Follow Slashdot blog updates by subscribing to our blog RSS feed

 


Forgot your password?
Close
typodupeerror
×
Graphics Hardware

Nvidia's GeForce RTX 5090 Laptop Graphics Benchmarks Revealed 30

Posted by BeauHD from the best-of-the-best dept.
MojoKid writes: Similar to Nvidia's recent desktop graphics launches, there are four initial GeForce RTX 50 series laptop GPUs coming to market, starting this month. At the top of the stack is the GeForce RTX 5090 laptop GPU, which is equipped with 10,496 CUDA cores and is paired to 24GB of memory. Boost clocks top out around 2,160MHz and GPU power can range from 95-150 watts, depending on the particular laptop model. GeForce RTX 50 series GPUs for both laptops and desktops feature updated shader cores with support for neural shading, in addition to 4th gen ray tracing cores and 5th gen Tensor cores with support for DLSS 4. The GeForce RTX 50 series features a native PCIe gen 5 interface, in addition to support for DisplayPort 2.1b (up to UHBR20). These GPUs are also fed by the latest high speed GDDR7 memory, which offers efficiency benefits that are pertinent to laptop designs as well. Performance-wise, NVIDIA's mobile GeForce RTX 5090 is the new king of the hill in gaming laptops, and it easily bests all other discrete mobile graphics options on the market currently.

Nvidia's GeForce RTX 5090 Laptop Graphics Benchmarks Revealed More | Reply

Nvidia's GeForce RTX 5090 Laptop Graphics Benchmarks Revealed

Comments Filter:

  • The target market will be AI programmers (Score:3)

    by rsilvergun ( 571051 ) on Thursday March 27, 2025 @04:46PM (#65263613)
    But at that market 24 GB doesn't seem like enough RAM. Really they want at least 64 and I don't understand why Nvidia refuses to give it to them except maybe they make some stupidly expensive devices I don't know about that that would be competing with.

    At the moment AMD's solution is to put a mid-range chip on a laptop and solder 128 gigs of RAM directly in. On first I didn't understand why they were doing that but one of the guys here explained it was for AI programmers.

    Meanwhile the number one GPU on steam last I looked was actually the laptop 4060 because with a little bit of hunting you can get them in sub $700 laptops.
    • The RAM limit means it won't be that useful for AI developers unless they have a really small model that can fit in 24 GB of VRAM. Gamers can rejoice for a change I guess. NVidia certainly does have cards with enough memory, but they cost more than a new car. The AMD APU (I wouldn't call a APU with 16 CPU cores, 2,560 GPU cores, other dedicated AI hardware, and a 256-bit memory bus midrange) uses LPDDR5x memory which is just a low power version of the standard DDR memory that PCs use. It is soldered so you

    • i think they are pretty specifically not targetting it at AI devs, and seem to be heavily hinting that AI devs should be gettign those fancy miniturzied blackwell things they are pitching for later this year.

      I wouldnt be surprised if they do a cut down card for that as well,later this year for the market that cant aford the filthy expensive datacenter cards but wont be satisfied with a 5090.

      Honestly, it sounds like a good opportunity for AMD to fill the gap, if it gets off its arse and fills the CUDA gap.

      • Especially with how chincy Nvidia is with RAM. I guess there are a handful of other workstation uses but they don't normally market the consumer cards to those folks. They get their own SKUs. I didn't say it was a good product for AI but I think Nvidia is marketing it as such or at least intending it to be

    • I don't understand why Nvidia refuses to give it to them

      Price their hardware that does have that much memory, and understanding will arrive.

      • The problem is that when it comes to AI and related tasks, AMD couldn't find their arse with two hands and a map. So NVidia can charge whateverthefuck they want. And they do.

        • The lack of competition certainly is the problem, and it in turn was created by not enforcing antitrust law and therefore allowing mergers which were not in the public interest.

          ATI was always shit at software, and now AMD is too. They honestly were never great at it either, though.

          • I'm all for anti monopoly abuse, but that's a different discussion too!

            Don't forget that the geniuses at AMD hobble the gaming cards for compute the ML so they didn't cannibalise their non existent pro cards sales. That's all on then.

            • NVidia hobbles their consumer cards in a bunch of ways too, sadly, like not giving you GPU virtualization... which is the key to maximum graphics performance for VMs. But you're right, they are smart enough to give everyone access to CUDA. I just wish they would stop with the low-VRAM bullshit, and make some higher-memory cards available as well. No 50xx desktop card should have less than 12GB, that's ridiculous. What year is it?

              • You're not wrong and I am not going to defend NVidia at all.

                I am firmly convinced that the reason NVidia are dominant is from their policy of having all cards do the compute stuff. I can train and test on my workstation and/or laptop, and then deploy on a cloud machine happy in the knowledge it'll work fine. And of course students starting out will work on cheap cards then move into industry and already be biased towards what they know works (Microchip did this 30 years ago with PIC microcontrollers).

    • Nvidia has loads of high end options at much higher price points with loads more memory. They don't want people to have a reason not to buy the expensive datacentre AI targetted cards, they want the gaming cards to be mostly gaming but kinda capable of playing with AI.
  • Nearly all laptop computer throttle, This wastes your hard-earned money. People who buy Gaming Laptops and Mobile Workstation don't have to keep putting up with bad cooling--when usually the cooling system just needs a larger fan or heatsink, which are a pittance compared to the price of a GPU. What we are left with is hot, throttling laptop computers that need a GPU repasting/recompunding every 2 years. All discrete laptops should have their CPU/GPU cooling capacity listed, and enforced.
    • Pretty sure proper cooling of a high end laptop with a 5090 doesnt exist because at some point it stops being a laptop when you try.

      I dont see why this laptop with inadequate cooling shouldnt exist. Do your research. If they arent selling what you want, dont complain if you purchase it anyways, because I got no sympathy.

      Some people do want this thing, for sure.

      The premise that adequate cooling for such a thing exists is highly dubious on the face of it. At some point its on you.

    • Not quite. It's not a waste of money when something is limited by some factor other than the one you're interested in on your desktop PC. I buy a laptop for its size and battery life. Throwing a bigger heatsink on it to improve performance would be in my opinion *THE WORST* way to do it. Improving part efficiency to increase speed while retaining the tiny and light form factor isn't a waste of money, it's a design trade off.

      that need a GPU repasting/recompunding every 2 years

      I've literally never repasted a laptop GPU, nor do I know a single person who has.

      All discrete laptops should have their CPU/GPU cooling capacity listed, and enforced.

      Wh

      • Re: (Score:1)

        by Anonymous Coward
        I think the point here is, there is no point buying a laptop like this as the capability of the GPU is artificially throttled by the cooling capacity. In effect you are paying significantly more for a machine that is not capable of fully utilising the GPU capacity outside of a few mins while it heats up. I guess if you need is "I need the GPU for 5 mins at a time a few times a day" then it probably will be fine, but I can't think of too many use cases for that.

        • Then you've missed the point along with maybe 16 years of semiconductor advancement. This isn't Intel ala Pentium 4 days. You don't get speed increase simply by increasing the power, speed, and thermal envelope linearly. The 50 series offers more performance per watt than the 40 series, meaning within the same thermal envelope you get a boost in speed. Even within throttling regime you have better performance.

          Sure it's not much of a boost in speed, and no one is going to upgrade their 40 series laptops to a

      • Why? And why would you limit it to laptops? Literally all CPUs and GPUs whether they are laptop or high end desktop PCs have this problem. It's why the watercooling crowd get a nice performance boost in benchmarks even without changing any settings on their PC. The thermal envelope is just a design feature in modern PC components. You want to know actual performance, go look up benchmarks for your particular device. If you're not happy with something return it. There's no reason to throw yet more meaningless confusing data at consumers who largely already don't know what they are looking at.

        Desktops for gamers etc are usually correctly specced, even if not, then adding a $100 3 fan radiator and cooling is easy. With a laptop this is almost impossible without creating somethign that is no longer a laptop. My desktop can fuilly utilise the GPU without the fans ever needing to go much beyond 50% regardless of how long I use it. many of these gaming laptops the GPU is limited within a few mins of heavy usage unless you have it sitting on some huge laptop cooler thus defeating the purpose of it be

        • Desktops for gamers etc are usually correctly specced

          They are specced the same way laptops are: stuck in a thermal envelope. Again this is why getting performance boosts by water cooling *without changing any system settings* is an actual thing that people experience. It's also why different cards with the same GPU are shipped with different cooling solutions.

          The spec is the spec on the laptop as it is with a desktop. When was the last time you saw a desktop advertise the PL1, PL2, PL3, or PL4 limit? Let me guess, you don't even know what they are or how they

  • laptop? (Score:3)

    by bloodhawk ( 813939 ) on Thursday March 27, 2025 @05:24PM (#65263673)
    I think they need a new name for these, unless your lap is made of asbestos or you enjoy 3rd degree burns I can't imagine every wanting to put one that is turned on, on my lap.

  • Easily beats? (Score:3)

    by alvinrod ( 889928 ) on Thursday March 27, 2025 @05:32PM (#65263695)
    ComputerBase [computerbase.de] has it losing to a 4090 mobile in some of the tests that they ran, so I don't know if I'd say that it's easily beating everything or that it's the top card. The lower TDP is clearly hurting it in some scenarios.

    • Re: (Score:3)

      by Xenx ( 2211586 )
      The problem is that when buying a laptop you're buying into the design of the laptop as a whole, and not just the individual components. For example, over at https://hothardware.com/review... [hothardware.com] it shows the MSI Raider A18 HX completely beating out the Razer 16 2025 in almost every test.
  • Why? Why would you do that? It's like running a marathon in combat boots [imgur.com]

    • On one hand, I agree. If you want to get into the big time AI stuff, you need the big time hardware.

      On the other hand, there's also a lot of demand for the lesser AI stuff, smaller models and so on, and this device is fairly future proof for that. With 24GB VRAM it will probably last through a whole other generation, in terms of what most people's machines will have by then. Keep in mind that they are still selling 8GB cards now. And then there's cellphones and tablets...

  • It is just a $6000 baby sitter.

  • Coming up next we reveal benchmarks for our V12 lawnmower.

  • If it compares favourably to a previous gen desktop then it could be an option for people upgrading from one of these if they fancy a laptop for portability but dont want to give up any performance.

  • Given the 5090 has over 21000 cuda cores this is more like a 5045 with half the compute capacity, a quarter of the power limits and around 70% the memory/bus.

    I guess on the plus side there's no 12v high fire connector to worry about because these chips are mb integrated and wouldn't draw enough power to worry about anyway.

Slashdot Top Deals

A successful [software] tool is one that was used to do something undreamed of by its author. -- S. C. Johnson

Close