NVIDIA GeForce RTX 4000 Series Release Date, Price, Specs

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!

Related:AMD RDNA 3 Release Date, Price, And SpecsIntel Arc Alchemist Release Date, Price, Performance, Specs

Release Date

Currently, there is no official information regarding the release date of the following NVIDIA GPU lineup. So, what do we know so far?

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.

There were several leaks about the release date of the RTX 4000 series, and most of them pointed towards Q3 or Q4 of 2022.

However, it seems that none of these leaks are now accurate.

Yes, only the AD102, and only one model.🤪

— Greymon55 (@greymon55) July 25, 2022

After several delay rumors, in the end, it seems that the only GPU NVIDIA will release in 2022 is an AD102 GPU (RTX 4090).

This is both unfortunate and unexpected. If AMD comes out with multiple SKUs in 2022, NVIDIA may have a hard time catching up with them.

NVIDIA also asked TSMC to cut down on orders for 5nm wafers. However, TSMC did not accept this agreement and only agreed to a 1 quarter delay on shipments which may be another reason why Nvidia is pushing for a 2023 release.

NVIDIA’s Stopgap

To reduce the time gap between GPU releases NVIDIA had to introduce a stopgap. Something to create better competition against AMD’s lineup and to improve market share.

NVIDIA achieved this by releasing three new GPUs: RTX 3090 Ti, RTX 3080 Ti, and RTX 3070 Ti.

Even though these cards aren’t exactly a great upgrade over previous RTX 3000 GPUs, they are still a good option for anyone that can’t wait for RTX 4000 GPUs anymore.

Specifications

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 5nm fabrication process, a solid 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 5nm 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.

GA102, AD102, AD103, and AD104 (rumored) specification comparison table:

Specifications	RTX 3090	RTX 4090 Ti/Titan	RTX 4090	RTX 4080	RTX 4070	RTX 4060
GPU Die	GA102	AD102	AD102	AD103	AD104	AD104
Process	Samsung 8nm	TSMC 5nm	TSMC 5nm	TSMC 5nm	TSMC 5nm	TSMC 5nm
Base Clock Speed	1395 MHz	?	2235 MHz	?	?	?
Boost Clock Speed	1695 MHz	?	2520 - 2750 MHz	?	?	?
CUDA Cores	10496	16384	16384	10240	7680	4608
Streaming Processors	82	128	128	80	60	?
Bus Width	384-bit	384-bit	384-bit	256-bit	192-bit	192-bit
Memory	24GB GDDR6X	48GB GDDR6X	24GB GDDR6X	16GB GDDR6X	12GB GDDR6X	8GB GDDR6/X
Memory Speed	24Gbps	21Gbps	21Gbps	21Gbps	21Gbps	?
Bandwidth	936.2 GB/s	?	1008 GB/s	676 GB/s	504 GB/s	?
L2 Cache	6MB	96MB	96MB	64MB	48MB	48MB
Total Board Power	350W	?	450W	420W	300W	200W
Release Date	September 1st, 2020	TBA 2023	TBA/Q3 2022	TBA/Q3 2022	TBA/Q4 2022	TBA/2022/23

We’re basing the specifications in the table above based on the latest leak from kopite7kimi.

Continue to update.
RTX 4090, AD102-300-A1, 16384FP32, 384bit 21Gbps 24G GDDR6X, 450W, base 2235 boost 2520 actual max >2750;
RTX 4080, AD103-300-A1, 10240FP32, 256bit 21Gbps 16G GDDR6X, 420W,
RTX 4070, AD104-275-Kx(x is a number)-A1, 7168FP32, 160bit 18Gbps 10G GDDR6, 300W.

— kopite7kimi (@kopite7kimi) July 4, 2022

If the RTX 4090 truly boosts up to 2520 MHz or even 2750 MHz, we might see some serious improvements in performance since the RTX 3090 boosts up to 1695 MHz. With AIB models, a 3000 MHz factory overclock is possible.

With a 60% improvement in boost clocks over its predecessor, NVIDIA might beat AMD’s RDNA 3 flagship SKU. But, this will come at a cost. With a 450W TDP and some of the highest clock speeds we’ve seen, we expect very high heat output.

Another thing to note is that the RTX 4080 will use the AD103 GPU die instead of AD102. In the previous generation, the RTX 3090 and 3080 used the same GPU die.

The RTX 4070 is where everyone’s focus should be, especially now with the newer specifications leaks from kopite7kimi.

There is an update of RTX 4070.
PG141-SKU331
7680FP32
12G 21Gbps GDDR6X
300W
TSE >11000

— kopite7kimi (@kopite7kimi) August 4, 2022

If the 4070 gets 12GB of GDDR6X VRAM @ 21Gbps instead of 10GB GDDR6 @ 19 Gbps (like previous leaks suggested), it could be as fast as the RTX 3090 Ti. At least, that’s what some leakers believe.

The 8GB of VRAM on the RTX 3070 are not enough in certain games at 4K or VR, so it’s definitely an upgrade we can appreciate.

The memory bus has also changed, from 160-bit to 192-bit. Power draw is still estimated at 300W, which is the RTX 4070’s only downside.

In February 2022, Nvidia went through a massive 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.

L2 Cache

It seems that 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.

The RTX 3060 has only 3MB of L2 memory, which means that the future RTX 4060 will have a ten 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 an L2 cache table comparing RTX 3000 and RTX 4000 GPUs:

RTX 3000	L2 Cache	RTX 4000	L2 Cache
GA102	6MB	AD102	96MB
GA103	4MB	AD103	64MB
GA104	4MB	AD104	64MB
GA106	3MB	AD106	48MB
GA107	2MB	AD107	32MB

It looks like boosting cache size is a trend these days, considering that 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.

Power Usage

You’re probably already pretty aware of all the news about RTX 4000 being too power hungry.

While we still have no official information from Nvidia regarding these specifications of Ada Lovelace, it seems that these rumors will be true.

Kopite7kimi comes out with a “final” version of the power limits of future RTX 4000 SKUs.

A truth. The power limits:
AD102, 800W;
AD103 (DT), 450W, AD103 (Mobile), 175W;
AD104 (DT), 400W, AD104 (Mobile), 175W;
AD106 (DT), 260W, AD106 (Mobile), 140W.
But I don't think we need to use the full power cap.

— kopite7kimi (@kopite7kimi) June 18, 2022

So, AD102, the flagship NVIDIA GPU will have a power limit of up to 800W. While this doesn’t necessarily mean that the RTX 4090 will come with such a TDP, but it is possible.

Kopite7kimi already shared that the RTX 4080 will have a 420W TDP, so, it’s not unexpected that AD103 has a 450W power limit.

However, a 400W power limit for AD104, or more accurately, an RTX 4070? We hope that the rumors for 300W TGP will be the final specification for this SKU.

We have to remember that the RTX 3070 has a 220W TDP which was already higher than previous NVIDIA xx70 GPUs.

At this point, people may start reconsidering Nvidia’s mid-range GPUs solely because of their high-power draw, especially when they consider that RDNA 3 may be more efficient.

TFLOPS

Allegedly, AD102’s target frequency is 2.2GHz, which is theoretically around 66 TFLOPS (FP32). That is almost double compared to Ampere’s flagship (RTX 3090).

However, Greymon55 has another say about the specifications of Ada Lovelaces, which counters 3DCenter’s rumors. In 2021, Greymon55 already posted a leak with AD102 specifications and mentioned that the 3090 successor could have up to 92 TFLOPS.

If lovelace gets up to 2.5GHz, you'll get about 90t of FP32, and 3090ti is just 40t.😬

— Greymon55 (@greymon55) January 25, 2022

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.

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.

Kopite7kimi revealed 3DMark Time Spy Extreme benchmarks with the RTX 4090, 4080, and 4070.

RTX 4080, TSE >15000,
RTX 4070, TSE ~10000.
These scores are based on specs I mentioned before.
Honestly, it's not quite sure yet except RTX 4090. Well, I don't care.

— kopite7kimi (@kopite7kimi) July 29, 2022

The RTX 4090 scores 19000 points in 3DMark, which is around 66% faster than the RTX 3090 Ti (~11400 points).

With 15000 points in Time Spy Extreme, the RTX 4080 is almost 25% faster than the 3090 Ti.

These kinds of performance jumps are expected, but what’s interesting is that the leak suggests that the RTX 4070 scores around 10000 points, making it almost as fast as a 3090 Ti.

It is also possible that this SKU might not be the real RTX 4070. The 4070 with 21Gbps will most probably be faster than the 3090 Ti.

This gives us an insight into the potential performance of Ada Lovelace, but don’t take these numbers as concrete evidence.

Originally, Greymon55 believed double performance for doubled power consumption and these benchmarks do point in that direction.

Double performance, double power consumption, can you accept it?

— Greymon55 (@greymon55) November 5, 2021

Will you accept 2x performance for such a high power draw?

Price

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, here’s what it should look like:

• RTX 4060 – $330
• RTX 4070 – $500
• RTX 4080 – $700
• RTX 4090 – $1500+?

However, this jump to TSMC’s 5nm process node could be the reason for far more expensive RTX 4000 GPUs

Scalping & Price Increases

Throughout the last couple of years, ever since the release of the RTX 3000 series, scalping has become a huge trend.

This new trend skyrocketed the prices of new GPUs while also considerably cut down the supply. This shortage continues in 2022 as it did in 2021 and 2020.

Even the world’s largest semiconductor manufacturers, TSMC and Samsung, struggle to keep up with the demand.

Right now, you might be wondering: will this kind of scalping affect the RTX 4000 series? Unfortunately, we cannot know for sure, but at least we know that both TSMC and Samsung are investing to improve their manufacturing capacities.

We’ll know the status of RTX 4000 availability once the GPUs are released.

Until then, you can check back for future developments on the RTX 4000 Series!