Steve Burke
speaker
449 appearances
1 recordings
1 series
first heard Jul 2026
last heard 15 Jul
Steve Burke’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Jul 2026 with 1.
Appearances
So yeah, to go back to your point of stacking, high bandwidth memory is physically vertically stacked.
As in they take silicon dyes for memory and they stack them atop each other and then they connect them all together as opposed to consumer memory where it's closer to, there's still layers, but it's closer to a flat plane.
Yeah, if you're on a consumer device, then the consumer memory, you'll need more physical area to put the same amount of capacity, whereas with the high-end server stuff, you just stack that capacity vertically.
And then it's also, HBM's really significant advantage is that it's very low power, comparatively.
Right.
Oh, yeah.
If you're talking about the fab space, then yes.
HBM requires a lot more space for a few reasons in terms of manufacturing area.
In terms of footprint on a product, it's less, but in a fab, it's more.
Why is that?
In the fab, one of the reasons... So there's a few, but one is that HBM requires additional silicon pieces to be able to function.
So with GDDR consumer GPU memory, you pretty much have the memory module.
It looks like a black rectangle.
And then if you were to...
use like acid or something to etch away the black plastic part, you'd be left with pretty much a singular silicon die underneath.
That's all it is for consumer.
For HBM, you also need a logic die.
So this is another piece of silicon that interfaces with the memory and the stack of memory because you need something to kind of
command and control and understand the vertical tower you've got of memory.
And then you also need an interposer die, which sits underneath everything.
Showing 81–100 of 449 · page 5 of 23
← Previous
Next →