti and xxx% over the 2080, only even with an Intel CPU you lot're still only looking at a 20% increase over the 2080 Ti.

Jumping to 4K does aid alleviate the CPU clogging and the RTX 3080 was 29% faster than the 2080 Ti. We saw the same frame rates on our 10900K test organisation. And so with an RTX 3080 you're looking at just 40 fps at 4K over large cities in Flight Simulator 2020.

Shadow of the Tomb Raider was one of the games that impressed me the virtually ii years ago when testing the 2080 Ti, the 4K feel was amazing. At 1440p the RTX 3080 is seen to exist 28% faster than the 2080 Ti and 57% faster than the 2080, at 154 fps on average under these conditions is very impressive.

Then at 4K we see that 87 fps on boilerplate is possible with the RTX 3080, pretty impressive stuff. That's a 67% increment over the 2080 and 30% over the 2080 Ti, and then both are quite impressive gains, particularly given this is a $700 GPU.

Moving on to Tom Clancy's Rainbow Six Siege, nosotros're looking at a 28% increase over the 2080 Ti and 63% over the standard 2080 or 54% over the Super version.

We're probably not going to reference the 2080 Super much since it offered a mere five% increment over the original model and most of you bought the original.

Now at 4K nosotros're looking at a 29% increase over the 2080 Ti and 71% over the 2080. Those are some pretty solid gains. It also ways whereas the 2080 Ti averaged 135 fps the 3080 saw 1% lows of 152 fps, providing a better feel for high refresh rate gaming at 4K.

Next up nosotros have F1 2020 and here we're seeing what has get a fairly typical 30-ish % boost over the 2080 Ti at 1440p, this time information technology'southward a 29% performance uplift to be exact. That meant the RTX 3080 was 57% faster than the vanilla 2080, and then another solid functioning uplift there.

Once again nosotros come across the margins abound a little at 4K and now the RTX 3080 is 36% faster than the 2080 Ti and 72% faster than the 2080, which is far more significant and much closer to the 80% that's been put out at that place since the proclamation.

It'south also worth noting that we're talking about 129 fps on average at 4K, that's staggering performance.

Moving on to Gears 5, we see "mild" performance gains at 1440p. An 18% boost from the 2080 Ti and 37% over the standard 2080. It will depend on your perspective only for this title, we're talking about a 34% comeback over the 3+ twelvemonth-old GTX 1080 Ti which also sold for $700.

Again increasing the resolution to 4K helps widening the gap and now the RTX 3080 is 47% faster than the GTX 1080 Ti.

This fourth dimension we're looking at a 33% performance uplift for the RTX 3080 over the 2080 Ti at 1440p, so that'south 1 of the amend functioning gains we've seen at this resolution so far.

Information technology also means the new Ampere GPU was 76% faster than the GTX 1080 Ti which is a pregnant performance uplift and nosotros also saw an almost lxx% increase over the RTX 2080, and so we're very impressed with these results.

And of course, they only become better at 4K.

The RTX 3080 was a whopping 43% faster than the 2080 Ti, a massive functioning uplift. It's likewise 83% faster than the 2080 and 92% faster than the 1080 Ti. Those are some of the best margins we've seen so far and certainly announced to be a all-time example scenario for the RTX 3080.

With Assassin'due south Creed Odyssey we're dorsum to milder gains at 1440p. The RTX 3080 leads the 2080 Ti by a 25% margin and the 1080 Ti and 2080 by 42%. Nosotros know it sounds foreign to say "just 42%" but remember the 1080 Ti toll the same amount as the 3080 when released 3.5 years ago.

At 4K, the RTX 3080 provides a 35% performance bump over the 2080 Ti and it was 51% faster than the 1080 Ti and 2080.

With no help from dynamic resolution scaling, the RTX 3080 pumped out 190 fps on average at 1440p in World War Z using the maximum in-game quality settings.

That's a stunning frame rate, though merely a ~17% boost over the 2080 Ti and 42% over the 2080. Nosotros should notation that the 3950X does create a small bottleneck in WWZ at 1440p, even so the margins are about identical on the Core i9-10900K, moving by a few percent in the 3080's favor.

Moving to 4K shows the RTX 3080's prowess, it's now 34% faster at 4K than the 2080 Ti which is a massive improvement.

Next upward we have Metro Exodus and the 3950X does become a problem in this championship. This is a rare case where the Ryzen CPU heavily limits performance at 1440p and as a result the RTX 3080 suffers. That's still a 25% boost over the RTX 2080, but let'due south come across what happens at 4K...

At 4K we're no longer CPU limited and there is a 35% uplift over the 2080 Ti and lxx% over the 2080.

Moving on to Resident Evil 3, nosotros run into an impressive 196 fps average from the RTX 3080 when using the best visual quality settings. That meant information technology was 31% faster than the 2080 Ti and 66% faster than both the GTX 1080 Ti and RTX 2080.

Jumping up to the 4K resolution and we're seeing just over 100 fps on boilerplate for a 30% boost over the 2080 Ti and a massive 76% improvement over the RTX 2080 and GTX 1080 Ti

Doom Eternal has been heavily touted by Nvidia to show off RTX 3080 functioning and you'll encounter why one time we movement to the 4K testing. At 1440p the results aren't overly impressive as the 3080 is ~20% faster than the 2080 Ti and 51% faster than the 2080. Again, l% remains an impressive gain.

Now moving to 4K we see some interesting stuff using the 'Ultra Nightmare' preset, which in our test sees 9GB of VRAM used.

Here the 3080 was 35% faster than the 2080 Ti, but quite incredibly 115% faster than the 2080. Too information technology was 93% faster than the 1080 Ti which does well here relative to the 2080 thanks to the extra VRAM.

This particular examination is why many are expecting/claiming upwards to 100% functioning gains over the 2080 later seeing Digital Foundry's Nvidia promo video. The information is accurate, but without more data can exist somewhat misleading.

And here is more than information. Past leaving all other options on 'Ultra Nightmare' with the exception of texture puddle size, which we dropped to 'Ultra' reducing VRAM usage to 7GB, this is what we find...

There is no change in operation with GPUs packing 10GB or more VRAM, but for those with 8GB like the RTX 2080, nosotros're seeing a 26% performance uplift and that means the 3080 is at present 70% faster, not 115% faster.

We'll discuss this farther, only for at present let's motility on to Wolfenstein: Youngblood.

Starting with the 1440p data we see that the RTX 3080 is good for 242 fps on average as it boosted functioning over the 2080 Ti past a 22% margin and 55% over the 2080. Adequately typical gains there for 1440p.

Now at 4K nosotros're looking at 142 fps on average with the 3080 which is extremely impressive, just 22% faster than the 2080 Ti, merely lx% faster than the 2080 which is a pretty solid proceeds.

Finally we have Hitman 2 results and this is another rare instance where the Ryzen 9 3950X limits performance at 1440p and as a issue the RTX 3080 is but 5% faster than the 2080 Ti. We also tested using the Core i9-10900K and with the Intel CPU the 3080 was marginally faster (10%) at this resolution.

Pumping the resolution up to 4K means we're back to a GPU limitation and the RTX 3080 performs 29% faster than the 2080 Ti and 69% faster than the 2080. Once more we can see the Ampere GPU come to life at the 4K resolution.

Functioning Summary

In that location is no escaping fact the RTX 3080 is very fast, especially when compared to the 2080 GPU it replaces, but it will besides demand a capable system to get along all this power. We tested many dissimilar titles and the margins were a flake all over the place, then allow'due south have a look at the 14 game average for all the GPUs tested and and then we'll do a few caput to heads looking at individual titles.

14 Game Average at 1440p

On average, the RTX 3080 is 21% faster than the 2080 Ti and 49% faster than the 2080 at 1440p. It's also 58% faster than the 1080 Ti and 113% faster than the vanilla 1080. So we're looking at roughly a 50% operation heave at the $700 cost betoken afterwards 2 years, at least for 1440p gaming.

14 Game Average at 4K

Playing games at 4K takes full advantage of what the GeForce RTX 3080 is capable of. At this resolution, it's 32% faster than the 2080 Ti, 71% faster than the RTX 2080 and 77% faster than the GTX 1080 Ti, on average. Those are far more impressive gains and makes Ampere a tremendous replacement at the aforementioned price point.

GeForce RTX 3080 vs 2080 Ti: 14 Game Average at 1440p

Here's a look at the margins on a per game basis and please annotation due to the way this data is calculated there tin be a small discrepancy in the totals when compared to the previous graphs. Every bit before, the 3080 is 21% faster than the 2080 Ti at 1440p and hither we see performance gains range anywhere from a few percent correct up to 33%.

GeForce RTX 3080 vs 2080 Ti: 14 Game Average at 4K

Nosotros run into far more predictable margins at 4K and this has little to do with the CPU used and much more to do with the Ampere compages -- we'll hash out this in better detail in the conclusion.

What we come across hither are gains of at least 22%, with the biggest margin seen in Horizon Zero Dawn where the RTX 3080 was up to 43% faster than the prev-gen 2080 Ti.

GeForce RTX 3080 vs RTX 2080: xiv Game Average at 1440p

At 1440p nosotros see less predictable gains fifty-fifty when compared to the RTX 2080. It's articulate however the 3080 is the more powerful GPU with margins going from 25% faster and then equally big every bit 69%.

GeForce RTX 3080 vs RTX 2080: fourteen Game Average at 4K

At 4K, the new Ampere GPU can be anywhere from 51 to 83% faster. Looking at this data yous could simply say the RTX 3080 is about seventy% faster when gaming at 4K.

Too note that we used the 7GB VRAM data for Doom hither as the 115% gain using the Ultra Nightmare preset was an outlier and not indicative of raw GPU performance.

Power Consumption

Now let's take a look at power consumption, starting with our total system ability draw, so literally everything in the PC is existence measured here.

This is useful for a few reasons: commencement, it lets you know roughly what kind of power supply yous'd need for a certain configuration. It also lets you lot know just how much difference in ability draw in that location is between say the 2080 Ti and 3080 when factoring in the entire gaming system.

We're looking at a big 21% increase in full organisation usage as the RTX 3080 pushes our examination arrangement upward to 523 watts. That's beyond what we saw with the GTX 1080 Ti and Vega 64 to proper noun two power hungry graphics cards. So the RTX 3080 does suck down a lot of ability to output all that raw performance.

Using Nvidia'south PCAT (Power Capture Assay Tool) we tin can accurately measure the power draw of the graphics card lone, including what it draws from the PCIe power connectors and the PCIe slot. Information technology'south a seriously impressive tool and kudos go to Nvidia for providing it to testers, that'south very cool of them.

Hither we can see that the RTX 3080 consumes 25% more ability than the 2080 Ti and 29% more than the RTX 2080. And while that'southward a lot of power, technically it is all the same more efficient. Encounter below:

Here we come across that the RTX 3080 offers the most performance per watt, so you can't deny its efficiency. This is a first generation node shrink GPU, meaning that if history serves whatsoever indication, in the future Nvidia will likely exist able to improve upon this 8% generational efficiency gain.

For testing thermal performance of the RTX 3080 Founders Edition, we installed it in our Corsair Obsidian 500D examination system and allow it run for an hr in F1 2020 while maintaining a room temperature of 21C. Later on that point the GPU peaked at merely 77C and maintained a cadre clock speed of around 1830 MHz.

This had the graphics card fans spinning at up to 1900 RPM where they could be heard merely nosotros're non overly loud. We don't take whatever sound measurements right now, but nosotros found the Iron model to perform well and given the performance and ability depict we were pleasantly surprised. We'll say the menu does tend to squeal a lot when pumping out over 300 fps, so if you are sensitive to coil whine, a frame cap will be in gild.

Cost Per Frame

One key metric we like to use is cost per frame and at $700 the RTX 3080 will exist delivering the goods, fifty-fifty at 1440p.

You all know we liked the value of the Radeon 5700 XT and here we're seeing that level of value despite offering a 64% operation boost. The RTX 3080 tin roughly match the toll per frame of the Navi GPU and it's too 33% cheaper per frame when compared to the RTX 2080.

All the same, if we look at the 4K performance, the toll per frame gets even better, much meliorate.

Here the RTX 3080 is the most affordable graphics card released in the concluding 4 years when information technology comes to cost per frame (value), beating even the base model RX 5700 which is about ridiculous.

DLSS / Ray Tracing

We program to follow upward with a more detailed analysis of DLSS and ray tracing on Ampere on a dedicated article, only for the time existence, here's a quick await at both in Wolfenstein Youngblood.

When enabling Ray Tracing the RTX 3080 suffers a 38% functioning hit which is better than the 46% operation hit the 2080 Ti suffers. Then if we enable DLSS with ray tracing the 3080 drops just xx% of its original performance which is marginally meliorate than the 25% drop seen with the 2080 Ti. The deltas are not that much unlike, the RTX 3080 is only faster to brainstorm with.

Using only DLSS sees a sixteen% performance boost in the RTX 2080. So let's see if things alter much at 4K.

Here the RTX 3080 was good for 142 fps when running at the native resolution without any RTX features enabled. Enabling ray tracing reduces performance by 41% to 84 fps on average, which is reasonable performance, only still a massive fps driblet. For comparing the RTX 2080 Ti saw a 49% drop.

When using DLSS, the 2080 Ti sees an 18% functioning boost whereas the 3080 sees a 23% spring. At to the lowest degree in this game implementation, it looks like the 3080 is faster at stuff like ray tracing because information technology's a faster GPU and non necessarily because the second-gen RT cores are making a divergence. We'll test more games in the weeks to come, of course.

Testing with AMD vs. Intel CPUs

In that location are a number of reasons why nosotros switched to a Ryzen CPU for game testing moving forwards (this was discussed and outlined in this article). But the well-nigh of import one was that the change wouldn't influence results or manipulate numbers in charts like our cost per frame summary.

Looking at how the 3950X and 10900K compare with the RTX 3080 and 2080 Ti at 1440p across our fourteen game sample, the 10900K was 4% faster with the 2080 Ti and half dozen% faster with the 3080, that 2% margin is non going to skew the results.

Then at 4K we can appreciate that there'south no departure at all between the two CPUs as both pushed the RTX 3080 to the same 108 fps average.

What Nosotros Learned

This is simply the offset of the journeying for Ampere GPUs and RTX thirty series graphics cards, but there you have it, that'southward the new GeForce RTX 3080 in a nutshell. We've got many more than tests that we want to evidence yous and many more nosotros want to run, but for now you lot should have a pretty good idea of what'southward on offering.

A big lxx% operation jump over the RTX 2080 at 4K is impressive, and a huge comeback in price per frame, so that's a job well done by Nvidia.

Now, that big jump won't be experienced across the lath, or more specifically, we saw less consequent improvements at 1440p where the RTX 3080 could beat the 2080 by as much as 70%, but also by every bit little as 25% (for a ~fifty% performance gain on boilerplate). That 50% boost is nonetheless great to see fifty-fifty though in terms of raw performance at the $700 cost point we tin look two years dorsum to the RTX 2080 release or three.5 years since the GTX 1080 Ti was released.

We've looked deeper at the RTX 3080's performance at 1440p, so we desire to comment on that. As far as we can tell, CPU bottlenecking alone is not the explanation. The reason comes down to Ampere's architecture and the alter to the SM configuration that we mentioned in the introduction. The 2x FP32 design can only be fully utilized at 4K and beyond. This is because at 4K the portion of the render time per frame is heavier on FP32 shaders. At lower resolutions like 1440p, the vertice and triangle load is identical to what nosotros run into at 4K, but at the higher resolution pixel shaders and compute effect shaders are more intensive and take longer and therefore can fill the SMs FP32 ALUs better.

We frequently see high performance GPUs better utilized at higher resolutions for similar reasons, then that in itself isn't unusual. Higher resolutions always practise a better job of putting core-heavy GPUs to piece of work, and they too help minimize other organization bottlenecks. The 2080 Ti, for case, extends its lead over the 2080 at 4K, going from 23% faster at 1440p to 28% faster at 4K and that'due south with the lower quality textures in Doom, where memory capacity is non a factor. We're talking about a small 5% discrepancy, the scaling at 1440p and 4K looks very similar, and the 3950X can also be blamed for a small part of that divergence. However with the RTX 3080 we're talking about a 21% margin over the 2080 Ti at 1440p and then a much larger 32% margin at 1440p (again, VRAM plays no role in the comparison). And our explanation hither is that Ampere is a compute heavy, datacenter and AI optimized architecture that goes beyond gaming needs.

Current RTX 2080 owners targeting high-refresh rates at 1440p volition encounter a 50% performance boost, which is however pretty sweet. The fact that the RTX 3080 simply excels at 4K gaming could be seen as the cherry on summit. Doesn't necessarily future proof the GPU however, as Nvidia paired it with 10GB of VRAM which might prove insufficient in a yr or two, though PCIe iv.0 will exist a lot more useful in mitigating those operation loses in modernistic platforms.

In the current market the GeForce RTX 3080 is an excellent value production.

We only have to expect at the cost per frame data to see that the RTX 3080 is a massive improvement over the RTX 2080 in terms value, it's fifty-fifty better value than the RTX 2060 and comparable to the 5700 XT which not everyone was fond of. In the current market place the GeForce RTX 3080 is an excellent value product, one of those rare instances where a high-finish product is likewise a great value. This may be helped past an overpriced GeForce twenty series (we chosen this 1 at launch), but the fact is, those RTX boards still flew off shelves no matter what.

In the coming months we're bound to come across more of Ampere, but likewise some competition from AMD and somewhen Intel. We're looking forward to that, though conspicuously Nvidia has upped the dues on this generation. Arguably Nvidia has given AMD an opportunity to strike with a gaming focused architecture, just we'll merely have to await and meet on that i.

As for ray tracing and DLSS, our opinion on that hasn't inverse. The engineering science is groovy, and we're glad it hasn't been used as cardinal selling points of Ampere, information technology's now only a overnice bonus and of course, it will matter more once more games bring proper support for them.

Bottom line, we're very pleased with what the GeForce RTX 3080 delivers at the $700 price point, and it's a hell of a lot improve than what nosotros got last generation. I personally would have not bought an RTX 2080, simply I would purchase an RTX 3080.