The Nvidia GeForce RTX 3000 series cards are coming up very soon. In fact, the RTX 3080 and RTX 3090 will be released later this month while the RTX 3070 will be released in October.
Seeing how these upcoming cards are a major upgrade for the RTX 2000 series, you should probably wait a little if you are looking to boost your PC’s performance.
AMD should be commended for its success in having its graphics chips included in next-generation gaming consoles, which will feature 4K gaming. However, it seems that Nvidia will offer PC gamers superior performance with the RTX 3090 being a viable option for 8K gaming.
If you wish to purchase one of the cards, you can check out our Pre-Order section at the bottom of the page. There are reports suggesting that the RTX 3000 GPUs will be in short supply until 2021 so you need to act fast if you wish to get one early.
- September 19, 2020: Updated outdated info and added extra specifications for the RTX 3080, RTX 3070 variant, and RTX 3060.
- September 18, 2020: Confirmed RTX 3070’s release date.
- September 16, 2020: Added benchmarks from reliable reviewers.
- September 14, 2020: Added the date for AMD’s Future of Radeon PC Gaming event (October 28, 2020).
- September 8, 2020: Added RTX 3070 AIB cards.
- September 4, 2020: Added RTX 3080 AIB cards.
- September 3, 2020: Added pre-order section with links to pages where you can pre-order the cards. Also added twelve RTX 3090 AIB cards, specifications will follow soon.
- September 2, 2020: Removed outdated info and updated the article with fresh info revealed at the September 1 Nvidia GeForce Launch Event. Removed a lot of rumors and speculations and mentions of DLSS 3.0.
- August 31, 2020: Clarified the naming of the flagship card. Updated the featured image with the ZOTAC GeForce RTX 3090.
- August 25, 2020: Added the water cooling news and the official name for the 12-pin power connector.
- August 23, 2020: Added the full price leaks, physical size rumors, and new power connector clarification.
- August 20, 2020: Clarified the price rumors.
- August 18, 2020: Added the rumored price.
- August 15, 2020: Confirmed GDDR6X and covered the PCB leak.
- August 11, 2020: Covered Nvidia’s “21 Days, 21 Years” announcement.
- August 9, 2020: Added speculation about professional cards and rumors about Nvidia sending a “better” review copy.
- August 8, 2020: Cleared up the RTX 2000 series rumors and added a reveal date.
- August 5, 2020: Added the dates for specific versions’ launch.
- August 1, 2020: Clarified the rumors regarding the performance increase over RTX 2080 Ti.
- July 30, 2020: Added the rumors about the RTX 3080 Ti (now RTX 3090) being 40% faster than the RTX 2080 Ti and news that Nvidia will start sampling their Ampere lineup in August.
- July 28, 2020: Added a link to our GPU hierarchy, which might be helpful to get an overview of the current GPU offerings from AMD and Nvidia.
- July 26, 2020: According to multiple sources, the RTX 3080 will be 20% faster than the RTX 2080 Ti.
Much like how they did with Volta, Nvidia announced Ampere for Data Centers without mentioning consumer GPUs at all. Although the RTX 2000 series wasn’t built on Volta, it still played a role in developing the Turing architecture, which was the first product to use Tensor Cores, which are instrumental in the RTX 2000 series.
For this reason, many believed that the RTX 3000 series wasn’t going to based on Ampere, but that was not the case.
Nvidia CEO Jensen Huang has previously stated that their next generation of GPUs will all use the 7nm manufacturing process, that’s not the case as Nvidia will be using Samsung’s 8nm process. This is the same process used for Exynos 9820 which was used in Samsung’s Galaxy S10 phones.
As this is an adapted 10nm process, it doesn’t increase transistor density by too much. Although this is still an improvement over the RTX 2000 series’ 12nm process, it’s still behind AMD’s 7nm. Regardless, even with a 12nm node, Nvidia managed to handily outperform whatever AMD and their 7nm process had to offer.
Table of ContentsShow
Nvidia seemed like they were determined to beat AMD’s RDNA 2 launch and officially announce their new flagship and enthusiast cards at the September 1st event.
The RTX 3080 and RTX 3090 have been in the DVT (Design Validation Test) for quite a while now so it’s fair to assume that the production started in August 2020.
Nvidia’s Jensen Huang confirmed in a virtual event that their flagship RTX 3080 will launch September 17, with their enthusiast-class RTX 3090 coming September 24.
Their mid-range representative, the RTX 3070, will be released on October 15. This information is pretty much in line with what was predicted earlier, although it was assumed that RTX 3080 and RTX 3090 will come on the same day.
Nvidia Ampere was originally intended to be presented to the world at Nvidia’s GTC conference in March this year, but the event was canceled. It was then tentatively moved to Computex which got delayed several times until it was fully canceled and that forced Nvidia to create its own virtual event.
After GTC was called off, it was reported that Ampere would be fully announced in August. There were rumors that Quadro RTX cards would be released first, with the RTX 3000 GPUs left for later on, but that was obviously incorrect.
Some leaks suggest that Nvidia will release yet another GTX 1650 card, which might have had interesting implications regarding the RTX 3000 series’ budget options, but it’s now been all but confirmed that there will be an RTX 3060 card, with an additional ‘Super’ or ‘Ti’ (or both) variants.
The latest rumors indicate that RTX 3060 will come in November 2020, and there are whispers too loud to ignore that say that RTX 3060 Ti will come in October 2020, so before the base 3060 version.
There have even been some shaky rumors regarding an RTX 3050 card, but it is believed that it won’t come until 2021.
As far as additional versions, like ‘Super’ or ‘Ti’ are concerned, it appears that Nvidia will release those variants and some rumors suggest that they might be really big upgrades over their base versions. Some suggest that there might be as much as two times the VRAM in these souped-up variants. That’s certainly an interesting thing to keep in mind when considering upgrading your GPU.
It is interesting that Nvidia followed the established release pattern that goes all the way back to 2014 with the GTX 980. Roughly two years later, in August of 2016, the world was introduced to the GTX 1080, and two years after that, we got the RTX 2080. They have kept up with the rough two-year space between the launches with RTX 3000 series coming in September 2020.
Apparently, the rumors were true and Nvidia started sampling their Ampere lineup in late August which helped a lot in nailing that September launch.
There were heavy speculations regarding the state of the world in 2020 and the way it would affect the tech industry, but it looks like Nvidia was able to muscle through largely unscathed and maintain business as usual.
However, you should still keep an eye out for AMD’s upcoming Future of Radeon PC Gaming event on October 28 to see what they have to offer.
The Nvidia GeForce September 1 Event
On September 1st, Jensen Huang hosted a virtual event where he presented us with new details for the RTX 3000 series, such as release date, pricing, specifications, and a host of other features.
Watch the full presentation below and you’ll get familiar with most of the things you need to know ahead of the launch.
The RTX 2000 Series Cards Are Now Deprecated
It goes without saying that Nvidia has stopped the production of the RTX 2000 cards and has told the vendors to empty out the shelves containing them to make room for the RTX 3000 series.
The possible consequence of this is that the retailers won’t raise the prices of the RTX 3000 cards in order to sway the consumers to buy more RTX 2000 cards so that their inventory can be cleared.
If you bought an RTX 2000 series card such as the enthusiast RTX 2080 Ti card and still have the option to return it, you should do it now and wait for the RTX 3090 or RTX 3080 for vastly superior performance.
Following Nvidia’s price bump trend from the GTX 1000 series to RTX 2000 series, it was widely assumed that the RTX 3000 series will be even more expensive.
The expected enthusiast-class was thought to be above the $2000 price point and likely closer to $3000. The presumed flagship RTX 3090 will likely take RTX 2080 Ti’s position at the consumer GPU throne with a likely price point in the range of $1000-$1500. Although this guess was correct, it is important to say that RTX 3090 will in fact be the enthusiast card, while 3080 will be the flagship.
With the news of RTX 3000 series easily outperforming even the best of the best from the 2000 series, and seeing how there was no mention of this generations Titan card, many are assigning the RTX 3090 as its replacement. Given the price of the previous generations’ Titan cards, if RTX 3090 is indeed taking its place on the market position, this a tremendous price cut.
However, because of that particular reason, there’s a strong belief that a Titan card still is coming, but at a later date.
On the September 1st event, Nvidia’s CEO Jensen Huang confirmed that Nvidia’s Ampere top model named RTX 3090 will cost $1499. This is some $300 more than the last generation’s best representative RTX 2080 Ti Founders Edition, but it’s also important to say that these cards are not equally positioned on the market with RTX 3090 being more comparable to Titan RTX.
There have also been rumors about an Ampere GPU coming in at a whopping $2000, but that is widely speculated to be a special edition type card or an Ampere Titan variable. There is no confirmation from Nvidia that this card even exists or that it was planned, but some reliable leakers have stuck to their guns regarding this one.
The apparent flagship of the Ampere series will be RTX 3080 and will cost $699 which is on the same level as Turing’s RTX 2080 on its launch. Another thing that is extremely interesting is the ‘flagship’ label. From the announcement of the previous generation, RTX 2080 Ti had that position, but at that ludicrous $1200 price point.
It appears that AMD’s last-generation undercutting of the prices in the mid-range didn’t really affect Nvidia as they announced that the RTX 3070 will cost $499, the same as their last-generation best mid-range option, the RTX 2070.
At the bottom of the price range, there’s just not a lot of wiggle room. Both companies will release their budget cards one way or the other, and the only question that remains is which will be best. So far, they’ve been relatively level in that area, but many are hoping that the next generation of GPUs brings another level in the budget category and that we get even more affordable graphics cards.
An important caveat – most of these predictions are largely dependent on AMD’s Big Navi. It may not seem that way since Nvidia has been the undisputed king of GPUs so far, but it can be said with some certainty that AMD’s price undercutting move destabilized its modus operandi.
Seeing how these prices were a little off from what was rumored and speculated, we’re not exactly sure how to handle the rumors of RTX 3060’s price. Both 3080 and 3070 proved to be $100 less than what was originally believed, but 3090 is actually $100 more than the original speculation.
With a $200 gap in prices between RTX 3080 and RTX 3070, both in pre-launch speculations and after the announcement, it’s fair to assume that RTX 3060 will cost $299. Although this may seem like a wild guess, it feels like it would be weird seeing only a $100 difference between the mid-range and budget options.
There are some speculations within the GPU community that these are pre-Big Navi prices and that come RDNA 2’s release, and especially come holiday season, we might see these prices dropped, first temporarily, but then (depending on how good Big Navi is) permanently.
If Big Navi is a hit that can challenge the upcoming RTX 3080, we might see Nvidia reduce its prices to stay competitive, and if Big Navi is just not in the same league, then Nvidia will likely stay the course.
Earlier this year, there was a massive leak that supposedly revealed ‘an unknown’ Nvidia GPU that was yet to be released. The most interesting point to take from this leak is the 33 teraflops of theoretical floating-point performance, which is utterly insane and completely blows the RTX 2080 Ti out of the water, which is at 13.45 TFLOPs. Interestingly enough, this is also far superior to what the Xbox Series X is rumored to be offering at 12 TFLOPs.
There was also a second GPU whose benchmark was leaked, but which appeared to be a slightly less powerful version. These two GPUs showed 118 and 108 streaming processors, respectively. Given Nvidia’s track record, this would equate to 7,552 CUDA Cores in the former and 6,912 CUDA Cores in the latter. This handily surpasses the RTX 2080 Ti’s 4,352 Cuda Cores and almost sounds too good to be true.
The bigger chip had 24GB of memory, ran at 1.11GHz, and earned an OpenCL score of 184,096 in Geekbench. The smaller one was shown to have mind-boggling 48GB of HBM2E VRAM although with a lower 1.01GHz clock and it scored 141,654 in Geekbench.
Both of these chips easily outperformed the current best offering from Nvidia (the Titan RTX and Quadro RTX 8000) which score around 130,000 in Geekbench.
The latest rumors indicate that these two are professional-grade Quatro cards, but the lesser one could also be an RTX 3000 Titan card. Given some rumors that AMD is preparing a ‘Biggest Navi’ card for the Q1 2021, it is certainly a strong possibility that Nvidia is working on its counter card.
There are several supposed specs that have been leaked to the public and these strongly suggest something similar to what Nvidia has done in the past with their RTX 2000 series as far as different classes of GPUs are concerned. It’s been heavily speculated that the three GA102 cards will be the 2nd Generation Titan RTX, the RTX 3090, and the RTX 3080.
Although there has been no mention of this generation’s Titan card, the RTX 3080 and RTX 3090 have officially been announced.
Other’s are taking a different approach and claiming that Nvidia might step away from the Ti/Super suffixes and that this rumored RTX 3090 is simply what we believed to be the RTX 3080 Ti.
After the official launch event, it’s being speculated that RTX 3090 is, in fact, this generation’s Titan variant. This likely shut down the door of Ti/Super suffixes losing their place, but nothing has been announced yet so we have to wait and see.
Some other information that came from the launch is that the RTX 3000 series will be equipped with GDDR6X memory, although RTX 3070 will stick with GDDR6.
Micron (memory manufacturer) came out in August with a document suggesting that RTX 3090 will be equipped with GDDR6X memory with 21 Gbps. It turned out to be 24GB of GDDR6X at 19.5 Gbps, which is still insane.
When GDDR5X first dropped, it was a significant improvement over GDDR5, but not quite a generational jump. If GDDR6X follows in the same steps, this could be important news.
Another thing that is overlooked a bit is that 19.5 Gbps memory speed which is more than 30% improvement over Nvidia’s previous generation’s best representative, the RTX 2080 Ti.
According to Nvidia itself, the RTX 3080 will be twice as fast as the RTX 2080, while the RTX 3070 will be able to outperform Turing’s best showing, which is the RTX 2080 Ti.
The actual specification for these cards is what blew us out of the water.
The RTX 3080 will feature 30 shader TFLOPs which is almost three times as much as the RTX 2080 Ti’s 11. Another big jump is in the RT TFLOPs area where 3080 will have 58 and outshine 2080 Ti’s 34.
However, the biggest leap appears to be in the AI tensor core department that will see the RTX 3080 with 238 TFLOPs which significantly outperforms the RTX 2080 Ti’s 89.
You can see the updated details of each GPU in the table below. An important note here is that this info was gathered during the official launch event, but also from some rumors and speculations.
|Model Name||Release Date||Specifications||Price|
|RTX 3090||September 24 2020||256 RT cores, 1024 Tensor cores, 1400MHz/1700 MHz base/boost clock, 21 TLFOPs, 10496 Cuda cores, 24GB GDDR6X, 19.5Gbps, 384bit||$1499|
|RTX 3080||September 17 2020||144 RT cores, 576 Tensor cores, 1440MHz/1710 MHz base/boost clock, 8704 Cuda cores, 10GB GDDR6X, 19Gbps, 320bit||$699|
|RTX 3070||October 2020||112 RT cores, 448 Tensor cores, 1500MHz/1730 MHz base/boost clock, 16Gbps, 5888 Cuda cores, 8 GB GDDR6, 256-bit||$499|
|RTX 3060||November 2020||80 RT cores, 320 Tensor cores, 2GHz boost clock, 16Gbps, 8 GB GDDR6||$299|
Disclaimer: there are rumors circulating that Nvidia may ship review samples of RTX 3080 with 12GB of VRAM and 384-bit bus width, but then limit it with a software solution to 10GB and 320-bit bus. However, these aren’t exactly substantiated, so it’s best to take them with a grain of salt.
Lenovo also leaked the existence of an unannounced RTX 3070 Ti which features 16GB of GDDR6 memory. It may also be referring to the unannounced RTX 3070 SUPER, but that remains unclear.
Another gaffe cam from Gigabyte who unwittingly informed us of a previously unannounced RTX 3080 variant with 20 GB. The previously mentioned RTX 3070 variant with 16 GB of GDDR6 VRAM is also a part of this leak, as well as the RTX 3060 with 8 GB, presumably GDDR6.
With the DisplayPort 2.0 standard getting its official and long-awaited release in 2019 and predictions that it will be available in late 2020, there are rumors that the RTX 3000 series will be offering DP 2.0 support.
As expected, they will support HDMI 2.1, as it was already available on Nvidia’s RTX 2000 series.
The NVIDIA Ampere Architecture supports the highest HDMI 2.1 link rate of 12Gbs/lane across all 4 lanes and supports Display Stream Compression (DSC) to be able to power up to 8K, 60Hz in HDR.Qi Lin – Nvidia Principal Product Manager
The latest information surrounding the power consumption is the 12-pin power connector required for the RTX 3080. Although it was originally speculated that two 6-pin power connectors would work just fine, it has now been confirmed that RTX 3000 series will require the brand new 12-pin power connector.
It has been confirmed that the new 12-pin connector will only be present in the RTX 3080 FE and the RTX 3090 FE and that it will be called “Nvidia 12-pin PCIe Molex Micro-Fit 3.0 Connector”. An adapter will be included so it’s still possible to use the cards without replacing your PSU. AIB cards will use traditional 8-pin connectors.
One thing that is sure is that these GPUs will require a PSU with at least 650W of power which means that some people will have the additional hassle of upgrading their PSU.
Nvidia has confirmed on its website that RTX 3090 and RTX 3080 will consume 350 and 320 Watts respectively and that both will need a total PC power of 750W while the RTX 3070 will need 220W and a total power supply of 650W.
With SLI slowly going its peaceful way and NVLink still failing to capture mainstream attention, we began to wonder if we’re ever going to see a fully realized multi-card setup. And it looks like we won’t get a proper answer to that with RTX 3000 series neither.
There has been an announcement from Nvidia in which they talk about the SLI support, but with little information given, it might sound confusing to some.
Basically, with DirectX 12 and Vulkan APIs, game developers no longer need to rely on a specific SLI driver profile and as such, Nvidia has decided to cease adding new SLI profiles for RTX 2000 series and earlier graphics cards and instead support game developers to implement SLI natively inside the games.
Although Nvidia considers this to be a step in right direction, so far the actual SLI support has been scarce.
Nvidia RTX 3000 series will have SLI support, but only in a way we just mentioned. Either way, we don’t think this should be an issue because GPUs nowadays have enough power to run latest games without the need for another card. And even if the card you have is on the lower end, most of the time the financial factor is minimal as getting a better GPU is better in the long run than getting the same cheaper card for the SLI setup.
The design of the new RTX 3000 series has caused some interesting discussions on the internet from the moment it leaked. Now that the designs have been revealed, the leaks proved to be true.
All three cards feature dual fans but the fans on Nvidia’s GeForce RTX 3090 and GeForce RTX 3080 cards feature independent push and pull configurations. So the fans are positioned on opposite sides of the heat sink.
Here’s an image of how it works:
How this approach will influence the airflow remains to be seen, but we’re certain that Nvidia has already addressed this and knows what they are doing. It does, however, raise the issue of cost, as this design may add an additional cost.
Even though it hasn’t been officially announced, it is believed that third-party cards will feature a more conventional two-fan or three-fan design. The image below shows Gainward’s RTX 3090 graphics card, which features a traditional three-fan design.
From the image above, it appears that the RTX 3090 will take up three slots and possibly requiring people to further upgrade their case to accommodate their new card’s size. Although this has now been confirmed to be true, we need to state that the RTX 3080 and RTX 3070 will take up less space and therefore “only” two slots.
It has recently been confirmed by EK Water Block (a well-known water cooling solutions company) that they are already working on the new water cooling systems for the next generation of graphics cards.
RTX 3000 Series Features
The September 1 RTX 3000 series launch event revealed many new features coming to the next-gen of Nvidia GPUs. Let’s talk a bit about them.
One of the first things Huang talks about is Nvidia’s foray into the streaming world. Although so far the details are scarce, it looks like Nvidia worked extra hard to make streaming video games a smoother experience.
This feature is called Nvidia Broadcast and the tech demo given during the event is more than impressive. Below is a decent video showing off some great AI-assisted features as well as a host of other options which will certainly make streamers’ jobs a lot easier.
The next thing that was announced was Nvidia Reflex. It was introduced as an eSports technology that “optimizes the rendering pipeline across CPU and GPU, to reduce latency by up to 50%”. It will be released in September with Nvidia’s Game Ready driver. As a sort of a continuation of the Reflex technology, Huang announced four brand new 360Hz monitors coming this fall.
Nvidia Omniverse Machinima
One of the really impressive things that Jensen Huang announced is Nvidia Omniverse Machinima.
It’s an app built on our Omniverse 3D workflow collaboration platform. Omniverse is a universal design tool asset exchange with a viewer based on photorealistic path tracing. The engine is designed to be physically accurate simulating light, physics, material, and artificial intelligence.Jensen Huang, during the RTX 3000 launch event
It’s important to note that this won’t be a launch day feature and that the beta is coming this October. To sign up for the beta go to nvidia.com/machinima.
Below is a demo of what this technology is capable of. For this video, the creators used assets from Mount & Blade II: Bannerlord video game.
As is the case with the previous Nvidia GPU series, the RTX 3000 will heavily feature ray tracing and there are some rumors suggesting that the new series is roughly four times faster than its predecessor.
This bodes very well for Nvidia, as AMD is still playing catch-up in that department, and with rumors suggesting that not every Big Navi card will support ray tracing.
The first rumors about RTX IO (then known as NVCache) started back in May, and as time goes on, they appear more and concrete. On paper, this looks to be the push that will launch the concept into the mainstream.
And that concept is something else. The idea is that the GPU will dynamically utilize the bandwidth from system RAM, VRAM, and SSD to execute multiple tasks simultaneously at a much higher speed.
This is presumably a response to Playstation 5’s custom memory solution that is said to be able to increase the loading speed a hundredfold (compared to PS4). Of course, those are Sony’s claims so we’ll have to see both how the console and the PC market will handle this technology jump.
However, the rumors indicate that we won’t see RTX IO in action until next year, and maybe not even in the Q1.
Tensor Memory Compression
This tech will apparently tensor cores for compression and decompression of VRAM stored items. Estimate suggest that this application of tensor cores could lead to anywhere between 20-40% less VRAM usage.
During the official September 1 launch event, Nvidia CEO Jensen Huang heavily touted DLSS 2.0 as the key for high-FPS 8K gaming. However, it should be noted that this technology still hasn’t proved itself, despite the lofty promises.
Some key rumors regarding DLSS 2.0 are that it should work with any game that has TAA (temporal anti-aliasing), but that a game ready driver will be required. It might be slow, but it’s still a work in progress.
According to Huang, the enthusiast RTX 3090 will feature 285 Tensor TFLOPs which is more than twice than that of the RTX Titan which it replaces. Despite RTX Titan featuring 130 Tensor TFLOPs, it still held a massive $2499 price tag and the fact that RTX 3090 comes at $1000 less is almost unbelievable.
As always, it’s best to wait for the benchmarks before making up your mind to buy a new graphics card.
Early looks suggest that based on benchmarks, the RTX 3080 offers around 80-100% better performance than the RTX 2080. Below is a review from JayzTwoCents which pretty much confirms that uptick in performance.
JayzTwoCents’ RTX 3080 Benchmarks
Another thing that Jay brings up that’s been bothering the internet community ever since the design leaked, is the temperature. Many were sceptical or downright worried that because of the fan placement (one on each side), there would be hot air being blown into the CPU.
However, Jay found that the air directed at the processor is warm-ish at best, and furthermore, he said that the overall GPU temperature never went above 72°C which is pretty good.
RTX 3080 Founders Edition Review By IGN
Because of IGN’s reputation to be more gaming oriented, we have chosen their review for this list as well. During the video, you can see different results across multiple games, but the overall feel is the same – RTX 3000 series has arrived and brought true 4K 60FPS gaming with it.
Linus Tech Tips
It would be hard to have a serious online discussion about technology without Linus’ opinion. As always, he’s been able to get the card on time and thoroughly test it.
It’s important to state that Linus and his team have also done tests for non-gamer GPU users. They’ve tested the RTX 3080 in professional programs like Blender, 3D Max, and Maya. As expected, the results are phenomenal.
One of the reasons why Linus is one of the best in the business, if not the best, is his ability to think outside the box. Much like JayzTwoCents, he wondered about the airflow and the temperature, but also took action to measure both the GPU and the overall system temperature.
Although RTX 3080 does use a lot more power, it’s truly amazing that they have not only managed to keep the GPU temperature down, but have also managed to lower the entire system’s temperature. Reminder – due to the RTX 3080 interesting double fan design, it was widely speculated that the airflow inside the case will be horrific and possibly heat up the entire case.
You can sign up to get a notification when the Founder’s Edition RTX 3000 series cards are available for pre-ordering.
You can not pre-order any AIB cards yet, and nor can you sign up anywhere to get an e-mail notification once they are available.
We’ll update this section here with links to product pages where you can pre-order the cards.