NVIDIA GeForce RTX 5080 to Stand Out with 30 Gbps GDDR7 Memory, Other SKUs Remain on 28 Gbps

[AleksandarK]

NVIDIA is preparing to unveil its "Blackwell" GeForce RTX 5080 graphics card, featuring cutting-edge GDDR7 memory technology. However, RTX 5080 is expected to be equipped with 16 GB of GDDR7 memory running at an impressive 30 Gbps. Combined with a 256-bit memory bus, this configuration will deliver approximately 960 GB/s bandwidth—a 34% improvement over its predecessor, the RTX 4080, which operates at 716.8 GB/s. While the RTX 5080 will stand as the sole card in the lineup featuring 30 Gbps memory modules, while other models in the RTX 50 series will incorporate slightly slower 28 Gbps variants. This strategic differentiation is possibly due to the massive CUDA cores gap between the rumored RTX 5080 and RTX 5090.

The flagship RTX 5090 is set to push boundaries even further, implementing a wider 512-bit memory bus that could potentially achieve bandwidth exceeding 1.7 TB/s. NVIDIA appears to be reserving larger memory configurations of 16 GB+ exclusively for this top-tier model, at least until higher-capacity GDDR7 modules become available in the market. Despite these impressive specifications, the RTX 5080's bandwidth still falls approximately 5% short of the current RTX 4090, which benefits from a physically wider bus configuration. This performance gap between the 5080 and the anticipated 5090 suggests NVIDIA is maintaining a clear hierarchy within its product stack, and we have to wait for the final launch to conclude what, how, and why of the Blackwell gaming GPUs.



View at TechPowerUp Main Site | Source

 

[john_]

It's funny. Nvidia users ask for more VRAM and Nvidia is giving them faster VRAM.
It's like a joke with deaf people.

 

[Fishymachine]

But are they 30GHz, or did SK/Samsung/Micron agree to bump to volts of 28G dies?

 

[Legacy-ZA]

It's funny. Nvidia users ask for more VRAM and Nvidia is giving them faster VRAM.
It's like a joke with deaf people.

They ask for an increase in capacity, the new speeds is just a bonus. Take the RTX3000 series, you can overclock the VRAM from 19000 to 21000, that is a 2000Mhz increase and yet, you won't even notice a 5% performance increase.

In my opinion, the best looking card, so far, is the RTX5070Ti, depending on the price, it will probably be the best buy from the lot if you look at the specifications across all the cards.

 

[Vya Domus]

These numbers might sound astronomical but from 28 to 30 Gbps is just a 7% increase.

 

[Firedrops]

Oh it'll stand out alright, with its price.

 

[The Quim Reaper]

E Waste GPU, filling a gap until 3GB modules enter mass production allowing for a 5080 Super refresh with 24GB of VRAM.

Avoid until then. It's only purpose for now is to serve as a cynical upsell tactic for the 5090.

 

[AcE]

Makes sense as it still uses 256 bit bus it will need the best vram to compensate it. Was the same with 4080 as well. Seeing this, it is quite possible 5080 will be as fast as 4090, maybe a bit faster or maybe a bit slower. Also I now expect a rather big jump from 4090 to 5090.

 

[kondamin]

E Waste GPU, filling a gap until 3GB modules enter mass production allowing for a 5080 Super refresh with 24GB of VRAM.

Avoid until then. It's only purpose for now is to serve as a cynical upsell tactic for the 5090.

Ti or super with a 384 bit bus?

 

[Outback Bronze]

Also I now expect a rather big jump from 4090 to 5090.

Hmm yeah but at what cost? Things are going to get scary.

 

[lepudruk]

And 5090 Ti with 30 Gbps as a cherry on the top, how sweet..

 

[londiste]

These numbers might sound astronomical but from 28 to 30 Gbps is just a 7% increase.

What are you comparing here? x080 GPUs have been most bandwidth starved for a couple generations now and if they can get that additional 7% relatively easily it definitely helps.
The fastest of current gen of VRAMs actually used for a GPU should be the 23Gbps on 4080S. A good 30% more bandwidth from memory change should be a good result. GDDR7 compared to GDDR6X also should have both simpler signalling as well as lower power consumption on top of that.

 

[Dimitriman]

Gotta justify that price increase from 999 to 1299 somehow...

 

[AcE]

Hmm yeah but at what cost? Things are going to get scary.

Yes 2000$ msrp I expect with scalping or let’s say “out of stock” prices up to 2500$ for normal models not overpriced ones like asus. With no one competing with Nvidia anymore they can basically do whatever they want, 2500-3000msrp is also possible but I think they won’t do that because then they sell less.

And 5090 Ti with 30 Gbps as a cherry on the top, how sweet..

Will also not exist like in 4090 because there is no competition, it will be used for Quadro card and be sold for double++ price with pro features activated.

 

[gffermari]

...all well but the core count compared to 5090 is disappointing.
It happened once with the Ada series but now it becomes ridiculous.

 

[Vya Domus]

x080 GPUs have been most bandwidth starved for a couple generations now

That's not really true of the last generation, Nvidia increased the cache by a lot, makes faster memory less relevant.

 

[londiste]

That's not really true of the last generation, Nvidia increased the cache by a lot, makes faster memory less relevant.

Instead of increasing the cache being a nice thing unfortunately this allowed them to make similarly placed GPUs with narrower memory buses. And not only Nvidia, AMD did the exact same thing one generation earlier.

 

[lukart]

They go all out on their 5090 and then if rumors are right, they have the 5070 with 12Gb...

 

[Steevo]

Its been a few years since memory OC gave more than a percent performance increase, what with all the fancy caches, compression and what not. Maybe if its to cover latency by matching cache/IO speeds.

 

[AcE]

Instead of increasing the cache being a nice thing unfortunately this allowed them to make similarly placed GPUs with narrower memory buses. And not only Nvidia, AMD did the exact same thing one generation earlier.

this is actually pretty good because it substantially lowers power consumption. All GPUs with huge cache are way more efficient. Just like Ryzen X3D

 

[TheinsanegamerN]

this is actually pretty good because it substantially lowers power consumption. All GPUs with huge cache are way more efficient. Just like Ryzen X3D

It's also cheaper to build, because with GDDR6/X the memory needs to be really close to the GPu, and busses larger then 256 bit become exponentially harder, and more expensive, to make.

i assume GDDR7 will be even worse. So anything that can reduce bandwidth needs is a good thing.

 

[Dr. Dro]

It's the same case with the RTX 4080. The 4080 (and Super) have 24 Gbps G6X modules while the RTX 4090 and 4070 (Ti and Ti Super included) use the same 21 Gbps modules that were first used in the RTX 3090 Ti. The original 3090's also 21 Gbps, but it's an earlier generation, lower capacity IC. The recent GDDR6 version of the 4070 was released so they could reallocate some of the supply to strengthen the 4090's production chain.

 

[londiste]

It's also cheaper to build, because with GDDR6/X the memory needs to be really close to the GPu, and busses larger then 256 bit become exponentially harder, and more expensive, to make.
i assume GDDR7 will be even worse. So anything that can reduce bandwidth needs is a good thing.

GDDR7 uses PAM3 signalling vs GDDR6X's PAM4. GDDR7 has lower voltages as well. Simpler (not by much but sill) and more efficient compared to GDDR6X.
GPUs are still often enough bandwidth starved. So anything that relieves pressure on that bandwidth is good but in the end it is still important to have.

 

[Guwapo77]

And 5090 Ti with 30 Gbps as a cherry on the top, how sweet..

Will there be a Ti without any competition?

 

[Cheeseball]

These numbers might sound astronomical but from 28 to 30 Gbps is just a 7% increase.

That 28 to 30 Gbps increase is the GDDR7 memory between the RTX 5080 (and lower SKUs) compared to the RTX 5090.

The RTX 4090 has 21 Gbps GDDR6X memory and the 7900 XTX has 20 Gbps GDDR6. That bandwidth increase is pretty substantial aside from the bus increase. I think the RTX 4080 SUPER had 23 Gbps GDDR6X chips.