// #eevblog — 2019-10-05
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
00:32:07 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
00:46:48 Quit: PsySc0rpi0n ([email protected]) Ping timeout: 121 seconds
02:24:20 Join: DHess_US ([email protected]) to #EEVblog
03:06:06 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
03:27:57 <rust_collector> aaah, IS WEEKEND!
03:39:17 Join: DHess_US ([email protected]) to #EEVblog
03:44:10 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
05:44:29 Join: DHess_US ([email protected]) to #EEVblog
05:49:07 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
05:49:23 Join: DHess_US ([email protected]) to #EEVblog
06:12:08 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
08:05:26 Join: DHess_US ([email protected]) to #EEVblog
08:25:06 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
08:37:39 <Halcyon_> YAY! Weekend
08:38:59 <rust_collector> end of week, sort of
09:20:58 Quit: cheater ([email protected]) Ping timeout: 121 seconds
09:23:05 Join: cheater ([email protected]) to #EEVblog
09:27:07 Quit: cheater ([email protected]) Ping timeout: 121 seconds
09:28:05 Join: cheater ([email protected]) to #EEVblog
09:50:30 Join: DHess_US ([email protected]) to #EEVblog
09:50:57 <DHess_US> I declare today a holiday. Linus himself responded to one of my posts.
09:54:53 <T3sl4> did he call you a pedo?
10:13:56 <Ampera> Hi Halcyon_
10:14:54 <DHess_US> Nope, and he didn't even call me an idiot.
10:15:08 <Ampera> Which Linus?
10:15:12 <Ampera> Torvalds or Sebastian
10:15:14 <DHess_US> Linus Torvalds.
10:15:22 <coppice> Snoopy's pall
10:15:31 <Ampera> That's pretty neat, what was it about?
10:16:40 <DHess_US> The discussion was about the economics of custom ARM core development. Samsung is apparently going to stop custom core development and switch to using standard designs from ARM.
10:17:22 <Ampera> Processor design has been pissing me off as of late
10:17:55 <DHess_US> How so? I'm more pissed off about "personal computers" which lack decent keyboards, decent monitors, and decent interface ports.
10:18:03 <DHess_US> i.e. Apple sells toys.
10:18:04 <Ampera> Even with the recent surge of processing power, it all still seems almost unused and just extra shit
10:18:13 <DHess_US> And everybody is copying Apple.
10:18:25 <Ampera> That's what you get for buying mobile devices.
10:18:29 <coppice> custom ARMs have always ended up a losing strategy, whether its custom architecture or people doing tight layouts. it remains to be seen how Apple's work plays out long term
10:18:36 <DHess_US> It applies to Apple PCs as well.
10:18:50 <Ampera> well that's also what you get for buying Apple, but I thought that was a given
10:19:08 <DHess_US> It wasn't losing at one time coppice but that was basically what the discussion was about.
10:19:29 <coppice> when was it not losing?
10:19:41 <DHess_US> My real complaint is that the other PC makers are copying Apple's bad designs.
10:19:48 <Ampera> I think Torvalds would actually be a pretty neat guy to know, because he sounds the exact sort of person I am, and that I make friends with easily
10:20:21 <DHess_US> Except for the being Finnish part.
10:20:38 <Ampera> One of my good friends is Finnish.
10:21:08 <DHess_US> That's why I mentioned him *not* calling me an idiot. Linus tends to be, um, confrontational?
10:21:15 * Ampera yeets Halcyon_ back to Suomi where he belongs
10:21:30 <DHess_US> For me it took some getting used to.
10:21:30 <Ampera> DHess_US, exactly, the sort of people I tend to become good friends with
10:21:41 <Ampera> since I have no problem with being called an idiot, nor calling other people idiots
10:21:52 <DHess_US> Yea, but I'm around Americans all the time and they tend not to do that whether it is true or not.
10:22:00 <coppice> Linus usually follows an abuse first and think later strategy, but it somehow ends up being more of a strength than a weakness
10:22:18 <Ampera> DHess_US, not sure what sort of Americans you've been with, but I find it's pretty universal within a subset of them.
10:22:27 <Ampera> As in, being a dick is something shared by a good number of people
10:22:54 <DHess_US> Here is the start of the discussion if you want to read it or participate. https://www.realworldtech.com/forum/?threadid=187531&curpostid=187531
10:24:16 <Ampera> that's okay
10:27:24 <coppice> whether its MCUs or MPUs or big SoCs, if you think the core is super important you've lost the plot. Even with the current Intel vs AMD battle the core design is less important than the decisions they are taking about how to use them
10:27:45 <Ampera> Because you simply can't pack more into a core anymore like you used to
10:28:17 <Ampera> architectural improvements, even though they don't mean too much compared to the previous leaps and bounds possible with fab tech shrinking, is all that's really left
10:28:26 <Ampera> and in that realm, AMD is kicking the fucking snot out of Intel.
10:28:49 <Ampera> I mean they're kicking the snot out of Intel in fab tech too, which is funny since they don't even own any fabs
10:28:51 <coppice> If I were building a phone SoC, I'd really want to use an available core, so I can put all my focus in getting the very best modem and other modules
10:29:00 <Ampera> ^
10:29:45 <Ampera> At the end of the day phones are at a point where they don't really benefit from more power anyways, besides letting developers be even lazier with optimizing their awful apps.
10:30:28 <coppice> right now being the first to have a proper 5G modem is where you can differentiate
10:30:50 <DHess_US> I wouldn't go that far. Intel still has better single thread performance unless meltdown mitigation is included. What AMD has done is catch up to the point where the difference is insignificant so AMD can compete on price and other features.
10:31:41 <Ampera> DHess_US, AMD's single threaded is actually quite close to Intel's, but their multithreaded and architectural implementations (Intel's 10th gen chips still have PCIe 3.0) do beat the snot out of Intel
10:31:44 <DHess_US> And Intel is constrained by its server business to not allow its desktop processors to compete with its server processors.
10:32:23 <Ampera> As far as servers, anecdotally AMD seems to have a bit of a trust issue with some people, but Rome chips are a whole new level of insane
10:32:23 <DHess_US> Which is why you find low end Intel processors with ECC and Xeon processors with ECC but not desktop processors.
10:33:12 <coppice> the low end intel processors go into infrastructure, so they must have ECC
10:34:36 <coppice> does intel ship anything with a long term supply commitment that doesn't have ECC?
10:35:33 <Ampera> Intel ships so many chips it'd probably take you a day to check
10:39:16 <coppice> they do love to turn a single die into a thousand variants
10:40:59 <DHess_US> That is an important part of their business plan.
10:40:59 <DHess_US> The embedded part is tricky because you *still* have to pay a premium for the south bridge to unlock ECC. I think that is how they differentiate on price.
10:41:00 <DHess_US> I suspect that market only exists to keep ARM out.
10:41:23 <DHess_US> If ARM gained a foothold there, and they are slowly doing so, then it is a short distance to laptops and desktops.
10:41:49 <Ampera> Why would it need a specific southbridge to support ECC?
10:42:23 <DHess_US> Because Intel requires it through some sort of authentication between the CPU and south bridge.
10:42:46 <Ampera> Sounds like Intel alright.
10:42:55 <DHess_US> That is why ECC on a Xeon processor is disabled if you use a consumer south bridge instead of a server or workstation south bridge.
10:43:48 <DHess_US> There is an overlap between standard and Xeon processors with socket 1150 or 1151 where you can use either processor in either motherboard but ECC requires both a Xeon processor and the server south bridge.
10:44:10 <DHess_US> The server south bridges all start with Cxxx or something.
10:44:26 <DHess_US> There might be an exception now.
10:49:21 <DHess_US> The exception I remember is a fairly recent highest end Intel desktop chipset but what supports ECC with what is so confusing that I just ignore them.
10:50:03 <DHess_US> On the other hand there is AMD where practically all CPUs support ECC and Pro GPUs support ECC.
10:50:13 <DHess_US> err, I mean Pro APUs.
10:53:41 <DHess_US> Hmm, past dinner time. No wonder I am hungry.
11:16:42 <Ampera> I would actually like to see a performant RISC CPU come into reasonable affordability
11:17:09 <coppice> isn't ARM RISCy enough for you?
11:17:12 <Ampera> If modern POWER implementations were cheaper, I'd actually be quite interested.
11:17:28 <Ampera> coppice, ARM isn't fast enough for me, at least not in its affordable implementations.
11:18:05 <coppice> RISC/V looks like it might well really get somewhere in the marketplace, and that will drive high performance versions
11:18:06 <Ampera> The server ARM implementations are sort of manycore solutions that aren't too fun to tinker around with if you have real work you want to give them
11:19:13 <Ampera> Indeed, but people say the exact same thing about ARM
11:19:23 <Ampera> and some people at IBM have wet dreams about that too
11:20:19 <coppice> The greater openness of RISC/V could well stimulate a broader range of developments than you see with ARM
11:20:55 <Ampera> The thing is while that's all well and good, will big iron actually want to replace their x86 gear at all?
11:21:03 <Ampera> If nobody's gonna buy it, nobody's gonna make it
11:22:19 <coppice> HPC absorbs 20% to 25% of all server chips, and they change every year
11:22:51 <Ampera> Yeah but changing x86 to x86 isn't really that much of a deal
11:23:21 <Ampera> particularly when modern workloads are often virtualized or containerized, so you can just commission one machine as you decommission another.
11:23:50 <coppice> So, you just go for the HPC business and try to grow out from there
11:24:04 <Ampera> I believe it was Torvalds that actually had the postulation that the reason x86 is so popular in the server and big iron markets is because all the developers ran x86 workstations
11:24:07 <coppice> there is very little lock in with HPC
11:24:10 <Ampera> so they could run their code native
11:24:23 <coppice> yep, he's right
11:24:35 <Ampera> I'd imagine it would depend on the application, since a lot of that would be fairly modular code that can be plopped to other machines, but that's time and money
11:24:37 <coppice> but that's much less an issue with HPC
11:25:23 <Ampera> The thing is though there have been plenty of RISC solutions in the past, and none of them (particularly after 2000) really took hold
11:25:38 <Ampera> if big iron really wanted RISC, surely they would have done it by now
11:25:59 <Ampera> with SPARC, Alpha, MIPS, POWER, etc.
11:26:01 <coppice> SPARC, PA, POWER and MIPS sold a LOT of kit in their day
11:26:09 <Ampera> I know they did.
11:26:12 <Ampera> Then they didn't.
11:26:39 <coppice> everything comes and goes
11:26:48 <Ampera> x86 didn't
11:26:58 <Ampera> x86 has been in use for one reason or another since the late 70's
11:27:20 <coppice> it did. everything is x86_64 now, and that is due to Intel's failings, not their success
11:27:48 <Ampera> that's kinda pedantic though, since it's still the same family, regardless of what's been added on to it or who makes/designs it
11:28:05 <Ampera> you can take a brand new Rome EPYC CPU off the shelf today and run code written for a 5150 on it
11:28:15 <coppice> but its the same family because of AMD, not Intel. Intel would have run it into the ground
11:28:47 <Ampera> So? It's still been around longer than any other architecture used today, unlike the other RISC offerings.
11:28:56 <coppice> it was an EPIC failure
11:29:40 <coppice> All the big successes are studen projects meant as a stopgap while the "serious" work went on in parallel
11:29:56 <Ampera> SPARC is mostly dead, MIPS is dead, Alpha is dead, POWER is in very limited use, PA is afaik dead. These aren't the sort of, modern versions are in use but the base idea is dead.
11:30:16 <Ampera> Those chips are D E A D for modern use in anything but legacy applications, whereas x86, in one form or another, is still here
11:30:29 <Ampera> and I just can't see people jumping ship from that to a RISC platform if they haven't already.
11:31:17 <coppice> a lot of the HPC people have jumped to GPUs, and they are a true PITA to switch to
11:31:32 <Ampera> That too, and yes.
11:32:04 <Ampera> GPUs are the one sort of GPIC I don't even remotely know how to program.
11:32:30 <Ampera> (not that I'm a programing genius by any means, the point being they are very different from normal CPUs)
11:34:04 <coppice> I was programming the forerunners of GPUs in the early 80s. I've always felt comfortable with that kind of environment
11:34:52 <Ampera> Tbh this is out of my league, I'm an awful programmer, and almost dislike doing it.
11:35:05 <Ampera> so before I say something retarded I'll just stop it and say idfk
11:35:31 <Ampera> I'd say my current best field in computing is systems configuration and management, as well as hardware configuration and management.
11:35:45 <Ampera> Not on an enterprise level, but on a personal one.
11:39:24 <DHess_US> I have read Torvalds say that and I agree. PC hardware is what made x86 take over and there is nothing similar for ARM which he has lamented multiple times over the years.
11:40:10 <DHess_US> Something with lots of PCIe slots and SATA ports and memory slots that fits in an ATX case and uses an ATX power supply.
11:40:30 <DHess_US> And that makes use of a 65 watt or higher CPU power budget.
11:41:37 <DHess_US> Apple has some ARM CPUs that might work for it but they sure are not going to sell them to others or make such hardware.
11:41:43 <coppice> he's obviously right for general purpose applications
11:42:02 <DHess_US> No, the hardware needs to cover the whole market for economy of scale.
11:42:58 <DHess_US> We already have dedicated ARM hardware for every market and it is junk for developement purposes.
11:43:02 <coppice> that doesn't stop GPUs moving into HPC. anything doing serious maths has always been apart from the general computer market
11:43:35 <DHess_US> What does that have to do with ARM becoming a desktop and thereby cloud standard?
11:44:37 <coppice> its the general purpose issue. anything general purpose needs to run everywhere, but they doesn't stop some key types of computation sitting apart from that need
11:46:09 <EEVBlog> New EEVBlog Tweet: That's some exodus from @Tesla https://t.co/XmBAXETx4y
11:46:52 <DHess_US> What I am getting at is that the original IBM PC was a general purpose design and could be used as the base for multiple markets including developement. Nothing currently using ARM like that exists.
11:47:11 <DHess_US> Serious ARM developement platforms run x86.
11:48:00 <coppice> interestingly the people developing the many ARM core server chips didn't seem to understand that they couldn't possibly succeed without getting some version of their devices onto desktops
11:49:06 <Ampera> ^
11:49:39 <Ampera> I like the idea of RISC because it's neat and different, but my x86 machine does what I need it to do. RISC machines, no matter how spectacular they are, unless they can emulate x86 faster than x86 native for the same price bracket, will never do that.
11:52:40 <coppice> the core idea of RISC is a fallacy. the biggest bottleneck is access to main memory. anything which increases the amount of stuff you need to move to or from main memory puts you at a disadvantage. RISC code always ends up bigger
12:19:10 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
12:47:23 Join: DHess_US ([email protected]) to #EEVblog
13:14:08 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
13:31:24 Join: DHess_US ([email protected]) to #EEVblog
13:31:29 <DHess_US> Christ, does every new movie include commentary about Trump?
13:40:24 <Ampera> DHess_US, are you complaining about the glorious representations about our great leader?
13:40:38 <Ampera> Maybe some concentrated reeducation would be appropriate...
13:42:34 <DHess_US> I nominate you to come reeducate me. Wear body armor so I can test penetration of different loads.
13:45:25 <Ampera> DHess_US, fill me with your load daddy
13:45:30 <Ampera> I want you to penetrate me so bad
13:45:44 <Ampera> uwu
13:45:59 <DHess_US> Yea, I knew after I wrote that that it was going to go too far.
13:46:18 <Ampera> please, daddy hessy, go too far inside of me
13:46:41 <DHess_US> Seriously, do people really expect Trump reeducation camps?
13:49:58 Join: ibiza_ ([email protected]) to #EEVblog
13:49:58 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
13:49:58 Quit: ibiza_ ([email protected]) Changing hosts
13:49:58 Join: ibiza_ ([email protected]) to #EEVblog
13:50:13 Nick: ibiza_ is now known as ibiza
13:56:00 Quit: ibiza ([email protected]) Ping timeout: 121 seconds
13:56:14 Join: ibiza ([email protected]) to #EEVblog
13:56:23 Quit: ibiza ([email protected]) Changing hosts
13:56:24 Join: ibiza ([email protected]) to #EEVblog
14:00:38 Join: ibiza_ ([email protected]) to #EEVblog
14:00:38 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
14:00:38 Quit: ibiza_ ([email protected]) Changing hosts
14:00:38 Join: ibiza_ ([email protected]) to #EEVblog
14:00:53 Nick: ibiza_ is now known as ibiza
14:14:15 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
14:14:15 Join: ibiza_ ([email protected]) to #EEVblog
14:14:30 Nick: ibiza_ is now known as ibiza
14:17:41 Join: ibiza_ ([email protected]) to #EEVblog
14:17:41 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
14:17:41 Quit: ibiza_ ([email protected]) Changing hosts
14:17:41 Join: ibiza_ ([email protected]) to #EEVblog
14:17:55 Nick: ibiza_ is now known as ibiza
14:31:41 Join: Agripa2 ([email protected]) to #EEVblog
14:32:07 Quit: DHess_US ([email protected]) Ping timeout: 121 seconds
14:48:44 Quit: ibiza ([email protected]) Ping timeout: 121 seconds
14:50:55 Join: ibiza ([email protected]) to #EEVblog
14:54:19 Join: ibiza_ ([email protected]) to #EEVblog
14:54:19 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
14:54:19 Quit: ibiza_ ([email protected]) Changing hosts
14:54:19 Join: ibiza_ ([email protected]) to #EEVblog
14:54:34 Nick: ibiza_ is now known as ibiza
14:59:09 Quit: Agripa2 ([email protected]) Ping timeout: 121 seconds
15:00:29 Join: Agripa2 ([email protected]) to #EEVblog
15:37:56 Join: ibiza_ ([email protected]) to #EEVblog
15:37:56 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
15:37:56 Quit: ibiza_ ([email protected]) Changing hosts
15:37:56 Join: ibiza_ ([email protected]) to #EEVblog
15:38:10 Nick: ibiza_ is now known as ibiza
15:43:01 Join: ibiza_ ([email protected]) to #EEVblog
15:43:01 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
15:43:01 Quit: ibiza_ ([email protected]) Changing hosts
15:43:01 Join: ibiza_ ([email protected]) to #EEVblog
15:43:16 Nick: ibiza_ is now known as ibiza
15:49:26 Join: ibiza_ ([email protected]) to #EEVblog
15:49:26 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
15:49:26 Quit: ibiza_ ([email protected]) Changing hosts
15:49:26 Join: ibiza_ ([email protected]) to #EEVblog
15:49:41 Nick: ibiza_ is now known as ibiza
15:55:11 Join: ibiza_ ([email protected]) to #EEVblog
15:55:11 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
15:55:11 Quit: ibiza_ ([email protected]) Changing hosts
15:55:11 Join: ibiza_ ([email protected]) to #EEVblog
15:58:15 Join: ibiza ([email protected]) to #EEVblog
15:58:25 Quit: ibiza ([email protected]) Changing hosts
15:58:25 Join: ibiza ([email protected]) to #EEVblog
15:59:20 Quit: ibiza_ ([email protected]) Ping timeout: 121 seconds
16:04:21 Join: ibiza_ ([email protected]) to #EEVblog
16:04:21 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
16:04:21 Quit: ibiza_ ([email protected]) Changing hosts
16:04:21 Join: ibiza_ ([email protected]) to #EEVblog
16:04:36 Nick: ibiza_ is now known as ibiza
16:19:07 Quit: Agripa2 ([email protected]) Ping timeout: 121 seconds
16:25:26 <DogWomble> quick! everybody do thje netsplti shuffle@
16:26:34 <T3sl4> (/^-^)/ netsplit dance!
16:32:03 <EEVBlog> New EEVBlog Tweet: RT @JohnCarreyrou: Elizabeth Holmes can't pay her lawyers. Attorneys from the firm Cooley argue in a court filing that you can’t get blood…
16:37:05 <EEVBlog> New EEVBlog Tweet: Being broke is just the start of her sweet comeuppance! Attorneys say disgraced Theranos founder Elizabeth Holmes isn’t paying them, and they don’t want to be her lawyers anymore. https://t.co/sBKpcIFvy5
16:51:10 <EEVBlog> New EEVBlog Tweet: EEVblog #1250 – World’s Thinnest Wallet Review – Stealth Razor https://t.co/PEwYNR64NM
17:27:37 Quit: ibiza ([email protected]) austnet.org!NickOP (Kill requested by ibiza_)
17:27:37 Join: ibiza_ ([email protected]) to #EEVblog
17:28:23 Join: ibiza ([email protected]) to #EEVblog
17:28:33 Quit: ibiza ([email protected]) Changing hosts
17:28:33 Join: ibiza ([email protected]) to #EEVblog
17:31:40 Quit: ibiza_ ([email protected]) Ping timeout: 121 seconds
18:02:33 Quit: steve30 ([email protected]) Quit: <Baljem> steve30: by the way, I had the most bizarre dream the other day that someone came in here and seriously flipped their shit over your quit message and why aren't the ops /doing/ something about it
18:14:23 <Stiff> hayclon life running race 3 at the gardens
18:14:37 <Stiff> paying $9, get on it
18:15:33 Join: Agripa2 ([email protected]) to #EEVblog
18:22:13 <Stiff> ah well ran second, ill pocket the dividends cheers halcyon
18:36:07 Quit: Agripa2 ([email protected]) Ping timeout: 121 seconds
18:37:51 <Stiff> halycon princess race 4 @ the gardens
18:38:05 <Stiff> 3.90 GET ONNNN ITTTT
18:40:13 <Stiff> winner winner chicken dinner
18:42:33 <Stiff> im sorry ive had way to many melbourne tinnies today cant wait to go to sleep
18:43:26 Join: Agripa2 ([email protected]) to #EEVblog
18:56:07 Quit: Agripa2 ([email protected]) Ping timeout: 121 seconds
18:57:03 Join: narayan ([email protected]) to #EEVblog
18:58:02 Nick: narayan is now known as Hello5hitty
20:52:28 Join: Agripa2 ([email protected]) to #EEVblog
21:01:14 Quit: DogWomble ([email protected]) Ping timeout: 121 seconds
21:23:20 Join: steve30 ([email protected]) to #EEVblog
21:25:32 Join: blackbird ([email protected]) to #EEVblog
21:33:09 Quit: Agripa2 ([email protected]) Ping timeout: 121 seconds
21:54:03 Quit: blackbird ([email protected]) Quit: Ik ga weg
22:47:21 <cheater> suddenly youtube ads seem to confuse me with my neighbors who are a middle aged indian guy married to an 80 year old austrian woman so i keep getting ads about indian music streaming services (ad in indian) and ads about denture adhesive (in german)
22:51:49 <steve30> You could stick your dentures in while listening to Indian music
22:57:12 <Ampera> Well they say love has no age
22:57:30 <Ampera> that being said the people who usually say that are in prison and on one or more publicly kept lists
23:29:30 Join: Agripa2 ([email protected]) to #EEVblog
23:38:07 Quit: Agripa2 ([email protected]) Ping timeout: 121 seconds
23:43:12 Quit: steve30 ([email protected]) Quit: Lost terminal
Search this day's buffer: