Nvidia GeForce RTX 5090: Two-minute overview

The Nvidia GeForce RTX 5090 is a troublesome GPU to strategy as knowledgeable reviewer as a result of it’s the uncommon shopper product that’s so highly effective, and so good at what it does, it’s important to actually study whether it is truly a helpful product for individuals to purchase.

Proper out the gate, let me simply lay it out for you: relying on the workload, this GPU can get you as much as 50% higher efficiency versus the GeForce RTX 4090, and that is not even factoring in multi-frame era with regards to gaming, although on common the efficiency remains to be a decent enchancment of roughly 21% total.

Merely put, no matter it’s you are wanting to make use of it for, whether or not gaming, artistic work, or AI analysis and improvement, that is the perfect graphics card for the job if all you care about is pure efficiency.

Issues get a bit extra sophisticated if you wish to carry power effectivity into the equation. But when we’re being trustworthy, when you’re contemplating shopping for the Nvidia RTX 5090, you do not care about power effectivity. This merely is not that form of card, and in order a lot as I need to make power effectivity a difficulty on this overview, I actually cannot. It is not supposed to be environment friendly, and people who need this card don’t care about how a lot power this factor is flattening—in reality, for a lot of, the large TDP on this card is a part of its enchantment.

Likewise, I am unable to actually argue an excessive amount of with the cardboard’s value, which is available in at $1,999 / £1,939 / AU$4,039 for the Founders Version, and which can probably be a lot increased for AIB companion playing cards (and that is earlier than the inevitable scalping begins). I might rage, rage in opposition to the inflation of the value of premium GPUs all I would like, however truthfully, Nvidia would not cost this a lot for this card if there wasn’t a line out the door and across the block filled with fanatics who’re greater than keen to pay that form of cash for this factor on day one.

Do they get their cash’s value? For essentially the most half, sure, particularly in the event that they’re not a gamer however a artistic skilled or AI researcher. In the event you’re within the latter camp, you are going to be very enthusiastic about this card.

In the event you’re a gamer, you may nonetheless get spectacular gen-on-gen efficiency enhancements over the celebrated RTX 4090, and the Nvidia RTX 5090 is absolutely the primary shopper graphics card I’ve examined that may get you constant, high-framerate 8K gameplay even earlier than factoring in Multi-Body Technology. That marks the RTX 5090 as one thing of an inflection level of issues to return, very like the Nvidia RTX 2080 did again in 2018 with its first-of-its-kind {hardware} ray tracing.

Is it value it although?

That, in the end, is as much as the fanatic purchaser who’s trying to make investments on this card. At this level, you in all probability already know whether or not or not you need it, and plenty of will probably be studying this overview to validate these selections which have already been made.

In that, relaxation straightforward. Even with out the bells and whistles of DLSS 4, this card is a hearty improve to the RTX 4090, and contemplating that the precise value of the RTX 4090 has hovered round $2,000 for the higher a part of two years regardless of its $1,599 MSRP, if the RTX 5090 sticks near its launch value, it is effectively definitely worth the funding. If it will get scalped to hell and sells for far more above that, you may want to think about your buy far more fastidiously to be sure you’re getting essentially the most on your cash. Be sure to take a look at our the place to purchase an RTX 5090 information that can assist you discover inventory when it goes on sale.

Nvidia GeForce RTX 5090: Worth & availability

• How a lot is it? MSRP is $1,999 / £1,939 / AU$4,039
• When are you able to get it? The RTX 5090 goes on sale January 30, 2025
• The place is it obtainable? The RTX 5090 will likely be obtainable within the US, UK, and Australia at launch

The Nvidia GeForce RTX 5090 goes on sale on January 30, 2025, beginning at $1,999 / £1,939 / AU$4,039 for the Nvidia Founders Version and choose AIB companion playing cards. Overclocked (OC) and different equally tweaked playing cards and designs will clearly run increased.

It is value noting that the RTX 5090 is 25% costlier than the $1,599 launch value of the RTX 4090, however in actuality, we are able to anticipate the RTX 5090 to promote for a lot increased than its MSRP within the months forward, so we’re actually an asking value nearer to the $2,499.99 MSRP of the Turing-era Nvidia Titan RTX (when you’re fortunate).

In fact, when you’re available in the market for the Nvidia RTX 5090, you are in all probability not squabbling an excessive amount of concerning the value of the cardboard. You are already anticipating to pay the premium, particularly the primary adopter premium, that comes with this launch.

That mentioned, that is nonetheless a ridiculously costly graphics card for anybody aside from an AI startup with VC backing, so it is value asking your self earlier than you affirm that buy if this card is actually the precise card on your system and setup.

• Worth: 3 / 5

Nvidia GeForce RTX 5090: Specs & options

• First GPU with GDDR7 VRAM and PCIe 5.0
• Barely slower clocks
• Obscene 575W TDP

There are loads of new architectural adjustments within the Nvidia RTX 50 sequence GPUs which can be value diving into, particularly the transfer to a transformer AI mannequin for its upscaling, however let’s begin with the brand new specs for the RTX 5090.

Initially, the flagship Blackwell GPU is the primary shopper graphics card to characteristic next-gen GDDR7 video reminiscence, and it’s considerably sooner than GDDR6 and GDDR6X (a roughly 33% improve in Gbps over the RTX 4090). Add within the a lot wider 512-bit reminiscence interface and you’ve got a complete reminiscence bandwidth of 1,790GB/s.

This, greater than even the will increase VRAM pool of 32GB vs 24GB for the RTX 4090, makes this GPU the primary actually succesful 8K graphics card in the marketplace. 8K textures have an infinite footprint in reminiscence, so transferring them by means of the rendering pipelines to generate playable framerates is not actually doable with something lower than this card has.

Sure, you can, possibly, get playable 8K gaming with some RTX 40 or AMD Radeon RX 7000 sequence playing cards if you employ aggressive upscaling, however you will not actually be getting 8K visuals that’ll be definitely worth the effort. In actuality, the RTX 5090 is what you need if you wish to play 8K, however good luck discovering an 8K monitor at this level. These are nonetheless years away from actually going mainstream (although there are a rising variety of 8K TVs).

In the event you’re settling in at 4K although, you are in for a deal with, since all that bandwidth means sooner 4K texture processing, so you may get very quick native 4K gaming with this card with out having to fall again on upscaling tech to get you to 60fps or increased.

The clock speeds on the RTX 5090 are barely slower, which is sweet, as a result of the opposite main top-line specs for the RTX 5090 are its gargantuan TDP of 575W and its PCIe 5.0 x16 interface. For the TDP, this thermal problem, in keeping with Nvidia, required main reengineering of the PCB inside the cardboard, which I am going to get to in a bit.

The PCIe 5.0 x16 interface, in the meantime, is the primary of its type in a shopper GPU, although you may anticipate AMD and Intel to rapidly observe go well with. Why this issues is as a result of various newer motherboards have PCIe 5.0 lanes able to go, however most individuals have been utilizing these for PCIe 5.0 m.2 SSDs.

In case your motherboard has 20 PCIe 5.0 lanes, the RTX 5090 will take up 16 of these, leaving simply 4 on your SSD. When you’ve got one PCIe 5.0 x4 SSD, you have to be fantastic, however I’ve seen motherboard configurations which have two or three PCIe 5.0 x4 m.2 slots, so when you’ve bought a kind of and you’ve got loaded them up with PCIe 5.0 SSDs, you are more likely to see these SSDs drop all the way down to the slower PCIe 4.0 speeds. I do not suppose it will be that massive of a deal, however it’s value contemplating when you’ve invested quite a bit into your SSD storage.

As for the opposite specs, they’re kind of much like what you’d discover within the RTX 4090, simply extra of it. The brand new Blackwell GB202 GPU within the RTX 5090 is constructed on a TSMC 4nm course of, in comparison with the RTX 4090’s TSMC 5nm AD102 GPU. The SM design is similar, so 128 CUDA cores, one ray tracing core, and 4 tensor cores per SM. At 170 SMs, you have bought 21,760 CUDA cores, 170 RT cores, and 680 Tensor cores for the RTX 5090, in comparison with the RTX 4090’s 128 SMs (so 16,384 CUDA, 128 RT, and 512 Tensor cores).

• Specs & options: 4.5 / 5

Nvidia GeForce RTX 5090: Design

• Slim, dual-slot type issue
• Higher cooling

So there is a vital change to this era of Nvidia Founders Version RTX flagship playing cards by way of design, and it isn’t insubstantial.

Holding the RTX 5090 Founders Version in your hand, you may instantly discover two issues: first, you may comfortably maintain it in a single hand due to it being a dual-slot card moderately than a triple-slot, and second, it is considerably lighter than the RTX 4090.

An enormous a part of that is how Nvidia designed the PCB inside the cardboard. Historically, graphics playing cards have been constructed with a single PCB that extends from the interior fringe of the PC case, down by means of the PCIe slot, and much sufficient again to accommodate the entire modules wanted for the cardboard. On prime of this PCB, you may have a heatsink with piping from the GPU die itself by means of a few dozen aluminum fins to dissipate warmth, with some form of fan or blower system to push or pull cooler air by means of the heated fins to hold away the warmth from the GPU.

The issue with this setup is that when you have a monolithic PCB, you may solely actually prolong the heatsinks and followers off of the PCB to assist cool it since a fan blowing air instantly right into a plastic wall does not do a lot to assist transfer sizzling air out of the graphics card.

Nvidia has a genuinely novel innovation on this account, and that is ditching the monolithic PCB that is been a mainstay of graphics playing cards for 30 years. As a substitute, the RTX 5090 (and presumably subsequent RTX 50-series GPUs to return), splits the PCB into three components: the video output interface on the ‘entrance’ of the cardboard dealing with out from the case, the PCIe interface phase of the cardboard, and the principle physique of the PCB that homes the GPU itself in addition to the VRAM modules and different essential electronics.

This segmented design permits a spot within the entrance of the cardboard under the fan, so moderately than a fan blowing air into an obstruction, it will possibly absolutely cross over the fins of the GPU’s heatsink, considerably enhancing the thermals.

Consequently, Nvidia is ready to shrink the width of the cardboard down significantly, transferring from a 2.4-inch width to a 1.9-inch width, or a roughly 20% discount on paper. That mentioned, it feels considerably smaller than its predecessor, and it is undoubtedly a card that will not utterly overwhelm your PC case the best way the RTX 4090 does.

That mentioned, the obscene energy consumption required by this card implies that the 8-pin adapter included within the RTX 5090 package deal is a comical 4-to-1 dongle that just about no PSU in anybody’s PC case can actually accommodate.

Most modular PSUs offer you three PCIe 8-pin energy connectors at most, so let’s simply be trustworthy about this setup. You are going to must get a brand new ATX 3.0 PSU with not less than 1000W to run this card at a minimal (it is formally really useful PSU is 950W, however simply spherical up, you are going to want it), so be sure you issue that into your price range when you decide this card up

In any other case, the appear and feel of the cardboard is not that completely different than earlier generations, besides the entrance plate of the GPU the place the RTX 5090 branding would have gone is now lacking, changed by a finned shroud to permit air to cross by means of. The RTX 5090 stamp is as an alternative printed on the middle panel, much like the way it was accomplished on the Nvidia GeForce RTX 3070 Founders Version.

As a remaining contact, the white back-lit GeForce RTX brand and the X strips on the entrance of the cardboard, when powered, add a pleasant RGB-lite contact that does not look too guady, however for RGB followers on the market, you may suppose it seems moderately plain.

• Design: 4.5 / 5

Nvidia GeForce RTX 5090: Efficiency

• Strongest GPU on the buyer market
• Considerably sooner than RTX 4090
• Playable 8K gaming

So how does the Nvidia GeForce RTX 5090 stack up in opposition to its predecessor, in addition to the perfect 4K graphics playing cards in the marketplace extra broadly?

Very rattling effectively, it seems, managing to enhance efficiency over the RTX 4090 in some workloads by 50% or extra, whereas leaving every little thing else just about within the mud.

Although when checked out from 30,000 ft, the general efficiency features are respectable gen-on-gen however aren’t the form of earth-shattering features the RTX 4090 remodeled the Nvidia GeForce RTX 3090.

Beginning with artificial workloads, the RTX 5090 scores anyplace from 48.6% sooner to about 6.7% slower than the RTX 4090 in varied 3DMark checks, relying on the workload. The one poor efficiency for the RTX 5090 was in 3DMark Evening Raid, a check the place each playing cards so utterly overwhelm the check that the distinction right here may very well be all the way down to CPU bottlenecking or different points that are not simply identifiable.

On each different 3DMark check, although, the RTX 5090 scores 5.6% higher or increased, as a rule by 20-35%. In the latest;y launched check, Metal Nomad, the RTX 5090 is sort of 50% sooner than the RTX 4090.

On the compute facet of issues, the RTX 5090 is as much as 34.3% sooner in Geekbench 6 OpenGL compute check and 53.9% sooner in Vulcan, making it an absolute monster for AI researchers to leverage.

On the artistic facet, the RTX 5090 is considerably sooner in 3D rendering, scoring between 35% and 49.3% sooner in my Blender Benchmark 4.30 checks. There’s little or no distinction between the 2 playing cards with regards to video enhancing although, as they primarily tie in PugetBench for Creators’ Adobe Premiere check and in Handbrake 1.7 4K to 1080p encoding.

The latter two outcomes is perhaps all the way down to CPU bottlenecking, as even the RTX 4090 pushes proper up in opposition to the efficiency ceiling set by the CPU in loads of circumstances.

On the subject of gaming, the RTX 5090 is considerably sooner than the RTX 4090, particularly at 4K. In non-upscaled 1440p gaming, you are a roughly 18% higher common body charge and a 22.6% higher minimal/1% framerate for the RTX 5090. With DLSS 3 upscaling (however no body era), you are 23.3% higher common and 23% higher minimal/1% framerates total with the RTX 5090 vs the RTX 4090.

With ray tracing activate with out upscaling, you are getting 26.3% higher common framerates and about 23% higher minimal/1% framerates, and with upscaling turned on to balanced (once more, no body era), you are about 14% higher common fps and about 13% higher minimal/1% fps for the RTX 5090 in opposition to the RTX 4090.

At 4K, nevertheless, the sooner reminiscence and wider reminiscence bus actually make a distinction. With out upscaling and ray tracing turned off, you are getting upwards of 200 fps at 4K for the RTX 5090 on common, in comparison with the RTX 4090’s 154 common fps, an almost 30% improve. The typical minimal/1% fps for the RTX 5090 is about 28% sooner than the RTX 4090, as effectively. With DLSS 3 set to balanced, you are a roughly 22% higher common framerate total in comparison with the RTX 4090, with an 18% higher minimal/1% framerate on common as effectively.

With ray tracing and no upscaling, the distinction is much more pronounced with the RTX 5090 getting simply over 34% sooner common framerates in comparison with the RTX 4090 (with a extra modest 7% sooner common minimal/1% fps). Activate balanced DLSS 3 with full ray tracing and also you’re about 22% sooner common fps total for the RTX 5090, however an unbelievable 66.2% bounce in common minimal/1% fps in comparison with the RTX 4090 at 4K.

Once more, none of this even components in single body era, which may already considerably improve framerates in some video games (although with the introduction of some enter latency). As soon as Multi-Body Technology rolls out at launch, you may anticipate to see these framerates for the RTX 5090 run considerably increased. Pair that with Nvidia Reflex 2 to assist mitigate the enter latency points body era can introduce, and the playable efficiency of the RTX 5090 will solely get higher with time, and it is ranging from a considerable lead proper out of the gate.

Ultimately, the general baseline efficiency of the RTX 5090 is available in about 21% higher than the RTX 4090, which is what you are actually searching for with regards to a gen-on-gen enchancment.

That mentioned, it’s important to ask whether or not the efficiency enchancment you do get is definitely worth the monumental improve in energy consumption. That 575W TDP is not a joke. I maxed out at 556W of energy at 100% utilization, and I hit 100% pretty typically in my testing and whereas gaming.

The twin flow-through fan design additionally does an excellent job of cooling the GPU, however on the expense of turning the cardboard into an area heater. That 575W of warmth must go someplace, and that someplace is inside your PC case. Be sure to have satisfactory airflow to vent all that sizzling air, in any other case every little thing in your case goes to slowly prepare dinner.

So far as performance-per-price, this card does barely higher than the RTX 4090 on worth for the cash, however that is by no means been a shopping for issue for this sort of card anyway. You need this card for its efficiency, plain and easy, and in that regard, it is the perfect there may be.

• Efficiency: 5 / 5

Do you have to purchase the Nvidia GeForce RTX 5090?

Purchase the Nvidia GeForce RTX 5090 if…

Do not buy it if…

Additionally think about

How I examined the Nvidia GeForce RTX 5090

• I spent a few week and a half with the RTX 5090
• I used my full GPU testing suite to research the cardboard’s efficiency
• I examined the cardboard in on a regular basis, gaming, artistic, and AI workload utilization

I spent a few week and a half testing the Nvidia GeForce RTX 5090, each operating artificial checks in addition to utilizing it in my day-to-day PC for each work and gaming.

I used my up to date testing suite, which makes use of trade normal benchmark instruments like 3DMark, Geekbench, Pugetbench for Creators, and varied built-in gaming benchmarks. I used the identical testbench setup listed to the precise for the needs of testing this card, in addition to the entire different playing cards I examined for comparability functions.

I’ve examined and retested dozens of graphics playing cards for the 20+ graphics card opinions I’ve written for TechRadar over the previous couple of years, and so I do know the ins and outs of those PC parts. That is why you may belief my overview course of that can assist you make the precise shopping for determination on your subsequent GPU, whether or not it is the RTX 5090 or any of the opposite graphics playing cards I overview.

• Initially reviewed January 2024