BBCube 3D Could be the Future of Stacked DRAM

[TheLostSwede]

Scientists at the Tokyo Institute of Technology have developed a new type of stacked or 3D DRAM that the researchers call Bumpless Build Cube 3D or BBCube 3D, which relies on Through Silicon Vias or TSVs to connect the DRAM dies. This is a different approach to HBM which relies on micro bumps to connect the layers together and the Japanese scientists are saying that their bumpless wafer-on-wafer solution should allow not only for an easier manufacturing process, but more importantly, improved cooling, as the TSVs can channel the heat from the DRAM dies down into whatever substrate the BBCube 3D stack is finally mounted onto.

If that wasn't enough, the researchers believe that BBCube 3D will be able to deliver higher speeds than HBM courtesy of a combination of the TSVs being relatively short and "high-density signal parallelism". BBCube 3D is expected to deliver up to a 32 fold increase in bandwidth compared to DDR5 memory and a four fold increase compared to HBM2E memory, while at the same time, drawing less power. The research paper goes into a lot more details for those interested at taking a closer look at this potentially revolutionary shift in DRAM assembly. However, the question that remains unanswered is if this will end up as a real world product some time in the near future, which is all based on how manufacturable BBCube 3D memory will be.



View at TechPowerUp Main Site | Source

 

[TumbleGeorge]

a four fold increase compared to HBM2E memor

Hmm, why they didn't make comparison with actual HBM3? In theory if they make to release their memory on market in 2025 or later, comparison will be with HBM4/E and I am sure that their numbers will looking not so good.

 

[zlobby]

A missed opportunity to call it 'BBQb'.

 

[wNotyarD]

Talk about power and temperature against HBM all you want, but the all-important question is: what about the price per Gb?

 

[TheLostSwede]

Talk about power and temperature against HBM all you want, but the all-important question is: what about the price per Gb?

Well, this is standard DRAM that they stack, so most likely DDR5, so the cost shouldn't be that much higher, but obviously the TSV process isn't free and I'm not sure if they can just take standard DDR5 dies and stack them.

Hmm, why they didn't make comparison with actual HBM3? In theory if they make to release their memory on market in 2025 or later, comparison will be with HBM4/E and I am sure that their numbers will looking not so good.

It might never make it into production, as it's a research project.

 

[Denver]

Isn't this even more complex and expensive to produce than HBM?

 

[TumbleGeorge]

Isn't this even more complex and expensive to produce than HBM?

No. In article

Japanese scientists are saying that their bumpless wafer-on-wafer solution should allow not only for an easier manufacturing process

Easier=cheaper

 

[TheLostSwede]

Isn't this even more complex and expensive to produce than HBM?

No, because each HBM layer have bumps that connects via TSV to the layer below, as per some of the drawings in the article.
It makes the stacks "unstable" because of the bumps and having single layer TSVs are not as benefitial as doing TSVs through the entire memory stack.
However, it doesn't mean that HBM couldn't copy this stacking approach in the future.

I guess the tricky part with BBCube 3D is that each layer of DRAM has to be as thin as possible from my understanding of skimming through the paper, which could lead to dificulties during production, but the rest seems pretty doable at fairly low costs.

Then again, this is a research paper, so it's possible that the researches has missed something and this simply won't be practical with today's manufacturing tools.

 

[wNotyarD]

Then again, this is a research paper, so it's possible that the researches has missed something and this simply won't be practical with today's manufacturing tools.

Well, part of research must be looking at technical and market feasibility.

 

[TheLostSwede]

Well, part of research must be looking at technical and market feasibility.

Of course, but you know how these things are, not all research ends up in successful products.

 

[AusWolf]

Considering how hot CPUs/GPUs and GDDR6(X) are, I have my concerns. Let them prove me wrong.

 

[TheoneandonlyMrK]

Verses HBM 2E, why.

it will be facing HBM3 by then in reality still more options can't be bad.

 

[TheLostSwede]

Verses HBM 2E, why.

it will be facing HBM3 by then in reality still more options can't be bad.

Most likely because that was the data that was available to the scientists when they started the research?

 

[Bones]

A missed opportunity to call it 'BBQb'.

Done that to a few sticks here and there before.
Got 'em Nice-N-Smokey with voltage.

Since this is a concept still on paper I woudn't get your hopes up too high over it, esp if expecting anything like it to appear soon because this will take time just to get it to the testing stage of it.
No telling if would be as good, better or worse than described.

 

[FierceRed]

A missed opportunity to call it 'BBQb'.

BYOBBQbRAM

 

[LabRat 891]

My inquiry on 'stacked' memory: How does it handle thermal-cycling?
We already have issues w/ cold solder joints from expansion-contraction cycles; I'm concerned that we might be building tech that implicitly will fail inside a decade of normal use.

 

[zlobby]

BYOBBQbRAM

 

[kondamin]

I hope this can get licensed quickly and really is easier to make than hbm.
IGPU with a couple of GBs of fast memory sitting right beside it.

 

[Minus Infinity]

Verses HBM 2E, why.

it will be facing HBM3 by then in reality still more options can't be bad.

HBM3 is massively dearer than HBM2e. This BBCube should be a lot cheaper than HBM2e let alone HBM3. The latter would neve rfind it's way into consumer applications. Compare it's price to the soon to be released GDDR7.