It’s also not very intensive, especially since video decode usually runs on the GPU.
(Or it should - browsers are not always very good about this but the tools are there, they just don’t want to pay licensing fees to use them.)
I always cringe a bit when people insist they need 6 cores “so they can have discord and Spotify open while they game”. People don’t generally have a very good intuitive sense of just how little cpu time it takes to run those (a lot of times the “at the desktop” measurements are not even with the processor clocked up, it’s running in the lowest active power state), and there is almost always a little cpu headroom due to non-perfect game scaling. Context switching is cheap, your OS is running thousands of threads at any given time, but people don’t have a good intuitive sense of the overhead of context switching either.
Playing a video and writing a document is something that can be trivially done on a core2duo era processor, or the modern equivalent of a low-end dual-core nuc with e-cores (atom cores) like a J4005. You don’t even need performance cores for that.
The tradeoff Intel is making is that some tasks need performance cores because they can’t really be parallelized all that well, and embarrassingly parallel stuff is more efficient to implement with lots of smaller cores. In principle this is also more power efficient, like big.LITTLE in phones, but Intel actually seems to have fairly poor power scaling - they are space efficient but not all that power efficient. This is also partially due to the fact that the little cores share a core voltage with the big cores - the big cores need higher voltage to clock up and that kills efficiency on the little cores. At that point you may as well clock the little cores up too since they have to share the voltage anyway, and that’s an inefficient way to run them.
As Agner Fog mentions, the AVX-512 is also present in silicon yet disabled, which hugely blows up core size for almost no benefit (it’s dark silicon which does have some minor thermal benefits). Alder Lake is in many ways not that well-designed at a high level, but it is magnificently fast on those performance cores and the e-cores would be very nice if they weren’t conjoined awkwardly to the p-cores. Gracemont-based server and atom chips are going to be nifty when they finally arrive.
Apple is still way ahead with Icestorm though - those cores are absolutely absurdly tiny for the performance they give. They are 0.5mm2 each iirc, even considering the node shrink that is teeny-tiny, they are half as many transistors as a Zen3 core and almost the same performance (meaning much higher IPC) or something along those lines. Firestorm is neat but Icestorm didn’t get nearly the attention and it is really insanely fast for how few transistors it uses.
> Playing a video and writing a document is something that can be trivially done on a core2duo era processor, or the modern equivalent of a low-end dual-core nuc with e-cores (atom cores) like a J4005. You don’t even need performance cores for that.
Playing a video and editing a document, sure. But change that to being on a video call and having a few moderately intensive web apps open (say Gmail, Google Docs and some kind of project management software) and your average dual core processor from a few years ago is going to start to struggle.
I do some work on my ThinkPad t420s. It's about a decade old now. I understand it's anecdotal but fwiw. Note that it's plugged into Ethernet.
I do ALL my work on a t25 which is just about 5 years old now. This includes daily calls with dozens of video participants (ms teams and WebEx mostly) while I have chrome and Firefox open with multiple tabs including outlook and Gmail and slack, couple of MS words, couple of PowerPoints, Visio, etc. That's a dual core processor from more than couple of years ago and it's nowhere near capacity with that workload.
It does have lots of ram and Ethernet to residential Fibe 1gbps. But CPU just hasn't been an issue and I plan to work on this another 5 years
With my t25? Participant, absolutely! As I said, it's my main daily workstation. Over last two years that included using built-in webcam, a dsrl, and currently Logitech external webcam.
Heck I've run full PeopleSoft multi vm setups on my ancient t420 back in the day. A little CPU really goes a longer way than people realize.
Video encode is also hardware-accelerated these days. J5005 may be Nehalem performance but unlike Nehalem it has a hardware video encoder onboard. Intel has had it for a long time, I think maybe back to Sandy or Ivy days so it’s basically in everything at this point. It’s not archival quality but it’s more than fine for videoconferencing starting from around Haswell. Goldmont actually has a fairly modern media block (newer than skylake - it has full hevc main10 decode and hdmi 2.0b support) that is backported from the Ice Lake architecture during the years Intel stalled out on 10nm.
I’m just using Nehalem and Conroe as a touchstone that people understand in terms of cpu performance. That is what you get in general cpu performance, but it’s also got better media support and crypto acceleration and so on, so it’s more usable today. Actually a closer look at the numbers puts it just below Sandy Bridge i5 actually.
AMD cpu users (non-apu) and RX 6500/6400 owners obviously need to look elsewhere however. NVIDIA’s 1030 chips are ancient pascal based crap but AMD dropping encoder support on a chip designed during the pandemic era of work-from-home is a bit tone-deaf.
(Yes, 6500 and 6400 were designed during the pandemic, the architecture took longer to develop but 6500 wasn’t the lead ship in its architecture, and taping out a new chip configuration takes much less time than designing a whole architecture from scratch… AMD definitely knew it would be used as a desktop chip too. It was a deliberate space-saving move in the face of their wafer shortages, not something they got blindsided by.)
> It’s not archival quality but it’s more than fine for videoconferencing starting from around Haswell.
My 2015 MacBook Pro with Broadwell i5 would indeed work "fine" for videoconferencing (in my case Google Meet), but it would also use most of a CPU core to do it. Which is fine if that's all you're doing. But if I wanted to share my screen and work on something else while doing so then it could quite quickly get to the point where I needed to start shutting down other programs, and even then it could cause lag when typing in less well-implemented apps.
My new MacBook with M1 Pro cpu it still uses most of a cpu core for videoconferencing, but it's a single E core and I have 9 others cores. Videoconferencing now makes zero difference to system responsiveness.
Most people won't need 8 cores, but I reckon the average office worker in 2022 would benefit from 4.
Sure chat apps and video playback are light, but what isn't is browsing the web. Web browsing has become a heavy task. If you tried to install a browser and then use it today on a core 2 duo, you'd have a very bad time. And everyone browses the web, usually quite a bit.
Does it seem strange to anyone that this might be the case, that web browsing is more resource intensive than streaming and playing video? If true, is it because everyone builds SPAs for everything using heavy frameworks like angular and react?
I can't speak for core 2 duo specifically, but I do tons of browsing on ThinkPad t420s (per parallel thread) which has Core i5-2540M. Depending on specific core 2 duo this shows as about 20 percent faster per interwebs - could be wrong. I find for browsing memory is king. Note I don't have any noscript or js killer, just pure browsers on windows 10.
I recently picked up a 2008 17" MacBook Pro. (2.5ghz core 2 Duo).
It browses the web perfectly fine, plays YouTube, and runs VS code, all that the same time. The only website I've notice it struggles on is Google maps, but Google maps has a 2D mode than runs great.
It did actually surprise me how well it runs and now I've been using it as my main machine for a few weeks.
Core 2 Duo on a late 2007 Macbook running Firefox on Linux Mint. It's doing surprisingly ok. It would not work as a daily driver because waiting one second for a moderately heavy page is not something anyone is willing to accept nowadays but totally functional whereas browsing on a single core 2004 iMac is excruciating.
> I always cringe a bit when people insist they need 6 cores “so they can have discord and Spotify open while they game”.
I mean, they're not wrong. I have Discord sitting open at the login screen and it's burning 10% of a core. Run a few Electron apps and your core2duo suddenly becomes a core1uno.
> Playing a video and writing a document
Using what? VLC + notepad, or Chrome + YouTube + Google Docs? There's a huge difference.
Discord gets a core, Spotify gets a core, Chrome with a few webapps open gets a few cores. That leaves 1 or 2 for a game if you're lucky. God forbid you also try to stream to Twitch. I'll cut the rant short here. You're right though; you should cringe if they insist on 6 cores, instead of 8 or more.
Bloated software is the problem. The worst part is that efficient alternatives exist (Discord -> Ripcord, Spotify -> spotify-tui) but they're against ToS and can get you banned from both services. From what I hear Discord is a lot more likely to ban you. I've not heard of any Spotify bans. But as a user, even if you know the problem, your only choices are to risk an account ban, or buy an "overpowered" CPU to run your chat client.
EDIT: Am I wrong? If you're skeptical, go find a Core 2 Duo and try watching a YouTube video while you're in a Discord call, and let me know how it goes. Those are the kinds of apps people use these days, like it or not. Don't shoot the messenger.
> I mean, they're not wrong. I have Discord sitting open at the login screen and it's burning 10% of a core. Run a few Electron apps and your core2duo suddenly becomes a core1uno.
That is what I said about “people look at the measurement at the desktop and freak out about single-digit utilization without even realizing their processor is sitting in the lowest frequency state”.
Is your processor even boosting all its cores when that 10% measurement is taken? In my experience usually it is not. Maybe different for Zen2/Zen3 but cpu measurement gets very complicated in general for those architectures since it can move between boost states so fast - it can boost so agilely that you can end up with garbage numbers because it’s moving between boost states faster than the task manager can enumerate the threads. It can move between boost states between the time you take the frequency measurement and the time you start enumerating the cpu load. This was a whole thing a couple years ago when people started looking at Zen2 and it turns out most third-party tools can’t handle this and only Ryzen Master gives you an accurate picture of what it’s doing.
In general it is such a small amount of cpu that it’s not worth caring about.
Also, typically while you game you’d have the overlay up and not the actual UI and that’s even faster - showing the login screen is likely slower than actually in game.
Load up a game and the cpu will be loaded up harder and will stay boosted and then take your measurement of discord.
People who are casually making observations about how intensive discord are never do that, it’s just “I see it using 6% at the desktop” [with the processor in the idle state rendering a full ui instead of just the overlay].
I ran a quick `while true {}` in the background. It used 7B cycles per second (twice the clock speed for some reason?), while Discord used about 700M. Discord does drop lower if I minimize the window. But still, thousands of context switches per second, while completely logged out. I have no idea what it could be doing.
You're right that I'm probably not benchmarking rigorously enough. But people don't care about any of this stuff. They care about "my FPS gets lower when Discord is open, pls fix". And they don't want to micromanage CPU usage either. Make sure you minimize Discord after starting the call, make sure you close that gmail tab during the game then open it back up when you're done so you don't miss notifications? No thanks. I'd rather just pay for a few more cores.
I guess what triggered me is you saying you cringe at the end users. They don't deserve that. They just want the stuff they buy to work reliably. And if they need a bunch of extra headroom for that to happen, well, that's just the sorry state of the industry. If you must cringe at someone, direct your cringe at developers.
I don't think Spotify will ban you for optional clients. Other clients are the whole point of their API. There's a bunch of rules you need to follow (largely about attribution) but I think at worst the dev will get their API key revoked - users should be perfectly safe.
Hardware decoding avoids license fees by anything that uses it. The fees are paid by the manufacturer of the hardware doing the encode/decode.
Even then it’s not like they’re paying per unit. All of the patent pools and licenses I know of have a cap and AMD, Intel, Nvidia, etc are easily already there.
Google makes enough money off Chrome and youtube they could afford to pay licensing fees so that end-users have a good experience. Software decoding vs hardware decoding is a massive difference in user experience.
I’ve seen degradation in game play when I have 20+ tabs open in chrome. I have a modern 8core 16thread cpu (ryzen 3700x) 32gb of ram, nvme, and a 3070. It’s noticeable, either bec the tabs are saturating the eth or bec they’re pegging the cpu, but you can clearly feel slower.
(Or it should - browsers are not always very good about this but the tools are there, they just don’t want to pay licensing fees to use them.)
I always cringe a bit when people insist they need 6 cores “so they can have discord and Spotify open while they game”. People don’t generally have a very good intuitive sense of just how little cpu time it takes to run those (a lot of times the “at the desktop” measurements are not even with the processor clocked up, it’s running in the lowest active power state), and there is almost always a little cpu headroom due to non-perfect game scaling. Context switching is cheap, your OS is running thousands of threads at any given time, but people don’t have a good intuitive sense of the overhead of context switching either.
Playing a video and writing a document is something that can be trivially done on a core2duo era processor, or the modern equivalent of a low-end dual-core nuc with e-cores (atom cores) like a J4005. You don’t even need performance cores for that.
The tradeoff Intel is making is that some tasks need performance cores because they can’t really be parallelized all that well, and embarrassingly parallel stuff is more efficient to implement with lots of smaller cores. In principle this is also more power efficient, like big.LITTLE in phones, but Intel actually seems to have fairly poor power scaling - they are space efficient but not all that power efficient. This is also partially due to the fact that the little cores share a core voltage with the big cores - the big cores need higher voltage to clock up and that kills efficiency on the little cores. At that point you may as well clock the little cores up too since they have to share the voltage anyway, and that’s an inefficient way to run them.
As Agner Fog mentions, the AVX-512 is also present in silicon yet disabled, which hugely blows up core size for almost no benefit (it’s dark silicon which does have some minor thermal benefits). Alder Lake is in many ways not that well-designed at a high level, but it is magnificently fast on those performance cores and the e-cores would be very nice if they weren’t conjoined awkwardly to the p-cores. Gracemont-based server and atom chips are going to be nifty when they finally arrive.
Apple is still way ahead with Icestorm though - those cores are absolutely absurdly tiny for the performance they give. They are 0.5mm2 each iirc, even considering the node shrink that is teeny-tiny, they are half as many transistors as a Zen3 core and almost the same performance (meaning much higher IPC) or something along those lines. Firestorm is neat but Icestorm didn’t get nearly the attention and it is really insanely fast for how few transistors it uses.