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

[80-watt Hamster]

no i have sources

Well, it's clear your sources are biased. It'll be 14.5677438%; they're rounding up because they're AMD shills.

 

[mechtech]

"Features 16128 CUDA Cores"​

Well that is a 60% increase, I would assume it to be at least that much faster as a minimum all other things considered equal.

 

[Count von Schwalbe]

Plus a testicle (or 2), your 1st borne, left arm & right leg.... PLUS both of your Bankers' Gold & Platinum cards, hehehe

And if you have none of these except left arm and right leg?

 

[Frank_100]

test is here

 

[Flydommo]

I hope both Nvidia and AMD come up with some efficient and powerful upper mid-range graphics cards at reasonable prices. Considering there will be serious changes both in architecture and fabrication process/node, there may be some justification to hope for that. Cards like the upcoming RTX 4070 will be expensive though, my expectation is around 800 €. The end of the mining boom and the massive amounts of current gen graphics cards flooding ebay and other reseller platforms will probably help alleviate the situation. Let's see how this pans out...

 

[Vya Domus]

I know many people who have bought 3080/3090/RX 6800 (XT) with high price, and they don't use those for toy money.

I sure there are people like that but it's likely a small minority.

 

[laszlo]

unfortunately for many of us it wouldn't matter the perf. increase of the new gen over previous one (both amd &nv) as prices will be too high...

 

[Quicks]

Seems AMD next GPU might not be a multi chip design:

AMD's top next-gen GPU may not be a multi-chiplet compute monster but that's better for everyone

Once rumoured to be a dual GPU chiplet design, it now seems as though Navi 31 is going to have a single, powerful compute die at its heart.

www.pcgamer.com

 

[Vayra86]

The second sentence starts with the word "Allegedly" which means that this is unverified and potentially utter bollocks.

I don't think TPU needs to spoonfeed people with nanny-state disclaimers on everything. If people can't interpret basic English and think with a little autonomy, perhaps they should get their news from somewhere more dumbed-down

I hear Twitter is a great place to go for that level of intelligence.

test is here

If you look at the moment that video was made, and how true it actually is, the man is a fucking genius

 

[Nephilim666]

I'm sure I'm not alone thinking of going AMD next round if it's competitive. Their FSR 2.0 is a viable competitor to DLSS and even though I have an ax1600i the rumoured higher efficiency is a big benefit in this age of crazy electricity prices.
Plus any opportunity to knock ngreedia down a peg

 

[arni-gx]

wow......so..... MSI rtx 4090 24gb trio gaming x, 1x16pin, with TDP 600w..... nice.....

but, what about rtx 4080, is it will using VRAM 16gb or 20gb ??

 

[Ruru]

wow......so..... MSI rtx 4090 24gb trio gaming x, 1x16pin, with TDP 600w..... nice.....

but, what about rtx 4080, is it will using VRAM 16gb or 20gb ??

4080 Ti is rumoured to have 20GB/320-bit and 4080 using a different GPU with 16GB/256-bit memory.

 

[Chrispy_]

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.

$1000 in 2013 is $1550 in today's money including the US-China trade tarrifs that didn't exist back then.

So they were effectively $1550 in a hypothetical world that doesn't have Etherium or COVID.

 

[Daven]

$1000 in 2013 is $1550 in today's money including the US-China trade tarrifs that didn't exist back then.

So they were effectively $1550 in a hypothetical world that doesn't have Etherium or COVID.

Don’t forget TSMC and Samsung raising foundry prices.

 

[Anymal]

2.0x means up to 1.7x in nvidia terms

 

[Assimilator]

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.

TPU hasn't bothered with the truth for a while now. bta is the king of yellow tech journalism and apparently other staff members are picking up his loathesome habits.

The second sentence starts with the word "Allegedly" which means that this is unverified and potentially utter bollocks.

I don't think TPU needs to spoonfeed people with nanny-state disclaimers on everything. If people can't interpret basic English and think with a little autonomy, perhaps they should get their news from somewhere more dumbed-down

No, TPU - as with any site that reports on the news - has a moral obligation to report truthfully. That means in the headline, not just the body. And it's especially important since most people only read the headline anyway. Don't make excuses for scummy behaviour.

 

[V3ctor]

I *would* sell my 3080 if the 4080 was in the same wattage ball park and have a noticeable improvement.

This is already a 320W gpu, Im not interested in anything that hogs much more power than this.

Lets see what appears from both AMD and nVidia, since Intel is *maybe* launching their Arc GPUs years late. Since they cant compete against the new stuff...

 

[GunShot]

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.

Cool. It is always great to witness someone being "right" about a /unconfirmed\ RUMOUR that has not been officially announced by its owner.

 

[razaron]

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

People still pay attention to MSRPs?

Hopefully these next gen cards drop the true/scalper price of current/prev gen cards. I can dream I guess...

 

[qubit]

I wish they'd make a bottom end "RT 4010". These tiddly cards are good for testing, general playing around with and comparing their performance with their big brothers, enthusiast style. Also, they'd be much improved over their older counterparts.

 

[Chrispy_]

People still pay attention to MSRPs?

Hopefully these next gen cards drop the true/scalper price of current/prev gen cards. I can dream I guess...

They always have done.
MSRPs tell you whether you're being scalped the retailer as well as the manufacturer!

 

[xorbe]

Power numbers are out of control, whatever the best card is under 200W is what I'll get for the 4000 series. I will revert from 2160p to 1440p if I have to.

 

[GunShot]

If there are twice as fast, well, the new refreshes are DOA for the next-gen games at 4K/60+ w/RT+ enabled.

The TRUE next-gen GPUs needs to be at least 175% faster just to get steady frames in today's games.

 

[gasolina]

i just need a 3090 performance level at 150w instead of 600w 4090 30%-40% performance leap from 3090 instead of making same performance 1/2 power consumption GPU makers like nvidia amd alike produce power hungry for what ? for sharing profit with Psu manufacturers...........

 

[THU31]

i just need a 3090 performance level at 150w instead of 600w 4090 30%-40% performance leap from 3090 instead of making same performance 1/2 power consumption GPU makers like nvidia amd alike produce power hungry for what ? for sharing profit with Psu manufacturers...........

Did you even read this "rumor" article?

I never understood where those 600 W or more rumors came from. TSMC 5N has triple the transistor density over Samsung 8N. And the 40 series might even be made on the 4N process.

So with "only" 16128 CUDA cores (much less than double over 3090), a 450 W TDP should mean a huge increase in the number of RT cores, and probably very high clocks.

Maybe the 4090 Ti can get to 600 W, with more shaders and even crazier clocks. But if it costs $3000, whether the TDP is 300 W or 600 W makes no difference, as only a few people will buy it anyway.