threadstomper58
|
|
# ? Sep 17, 2020 17:17 |
|
|
# ? Jun 12, 2024 12:16 |
|
Laslow posted:threadstomper58
|
# ? Sep 17, 2020 17:40 |
|
https://www.youtube.com/watch?v=TZRgnlRPTTQ&t=4s
|
# ? Sep 17, 2020 18:32 |
|
threadthreader
|
# ? Sep 17, 2020 19:34 |
|
double threader yes, this is threading close to the line
|
# ? Sep 17, 2020 19:56 |
|
pram posted:have you literally never heard of SGI lol The Management posted:lmao yes I’m aware of all that, and? I’m not disputing that mips was used in supercomputers, I’m disputing that it was “really for unix servers”, when the vast majority of mips devices have always been consumer electronics, embedded devices, etc. there’s no way mips belongs in a list of unix server chips if x86 doesn’t.
|
# ? Sep 18, 2020 01:49 |
|
echinopsis posted:threadthreader Needlethreader Spending extra cycles to get your exact results
|
# ? Sep 18, 2020 01:55 |
|
camelthreader, for all your biblical rich person apologia needs
|
# ? Sep 18, 2020 03:45 |
|
Soricidus posted:yes I’m aware of all that, and? mips was born as a unix server chip and for like a decade plus as a unix server chip before the rest of all that
|
# ? Sep 18, 2020 03:54 |
|
ah, but you see by sheer numbers it's a faaaaaaart
|
# ? Sep 18, 2020 03:57 |
|
hurf durf yes 1987s $10k workstation bespoke 32 bit processor on the cutting edge of chip design is has always been an embedded chip
|
# ? Sep 18, 2020 03:58 |
|
mips and amd64 adjacent story one of the first fast amd64 compilers was from a company called pathscale which was originally a mips compiler for sgi they also developed an 10g infiniband card qlogic bought them for that card (and later sold to intel who canceled the tech a decade later) and then sold the compiler to sicortex which was an supercomputer company that glued a bunch of mips cores together who went out of business because it was weird cray bought the assets and sold it to some of the original team but everyone had moved on to llvm or accelerators anyway
|
# ? Sep 18, 2020 04:19 |
|
Laslow posted:intel has the cash to ad blitz the hell out of some nonsense like avx 512 but they lack their cold cynical hearts that they had in the 90’s. breaking out the breakdancing clean suit guys and hyping up “Advanced Vector X” technology instead of Core i7 n+1 gen will move way more units because people are still idiot rubes, now more than ever. i know this is from a couple pages back but i don't think they're really in a position to need to do this. last i heard they're still selling chips as fast as they can make them.
|
# ? Sep 19, 2020 09:21 |
|
intel chips might as well be cheetos
|
# ? Sep 19, 2020 11:01 |
|
Farmer Crack-rear end posted:i know this is from a couple pages back but i don't think they're really in a position to need to do this. last i heard they're still selling chips as fast as they can make them. if my old co-workers are anything to go by, i'm not surprised. "i don't trust that AMD - their stuff always has so many issues. if you want something that just works, you buy intel", said one of them a month or so ago
|
# ? Sep 19, 2020 13:48 |
|
echinopsis posted:gently caress suppose it’s another one of those learning moments isn’t it E: gently caress, now I want kung pao tofu. Outpost here makes pro-tier kung pao tofu.
|
# ? Sep 19, 2020 15:23 |
|
Instant Grat posted:if my old co-workers are anything to go by, i'm not surprised. "i don't trust that AMD - their stuff always has so many issues. if you want something that just works, you buy intel", said one of them a month or so ago it's not that, it's just that demand is through the roof. AMD can't keep up with orders either.
|
# ? Sep 19, 2020 16:24 |
|
if/when demand craters, then we might see more cynical advertising tricks out of intel again to hold off AMD. but for now they've got plenty of business
|
# ? Sep 19, 2020 16:26 |
|
threadhugger 🤗
|
# ? Sep 21, 2020 05:24 |
|
threadrespecter
|
# ? Sep 21, 2020 06:03 |
|
threadsupporter
|
# ? Sep 21, 2020 06:08 |
|
threadambivalence
|
# ? Sep 21, 2020 07:12 |
|
threadthreader
|
# ? Sep 21, 2020 07:51 |
|
Department of threadland security
|
# ? Sep 21, 2020 07:57 |
|
threadward ripperhands
|
# ? Sep 21, 2020 08:01 |
|
Wet rear end Threads
|
# ? Sep 21, 2020 08:25 |
|
Sereri posted:Department of threadland security now that's savage
|
# ? Sep 21, 2020 08:26 |
|
threadatomizer straight into my systems lungs Private Speech fucked around with this message at 08:30 on Sep 21, 2020 |
# ? Sep 21, 2020 08:26 |
|
threadfrayer
|
# ? Sep 21, 2020 09:04 |
|
Private Speech posted:threadatomizer similarly, the threadomizer, named after the late thread simmons
|
# ? Sep 21, 2020 11:29 |
|
Thredanos
|
# ? Sep 21, 2020 19:49 |
|
I'm jizzin' for Ryzen!
|
# ? Sep 21, 2020 19:52 |
|
Pittsburgh Fentanyl Cloud posted:I'm jizzin' for Ryzen!
|
# ? Sep 21, 2020 23:30 |
|
https://youtu.be/1kQUXpZpLXI haven’t had a chance to look at this but may be interesting?
|
# ? Sep 22, 2020 02:00 |
|
echinopsis posted:https://youtu.be/1kQUXpZpLXI it’s pretty accurate. it confirms that the 6T SRAM cell gate fin is a little smaller in the tsmc 7nm process than the 14nm intel process. looks like the tsmc process is somewhere between 10 and 20% smaller, but again, thats only one dimension of one kind of transistor. sram cells are usually good to compare because they’re generally the most optimized cell in terms of area. you have a limited 2D space on the surface of the wafer and you have 6 transistors per SRAM cell (usually just called a 6T SRAM cell) which translates directly to those exact same 6 transistors repeated over and over again in ever cache on the die, (one 6T cell for each but in the cache bit) so putting a little more work into the size of those transistors translates directly into manufacturing cost savings. and yeah, you can’t really compare them by the marketing name, but you can average out what all the different areas of the chip (caches, logic, etc) and come up with some average densities. the point that you could call it blueberries or strawberries is pretty true, apples to oranges, etc. they try to scale the overall transistor dimensions by around .7, because that’s the sqrt(2)/2 and roughly means “half the size” in terms of wafer area. if you have a square on the chip where each side is 1 hotdog long, that square has an area of 1 square hotdog. replacing it with a square that does the same function but is .7 hotdogs by .7 hotdogs gets you a square that is .7 * .7 = 0.5 square hotdogs, and you can now put twice as many squares on the same wafer. but no company started with a really accurate standard for how big a hotdog was, and sometimes your new hotdog machine can make the same thing the last one did in a rectangle that is actually .5 hotdogs by .8 hotdogs, or something like that. you can compare them in other ways, but using each company’s name for how small their chip is relative to their last one doesn’t really mean anything because they all just pick a number that is .7 times whatever their last number was and call it that shrinking from “10nm” to “7nm” is just changing a hotdog marketing number from 1 to 0.7. it just means: “this process is smaller *enough*, when compared to the last one we made, that we decided to run with it and say it’s about half the area of our last one” 22nm * 0.7 is 14nm 14nm * 0.7 is 10nm 10nm * 0.7 is 7nm
|
# ? Sep 22, 2020 03:23 |
|
I thought it was common knowledge to anyone that wasn’t a dumb gamer that the fabs are roughly on par with each other, and one fab’s 10nm is another’s 7. It’s the same marketing fluff that led to companies crowing about their console being 64-bit* * actually 2 32-bit cpus glued together
|
# ? Sep 22, 2020 04:01 |
|
eide, please write a book that explains professor manufacture but use processed meats and other foods instead of real jargon. you’ll get famous
|
# ? Sep 22, 2020 04:07 |
|
well basically you have to be smart enough to get a phd but dumb enough to stay in academia
|
# ? Sep 22, 2020 04:09 |
|
Raere posted:I thought it was common knowledge to anyone that wasn’t a dumb gamer that the fabs are roughly on par with each other, and one fab’s 10nm is another’s 7. It’s the same marketing fluff that led to companies crowing about their console being 64-bit* i think gamers are still stuck on brand warfare but youll see plenty of dumb fab speculation on macrumors
|
# ? Sep 22, 2020 05:15 |
|
|
# ? Jun 12, 2024 12:16 |
|
are we talking about the same people who returned their iPhone because it had the wrong cpu fabricator?
|
# ? Sep 22, 2020 06:19 |