Nvidia GeForce RTX 4080: everything we know about the new GPU

Audio player loading…

THE BEST RTX 4080 LEAKS

– We’re expecting two variants of the RTX 4080: one with 12GB of video memory, and one with 16GB
– ‘Full-fat’ 16GB model will feature 9728 CUDA cores, 23Gbps memory speed
– Base TGP is 285W/240W, lower than previous predictions
– Max power draw is expected to be 516W
– 12GB model actually has a slightly higher clock speed

We’re frighteningly close now to the launch event for Nvidia’s ‘Lovelace’ RTX 4000 graphics cards, which will be taking place on September 20 and is expected to feature the Nvidia GeForce RTX 4080. As of right now, we’re expecting to see three GPUs properly showcased, though sadly this doesn’t include the RTX 4070.

What it should include is the flagship RTX 4090, along with two different flavors of RTX 4080; one with 12GB of GDDR6X VRAM, and a full-fat version packing a more powerful 16GB.

Since the Nvidia GeForce RTX 3080 was such an incredible GPU, introducing 4K gaming to the mainstream PC space, we’ve got high hopes for the RTX 4080. As far as we’re concerned, it has the potential to be one of the best Nvidia graphics cards on the market. 

Proceed with caution as you continue to scroll; much of what we ‘know’ so far is actually based on leaks, rumors, and reasonable logical extrapolation. The information contained within this article could be right on the money, or it could be completely wrong. There’s no way to know until September 20, so we await with bated breath…

Cut to the chase

  • What is it? Nvidia’s next xx80-series graphics card
  • When is it out? TBD, likely October or November 2022
  • What will it cost? TBD

Nvidia GeForce RTX 4080 release date

  • No official release date yet, should be revealed September 20
  • Current leaks put on-sale date in October or November
  • Unclear whether new cards will launch at the same time

Debate has been rife for weeks regarding when the RTX 4080 will actually be released, even this close to the Nvidia presentation where it will be revealed. Rumors have been flying in every direction, but the main one is from famed Twitter leaker @kopite7kimi, who suggested that Lovelace, the architecture the RTX 4080 will be use, will be launching “a little bit earlier”

This was in reference to earlier rumors that suggested the next-generation graphics architecture would be launching at the end of 2022. Couple that with a different leaker previously suggesting that the next graphics card will be launching in October 2022, and it’s starting to look like we may see the Nvidia GeForce RTX 4080 almost exactly two years after the launch of the RTX 3080. However, VideoCardz has suggested that only the RTX 4090 will launch in October, with the 4080 following a few weeks later in November.

Either way, this would be par for the course, as most graphics card generations last for roughly two years before being replaced. The RTX 2080, for instance, launched in September 2018, a little over two years after the May 2016 launch of the GTX 1080. And, of course, the RTX 3080 launched in September 2020, which was almost two years on the dot. So, if Nvidia is going to keep up its regular cadence with its next GPUs, we can assume that current leaks are accurate and we’ll see at least one RTX  4000 card launching in October 2022.

Cyberpunk 2077 RTX On screenshot

(Image credit: Nvidia/CD Projekt Red)

Nvidia GeForce RTX 4080 price

  • Very little info on pricing at this stage
  • Current-gen RTX 3080 originally retailed at $699
  • Two different models make guessing prices tricky

Back when we originally reviewed the Nvidia GeForce RTX 3080, we praised it for bringing such a huge improvement in performance at a notably lower price point than the RTX 2080 Super. Of course, the rise of crypto mining saw GPU prices reach some frankly insane prices, so that point feels a bit double-edged in hindsight.

However, the crypto crash happened: chunky mining rigs with multiple graphics cards are being sold off en masse,  as the value of cryptocurrencies plummeted and Ethereum’s great ‘Merge’ saw conventional mining become redundant. Most current-gen GPUs are actually now available at (or close to) MSRP once again as the market stabilizes.

Still, the RTX 3080 carried an MSRP of $699 (£649) when it first launched, and generational pricing tends not to shift too much where Nvidia is concerned, so we can likely see a price tag somewhere in that region. Pricing information has yet to be revealed (or leaked), so this is all pure speculation on our part.

With two different versions of the RTX 4080 arriving, though, it’s going to be tricky to accurately predict pricing. While it’s a fairly safe bet that one of these two variants will stick close to the 3080’s $699 retail price, which one that’ll be is complete guesswork at this stage.

Our personal prediction – based somewhat on the piggy-bank-shattering price rumors around the RTX 4090 – is that the 12GB model will arrive at $699 while the 16GB version will sit somewhere around the $800-900 mark. We hope we’re wrong and the cards are cheaper than that, but we’re trying to be realistic with our expectations.

Nvidia RTX 3080 Ti on a black surface

(Image credit: Nvidia)

Nvidia GeForce RTX 4080 specs

  • 16GB model will feature 9728 CUDA cores, 2,505MHz boost clock, 256-bit memory bus and memory bandwidth of 736GB/s
  • 12GB model will feature 7680 CUDA cores, 2,610MHz boost clock, 192-bit memory bus and memory bandwidth of 504GB/s
  • Power demands won’t be as severe as originally feared

The 12GB model of the RTX 4080 was, in fact, what many leakers had previously referred to as the RTX 4070 or 4070 Ti, since it is believed to use a slightly different central GPU (though still manufactured on Nvidia’s 4nm Ada Lovelace process).

Given that it was previously mistaken for a lower-tier card, it unsurprisingly sits beneath the 16GB version in just about every area, not just VRAM. There isn’t just less memory; it’s less powerful, too, with a smaller bus and a bandwidth of 504GB/s as opposed to the 16GB model’s 736GB/s.

The 16GB 4080 also has faster memory than even the RTX 4090, with a speed of 23Gbps compared to 21Gbps for the other two cards. It’s worth noting that both leaked iterations of the RTX 4080 actually have slightly less memory bandwidth than the original RTX 3080, but the memory is faster across the board and that card only had 10GB of VRAM to start with.

The previous-generation Ampere architecture didn’t allow for as many CUDA cores. With Lovelace, we’re getting 9728 cores (and 7680 for the slimmed-down 12GB model). This means that the 16GB model of the RTX 4080 has more than the 3080 did, but the 12GB model does not. For the uninitiated, ‘CUDA cores’ is just Nvidia’s name for graphics processor cores; naturally, more is better.

When it comes to power requirements, there has been a lot of fear around these cards being seriously power-hungry. Older leaks pointed to power requirements upwards of 650W, but it now looks like the 16GB RTX 4080 will have a much more reasonable base TGP of 340W (285W for the 12GB model). When it comes to overclocking, the ceiling appears to be 516W.

Speaking of overclocking: the factory boost clock for these two GPUs will be 2,505MHz and 2,610MHz. Surprisingly, the faster card is actually the 12GB version, though this is likely because the beefier 16GB model will need to throttle sooner to avoid overheating. Either way, these are massively higher than the original 1710MHz boost clock of the RTX 3080.

box of a rtx 4090

(Image credit: Wccftech)

Nvidia GeForce RTX 4080 design

  • Quad-slot cooling design with three fans as standard
  • No leaked images of Founder’s Edition cards yet
  • Expected to be larger and heavier than anything we’ve seen yet

It’s impossible to judge at this point which image leaks of purported RTX 4000-series GPUs are real – and so far, they’ve all been focused on the flagship RTX 4090. Our US computing editor John was rather suspicious of the fonts used on some supposed Zotac Gaming RTX 4090 packaging.

Still, we can expect a seriously chunky RTX 4080, which will presumably use the same new quad-slot design seen in leaked RTX 4090 pictures. One leaked RTX 4090 from Galax supposedly uses four fans (a standard three-fan design, with an extra fan attached to the rear) along with a meaty heat sink.

We’ve yet to see Founder’s Edition (FE) versions of these cards leak properly, so the exact design of these is unclear at this point (although various renders have been circulating).

Naturally, every third-party manufacturer will have their own style when it comes to the exterior casing of the card, so if you think Gigabyte GPUs look better than MSI ones, go with your gut. Either way, these new GPUs are going to mean trouble for PC builders, as they’ll be bigger, heavier, and more cumbersome than anything we’ve seen before.

Remember that everything we’ve discussed here is pure conjecture at this point; we’ve collated our most trusted and believable leaks to form this report, but there’s no way of confirming which nuggets of information are true and which are hogwash. We’ll be back to update this page on September 20 once we have all the news from Nvidia!

Source

Be the first to comment

Leave a Reply

Your email address will not be published.


*


twelve + thirteen =