Now that it’s been over two years because the mighty RTX 4090 set a brand new normal for high-end gaming GPUs, we count on the Nvidia GeForce RTX 5090 to reach quickly, and unseat it because the quickest graphics card you should buy. As at all times within the lead-up to a brand new graphics card launch, we’ve been monitoring the regular drip of leaks and rumors about specs, value, and efficiency, which have solely accelerated in latest months as we transfer towards the anticipated launch date of the cardboard.
Whereas we look ahead to the brand new flagship GPU from Nvidia to reach, and little question take the highest award as the most effective graphics card cash can purchase, you’ll discover all the knowledge we at present know in regards to the new graphics card under.
Nvidia GeForce RTX 5090 specs estimate
The newest Nvidia GeForce RTX 5090 specs rumors counsel the GPU could have 21,760 CUDA cores and 170 RT cores, with the best reminiscence bandwidth of any GeForce GPU.
| Nvidia GeForce RTX 5090 | |
| GPU | GB202-300-A1 |
| CUDA cores | 21,760 |
| SMs | 170 |
| RT cores | 170 |
| Increase clock | 2.9GHz |
| Interface | 16x PCIe 5.0 |
| VRAM | 32GB GDDR7 |
| Reminiscence interface | 512-bit |
| Reminiscence bandwidth | ~1.7TB/s |
| L2 cache | 128MB |
| Energy draw | 600W |
| Energy connector | 1 x 16-pin |
Because the RTX 5090 might be constructed on a brand new structure, it is going to be powered by one of many new GeForce RTX 50 GPU dies. Naturally, because the flagship of its household, it’s prone to be based mostly on the biggest and strongest of your complete stack, which is predicted to be referred to as GB202, with the variant within the RTX 5090 anticipated to be referred to as the GB202-300-A1.
The complete Nvidia GB202 specs are anticipated to incorporate 24,576 CUDA cores if all its components are absolutely enabled, unfold throughout 192 streaming multiprocessor (SM) models, giving it 192 RT cores. Nonetheless, in accordance with a leak from common tech leaker kopite7kimi, the RTX 5090 GPU could have 22 of its SMs disabled, giving it 21,760 CUDA cores and 170 RT cores. This leaves room for Nvidia to probably launch a brand new GeForce RTX 5090 Ti, or 5090 Extremely, card at a later date, with extra SMs enabled.
As some extent of comparability, the GeForce RTX 4090 has 128 SMs, giving it 16,384 CUDA cores, which means the RTX 5090 probably has much more parallel processing energy. Nonetheless, this isn’t an apples-to-apples comparability, because the RTX 5090 might be constructed utilizing the Blackwell structure, reasonably than the Ada Lovelace structure utilized by the 4090, and it might not have the identical efficiency per core.
One other huge rumor in regards to the RTX 5090 specs considerations the reminiscence bandwidth. In keeping with the rumor mill, the RTX 5090 reminiscence might be GDDR7, and hooked up to a super-wide 512-bit bus, which is way wider than the RTX 4090’s already fats 384-bit interface. There’s potential for an unlimited quantity of reminiscence bandwidth right here. If the GDDR7 VRAM runs at 28Gbps, for instance, then the reminiscence bandwidth may very well be 1.79TB/s, which is large in comparison with the RTX 4090’s bandwidth of simply over 1TB/s.
Not solely that, however the RTX 5090 can also be rumored to be coming with an enormous 32GB of GDDR7 VRAM, giving it loads of headroom for skilled work, in addition to gaming at very excessive resolutions with all of the settings maxed out.
There’s a potential value to all this energy, although, which is the RTX 5090 energy draw. The GPU is rumored to attract round 600W, which is identical restrict as a single 16-pin energy cable, resulting in hypothesis that the RTX 5090 wants two 16-pin energy cables. Nonetheless, that is unlikely, on condition that the cardboard may also get some energy from the PCIe slot. That stated, the additional energy required by severely overclocked RTX 5090 playing cards could push them over that cliff, and require two sockets – we’ll have to attend and see.
The GeForce RTX 5090 can also be anticipated to be Nvidia’s first gaming GPU to make use of the brand new 16x PCIe 5.0 interface. If that’s the case, you’ll nonetheless have the ability to set up it in a PCIe 4.0 and even PCIe 3.0 slot, however you gained’t get the total quantity of bandwidth, which is prone to prohibit efficiency on a PCIe 3.0 motherboard.
There also needs to be RTX 5090 DisplayPort 2.1 outputs, permitting it to correctly help gaming screens with ever larger resolutions and refresh charges – akin to the most recent 240Hz, 4K OLED screens. Controversially, nevertheless, it seems that the RTX 5090 energy connector will proceed to be the 12VHPWR energy enter discovered on the present RTX 4000 collection, albeit with the extra trendy 12V-2×6 design.
Lastly, the RTX 5090 clock velocity has been rumored to be considerably larger than that of the RTX 4090, with a possible enhance clock of two.9GHz, whereas the GPU is predicted to be geared up with 128MB of L2 cache.
Nvidia GeForce RTX 5090 launch date window
The RTX 5090 launch date window is predicted to be January 2025, based mostly on rumor and hypothesis, as Nvidia hasn’t formally introduced a date.
The RTX 5090 was initially anticipated to launch in This autumn 2024, together with the RTX 5080, as this may be round two years after the October 22 launch date of the RTX 4090. Nonetheless, extra lately we’ve seen stories that counsel the RTX 5080 launch may very well be sooner than the RTX 5090, although equally different rumors have hinted at each playing cards launching at related instances.
Currently, although, there have been rumors that each GPUs might be popping out later, leading to an RTX 5090 launch date in 2025, with rumors that Nvidia will announce the GPU in January 2025 at CES. That is the discharge window we’re anticipating proper now.
The explanation for this significantly unclear scenario proper now could be that Nvidia merely doesn’t must launch a sooner product to compete with its competitors. The RTX 4090 stays far and away the quickest graphics card you should buy on common – the AMD RX 7900 XTX might be as quick or sooner in some video games nevertheless it’s not pretty much as good at ray tracing – and AMD’s next-generation RDNA 4 graphics playing cards aren’t anticipated to reach till subsequent yr, in accordance with the most recent rumors. As such, Nvidia can simply sit fairly persevering with to promote all of the 4000 collection playing cards it likes earlier than dropping its RTX 5000 collection merchandise.
Nvidia GeForce RTX 5090 value estimate
The RTX 5090 value is estimated to be between $1,999 and $2,499, making it rather more costly than its predecessor, the RTX 4090.
The newest RTX 5090 pricing leak offers a spread of $1,999 to $2,499 for the brand new GPU, with the Founders Version presumably costing $1,999, whereas overclocked playing cards from board companions with fancy coolers will value extra.
Though this rumored value is considerably larger than the $1,599 MSRP of the RTX 4090 Founders Version, you possibly can very not often purchase this card at this value, and also you’re normally a extra practical value of at the least $1,799 for an RTX 4090 card, and extra like $2,000 for an overclocked mannequin.
What’s extra, now that AMD has deserted high-end GPUs, there’s little incentive for Nvidia to rein within the value of its flagship. If persons are keen to pay two grand to get the most effective graphics card, then Nvidia will promote it at that value. That is all based mostly on rumor proper now, although – we’ll have to attend for the RTX 5090 to be formally launched by Nvidia earlier than we all know the value for sure.
In comparison with the costs of present graphics playing cards, GeForce RTX 50 costs could also be larger owing to elevated manufacturing prices imposed by TSMC, in addition to the additional value of including the large 32GB of GDDR7 reminiscence rumored to be discovered on RTX 5090 graphics playing cards.
Nvidia GeForce RTX 5090 benchmark estimates
Our Nvidia GeForce RTX 5090 benchmark estimates put the GPU forward of each graphics card at present out there in the marketplace, together with the RTX 4090.
Whereas it’s one thing of a on condition that the GPU would be the quickest round, we gained’t understand how significantly better it’ll be versus its counterparts till launch. Nonetheless, early rumors of its CUDA core rely and sooner reminiscence would trace at it being over 50% sooner.
In help of that, some rumormongers counsel that RTX 5090 efficiency might be 60 or 70% sooner than the RTX 4090, nevertheless it’s unclear whether or not that is with ray tracing enabled or disabled. Complicating issues additional, this is also with Nvidia DLSS 3 enabled or as-yet unannounced performance-enhancing options, as Nvidia likes to do when it releases its personal benchmark figures.
Within the meantime, if you wish to purchase a brand new graphics card now, take a look at our GeForce RTX 4080 Tremendous assessment, the place we take Nvidia’s newest high-end GPU for a spin. It corrects the massive pricing fake pas Nvidia made with the unique RTX 4080, whereas additionally providing first rate 4K gaming efficiency for $999.
