Hardware leaker Harukaze5719 has shared the alleged Ada Lovelace dies that will power Nvidia's next-generation RTX 40-series Mobile GPUs and their associated power targets. Most interestingly, the leak claims that Nvidia could prepare a GeForce RTX 4090 Mobile, the first time this model number has appeared in the mobile segment. But, of course, treat the leaked information with a grain of salt.
The leak (via REHWK) reveals five GeForce RTX 40-series Mobile GPUs in total, including the GeForce RTX 4090 Mobile, GeForce RTX 4080 Ti Mobile, GeForce RTX 4070 Mobile, GeForce RTX 4060 Ti Mobile, and GeForce RTX 4060 Mobile. In addition, the mobile variant of the GeForce RTX 4090 will reportedly utilize the AD103 die, while the GeForce RTX 4080 Ti will rely on the AD104 die, whereas the AD106 silicon will serve the GeForce RTX 4070 Mobile. Finally, the GeForce RTX 4060 Mobile models could sport the AD107 die.
Power consumption for the GeForce RTX 4090 Mobile and GeForce RTX4080 Ti Mobile allegedly peak at 175W, and the GeForce RTX 4070 Ti Mobile tops out at 140W. Unfortunately, the leaker didn't provide the TDP for the GeForce RTX 4060 Mobile and GeForce RTX 4060 Ti Mobile SKUs. If these figures are accurate, the GeForce RTX 4090 should be a massive upgrade in the mobile space, providing desktop GeForce RTX 3090 or GeForce RTX 3080-like performance at just 175W. To keep performance up at these lower TDPs, Nvidia may have to unlock all cores on the AD103 for maximum performance and efficiency compared to the desktop GeForce RTX 4080 16GB, which will come with a slightly neutered AD103 die.
As for the GeForce RTX 4080 Ti Mobile, we should expect this GPU to be a significant downgrade in performance compared to the GeForce RTX 4090 Mobile due to the rumored die change from AD103 to AD104. Since AD104, with all cores enabled, features a 40% reduction in core count, essentially, the GeForce RTX 4080 Ti Mobile will be the mobile variant of the now deleted GeForce RTX 4080 12GB SKU. However, performance will be dictated heavily by clock speed in the mobile segment since clock speeds have the chance to fluctuate significantly due to the considerably lower TDPs required on mobile SKUs. With a 40% reduction in core count on AD103, there's a likely chance we'll see far higher clock speeds on the GeForce RTX 4080 Ti Mobile compared to the GeForce RTX 4090 Mobile.
GeForce RTX 40-Series Mobile Specifications*
|GeForce RTX 4090 Mobile
|GeForce RTX 4080 Ti Mobile
|GeForce RTX 4070 Mobile
|GeForce RTX 4060 Ti Mobile
|GeForce RTX 4060 Mobile
*Specifications are unconfirmed.
Unfortunately, we don't have enough details on AD106 or AD107 to know how fast or slow the GeForce RTX 4060 (Ti) Mobile and GeForce RTX 4070 Mobile will be. But we expect all to pack a noticeable performance upgrade over their GeForce RTX 30-series counterparts. What is interesting, though, is the potential introduction of the GeForce RTX 4060 Ti Mobile so early in the mobile pipeline, with both the GeForce RTX 4060 Mobile and Ti variant GPUs using the entry-level AD107 die instead of AD106. It differs from 30 series models, where the xx60 series cards operated on the GA106 die.
It would suggest that AD107 will be an exceptionally powerful die on Ada Lovelace. Still, we suspect this die change is only due to the GeForce RTX 4090 Mobile being introduced to the product stack, forcing other models to go down a notch in core count and GPU dies.
Overall, Ada Lovelace mobile is looking very strong based on the power consumption targets and die models suggested in the leaks. GeForce RTX 4070 Mobile and higher SKUs will feature noticeably boosted power targets over GeForce RTX 30-series GPUs, with the GeForce RTX 4090 Mobile having the possibility to match the desktop GeForce RTX 3090 or GeForce RTX 3080 gaming performance. But we'll have to wait and see the spacing between these GPUs once Nvidia releases them. Nvidia's GeForce RTX 40-series Mobile GPUs still have a 50% effective nerf to power targets over RTX 30-series desktop GPUs, which can hugely impact performance. It's all down to Ada Lovelace's efficiency on the TSMC 4n node at these "very low" power targets.
Stay on the Cutting Edge
Join the experts who read Tom's Hardware for the inside track on enthusiast PC tech news — and have for over 25 years. We'll send breaking news and in-depth reviews of CPUs, GPUs, AI, maker hardware and more straight to your inbox.
Aaron Klotz is a freelance writer for Tom’s Hardware US, covering news topics related to computer hardware such as CPUs, and graphics cards.
I was considering building another desktop computer but if they're going to come out with a 4090 laptop version then I might just buy a gaming laptop. They're easy to move around and they're usually much more efficient albeit being less powerful.Reply
But from what we've seen from reviews they really push the 4090 graphics card pretty high and there's room to cut off and still be a very good performer.
But when they say 4090 laptop version don't always think that you're going to get a 4090 chip.
I bought an RTX 3080 laptop only to find out that it was just a RTX 3060 full chip which they just called a 3080 laptop GPU version.
I wasn't sure how to feel about that when I found out but the thing performs as good as a RTX 2080 Ti, in a laptop, so I was quite happy with the performance.
My laptop kills it 1440p and does well in 4k. That's been hard to do even with gaming laptops for years.
Therefore, I hope these RTX 4090 laptops come out soon because it might be worth investing in that rather than build a $4,500 desktop. The main reason is not because the performance will be as much as a 4090 desktop but because it will have all the features of the 4090 in a laptop. Dlss 3, frame generation, and all the encoders that you would want plus hopefully a whole bunch of vram. I hope they don't cut back on the vram like they did with the RTX 3080 mobile.
Anyway, that's all I got to say for now. I don't really like leak information but this has got me a little excited.
legit dumb imho.Reply
you know laptop makers wont go back to dummy thicc laptops for betetr cooling. (thin is i beign the thing)
the pwr/thermals of 4090 would be unmanageable in a laptop (w/o severely cutting down on performance) and you'd be pwr limited as well as thermal limited.
der8auer EN did some testing by lowering the power limit and at 50% power is only loses around 20% in 3DMTSE.Reply
If you reduce voltage, and lose 20% performance, WHY O WHY are you wanting a 4090?thisisaname said:der8auer EN did some testing by lowering the power limit and at 50% power is only loses around 20% in 3DMTSE.
How does the actual user facing performance compare to a "lesser" GPU?
60yFji_GKak:1180USAFRet said:If you reduce voltage, and lose 20% performance, WHY O WHY are you wanting a 4090?
How does the actual user facing performance compare to a "lesser" GPU?
If they had kept power to 300W it would still be 35-75% better, rather than 40-80% better with a 450w power draw.
Plus I was showing how a 4090 mobile is not so mad as the power draw of the 4090 desktop version would indicate.
thisisaname said:der8auer EN did some testing by lowering the power limit and at 50% power is only loses around 20% in 3DMTSE.
Even with a 50% power drop, its still well over 200W..... Vast majority of laptops do not have such cooling capacity.... Anyone that does will have the size and weight of a desktop instead. I guess the power supply has to be at least 400W and would be huge....
I was just pointing out it was possible to do not weather it was a good idea.escksu said:Even with a 50% power drop, its still well over 200W..... Vast majority of laptops do not have such cooling capacity.... Anyone that does will have the size and weight of a desktop instead. I guess the power supply has to be at least 400W and would be huge....
Comparing broadly the desktop/laptop version of each card 3060 is a bout 10% faster, 3070 about 40% faster and the 3080 about 60% faster.
Could not find anything for the 3090 way to much power.
If it scale like that a 4090 in a laptop is going to be so cut down and power restricted as to be only good in terms of marketing.
Looking at the power and assuming that 40% cut down die and 175 would scaling that up laptop>desktop we get about 437.5w (which is less that a 4090 desktop card of 422w)
Or we go with 70% power (drops desktop power from 422 to 329 score drops from 123,49 to 117,08 about 5%) the die only needs to be cut down to 55%
If we go down to 60% power(or 284w) the die would only need to be cut down to about 60%. This could have about 54% of the desktop 4090 performance
All the way down to 50% ( 234w) and the die cut down is 75% and this could be around 60% of the desktop part.
While the 50 and 60 power limit end up around the nearly the same assuming my assumption are right the core count of the 50% reduction would be higher and Nvidia does have a history of hiding the power of the GPU on mobile parts. That core count is going to be in big number the power in the small print.