Thursday, March 9th 2023
NVIDIA GeForce RTX 50-series and AMD RDNA4 Radeon RX 8000 to Debut GDDR7 Memory
With Samsung Electronics announcing that the next-generation GDDR7 memory standard is in development, and Cadence, a vital IP provider for DRAM PHY, EDA software, and validation tools announcing its latest validation solution, the decks are clear for the new memory standard to debut with the next-generation of GPUs. GDDR7 would succeed GDDR6, which had debuted in 2018, and has been around for nearly 5 years now. GDDR6 launched with speeds of 14 Gbps, and its derivatives are now in production with speeds as high as 24 Gbps. It provided a generational doubling in speeds from the preceding GDDR5.
The new GDDR7 promises the same, with its starting speeds said to be as high as 36 Gbps, going beyond the 50 Gbps mark in its lifecycle. A MyDrivers report says that NVIDIA's next-generation GeForce RTX 50-series, probably slated for a late-2024 debut, as well as AMD's competing RDNA4 graphics architecture, could introduce GDDR7 at its starting speeds of 36 Gbps. A GPU with a 256-bit wide GDDR7 interface would enjoy 1.15 TB/s of bandwidth, and one with 384-bit would have a cool 1.7 TB/s to play with. We still don't know what is the codename of NVIDIA's next graphics architecture, it could be any of the ones NVIDIA hasn't used from the image below.
Source:
MyDrivers
The new GDDR7 promises the same, with its starting speeds said to be as high as 36 Gbps, going beyond the 50 Gbps mark in its lifecycle. A MyDrivers report says that NVIDIA's next-generation GeForce RTX 50-series, probably slated for a late-2024 debut, as well as AMD's competing RDNA4 graphics architecture, could introduce GDDR7 at its starting speeds of 36 Gbps. A GPU with a 256-bit wide GDDR7 interface would enjoy 1.15 TB/s of bandwidth, and one with 384-bit would have a cool 1.7 TB/s to play with. We still don't know what is the codename of NVIDIA's next graphics architecture, it could be any of the ones NVIDIA hasn't used from the image below.
27 Comments on NVIDIA GeForce RTX 50-series and AMD RDNA4 Radeon RX 8000 to Debut GDDR7 Memory
Also while I know we get this kind of information all the time, just shove it over one gen, inb4 future message of GDDR8 memory being used in the gen after this one or the gen after that. shocking news I know, ANYWHO it is in a way still weird to read about this when the current gen isnt even fully out yet: 7900x, 7900, 7800, 7600, rtx4070, 4060, 4050....still MIA.
Who would have thought....
HBM3 is up to 1 TB/s per 1024bit interface, so 4 chips of HBM3 for example would equate to 4 TB/s of bandwidth, there is absolutely no way to match that with GDDR7 modules in a real product.
The doubling of capacity per package, general cooler improvements, and not having packages on the back of the PCB all help.
AMD's new 20Gbps GDDR6 for RDNA3 does run quite warm, but it's hard to say for sure since
- Memory temp is not officially reported by AMD
- Reviews mostly do not report on memory temp
- MBA coolers are unimpressive and also pretty bad at memory cooling
- 10-12 packages is still a lot for smaller not-4090 coolers
All in all, not really any reason to assume that faster product runs hotter just because, remember speculation about 21Gbps G6X just because Micron's 19Gbps product was hard to cool?Granted, it's still 3rd party software only for Radeon (HWInfo) so the jury is still out, but I find it highly unlikely that the move from 16Gb 16Gbps G6 > 16Gb 20Gbps G6 completely broke memory temp reporting.
It also has iteration/generation that are faster and more energy efficient.
I mean purely talking physics, distance between objects = delay, then GDDR per definition loses out because its further away.
12 year olds hang out in wccftech comment section where they use terms like AMDone etc.
Will prices ever come back to 'normal' for mid-range and top tier GPU's? I think the days of $700 to $750 MSRP top tier videocards are gone forever...
2019 the Radeon VII has HBM2 1,024GB/s.
2020 the MI100 has HBM2 1,228.8GB/s
2022 the MI210 has HBM2e 1,638.4GB/s
Nvidia has an H100 with HBM3 that will do 3,000GB/s.
When you consider that the Vega 56/64 back in 2017 had 8GB of HBM2 starting at $400 and the VII having 16GB for $700. I don't think that HBM is crazy expensive. For RDNA3 there is the packaging issue. Would be coold if AMD could 3D stack the MCDs onto the GCD so they have room for HBM3 :D
Stop the bickering.
Or as low as $1519 @ microcenter with 5% discount but the later might trigger those that paid over $2000 plus for theirs.