We don’t know much about the NVIDIA GeForce RTX 4000 series cards. Unfortunately, no one does at the time of writing this article.
However, there is some information we’ve managed to gather about this new generation of GPUs.
We also consider NVIDIA’s previous GPU lineup, the RTX 3000 series, and the performance improvements over the RTX 2000 series. This helps us visualize the potential of the RTX 4000 GPUs.
If you are interested in what the next generation of NVIDIA graphic cards will bring, this article contains the information you are looking for.
Let’s get right into it!
AMD RDNA 3 Release Date, Price, And SpecsIntel Arc Alchemist Release Date, Price, Performance, Specs
May 11, 2022: A couple of grammar fixes, no significant updates.May 4, 2022: Added and replaced a lot of information to coordinate with news that Ada Lovelace is on TSMC 4nm.April 22, 2022: Added new specification table, new power-related leaks and more.April 12, 2022: Added the possibility of Nvidia going with an RTX 5000 naming scheme.March 24, 2022: Updated information about Ada Lovelace’s potential power usage.March 15, 2022: Added power usage leaks for the RTX 4090.March 4, 2022: Added new release date, specifications leaks, L2 cache table, and a lot more.March 1, 2022: Minor information updates, nothing significant.February 22, 2022: Grammar updates.January 31, 2022: Added additional specifications leaks on the potential RTX 4090 Ti and more.January 25, 2022: Added information on RTX 3090 Ti and fixed some grammar issues.January 13, 2022: Added new specification-related leaks.December 22, 2021: Added a bit more information on RTX 3000 Super/Ti.December 14, 2021: A few updates, nothing notable.December 6, 2021: Added new information about RTX 40, TSMC, and the Metaverse.November 29, 2021: Reworked grammar throughout the entire article.November 22, 2021: Added a cleaner introduction.November 15, 2021: Added new release date information.November 12, 2021: Added performance leaks.November 3, 2021: Optimized readability. October 28, 2021: Reworked certain paragraphs for clarity.October 18, 2021: Added new information for the RTX 3090 Ti and updated some RTX 4000 related information.October 12, 2021: Added new information for RTX 3000 Super SKUs and RTX 4000 release date.October 4, 2021: Added specifications table for AD102 and other useful information.September 28, 2021: Made a few grammar corrections.September 20, 2021: Added more information regarding the release date.September 2, 2021: Added new leaks regarding the release date of Ada Lovelace.
Table of ContentsShow
Currently, there is no official information regarding the release date of the following NVIDIA GPU lineup. So, what do we do know so far?
Well, there is something in the works for sure. Judging by NVIDIA’s GPU history and the pattern of their GPU releases, the RTX 4000 series is almost certainly already in development.
Based on NVIDIA’s official architecture roadmap, Ampere Next (or Ada Lovelace) has a planned 2022 release. By 2024, they believe they will have Ampere Next Next ready, which is supposedly the RTX 5000 series.
However, there are also talks that Nvidia might skip the 4000 naming scheme altogether and jump to 5000. So, RTX 5000 could be Ada Lovelace which means Ampere Next Next will be RTX 6000.
The latest leak by kopite7kimi claims that the RTX 4000 series could show up in Q3 of 2022. However, we are unsure whether kopite7kimi refers to a summer release or just an announcement. It could be either.
Although Greymon55 mentioned on Twitter that Ada Lovelace will release in September of 2022.
It still feels early for the RTX 4000 series, considering the relatively recent 3070 Ti and 3080 Ti releases.
But wait, NVIDIA is planning more!
Circulating rumors hint towards RTX 3000 Super GPUs that will release at the end of 2021 or early 2022.
The above leak by kopite7kimi shows the specs of the entire lineup. Highlights include increased CUDA cores, VRAM memory, and an upgrade to GDDR6X from GDDR6 non-X (excluding the 3060S).
There are some rumors that the RTX 3090 Ti’s production is at a halt because of a hardware problem. A hardware problem during production can cause a several-week or several-month delay. It could even completely suspend the release of the GPU.
Either way, these new Super or Ti variants of the RTX 3000 Series won’t be a significant upgrade from the current lineup. In fact, the upgrade will be minimal. So, if you are already on an RTX 3000 SKU, we recommend waiting for RTX 4000.
As mentioned previously, no official info is available for the RTX 4000 series, but there have been some leaks. Currently, the only concrete information is its codename: Ada Lovelace.
The Ada Lovelace microarchitecture will be on a 4nm fabrication process, a giant leap from the 8nm Ampere (Samsung) chips. A smaller microarchitecture will result in two significant benefits: more performance for a lot less power.
Thanks to TSMC’s 4nm process, Nivida gets those benefits by cramming more transistors onto the chip.
Note that most of the first leaks we had on the RTX 4000 suggested that Nvidia will use TSMC’s 5nm process.
Leaks suggest an increase of more than 7500 CUDA cores from the GA102 (3090, 3080) chip. This includes almost double GPC (Graphics Processing Clusters), 7 to 12, and 30 more TPC (Texture Processing Clusters).
That’s a massive bump in specifications, considering the Ampere chip had an increase of just 1 GPC from the Turing 102 chip.
GA102, AD102, AD103, and AD104 (rumored) specification comparison table:
SpecificationsRTX 3090RTX 4090 Ti/TitanRTX 4090RTX 4080RTX 4070
ArchitectureAmpereAda LovelaceAda LovelaceAda LovelaceAda Lovelace
ProcessSamsung 8nmTSMC 5nmTSMC 5nmTSMC 5nmTSMC 5nm
Memory24GB GDDR6X48GB GDDR6X24GB GDDR6X16GB GDDR6X12GB GDDR6X
Total Graphics Power350W900W600W350W300W
Release DateSeptember 1st, 2020TBA 2022TBA 2022TBA 2022 TBA 2022
We’re basing the specifications in the table above based on a leak from Twitter user 3DCenter_org.
In February 2022, Nvidia went through a huge data breach which is unfortunate and we do not condone such attacks/hacks, however, this information is already out there, so here’s what we managed to find related to RTX 4000.
CUDA Cores and L2 Cache
Interestingly, 3D Center’s December leaks about the AD102 CUDA cores are actually correct. The leaks straight from Nvidia suggest 18,432 CUDA cores which is about 70% more than the RTX 3090. Already, those are some impressive numbers.
Additionally, Nvidia looks to increase L2 cache memory on all Ada Lovelace SKUs. Even the low-end GPUs (AD107) will get about 32MB of L2 cache. The RTX 4090 card will max out at 96MB of L2 cache.
Information extracted from the data breach – Source: xinoassassin1
The RTX 3060 has only 3MB of L2 memory which means that the future RTX 4060 will have 10 times bigger L2 cache. To make this change even crazier, this means that the 4060 will have more L2 than the RTX 3090.
Here’s a L2 cache table comparing RTX 3000 and RTX 4000 GPUs:
RTX 3000L2 CacheRTX 4000L2 Cache
It looks like boosting cache size is a trend these days considering that both AMD and Intel increased their L2 and L3 cache sizes on their processors to boost gaming performance.
We’ll see whether Nvidia’s memory boost will improve the GPU’s ability to tackle games.
With these specifications in mind, it’s definitely possible that the RTX 4000 series graphics cards can be twice as fast as RTX 3000 cards. Maybe even faster.
However, not all news from these leaks is good. There are some rumors circulating about a considerable increase in power consumption for AD102. Greymon55 mentioned that the TGP (total graphics power) for AD102 can be up to 850W.
However, with Nvidia’s announcement of H100 Hopper compute GPU, it seems those rumors have died down a bit. Tom from Moore’s Law Is Dead claims that Ada Lovelace GPUs will not be able to draw more than 700W.
So, a realistic assumption is around 600W for an RTX 4090/4080.
These power usage-related assumptions are also supported by the fact that many power supply manufacturers are now switching over to 12-pin PCIe 5.0 power connectors with the capability to deliver 600W of power.
Kopite7kimi also mentions that it’s been confirmed that the RTX 4090 will be a 600W TGP GPU.
While these kinds of changes are unfortunate, it seems they are necessary. Even the switch to 4nm might not be enough to tackle AMD’s new MCM-based GPUs.
AMD’s Navi 31 SKUs will also have higher power draw, but the limit should be around 400W which is considerably lower than Nvidia’s future flagship.
There are also rumors circulating about a new RTX Titan which could draw over 900W.
Kopite7kimi shares that the RTX 4070 will have 300W TGP. While 300W is very far away from 600W, that’s still a large power draw for an xx70 card. The 3070 has a TDP of 220W which is miles better.
Interestingly, Kopite7kimi mentions that the RTX 4080 will use AD103 chips rather than the AD102 chip.
That’s unexpected considering that the RTX 3080 shared the same chip (GA102) as the RTX 3090. Many of us used this same logic for the RTX 4000 series, but, looks like we were wrong.
Allegedly, AD102’s target frequency is 2.2GHz, which is theoretically around 66 TFLOPS (FP32). That is almost double in comparison to Ampere’s flagship (RTX 3090).
However, Greymon55 has another say about the specifications of Ada Lovelaces, which counters 3DCenter’s rumors. Back in 2021, Greymon55 already posted a leak with AD102 specifications and mentioned that the 3090 successor could have up to 92 TFLOPS.
Now, the leaker doubles down on the idea and still believes that AD102 will have up to 90 TFLOPS compared to the 66 TFLOPS rumored by 3DCenter. That’s double compared to the 3090 or even the 3090Ti.
However, remember that TFLOPS isn’t the most accurate depiction of in-game performance.
There are currently no benchmarks open to the public, so we can’t really determine the performance of the RTX 4000 series, but we have a few rumors to share.
Greymon55 has again provided some insight. Based on their claims, these new GPUs should bring double the performance and double the power consumption.
Should this be acceptable? We welcome 2x performance, but doubling the power consumption can be a problem.
Only a small percentage of PC users have 1000W+ PSUs that can handle a 600W GPU.
However, it’s not all bad. Previously, most of us tech sites/reviewers assumed that Nvidia’s Ada Lovelace will be based on TSMC 5nm fabrication process.
But, it seems that we’ve all been wrong because RTX 4000 GPUs will actually use a 4nm node. Before this news, the favor was heavily shifted towards AMD’s RDNA 3, but the scales are finally tipping.
So, even with these huge power draw spikes, Ada Lovelace should be very competitive.
It is impossible to guess how the pricing will look with the current info. However, since AMD’s RX 6000 series is on par with the RTX 3000 series in terms of performance (in some cases even better), NVIDIA will most likely stick with the exact pricing as Ampere.
So, the 4060 should be around $330 while the 4070 and 4080 for $500 and $700 respectively. We’re unsure about the 4090/Titan, though. Those two would cost more than $1000.
Unfortunately, there are some talks circulating that the jump to 4nm from TSMC could be a lot more expensive than we expected.
For example, the previously mentioned Twitter leak by kopite7kimi was a response to nerdtechgasm’s suggestion that 4nm NVIDIA is going to be very expensive for consumers.
Additionally, Nvidia is reportedly spending tens of billions of dollars to help the supply of RTX 4000 and future cards.
Scalping & Price Increases
Image credit: Gizmodo
If you have recently been searching for a new graphics card for your system, you have probably been disappointed several times. GPUs with a $500 MSRP are on retail stores like Amazon or Newegg for more than $1000.
The $700 RTX 3080 is frequently on sale by scalpers for $1500 on the second-hand market.
Such a level of scalping is possible due to the global shortage of semiconductor chips in 2021. The world’s largest semiconductor manufacturers, TSMCSamsung,
Already this shortage continues in 2022 as it did in 2021 and 2020.
Right now, you might be wondering: will this kind of scalping affect the RTX 4000 series?
This is a great question and quite a concerning one, but when the semiconductor industry goes back to normal, we cannot know. Fortunately, TSMC has already begun investing to increase its manufacturing capacities. In addition, Samsung is also planning to build a new chip manufacturing factory in the US.
There are also rumors that the USA is planning to invest $37 billion in the chip-making industry.
We will keep you updated if there is any further information released about the future NVIDIA GPU lineup.