If the lately rumoured specs for the brand new Nvidia RTX 5090 and RTX 5080 graphics playing cards are certainly true we’re a flagship GPU with a terrifying quantity of processing energy and a second-tier chip that is already pissing off lots of gaming folks.
GeForce RTX 5090PG144/145-SKU30GB202-300-A121760FP32512-bit GDDR7 32G600WSeptember 26, 2024
GeForce RTX 5080PG144/147-SKU45GB203-400-A110752FP32256-bit GDDR7 16G400WSeptember 26, 2024
The most recent spec ‘leak’ has come from kind of dependable X leaker, kopite7kimi, and tells a narrative of 1 almighty RTX 5090, with a specs bump that provides it a mouth-watering quantity of TSMC-built tech in a single graphics card. And reportedly nonetheless a two-slot reference card at that.
It’ll make the RTX 4090 appear like a clown card in additional methods than one. That Founders Version is a mammoth graphics card that also makes me snicker each single time I slot it into the check rig within the workplace, such is its huge, novelty-GPU measurement. That is going to be going in opposition to what’s reportedly going to be a dual-slot card will make it look much more outsized. And that is earlier than you get to speak in regards to the conversely mammoth spec of the reported new GPU. With 21,760 CUDA cores (efficient FP32 models)—that is nonetheless not utilising the total GB202 complement of cores—it is a hefty 33% improve in core depend over the RTX 4090, which itself had over 50% extra cores than the RTX 3090.
Then the brand new Nvidia Blackwell flagship GeForce card is reportedly going to be flexing with a 512-bit aggregated reminiscence bus and a full 32 GB of GDDR7 reminiscence. And, guess what, that makes for a 33% improve in each reminiscence and reminiscence bus, too. The elevated energy numbers (presumably the TGP of the cardboard) suggests a 600 W peak, which, for those who’ve been following the maths, additionally hits that seemingly magic 33% mark.
I’m wondering what kind of efficiency enchancment Nvidia goes to be touting with its new card? Certainly it is bought to do higher than 33%? In our testing, the RTX 4090 was, on common, 72% quicker than the RTX 3090 in our 4K gaming assessments. So, I am kinda hoping it is extra in keeping with the reminiscence bandwidth uplift as an alternative.
As a result of one of many few numbers which does not comply with the plus-one-third pattern is reminiscence bandwidth the RTX 5090 is supposedly going to have at its disposal. With each that 512-bit bus and the usage of 28 Gbps GDDR7, we’re 1,792 GB/s when it comes to uncooked reminiscence bandwidth, that is a reasonably incroyable 78% improve over the RTX 4090.
Fairly how Nvidia goes to squeeze all that additional energy right into a dual-slot GPU design, I actually do not know. The rumours are that we’re solely a modest change in course of node—N4P versus N4 for the Ada technology—in order that’s not going to get you operating a lot cooler. So, we might be a really particular Founders Version certainly, doubtlessly some sort of liquid-cooled beastie.
Probably the third-party GPU distributors are nonetheless more likely to be tacking monstrous coolers to their very own variations, and I doubt subsequent RTX 50-series Founders Version playing cards would get an analogous therapy. Although, who is aware of.
I kinda do not wish to take into consideration how a lot this factor goes to value, although. The RTX 4090 was a relative discount in comparison with the pricing will increase that occurred throughout the remainder of the RTX 40-series vary, however I wrestle to imagine that in a world the place it has zero competitors Nvidia is not going to cost its flagship GPU anyplace south of $2,000. If we’re going by the 33% numbers round $2,100 makes some sense…
On condition that it is twice the GPU of the rumoured RTX 5080, which is more likely to ship on the similar $1,000 mark because the RTX 4080 Tremendous, that every one kinda follows.
| Header Cell – Column 0 | Nvidia RTX 5090* | Nvidia RTX 5080* | Nvidia RTX 4090 | Nvidia RTX 4080 Tremendous |
|---|---|---|---|---|
| GPU | GB202-300 | GB203-400 | AD102-300 | AD103-400 |
| CUDA Cores | 21,760 | 10752 | 16384 | 10240 |
| SMs | 170 | 84 | 128 | 80 |
| Reminiscence | 32 GB GDDR7 | 16 GB GDDR7 | 24 GB GDDR6X | 16 GB GDDR6X |
| Reminiscence bus | 512-bit | 256-bit | 384-BIT | 256-bit |
| Reminiscence bandwidth | 1792 GB/s | 896 GB/s | 1008 GB/s | 736 GB/s |
| TGP | 600 W | 400 W | 450 W | 320 W |
| Row 7 – Cell 0 | * rumoured specs | Row 7 – Cell 2 | Row 7 – Cell 3 | Row 7 – Cell 4 |
So sure, that RTX 5080. The rumours will not be type, and neither has been the response to its personal specs. With a reported miserly 5% improve in CUDA core depend over the RTX 4080 Tremendous this new card is not trying notably thrilling. With the identical 256-bit bus and 16 GB of GDDR7 VRAM there’s little or no that is getting my techie coronary heart pounding about that make up.
It appears prefer it’s relying nearly solely on a rise in reminiscence bandwidth and a ton extra energy jammed by means of its silicon to have the ability to achieve the efficiency lead over the RTX 4080 Tremendous it must ship a extremely tantalising improve prospect. I imply, the scuttlebutt means that it must beat the gaming efficiency of the RTX 4090 by a contact—although given these specs I am struggling to see it—but it surely’s nonetheless going to be a tricky promote given the hole between the RTX 5080 and RTX 5090 is extensive sufficient you might toss an asteroid by means of.
As ever, we have got to take these specs with a dose of the salty stuff till we get official affirmation from Nvidia itself on its new playing cards, and it performs them classically near its chest proper as much as launch. However they really feel proper. In a method that, y’know, feels kinda flawed. No-one desires the RTX 5080 to look so gimped, with the doable exception of AMD who is likely to be tempted out of high-end GPU retirement if the goal is doubtlessly achievable.
Although we nonetheless do not know when the playing cards themselves will launch. From the individuals I’ve spoken too, I’ve lengthy had a sense it will be this 12 months in some type, however then Nvidia has no purpose to must launch the RTX 50-series in 2024. It has no competitors, and the RTX 40-series are nonetheless doing the products proper now. However even when the RTX 5090 and RTX 5080 are to launch within the subsequent three months there’s an opportunity that these numbers aren’t remaining, particularly these energy figures. We noticed an entire host of TGP hypothesis forward of the ultimate 450 W reveal of the RTX 4090, so there might nonetheless be some motion even when these leaked specs are appropriate as of proper now.
