GeForce RTX 4090 was overclocked to 3.0GHz in Nvidia's lab

Daniel Sims

Posts: 1,335   +43
Staff
In context: A brief comment during Nvidia's announcement of the RTX 4000 series graphics cards sheds some light on the flagship model's overclocking potential. It isn't unprecedented, strictly speaking, but it could push some new boundaries for the latest GPU tech, depending on future details, such as what kind of cooling it required.

While presenting the new Ada Lovelace GPU series, Nvidia CEO Jensen Huang claimed the company overclocked the GeForce RTX 4090 to 3.0GHz in its labs. What that means could depend on information that might not come out until the card is in users' hands. Leaked 4090 benchmarks reported the same number earlier this month without further details.

Huang didn't dwell on the comment before moving on (about 16 minutes into the video above), so we don't know what kind of cooling or how much wattage Nvidia's labs used to reach 3.0GHz. That number is slightly under last year's standing GPU overclock world record of 3.2GHz that an AMD Radeon RX 6900 XT managed through liquid nitrogen cooling.

Older GPUs have topped 3.0GHz using similar methods, but the 4090 reaching that point with its incredible 76 billion transistors is an outstanding milestone. It would be even more impressive if it turns out that Nvidia got there with air cooling.

Overclocking also raises some power-draw concerns that the Lovelace announcement hushed initially. Many users feared the RTX 4090 could require 600W — a worrying number for those living under rising energy bills. Later rumors brought that number down before Nvidia confirmed the 4090 only needs 450W — the same as the 3090Ti. However, could the 4090 need 600W to reach 3GHz? Its official boost clock is 2.52GHz.

The GeForce RTX 4090 hits shelves on October 12, starting at $1,599. The 12GB and 16GB variants of the RTX 4080 arrive in November for $899 and $1,199, respectively. The cards feature third-generation RT cores, AV1 encoding (this year's new GPU generations from all three manufacturers will add the feature), DLSS3, and other improvements over the RTX 30 series.

Critics immediately challenged Nvidia over Lovelace's high prices, as overall GPU demand is trending downward after remaining at historic levels over the last two years. However, Huang defended the MSRP, declaring Moore's Law dead.

"The idea that the chip is going to go down in price is a story of the past," he said.

Permalink to story.

 
"The idea that the chip is going to go down in price is a story of the past," he said.
I understand that prices are going up, everythings has been going up. However, the price performance for this new series is lower than what the 30 series is at right now at powerlevels no one wants to deal with. They're making a product that many people don't want and certainly can't afford. I know that they will sell many of them, but I doubt these halo products are going to sell like crazy. We are entering a recession and people just don't have money for $1000 video cards that are going to require higherend powersupplies.

now, AMD has been quiet about their prices, performance and power consumption numbers so I have a feeling they are in a similar boat as nVidia. It's just that nVidia has been taking all the heat and I feel AMD is more than happy to let them take all the bad PR.
 
TL;DR: "We can OC it to 3GHz but we can't release it because high quality PCIe 5.0 PSUs are not readily available on the market to power them. "
 
I understand that prices are going up, everythings has been going up. However, the price performance for this new series is lower than what the 30 series is at right now at powerlevels no one wants to deal with. They're making a product that many people don't want and certainly can't afford. I know that they will sell many of them, but I doubt these halo products are going to sell like crazy. We are entering a recession and people just don't have money for $1000 video cards that are going to require higherend powersupplies.

now, AMD has been quiet about their prices, performance and power consumption numbers so I have a feeling they are in a similar boat as nVidia. It's just that nVidia has been taking all the heat and I feel AMD is more than happy to let them take all the bad PR.
Looking at the RDNA2 launch, it is possible that AMD is gauging performance and price from Nvidia (since Nvidia tends to launch earlier), before determining how they should price their products. Cost wise, I believe it may be cheaper for AMD to manufacture their chips as compared to the likes of Nvidia. The reasons are,
1. AMD's been a longer and more stable strategic partner, whereas Nvidia tend to source for the cheaper alternative. By having to squeeze their way into the backlog of orders, I am sure they paid quite a bit to jump the queue,
2. Monolithic vs chiplets - While both are complex chips, the chiplet design is likely easier to make and lower in cost.
Thus, I am not expecting AMD to price their RDNA3 GPUs more than Ada Lovelace, but that may depend on the performance level it offers. For me, most Ampere users may not find the 4000 series attractive, mostly because of the price. The numbers flashed out by Nvidia are really shady when they generally show scenarios where DLSS 3.0 is enabled to obtain an impressive x times performance improvement. When marketing deliberately tries to skew the results, it just creates more doubts about the actual performance without these bells and whistles. A decent performance uplift is expected, but I don't believe it is 2x or more as painted by Nvidia in their marketing slides.
 
Funny how many people believe or even think AMD is waiting for anything. Decisions like pricing was made long ago. They aren't waiting for anything. They know what the price will be n when. All of this has been factored in for months if not close to a year. No manufacturer waits till the last minute to make a decision on pricing n when.

They are as most do, get some benefits when dealing with new tech when a competitor comes out first. Sure they get to watch n see but again they already have a idea of what their competitor is doing n they have a answer to it. Will it be the same or better than, that is what everyone is waiting n wanting to know. AMD knows n I bet nVidia does too. AMD will likely have similar pricing, maybe a little cheaper but nothing crazy. It's not going to happen cause AMDs pricing isn't going to be $200 cheaper with similar tech. If AMDs cards turn out to be better even just a lil bit that may make nVidia drop prices but if nVidia still holds the KING crown, don't expect any price drops anytime soon.
 
Funny how many people believe or even think AMD is waiting for anything. Decisions like pricing was made long ago. They aren't waiting for anything. They know what the price will be n when. All of this has been factored in for months if not close to a year. No manufacturer waits till the last minute to make a decision on pricing n when.

They are as most do, get some benefits when dealing with new tech when a competitor comes out first. Sure they get to watch n see but again they already have a idea of what their competitor is doing n they have a answer to it. Will it be the same or better than, that is what everyone is waiting n wanting to know. AMD knows n I bet nVidia does too. AMD will likely have similar pricing, maybe a little cheaper but nothing crazy. It's not going to happen cause AMDs pricing isn't going to be $200 cheaper with similar tech. If AMDs cards turn out to be better even just a lil bit that may make nVidia drop prices but if nVidia still holds the KING crown, don't expect any price drops anytime soon.
AMD isn't going to be $200 cheaper on parts of performance level parts, but they would happily match nVidia's price at different performance levels. AMD likely has a minimum and median price they'd like to sell at but if nVidia starts selling cards for $200 more than those price they'd be fools not to raise theirs. They are a business after all.

And there are plenty of last minute changes of all kinds of things in products. Just like if AMD was $200 higher than nVidia they'd price themselves out of the market.
 
I understand that prices are going up, everythings has been going up. However, the price performance for this new series is lower than what the 30 series is at right now at powerlevels no one wants to deal with. They're making a product that many people don't want and certainly can't afford. I know that they will sell many of them, but I doubt these halo products are going to sell like crazy. We are entering a recession and people just don't have money for $1000 video cards that are going to require higherend powersupplies.

now, AMD has been quiet about their prices, performance and power consumption numbers so I have a feeling they are in a similar boat as nVidia. It's just that nVidia has been taking all the heat and I feel AMD is more than happy to let them take all the bad PR.
I am sure AMD will match Nvidia on performance tiers and their power consumption according to all the rumors will be much lower than Nvidia.

I don't expect any better pricing though.....sigh.
 
They are really looking for anything to make this upcoming generation of cards look good.

"B..but it's overclocked in a lab"

My GPU is overclocked to 6GHz in my lab, what now?
 
The most disappointing part is that nVidia seem to think that they can still charge mining boom prices in the current situation.

Heck, Ada‘s hash rate might justify prices, just that it doesn‘t matter (as much) now.

Thinking about it, the most disappointing part will be loyal customers still buying.
 
2. Monolithic vs chiplets - While both are complex chips, the chiplet design is likely easier to make and lower in cost.

Two factors here:

- smaller dies, so lower cost and higher yields
- only the graphics die (GCD?) is on 5nm, the other dies are on a cheaper 6nm process, so this should lower cost even more.

If RDNA3 does use less power (we‘ll see) for the same performance, this will also mean overall lower build cost.
 
AMD isn't going to be $200 cheaper on parts of performance level parts, but they would happily match nVidia's price at different performance levels. AMD likely has a minimum and median price they'd like to sell at but if nVidia starts selling cards for $200 more than those price they'd be fools not to raise theirs. They are a business after all.

And there are plenty of last minute changes of all kinds of things in products. Just like if AMD was $200 higher than nVidia they'd price themselves out of the market.
The point is these things are done long before a product comes out no matter what the competition does or is. And no changes do not happen right away. Again AMD already has a plan and won't be changing it. Maybe a couple months in but nothing right away. Which we have seen this happen before, that's nothing new.
 
The point is these things are done long before a product comes out no matter what the competition does or is. And no changes do not happen right away. Again AMD already has a plan and won't be changing it. Maybe a couple months in but nothing right away. Which we have seen this happen before, that's nothing new.
that's not how business works, adapt or die. Companies will often sell products at a loss rather than price them too high and lose their entire investment
 
that's not how business works, adapt or die. Companies will often sell products at a loss rather than price them too high and lose their entire investment
Not about selling at a loss. MS sold its consoles at loss to get its foot into the console market. Which it had the money n brains to do so. Not everyone can or have the money to do so. This is about they already have a set plan. That is how business works. They just don't go willy nilly, oh my competitor did this let's rush to change it right away. They look at everything n see if changes need to be made but these are done afterwards not some last second decision. No major company works this way.
 
Not about selling at a loss. MS sold its consoles at loss to get its foot into the console market. Which it had the money n brains to do so. Not everyone can or have the money to do so. This is about they already have a set plan. That is how business works. They just don't go willy nilly, oh my competitor did this let's rush to change it right away. They look at everything n see if changes need to be made but these are done afterwards not some last second decision. No major company works this way.
MS sold the xbox at a loss because it's a software company. The xbox locked users into their software platform and created a reoccurring payment system. Everything you're saying is basically nonsense. There are long term targets but there are market outliers that need to be addressed don't hours before a product launch. A company's most important asset is it's ability to adapt to market changes. Just look at Blockbuster, they made fun of Netflix. Kodak used to be one of the most profitable companies in the world and they made fun of digital cameras. Heck, even blackberry missed the smartphone market by a year and they couldn't even give them away.

I could go on, I even have personal examples I could give within my industry of stuff literally being changed within minutes of completion. We had supply chain issues when we were suppose to complete a project that day and everything that showed up was unusable. Even things planned years in advance change constantly for reasons outside of their control. You either adapt and fix the issue or you fail, it's as simple as that.
 
MS sold the xbox at a loss because it's a software company. The xbox locked users into their software platform and created a reoccurring payment system. Everything you're saying is basically nonsense. There are long term targets but there are market outliers that need to be addressed don't hours before a product launch. A company's most important asset is it's ability to adapt to market changes. Just look at Blockbuster, they made fun of Netflix. Kodak used to be one of the most profitable companies in the world and they made fun of digital cameras. Heck, even blackberry missed the smartphone market by a year and they couldn't even give them away.

I could go on, I even have personal examples I could give within my industry of stuff literally being changed within minutes of completion. We had supply chain issues when we were suppose to complete a project that day and everything that showed up was unusable. Even things planned years in advance change constantly for reasons outside of their control. You either adapt and fix the issue or you fail, it's as simple as that.
You just spoke nonsense. That is not why MS sold them at a loss. I literally told you why. They had the money from their software to do a console, take a the loss so that they could show everyone they could do it n compete. They did. Game Pass is one of the biggest things to happen in gaming.

Sounds like you just don't like MS. That's fine if you don't, nothing to do with anything tho.
 
Back