Skip to main content

Next-gen GPU memory will achieve unheard-of speeds

Nvidia GeForce RTX 4090 GPU.
Jacob Roach / Digital Trends

Faster graphics memory is on the horizon. Samsung is set to unveil 37Gbps GDDR7 VRAM, meaning the fastest memory of this type that we’ve seen to date. This advancement could bring some serious gains to next-gen top graphics cards, especially if it manages to be featured in Nvidia’s next-gen RTX 50-series GPUs.

Recommended Videos

Samsung is holding a session at the five-day 2024 IEEE International Solid-State Circuit Conference, titled “A 16Gb 37 Gb/s GDDR7 DRAM with PAM3-Optimized TRX Equalization and ZQ Calibration,” first spotted by TechRadar. That’s how we know this is happening, as Samsung hasn’t officially announced these faster memory variants yet.

GDDR7 memory is currently a work in progress for Samsung, but also for companies like SK Hynix and Micron. We’ve previously heard of VRAM reaching speeds up to 32Gbps, but now, those speeds are inching much higher with the latest developments.

To achieve these bleeding-edge speeds, Samsung is said to utilize PAM3 and NRZ signaling. PAM3 encoding will allow for higher data transmission rates per cycle, ultimately delivering lower power consumption while upping the memory speed.

VRAM with 37Gbps speeds, once we get there, will mark a significant upgrade over what we’re currently seeing in GDDR6X. For instance, Nvidia’s RTX 4090 provides bandwidth of 1TB/s thanks to its 21Gbps memory across a 384-bit bus. However, next-gen GDDR7 memory will provide that kind of bandwidth across a 256-bit interface when clocked at 32Gbps.

Samsung’s latest 37Gbps invention, when paired with a 384-bit bus, will translate to a whopping 1.79TB/s. That’s nearly double what the RTX 4090 can provide.

Samsung's GDDR7 memory.
Samsung

However, it has already announced 32Gbps GDDR7 memory in the summer of 2023. Samsung isn’t alone in pushing for faster VRAM — SK Hynix is also reportedly set to announce its own updates, with a session titled “A 35.4Gb/s/pin 16GB GDDR7 with a Low-Power Clocking Architecture and IO Circuitry.”

This is pretty exciting news for the GPU industry, showing us that the future of GPUs will greatly increase bandwidth. However, even though the speculation is that the next-gen graphics cards from AMD and Nvidia will utilize GDDR7, it’s unlikely that these fast memory modules will be available by then.

We might see the RTX 50-series and AMD RDNA 4 launch with GDDR7 but with lower speeds. They’ll still be an improvement over current-gen variants, but 37Gbps VRAM may have to wait until the generation after the one that’s coming next. In other words, we’ll most likely see it arrive in 2026.

Monica J. White
Monica is a computing writer at Digital Trends, focusing on PC hardware. Since joining the team in 2021, Monica has written…
Next-gen GPUs are coming ‘later this year’ — but which?
RX 7900 XTX slotted into a test bench.

What's going on with next-gen graphics cards? I've been asking myself that question for months now. Reports about Nvidia's RTX 50-series and AMD's RDNA 4 first pointed to a 2024 release, but most sources now agree that we won't see any new GPUs until 2025. Except EK Water Blocks, a company that now claims that we'll see an announcement "later this year."

EK Water Blocks makes liquid cooling solutions, and it's partnered with both Nvidia and AMD, which makes it harder to determine which GPU manufacturer it's talking about here. According to the latest leaks, both GPU makers aren't launching their new products this year, although one source (admittedly uncertain) claimed that we'd have an announcement this month. This is now the second leak in as many days that implies good news in 2024.

Read more
Nvidia is launching a new GPU, but read this before buying
Two RTX 4070 graphics cards sitting side by side.

Nvidia is launching a new version of one of its best GPUs -- the RTX 4070 (non-Super). Normally, this would be great news. However, there's one downside: Nvidia is equipping the new RTX 4070 with GDDR6 memory, which is a step down from the GDDR6X that the GPU usually comes with. This wouldn't be the end of the world, but there are a couple of red flags.

For starters, Nvidia itself lists the GPU alongside its previous version. It seems like the two RTX 4070s will only differ by memory type, all the while retaining the same clock speeds, bus width, and CUDA core counts. This is good news, but switching from GDDR6X to GDDR6 means a drop in bandwidth.

Read more
Nvidia’s next-gen GPUs may be delayed due to ‘design flaws’
Nvidia introducing its Blackwell GPU architecture at GTC 2024.

Dark clouds are looming over the future of Nvidia's best graphics cards. According to a new report, Nvidia told some of its partners that it will be delaying its upcoming Blackwell GPUs, and is now aiming for an early 2025 release instead. Delays are one thing, but the cause is perhaps the most worrying part of it all -- design flaws. What does this mean for Nvidia's RTX 50-series?

This worrying report originates from The Information, which cites two sources who helped produce the Blackwell chip, as well as its server hardware. Bloomberg recounts that the chips may be delayed by over three months at this point. Nvidia is preparing B100 and B200 chips for some of the world's most prominent tech companies, including Meta, Google, and Microsoft, so these delays could hit pretty hard.

Read more