Re:Arm and TSMC Showcase 7nm Chiplet, Eight A72 at 4GHz on C...
by phkahler on Saturday September 28, 2019 @08:00PM
(#21103957)
Did I miss it, or was there no mention of power consumption? We know they have lower perf/mhz but if they're low enough power they might be decent per watt.
Re:Arm and TSMC Showcase 7nm Chiplet, Eight A72 at 4GHz on C...
by DuskStar on Saturday September 28, 2019 @08:09PM
(#21104002)
> When it comes to overall efficiency of LIPINCON on CoWoS, TSMC says that it features a 0.56 pJ/bit (pico-Joules per bit) power efficiency as well as a 1.6 Tb/s/mm2 (terabits per second per square millimeter) bandwidth density.
Re:Arm and TSMC Showcase 7nm Chiplet, Eight A72 at 4GHz on C...
by loeg on Saturday September 28, 2019 @08:33PM
(#21104107)
It's not a product you'll ever be able to buy and AFAIK it's a press piece, not something every released to 3rd parties for review:
> The chip will never be sold in volume, but it proves that technologies by the two companies can enable designers to build complex chiplet-based products with unique characteristics.
So it's kind of interesting from a "is it possible to take the AMD approach to ARM cores" but it's not worth evaluating too hard because, again, you can't buy it.
I think ARM hopes this inspires one of the less creative licensees (Qualcom? Who else makes ARM besides Apple?) to pursue this as a product.
Re:Arm and TSMC Showcase 7nm Chiplet, Eight A72 at 4GHz on C...
by reportingsjr on Saturday September 28, 2019 @10:56PM
(#21104738)
There are hundreds of companies that make or have made ARM cores, here are a few off of the top of my head: Intel, Xilinx, STMicro, NXP, silicon labs, broadcom, qualcomm, Samsung, Atmel (now microchip), etc etc.
Re:Arm and TSMC Showcase 7nm Chiplet, Eight A72 at 4GHz on C...
by benj111 on Sunday September 29, 2019 @04:29AM
(#21105874)
Why can Arm not pursue it as a product?
They already create and license cores, why not the substrate too? At the moment theres almost a binary choice for Arm using customers, either buy complete chips, or take Arm ip and roll your own, this seems like a useful third option.
Re:Arm and TSMC Showcase 7nm Chiplet, Eight A72 at 4GHz on C...
by qaute on Sunday September 29, 2019 @01:07AM
(#21105217)
To elaborate:
A single IC die (a "chiplet" when stacked on others) is made of a layer of transistors and wiring up to several hundred nanometers thick on top of a much thicker (several hundred micrometer) silicon wafer substrate.
Yep, stacking these dies ("chiplets") on top of one another is one form of 3D, and is definitely useful: IIRC, AMD's recent popular Ryzen 3000 line uses chiplets, and the Raspberry Pi's RAM chip is stacked on top the CPU (in a much cruder post-IC-manufacturing assembly process). The shorter distance between dies (compared to placing them in separate plastic packages on a PCB) can lead to maybe an order of magnitude improvement (in speed/power/etc). It's hard to stack more than a few layers.
The most ambitious form of 3D design (and the one most people probably think of) is multiple thin layers of transistors on a single die ("monolithic 3D"); this would give maybe another order of magnitude improvement. Monolithic 3D memory chips are becoming popular (V-NAND, etc), with the most recent at ~100 layers. Monolithic 3D CPUs are still an unsolved problem because they need different, more difficult process steps and better heatsinks (but we're close!).
Re:Arm and TSMC Showcase 7nm Chiplet, Eight A72 at 4GHz on C...
by hinkley on Sunday September 29, 2019 @07:15PM
(#21110258)
I can't imagine the kind of quality control you have to have to have to achieve a useful yield, performing a lithography operation one hundred times on the same piece of silicon.
Re:Arm and TSMC Showcase 7nm Chiplet, Eight A72 at 4GHz on C...
by Bayart on Sunday September 29, 2019 @10:31AM
(#21107239)
Yeah it's 3D-stacked. 4 layers for HBM, 4 for HBM and 12 for HBM2E.
It's neat but really expensive, and I don't think the gains for RAM in that sort of designs are that big compared to computing cores outside of edge cases.
Not the total power, but presumably still useful.
> The chip will never be sold in volume, but it proves that technologies by the two companies can enable designers to build complex chiplet-based products with unique characteristics.
So it's kind of interesting from a "is it possible to take the AMD approach to ARM cores" but it's not worth evaluating too hard because, again, you can't buy it.
I think ARM hopes this inspires one of the less creative licensees (Qualcom? Who else makes ARM besides Apple?) to pursue this as a product.
Ampere, Cavium, HiSilicon, Fujitsu are the ones with actually custom cores.
They already create and license cores, why not the substrate too? At the moment theres almost a binary choice for Arm using customers, either buy complete chips, or take Arm ip and roll your own, this seems like a useful third option.
Amazon (via Annapurna labs) with their Nitro chips. I'm pretty sure there are more, but that's the first one that came to mind.
If the chips are stacked on top of a communications backplane it would seem like that would be the case.
A single IC die (a "chiplet" when stacked on others) is made of a layer of transistors and wiring up to several hundred nanometers thick on top of a much thicker (several hundred micrometer) silicon wafer substrate.
Yep, stacking these dies ("chiplets") on top of one another is one form of 3D, and is definitely useful: IIRC, AMD's recent popular Ryzen 3000 line uses chiplets, and the Raspberry Pi's RAM chip is stacked on top the CPU (in a much cruder post-IC-manufacturing assembly process). The shorter distance between dies (compared to placing them in separate plastic packages on a PCB) can lead to maybe an order of magnitude improvement (in speed/power/etc). It's hard to stack more than a few layers.
The most ambitious form of 3D design (and the one most people probably think of) is multiple thin layers of transistors on a single die ("monolithic 3D"); this would give maybe another order of magnitude improvement. Monolithic 3D memory chips are becoming popular (V-NAND, etc), with the most recent at ~100 layers. Monolithic 3D CPUs are still an unsolved problem because they need different, more difficult process steps and better heatsinks (but we're close!).
Fyi, I don't think this is done in the most recent Pi 4
It's neat but really expensive, and I don't think the gains for RAM in that sort of designs are that big compared to computing cores outside of edge cases.