Micron to Launch GDDR7 Memory Chips in 2024, Promising Enhanced Performance

Published by

Click here to post a comment for Micron to Launch GDDR7 Memory Chips in 2024, Promising Enhanced Performance on our message forum
https://forums.guru3d.com/data/avatars/m/248/248627.jpg
I'm going to guess we won't see gddr7 on any mainstream gpus till after rtx 5000 and Radeon 8000 series cards
data/avatar/default/avatar08.webp
icedman:

I'm going to guess we won't see gddr7 on any mainstream gpus till after rtx 5000 and Radeon 8000 series cards
Yes, most likely series after next series. Depends if Nvidia and AMD release new GPUs next year
https://forums.guru3d.com/data/avatars/m/38/38428.jpg
Given how much they need each other, it could be almost like AMD and nVidia are part of the development process, and maybe even the opposite is true to an extent, with Micron providing input as to how they should develop their memory controllers. I'd be amazed if AMD and nVidia weren't getting some of the first engineering samples "out of the oven" to play with. But I also wouldn't be surprised if I'm way off base, and there are good reasons for none of above being true.
https://forums.guru3d.com/data/avatars/m/282/282473.jpg
icedman:

I'm going to guess we won't see gddr7 on any mainstream gpus till after rtx 5000 and Radeon 8000 series cards
for sure. imo that could be the reason nv is moving ada next to 2025, better g7 availability. they can wait, ada has no competition in high end so far, and when amd release theirs, they can just make new skus or cut price on the existing ones. wonder what amd will do. fix rdna3 on g6 ? or abandon it and go straight to rdna4 on g7. they're in a pickle.
https://forums.guru3d.com/data/avatars/m/255/255510.jpg
This is great news. 😀 So AMD and Nvidia can cripple it with a memory bus width of 8bit. 😛
https://forums.guru3d.com/data/avatars/m/189/189980.jpg
vestibule:

This is great news. 😀 So AMD and Nvidia can cripple it with a memory bus width of 8bit. 😛
Sssh, don't give them ideas.
https://forums.guru3d.com/data/avatars/m/198/198862.jpg
128bit rtx5070 incoming. 😀
https://forums.guru3d.com/data/avatars/m/248/248994.jpg
If they made a GDDR7 chip with a capacity of 24Gb, it would allow a 128-bit bus to cheaply handle 12GB of VRAM. Alternatively they could even use a 96-bit bus and have 9GB of VRAM for the cheapest gaming cards, allowing them to get over 8GB, even if marginally.
https://forums.guru3d.com/data/avatars/m/255/255510.jpg
@Kaarme lol good stuff. You may not have been making a joke, but it was funny. 🙂
https://forums.guru3d.com/data/avatars/m/271/271560.jpg
cucaulay malkin:

for sure. imo that could be the reason nv is moving ada next to 2025, better g7 availability. they can wait, ada has no competition in high end so far, and when amd release theirs, they can just make new skus or cut price on the existing ones. wonder what amd will do. fix rdna3 on g6 ? or abandon it and go straight to rdna4 on g7. they're in a pickle.
they're going to add fgpa to die almost certainly to address rt, otherwise...
https://forums.guru3d.com/data/avatars/m/198/198862.jpg
Kaarme:

If they made a GDDR7 chip with a capacity of 24Gb, it would allow a 128-bit bus to cheaply handle 12GB of VRAM. Alternatively they could even use a 96-bit bus and have 9GB of VRAM for the cheapest gaming cards, allowing them to get over 8GB, even if marginally.
16gb/128bit is also now possible and i can easily see upcoming x70 class gpus using this configuration.
https://forums.guru3d.com/data/avatars/m/252/252758.jpg
So that's why RTX 5000 is coming in 2025, they're testing GDDR7 in 2024 until it's ready?!!
https://forums.guru3d.com/data/avatars/m/56/56686.jpg
This could be good thing, no more castrated bandwidths in theroy 128bit with 576Gbps
https://forums.guru3d.com/data/avatars/m/165/165326.jpg
vestibule:

This is great news. 😀 So AMD and Nvidia can cripple it with a memory bus width of 8bit. 😛
Sad , but so true lol 😛
https://forums.guru3d.com/data/avatars/m/273/273838.jpg
It would be very interesting to see how this goes. From what I can tell, most AMD architectures were memory starved (which is why AMD exploerd HBM in the first place). I would expect core power to have advanced to a point that even GDDR7 is saturated, so we might see the true potential of AMD GPUs. That being said, I don't expect AMD to overtake nVidia, since they've been collaborating with Micron for years (and had exlcusive access to GDDR6x memory) and might have a head start in those memory controller specs.
https://forums.guru3d.com/data/avatars/m/255/255510.jpg
If I remember correctly the radeon HD 4870 (2008) of the time performed V well because of the use of next gen GDDR. It was real wining trick for AMD, imo some thing of the like they have not experienced since despite trying?
https://forums.guru3d.com/data/avatars/m/268/268248.jpg
vestibule:

If I remember correctly the radeon HD 4870 (2008) of the time performed V well because of the use of next gen GDDR. It was real wining trick for AMD, imo some thing of the like they have not experienced since despite trying?
Yeah the gddr5 really helped em there , Nvidia might had a "this will never happen again moment"
https://forums.guru3d.com/data/avatars/m/255/255510.jpg
@Venix Yeah, ty for verification. 🙂
https://forums.guru3d.com/data/avatars/m/248/248994.jpg
Mpampis:

That being said, I don't expect AMD to overtake nVidia, since they've been collaborating with Micron for years (and had exlcusive access to GDDR6x memory) and might have a head start in those memory controller specs.
I can't find anything but random forum posts claiming that Nvidia exclusivity. Other posts say AMD doesn't want to use it because it's more expensive and uses more power. I wouldn't call AMD cards' current power consumption bad, but maybe they don't feel like stretching it. Furthermore, since AMD is the one who pioneered the extra large cache to compensate for VRAM bandwidth, I guess it makes sense to stick to it. Nvidia copied the technique despite still using GDDR6X as well. Less chips (with reduced bus width) is better than more chips for the manufacturers, even if the chips happened to be a little bit more expensive. At the end of the day, AMD and Hynix developed HBM, but these days even Micron manufactures it, and Nvidia is the one benefitting the most due to the massive demand in server/pro cards. So, I wouldn't be so quick to believe weird exclusive deals exist in the memory market.
https://forums.guru3d.com/data/avatars/m/273/273678.jpg
People keep reporting on "projected" end game clocks, as if they are going to be achieved day one, GDDR6x is going to be king for a long while even after 7 is out.