304 North Cardinal St.
Dorchester Center, MA 02124
September will mark two years since NVIDIA launched its Ampere series with the RTX 3090 leading the way. While impressive from a technical standpoint, the generation was plagued by accessibility issues and ridiculous price tags. These are finally disappearing, just in time for the RTX 4090 to drop.
What do we know about the 4090? Not that much yet, but enough to whet the appetite of enthusiast players and serious creators alike!! We’ve collected all credible rumors and leaks in one place, so it’s easier for you to follow.
GPU manufacturers try to keep their cards’ technical specifications under wraps until pre-launch events, so almost nothing about the 4090 RTX is officially known. The only two pieces of information we don’t need the rumor mill for are that the generation’s microarchitecture is called Ada Lovelace and is based on TSMC’s 5nm production process.
Fortunately, several tasters have shed light on a handful of details. Here’s a table comparing the RTX 3090 to what’s speculated about its successor for a better overview:
|RTX 3090||RTX 4090|
|Manufacturing process||Samsung 8nm||TSMC 5nm|
|Memory||24 GB GDDR6X||24 GB GDDR6X|
|Memory speed||19.5 Gbps||21 Gbps|
|Basic watch||1395 MHz||2235 MHz|
|Boost clock||1695 MHz||2720 MHz|
|TFLOPs||35.6||Up to 90|
There’s a lot to unpack here, so let’s start with the new die. Moving to 5nm will certainly increase the transistor count significantly from the 3090’s 28 million. However, no one has hinted at how much yet.
The increases to base and boost clocks are even more indicative of a large expected performance bump. Initial leaks claimed the boost clock would go up to 2500MHz, which credible leakers Greymon55 and Kopite7kimi later revised to around 2750MHz. That’s more than a 1000MHz increase for the Founder’s Edition alone, not counting what board partners will come up with. Could we see the ROG or FTW models break the 3GHz mark?
Then there is the matter of CUDA kernels. Kopite7kimi’s original estimate was 16,128, revised to 16,384. Whatever turns out to be true, the 60% upgrade is a strong indicator of excellent performance.
The cached information occurred as a result of a data breach, although we cannot be sure that it is trustworthy. If it is, it would be one of the most interesting things we know. The fifteenfold increase in L2 cache was probably inspired by AMD’s attempt to mitigate the lack of memory bandwidth that the 6000 series cards would otherwise have.
For example, the RX 6900 XT has about half the memory bandwidth of the RTX 3090, yet 128 MB of its Infinity Cache smooths the communication between the first two cache levels and the memory. AMD has also successfully applied a similar approach to the new Ryzen 7 5800X3D, so it’s likely that NVIDIA has been paying close attention and will capitalize on this development.
Calculations suggest that the new card can achieve around 66 TFLOPs, but the leak Greymon55 has continued to insist for months that it’s a whopping 90. TFLOPs aren’t as clear an indicator as they once were, but such a drastic generational increase still warrants attention.
So what can we infer about the real world of the 4090s from all of this? It will undoubtedly be a big step up from the 3090, which follows the trend of Ampere and Turing. We can only speculate how big the increase will be.
Another of Kopite7Kimi’s more cryptic tweets says doubling the 3090’s output is not difficult to achieve. If that is indeed the case, the 4090 will be a great upgrade for content creation, video encoding and 3D modeling. On the gaming front, you can expect it to run even the most demanding or poorly optimized titles at or close to triple-digit framerates in 4K Ultra.
A few days after our article was launched, Kopite7Kimi had something more substantial to say. They claim that the RTX 4090 can reach a score of more than 19,000 in the 3D Marks Extreme benchmark. That’s a third more than the RTX 3090 Ti’s hall of fame results when paired with a HEDT CPU. The RTX 3090 paired with something like the i9-12900K scores in the 10-12K range, meaning the 4090 could effectively double its performance.
There’s no such thing as a free lunch, and it looks like 4090 owners will have to pay through their utility bills. Kopite7Kimi informs that the card’s TBP is 450W, which is in line with the RTX 3090 Ti’s requirements. However, they also suggest that the power limit for the AD102 will be 800W.
It’s not clear if this refers to the 4090 or a potential Ti version allegedly built on the same die. We’re inclined to believe the latter, so a 600W limit for the regular card sounds more reasonable. Even so, it’s still a significant increase that will make enthusiast PC builders rethink their PSU choice. You’ll want a 1,000+W PSU to run such a system comfortably, and even that isn’t sure enough until advanced Raptor Lake requirements become known.
A more dubious rumor that surfaced on the Chiphell Forums claims that the RTX 4090 will be available in 450W and 600W variants. Does this mean the card will be available in two versions? Will the lower power limit refer to the Founder’s Edition and the higher one reserved for the most advanced partner cards? It is not clear at this time.
Nothing is known yet about how much the RTX 4090 will cost at release. Crypto market volatility has left many scalpers with a sour taste in their mouths as they struggle to offload used 3000 series cards, people wary of buying even at a discount. Since the incentive to mine is severely depleted, the launch of 4090s should not be as plagued by shortages and price increases as its predecessors.
We assume it will be easier to find the RTX 4090 at MSRP, but what will that MSRP be? NVIDIA will have to balance being competitive with covering the costs associated with Ada Lovelace’s added demands. They will need to account for the much more powerful cooling needed to handle a 450W TGP GPU, for example.
One thing we can say with some certainty is that the RTX 4090 won’t cost less than the 3090, so expect to pay no less than $1,500. The most popular MSRP theories put the price between $1,699 and $1,799. We wouldn’t be surprised if the actual price rises even higher.
There is no longer any need to speculate about the 4090’s release date, as many signs now point to an October release.
Wccftech has the latest scoop on NVIDIA’s plans. According to one of their sources, the RTX 4090 will make its debut at the GTC developer conference on September 20. Its real launch should follow soon, sometime in October.
There is more evidence to support this, tiah just hearsay. The same site has leaked several images of Zotac’s AMP Extreme version, which is usually one of the most advanced models. To say the map pictured is massive is an understatement. Get your PC cases ready folks, because it’s going to take a whopping four PCIe slots to house that giant cooler!
More news came at the same time via Videocardz. They confirmed that Gigabyte has submitted no less than 11 RTX 4090 models to the EEC. Three of them have water cooling, while the rest are air-cooled variants in the AORUS, Eagle, Windforce and Gaming lines.
Although much about the NVIDIA 4090 remains shrouded in mystery, the information that is out there suggests that we are in for a treat! More informed speculation and facts will be available as the release date approaches. We will make sure to keep the overview fresh so you don’t miss any updates.