hmmmm, hard (not elastic, true hard) shell briefcase, foam padding behind, something to keep it from sliding out of the center, and if the top needs to stay scratch-free a display-grade microfiber towel on top; then foam for the other half of the case.... well, that's what I'd expect to rig up from like local craigslist and hardware store parts.
3:23 a.m.
(car cleaning/polish-applcation cloth that never saw polish (microfiber kind) might also be quite appropriate for the front anti-scratch layer)
for stacking I'd expect a freezer bag sliced open would give a nice comfortably-thick sheet of continuos/crinkle-free polyethylene foil to pad between stacked wafers
If they're not complex and from polycarbonate I'd eyeball 10$/kg as reasonable based on cost to make and them being a generic thing (I think they are generic?).
Something like ~$10 USD per shell (they seemed very over priced for what they where), but I'm pretty sure I was ordering like 100 at the same time or something....
12:50 a.m.
@Andrew Wingate / @stuart - Actually, didn't we ask Ken (in China) to source pricing on some of these types of things for us?
12:52 a.m.
@carlfk - FYI - I ended up paying something like $70 USD for the square ones the GF180MCU Run #1 wafers are in.
Tim 'mithro' Ansell
@Andrew Wingate / @stuart - Actually, didn't we ask Ken (in China) to source pricing on some of these types of things for us?
Explore detailed specs for every Raspberry Pi processor from BCM2835 to BCM2712—die area, CPU cores, cache, clock speed, and process nodes in one comprehensive table.
9:48 a.m.
It seems to pass the sniff test. I wonder if the following is true:
Dark silicon is expensive. BCM2712 C1 carried Broadcom set-top-box blocks that no Pi ever enabled. Deleting them for the D0 stepping cut 18 mm², a third of the die, with no change to the CPU, GPU or any advertised feature.6
Raspberry Pi launched the 2 gig Pi 5 for $50, and besides half the RAM and a lower price, it has a new stepping of the main BCM2712 chip.
This is the BCM2712 D0 stepping. Older Pi 5's shipped with a C1. In their blog post, they said:
The new D0 stepping strips away all that unneeded functionality, leaving only the bits we need.
Steppings are bas...
I do remember asking that. This is a neat compilation, but does seem moore's law-esque, which I don't think answers my question in the spirit of the way I meant it.
I think a refined version sounds like:
What [novel and useful] things will people do with [trailing edge] silicon.
Admittedly the ignorance I had (and still have) didn't allow me to comprehend the landscape, let alone understand what the possibilities or challenges really are.
Now we're starting to see extra flashy things like microTheia and I always love novel architectures like @BreakingTaps. As I learn about the sweet spot for gf180mcu I am super excited to see what comes from @Tim Edwards Chipalooza, and more to the point; things people will do with those analog blocks, and I literally have given talks about many of the other awesome ones [coming soon I hope].
So, I'll never stop asking "What will people do with silicon?", but will say I'm honored to be a part of this with you all, and thank you for everything you've shared with me and with us all!
And can't wait to see what you all come up with next!
-# p.s. I can't really see myself being interested in any press release broadcom will ever make again ;)(edited)
Event-based, general-purpose machine vision ASIC. Contribute to dhgaddy/microTheia development by creating an account on GitHub.
10:42 a.m.
Andrew Wingate
I do remember asking that. This is a neat compilation, but does seem moore's law-esque, which I don't think answers my question in the spirit of the way I meant it.
I think a refined version sounds like:
What [novel and useful] things will people do with [trailing edge] silicon.
Admittedly the ignorance I had (and still have) didn't allow me to comprehend the landscape, let alone understand what the possibilities or challenges really are.
Now we're starting to see extra flashy things like microTheia and I always love novel architectures like @BreakingTaps. As I learn about the sweet spot for gf180mcu I am super excited to see what comes from @Tim Edwards Chipalooza, and more to the point; things people will do with those analog blocks, and I literally have given talks about many of the other awesome ones [coming soon I hope].
So, I'll never stop asking "What will people do with silicon?", but will say I'm honored to be a part of this with you all, and thank you for everything you've shared with me and with us all!
And can't wait to see what you all come up with next!
-# p.s. I can't really see myself being interested in any press release broadcom will ever make again ;) (edited)
I think there is some interesting things there -- like the mm2 went up despite the process node going down, the different revisions on the same process node, & the fact that RPi -- which isn't that of a small customer only went with custom silicon as a last resort despite having 1/3rd of their silicon take up by stuff they didn't use.
I wonder what the world looks like if their second / third design used custom silicon.....
like the mm2 went up despite the process node going down
This being the more memory and/or dark silicon?
I guess I don't really understand this completely. Like broadcom is using a section of the chip for set top boxes that is unavailable on the rpi varieties?
And how the economics may change if they got more die/wafer going it alone?
Sorry, not really following?
Andrew Wingate
like the mm2 went up despite the process node going down
This being the more memory and/or dark silicon?
I guess I don't really understand this completely. Like broadcom is using a section of the chip for set top boxes that is unavailable on the rpi varieties?
And how the economics may change if they got more die/wafer going it alone?
Sorry, not really following?
My current theory is that as the cost to make the cheap went up (because they are using a more advanced process node) they needed to cramp more stuff into the chip to justify the expense -- so even though the node has a lot more transistors per mm2 they end up using more mm2 in total.
Tim 'mithro' Ansell
My current theory is that as the cost to make the cheap went up (because they are using a more advanced process node) they needed to cramp more stuff into the chip to justify the expense -- so even though the node has a lot more transistors per mm2 they end up using more mm2 in total.
Yeah, it's unfortunate there was no data for cache mm^2
they needed to cramp more stuff into the chip to justify the expense
You are thinking this is more than just the cache upgrades?
Tim 'mithro' Ansell
I think there is some interesting things there -- like the mm2 went up despite the process node going down, the different revisions on the same process node, & the fact that RPi -- which isn't that of a small customer only went with custom silicon as a last resort despite having 1/3rd of their silicon take up by stuff they didn't use.
I wonder what the world looks like if their second / third design used custom silicon.....
I think defining what "set top blocks" means is important. RPi might have thought they were great features, then after tapeout some management structures changes, and the new folks said "don't waste money implementing the firmware and software, it isn't worth how much $$ it will rake in from customers"
nmz787
I think defining what "set top blocks" means is important. RPi might have thought they were great features, then after tapeout some management structures changes, and the new folks said "don't waste money implementing the firmware and software, it isn't worth how much $$ it will rake in from customers"
@nmz787 - I'm pretty sure that it is "set top box" -- IE Things needed for cable boxes, DVRs, android tv sticks, etc. -- Not a bunch of functionality requested by RPi.
11:28 p.m.
@nmz787 - IIRC There was some stuff from when RPi was originally starting that they were just using some SoC Broadcom had made for another purpose that was extremely cheap due to massive orders never happening.