NVIDIA GeForce RTX 4090 Twice as Fast as RTX 3090, Features 16128 CUDA Cores and 450W TDP

[AleksandarK]

NVIDIA's next-generation GeForce RTX 40 series of graphics cards, codenamed Ada Lovelace, is shaping up to be a powerful graphics card lineup. Allegedly, we can expect to see a mid-July launch of NVIDIA's newest gaming offerings, where customers can expect some impressive performance. According to a reliable hardware leaker, kopite7kimi, NVIDIA GeForce RTX 4090 graphics card will feature AD102-300 GPU SKU. This model is equipped with 126 Streaming Multiprocessors (SMs), which brings the total number of FP32 CUDA cores to 16128. Compared to the full AD102 GPU with 144 SMs, this leads us to think that there will be an RTX 4090 Ti model following up later as well.

Paired with 24 GB of 21 Gbps GDDR6X memory, the RTX 4090 graphics card has a TDP of 450 Watts. While this number may appear as a very power-hungry design, bear in mind that the targeted performance improvement over the previous RTX 3090 model is expected to be a two-fold scale. Paired with TSMC's new N4 node and new architecture design, performance scaling should follow at the cost of higher TDPs. These claims are yet to be validated by real-world benchmarks of independent tech media, so please take all of this information with a grain of salt and wait for TechPowerUp reviews once the card arrives.


View at TechPowerUp Main Site | Source

 

[Chrispy_]

MSRP is going to be a shocker, I'm sure.
I'm going to guess $2499, like the old Titan RTX....

 

[Dr. Dro]

I'm taking this as a massive rumor, and I mean absolutely massive. The "leakers" are working in the dark and their claims to specifications seem to change practically every day.

I do believe that Ada will be ~40-50% faster than Ampere, but mostly because Ampere is also ~40-50% faster than Turing (GA102 to TU102 comparison), which in turn is about ~40% faster than Pascal, as well (assuming APIs that Pascal can run at all and excluding advanced DX12 features). Basically, 2080 Ti = 3070 (or a really good 3060 Ti model) = 4050.

MSRP is going to be a shocker, I'm sure.
I'm going to guess $2499, like the old Titan RTX....

That will depend entirely on how Navi 31 is priced and whether AMD can meet performance expectations and feature set on Radeon, though. NVIDIA isn't as invincible as it once was.

 

[R0H1T]

I'm going to guess $2499, like the old Titan RTX....

Or how about "just" 6k for the 2020 era TitanicZ

NVIDIA Announces the GeForce GTX TITAN-Z

Here it is, folks, the fabled monster dual-GPU graphics card from NVIDIA, based on its GK110 silicon, the GeForce GTX TITAN-Z (sounds like "Titans"). The first reference-design graphics card to span across three expansion slots, the GTX TITAN-Z features a cooler design that's an upscale of the...

www.techpowerup.com

 

[thesmokingman]

Not one but two kidneys...

 

[Punkenjoy]

It's important for Nvidia to have the fasted GPU on the market. The price do not really matter.

 

[Colddecked]

I'm taking this as a massive rumor, and I mean absolutely massive. The "leakers" are working in the dark and their claims to specifications seem to change practically every day.

I do believe that Ada will be ~40-50% faster than Ampere, but mostly because Ampere is also ~40-50% faster than Turing (GA102 to TU102 comparison), which in turn is about ~40% faster than Pascal, as well (assuming APIs that Pascal can run at all and excluding advanced DX12 features). Basically, 2080 Ti = 3070 (or a really good 3060 Ti model) = 4050.



That will depend entirely on how Navi 31 is priced and whether AMD can meet performance expectations and feature set on Radeon, though. NVIDIA isn't as invincible as it once was.

Doesn't seem like Nvidia cares too much about how their competitors are pricing. 3090s still sell for >1.5x more than 6900xt. Top end 30 series has been pretty resistant to AMD pricing due to their potential uses outside of gaming.

 

[Arkz]

Notice how your source says " Please note that this post is tagged as a rumor." at the top? Maybe you should do the same, as this could be utter bollocks.

 

[Metroid]

I have been saying in the past 12 months or so this would be on 4nm and trolls said 5 or 6 nm. I was right once again.

 

[ModEl4]

In what way is he disappointed with RDNA3?

 

[Zareek]

MSRP is going to be a shocker, I'm sure.
I'm going to guess $2499, like the old Titan RTX....

Is that all, I was thinking at least $3000. Knowing Ngreedia, I'd be shocked if it was less $2500.

 

[ARF]

In what way is he disappointed with RDNA3?

Probably has no clue, just spamming the twitter space...

 

[Ruru]

MSRP is going to be a shocker, I'm sure.
I'm going to guess $2499, like the old Titan RTX....

Funny how back in the day almost nobody bought Titans when they were 1000EUR/USD, but now a the cut-down cards (even 3090) sell like hot cakes for far beyond that grand.

 

[Legacy-ZA]

I have been saying in the past 12 months or so this would be on 4nm and trolls said 5 or 6 nm. I was right once again.

I am expecting around the 3000Mhz clock frequency.

 

[Solaris17]

Compared to the full AD102 GPU with 144 SMs

Is this confirmed?

 

[windwhirl]

It better be twice as fast with 50% more cores and the TDP going sky high.

 

[ARF]

Don't forget these numbers:

NVIDIA AD102 'Ada Lovelace' & AMD Navi 31 'RDNA 3' GPUs To Be Power Chugging Beasts, Over 400W Rumored (wccftech.com)

 

[DeathtoGnomes]

Wasnt there a leak on this very thing not too long ago with the same information, the news that got everyone's panties in a bunch over 900w power draw.

So how is this different than that? both are still subject to speculation.

 

[Tomgang]

Rumors should all ways be taken with a grain of salt.

But no matter what is true and what is not. I am looking for to see some real numbers on spec and performance, not just rumors.

If I will be getting a rtx 4080 none TI properly over my current rtx 3080, depending on price, performance and the economic outcome in the coming months ahead.

 

[Daven]

In what way is he disappointed with RDNA3?
View attachment 247663

I was wondering the same thing. Pretty random statement embedded in that tweet.

 

[Frick]

I was wondering the same thing. Pretty random statement embedded in that tweet.

RDNA3 will beat Lovelace with about 15%.

 

[HisDivineOrder]

I was wondering the same thing. Pretty random statement embedded in that tweet.

I suspect it has to do with the original promise of dual-chips and now rumors are saying it's back to plain old monolithic just like always. If you were on the cusp of something cool and interesting but then watch as it returns back to business as usual...

Well, that'd be disappointing imo.

 

[mouacyk]

In what way is he disappointed with RDNA3?
View attachment 247663

It's one of the 5% bots

 

[ModEl4]

Don't forget these numbers:

View attachment 247674
NVIDIA AD102 'Ada Lovelace' & AMD Navi 31 'RDNA 3' GPUs To Be Power Chugging Beasts, Over 400W Rumored (wccftech.com)

While i agree with the 2.2X figure for AD102 vs GA102 (4090Ti vs 3090Ti), i think the 2.78X (2.5X/0.9) for RDNA3 vs RDNA2 will die a miserable death imo.
This will be just the TFlops difference, it has nothing to do with performance difference.
I would be very surprised if full Navi 31 at 4K is more than 2.3X-2.4X vs full Navi 21 at launch (although after 1-1.5 year if they refresh with 24Gbps Samsung ICs we should see a highly OC 7950X to reach 2.4X-2.5X vs a highly OC 6950X so essentially matching the highest AD102 refresh part of that time, at launch Nvidia will not be matched imo)

 

[Vya Domus]

Funny how back in the day almost nobody bought Titans when they were 1000EUR/USD, but now a the cut-down cards (even 3090) sell like hot cakes for far beyond that grand.

Back then they weren't bought by miners.