|
Paul MaudDib posted:The idea on 4+1 is that you build it using slower, lower-power processes than the big cores. Which is completely unnecessary in the x86 world. It doesn't get you anything. You simply clock down one of your cores and it uses much less power naturally. The extra 0.1 watt power saving or whatever isn't worth the significantly more complex situation of having a different architecture core (say atom maybe) on the same die.
|
#
?
Apr 18, 2015 05:54
|
|
|
|
| # ? May 21, 2024 16:07 |
|
|
Paul MaudDib posted:The tradeoff is that your software has to be smart enough to take advantage of it. If your kernel treats the battery-saver core like a normal core you're going to have issues.
|
|
#
?
Apr 18, 2015 06:31
|
|
|
Professor Science posted:No software is smart enough for this. Also, 4+1 is dead in favor of 4+4 A53+A57 (or A53+A53 if you're Huawei lol), which software is even less equipped to handle. I would think it's a pretty straightforward fix - you tweak your kernel scheduler and power manager to prefer the battery-saver core when load is below some threshold. I guess I shouldn't have said "software" - that's a kernel thing. Userland software shouldn't handle processor management. Guess I'm behind the times on that. In terms of being "equipped to handle that" my intuition would be that it's a lot simpler to write a rule for handling one low power core when threshold < X (let's say sysload < ~0.1) than 4 cores, for various reasons.
|
|
#
?
Apr 18, 2015 07:49
|
|
|
The whole 'power saver core' thing really only seems to matter for form factors that are extremely power limited like smartphones. For laptops, desktops, and servers current x86 CPU power saving tech can do a pretty good job in low performance or idle situations. http://techreport.com/review/24879/intel-core-i7-4770k-and-4950hq-haswell-processors-reviewed/7 At idle or in low performance situations the PSU inefficiency is probably a bigger problem for most desktops since most of the cheap 80+ PSU's have crap efficiency under 20% of capacity. Only relatively recently have even the more expensive 80+ Gold PSU's been even trying to resolve that issue.
|
|
#
?
Apr 18, 2015 09:52
|
|
|
Paul MaudDib posted:I would think it's a pretty straightforward fix - you tweak your kernel scheduler and power manager to prefer the battery-saver core when load is below some threshold. I guess I shouldn't have said "software" - that's a kernel thing. Userland software shouldn't handle processor management. Google wasn't able to do it (Android never really took advantage of the 4+1 arrangement properly), which means either they didn't care to because the gains were small, or weren't able to because of limitations with what they could do with the software.
|
|
#
?
Apr 18, 2015 17:36
|
|
|
There was both Samsung Galaxy S 4 phones with quad core "high power usage" cores and quad core "low power usage" cores, and Samsung Galaxy S 4 models with just a single quad core (and faster) CPU. The one with the high and low power switchable set had neglibly more battery life in real world use (something like a few minutes on hours and hours of active use and also a few extra minutes onto the days of standby).
|
|
#
?
Apr 18, 2015 18:09
|
|
|
Paul MaudDib posted:I would think it's a pretty straightforward fix - you tweak your kernel scheduler and power manager to prefer the battery-saver core when load is below some threshold. I guess I shouldn't have said "software" - that's a kernel thing. Userland software shouldn't handle processor management.
|
|
#
?
Apr 18, 2015 19:11
|
|
|
FaustianQ posted:Hey come on now, HP-Compaq recovered I love the words of Sun's ex-CEO describing the merger as "two garbage trucks colliding"
|
|
#
?
Apr 18, 2015 21:54
|
|
|
FaustianQ posted:Hey come on now, HP-Compaq recovered u wot m8 Are we talking about the same HP that's splitting into two companies by the end of this year?
|
|
#
?
Apr 18, 2015 22:52
|
|
|
SwissArmyDruid posted:u wot m8 Which you think will perform better, H or P?
|
|
#
?
Apr 18, 2015 23:34
|
|
|
Wheany posted:Which you think will perform better, H or P? Actually, it's going to be a enterprise/consumer split. I can't remember who holds onto the printer business, but I fear it's the Meg Whitman-led consumer division.
|
|
#
?
Apr 19, 2015 00:30
|
|
|
SwissArmyDruid posted:Actually, it's going to be a enterprise/consumer split. I can't remember who holds onto the printer business, but I fear it's the Meg Whitman-led consumer division. Printers are going to consumer. Whitman's running Enterprise, and the executive VP of printing and PCs will be running conumer.
|
|
#
?
Apr 19, 2015 01:19
|
|
|
Garbage all around
|
|
#
?
Apr 19, 2015 01:53
|
|
|
Don't forget the HP-EDS merger too. Now it's three garbage trucks. HP is a microcosm of tech company mismanagement and a godawful place to work. It survived because it's a giant but its inertia is finally running out. In no way should it be viewed as a successful anything.
|
|
#
?
Apr 19, 2015 02:15
|
|
|
Wikipedia posted:The Wall Street Journal and confirmed by other media, will result in two publicly traded companies: Hewlett-Packard Enterprise and HP, Inc. Meg Whitman will serve as chairman of HP, Inc. and CEO of Hewlett-Packard Enterprise, Patricia Russo will be chairman of the enterprise business, and Dion Weisler will be CEO of HP, Inc. HP: Chair: Meg Whitman CEO: Dion Weisler HPE: Chair: Patricia Russo CEO: Meg Whitman Worse than I thought.
|
|
#
?
Apr 19, 2015 02:37
|
|
|
Professor Science posted:everybody seems to think this (including _lots_ of people in industry) and it's magical thinking, that you can somehow know the load at an exact point in time on the small core and instantly migrate. migration is not free power-wise, it's not free latency-wise, and there's no way to predict the need to migrate from one to the other Nintendo Kid posted:Which is completely unnecessary in the x86 world. It doesn't get you anything. You simply clock down one of your cores and it uses much less power naturally. The extra 0.1 watt power saving or whatever isn't worth the significantly more complex situation of having a different architecture core (say atom maybe) on the same die. And I really think that you want as different an ISA as possible. With similar-enough ISA's people hide behind ACPI and try to take this difficult real-time problem that's very hardware dependent and leave it up to an OS policy with half of those hardware details unavailable. I really don't think that's going to solve it. If you break it and have a completely different ISA, you're ensuring that anyone who wants those power savings knows what they're doing and has invested enough time to understand it.
|
|
#
?
Apr 19, 2015 04:19
|
|
|
Reading an issue of Maximum PC online, back from when the Phenom 9600 was about to be released. http://books.google.com/books?id=bgIAAAAAMBAJ&printsec=frontcover&rview=1#v=onepage&q&f=true It makes me sad to see AMD having spent practically the last 7 years trying to play off their misfortunes as a potential value-market boon. How has AMD survived for so long, especially after being practically ignored by OEMs for so long? Is it inertia? Are we watching a slow death? VIA didn't take as long to slow down to a 14mil/yr revenue business, but then again they didn't have any viable offerings beyond a brief blip with the Nanos back when the industry was saying the netbook was the future. Anyways, these back issues are fun. "Powerline-based networking is the future?" "Hey, those GeForce 8800GT's are good!" "What is this Windows Home Server thing? Oh, it's pretty neat!" A Bad King fucked around with this message at 15:44 on Apr 19, 2015 |
|
#
?
Apr 19, 2015 15:42
|
|
|
Angry Fish posted:How has AMD survived for so long, especially after being practically ignored by OEMs for so long? AMD was never good, just competitive and if only thanks to Netburst head trauma that Intel got over. The P-Ms were competitive with 754 and 939 single cores, and used like 60% less power. Imagine a timeline where Intle never goes netburst and just focuses on releasing updated P-Ms, never losing the performance crown to AMD and beating AMD to any important milestone. Imagine 2004 expect instead of P4Es barely meeting an Athlon 3800+ in performance, some Socket P monster destroys AMDs offerings, 2005 drops C2D, and from 2001-2006 never needs to change the socket undercutting any value capability AMD has. AMD lives because Intel had a "Bulldozer", and it colored consumer perception enough. AMD might also be alive because the smartest thing they ever did was acquire ATI (which, rip Radeon, Samsung please buy!). SwissArmyDruid posted:u wot m8 I wasn't entirely serious, still they did bounce back a little after Miss Fiorina. Hey, did you know she wants to run for POTUS? America, the next HP-Compaq.
|
|
#
?
Apr 19, 2015 15:59
|
|
|
FaustianQ posted:AMD was never good, Now chill that kind of rhetoric. drat. Whose 64-bit instruction set became de facto? Intel's IA-64? HalloKitty fucked around with this message at 16:35 on Apr 19, 2015 |
|
#
?
Apr 19, 2015 16:31
|
|
|
HalloKitty posted:Now chill that kind of rhetoric. drat. Whose 64-bit instruction set became de facto? Intel's IA-64? Correct me if I am wrong, but isn't that mainly because IA-64 was 64bit only and AMD's was 32 and 64bit? Like if you bought an Itanic, you would also be required to get a 64bit OS (which did not exist in the consume world at the time) and upgrade all of your software to 64bit versions, and even today a lot of software is 32bit. Or where there other bigger problems with the Itanium?
|
|
#
?
Apr 19, 2015 16:54
|
|
|
Crotch Fruit posted:Correct me if I am wrong, but isn't that mainly because IA-64 was 64bit only and AMD's was 32 and 64bit? Like if you bought an Itanic, you would also be required to get a 64bit OS (which did not exist in the consume world at the time) and upgrade all of your software to 64bit versions, and even today a lot of software is 32bit. Or where there other bigger problems with the Itanium? Yeah IA-64 was completely unrelated to x86.
|
|
#
?
Apr 19, 2015 17:11
|
|
|
Crotch Fruit posted:Correct me if I am wrong, but isn't that mainly because IA-64 was 64bit only and AMD's was 32 and 64bit? Like if you bought an Itanic, you would also be required to get a 64bit OS (which did not exist in the consume world at the time) and upgrade all of your software to 64bit versions, and even today a lot of software is 32bit. Or where there other bigger problems with the Itanium? AMD added 64 bit instructions to X86, Itanic was a totally different architecture. "AMD was never good" - "some alternate timeline where they weren't"... that's kinda bad posting right there. GRINDCORE MEGGIDO fucked around with this message at 17:15 on Apr 19, 2015 |
|
#
?
Apr 19, 2015 17:11
|
|
|
HalloKitty posted:Now chill that kind of rhetoric. drat. Whose 64-bit instruction set became de facto? Intel's IA-64? And? Yea, AMD definitely came out with a 64-bit instruction set, but still need Intels permission to use x86. Their processors are still poo poo, and their GPUs are hot, and they'll run out of money where Intel just gobbles up the AMD64 license and locks everyone out of x86-64 forever. Hey, VIA isn't total poo poo, they made good chipsets once. Hey, Voodoo isn't bad, they came up with SLI! Hey, Cyrix had a lot of really smart people on their bench, stop calling them bad 
|
|
#
?
Apr 19, 2015 17:14
|
|
|
Crotch Fruit posted:Correct me if I am wrong, but isn't that mainly because IA-64 was 64bit only and AMD's was 32 and 64bit? Like if you bought an Itanic, you would also be required to get a 64bit OS (which did not exist in the consume world at the time) and upgrade all of your software to 64bit versions, and even today a lot of software is 32bit. Or where there other bigger problems with the Itanium? A combination of x86-64 being backwards compatible and the original Itanium's being a dumpster fire sealed the deal for the Itanic. My understanding is that Microsoft in particular made it very clear they supported x86-64 over Itanium because of BC even if IA-64 was the technically superior instruction set. Then came the first Itanium stinking up the joint leaving a bad taste in people's mouths and the first Opteron's being made available soon after that were a much more desirable choice in the server market. That being said, saying AMD was never good was straight up wrong. They were clearly superior to Intel's offerings during the early Athlon 64 era when its competition were Netburst CPU's. 200 dollar Athlon 64's were generally faster than the 1000 dollar Pentium 4 Extreme Edition available at the time.
|
|
#
?
Apr 19, 2015 17:19
|
|
|
FaustianQ posted:And? Yea, AMD definitely came out with a 64-bit instruction set, but still need Intels permission to use x86. Their processors are still poo poo, and their GPUs are hot, and they'll run out of money where Intel just gobbles up the AMD64 license and locks everyone out of x86-64 forever. Are you off your meds?
|
|
#
?
Apr 19, 2015 17:20
|
|
|
It's strange to hear people define good as "not poo poo". AMD not making GBS threads their pants while Intel fucks up with Netburst doesn't make AMD good. If Intel never went netburst and just stayed the course with P-M, AMD would never even have a golden age people could pine and weep for, instead AMD might be competitive up until 2005 instead of 2009.
|
|
#
?
Apr 19, 2015 17:54
|
|
|
FaustianQ posted:they'll run out of money where Intel just gobbles up the AMD64 license and locks everyone out of x86-64 forever. That's not a viable long term play for Intel though. A lot of AMD's existence is owed to second sourcing, and the bigger customers would start funding any alternative they could find if Intel suddenly found themselves sole owners of x86. Intel would have a great run for ~5 years until all the money being dumped into obviating them caught up and they find themselves begging for fab work.
|
|
#
?
Apr 19, 2015 18:21
|
|
|
FaustianQ posted:Bunch of stuff VIA NEVER made good chipsets. Every loving one of them was either unstable, had lovely disk throughput, conflicts with hardware, or some unholy combination of the above. The best thing that could be said about them is that they were usually fairly cheap. JnnyThndrs fucked around with this message at 19:33 on Apr 19, 2015 |
|
#
?
Apr 19, 2015 18:30
|
|
|
JnnyThndrs posted:VIA NEVER made good chipsets. VIA and Creative Labs arguing over whose fault it was that audio was scratchy was fun
|
|
#
?
Apr 19, 2015 18:33
|
|
|
JawnV6 posted:That's not a viable long term play for Intel though. A lot of AMD's existence is owed to second sourcing, and the bigger customers would start funding any alternative they could find if Intel suddenly found themselves sole owners of x86. Intel would have a great run for ~5 years until all the money being dumped into obviating them caught up and they find themselves begging for fab work. So where is all that funding for the current alternative known as AMD?
|
|
#
?
Apr 19, 2015 19:00
|
|
|
FaustianQ posted:It's strange to hear people define good as "not poo poo". AMD not making GBS threads their pants while Intel fucks up with Netburst doesn't make AMD good. They seemed pretty "good" to me at the time, because the price/performance ratio was the best around. Your "good" seems to be some theoretical target that even the best R&D on the planet wasn't able to produce at the time. FaustianQ posted:If Intel never went netburst and just stayed the course with P-M, AMD would never even have a golden age people could pine and weep for, instead AMD might be competitive up until 2005 instead of 2009. If. I'm not sure how much partisan posting is really going on here. I think I speak for the vast majority in saying that competition is good for driving innovation and lower prices. If anyone misses AMD's "golden age" it's for that reason rather than for AMD brand loyalty. So surely that's understandable.
|
|
#
?
Apr 19, 2015 19:00
|
|
|
WhyteRyce posted:VIA and Creative Labs arguing over whose fault it was that audio was scratchy was fun Considering I had a SB Platinum and a dual PIII w/Via 694x chipset at the time, yeah, that was loads of fun 
|
|
#
?
Apr 19, 2015 19:32
|
|
|
FaustianQ posted:So where is all that funding for the current alternative known as AMD? Half of it's in gcc/llvm/compiler research, the other half is in ARM? AMD's limping along just fine, but dump everything going to them into those two areas and it changes things. I'm confused why you think "second source" apparently implies gangbuster profits for AMD instead of subsistence-level they're currently getting.
|
|
#
?
Apr 19, 2015 19:45
|
|
|
LLCoolJD posted:They seemed pretty "good" to me at the time, because the price/performance ratio was the best around. Your "good" seems to be some theoretical target that even the best R&D on the planet wasn't able to produce at the time. That theoretical target was the PIII-M and P-M, both of which could compete with the Athlons just for being mobile chips. It's kind of hard to say that making PIII-M and P-M desktop parts was unachievable. Has nothing to do with brand loyalty, but exactly as you said - AMD's golden age was the height of real competition between the companies, and then everything was downhill from Bulldozer. Wanting that again is something I hope for, but I'm under no illusion that such a time was because of AMD's genius but rather Intel's idiocy. JawnV6 posted:Half of it's in gcc/llvm/compiler research, the other half is in ARM? AMD's limping along just fine, but dump everything going to them into those two areas and it changes things. I'm confused why you think "second source" apparently implies gangbuster profits for AMD instead of subsistence-level they're currently getting. So they're pretty much okay with someone holding onto the AMD64 license that's not Intel, regardless of that companies competitive capability, as every scrambles to try and obsolete x86-64 to begin with? Also, 2016 will be hilarious for AMD stock. I can't wait for their next noncompetitive products in CPUs and GPUs.
|
|
#
?
Apr 19, 2015 19:56
|
|
|
FaustianQ posted:So they're pretty much okay with someone holding onto the AMD64 license that's not Intel, regardless of that companies competitive capability, I mention compilers because I think ISA's matter less and less. Apple could come out with a new phone on a new ISA and hardly mention it to anyone. They control the entire toolchain that gets HLL code onto their mobile platforms. It wouldn't be free, but if some ISA had 50% gains on power or perf they could eat that cost. FaustianQ posted:as every scrambles to try and obsolete x86-64 to begin with?
|
|
#
?
Apr 19, 2015 20:22
|
|
|
JawnV6 posted:This is pretty much happening. Your phrasing acts like it isn't, which is sorta confusing? Intel's doing their best to jam it into the mobile segment but everyone else is counting down the days until ARM scales up to servers.
|
|
#
?
Apr 19, 2015 21:06
|
|
|
JawnV6 posted:Dunno who "they" is, but the mobile segment is where the money and attention is. There's a long history of companies competing over the bargain-segment and learning enough iterating there to eat up the higher-end segments. That's all I'm saying will happen. There's no need to invoke a cabal of shadowy figures. Uh, I wasn't referencing a shadowy cabal either, just using they in the same sense as you were here JawnV6 posted:...A lot of AMD's existence is owed to second sourcing, and the bigger customers would start funding any alternative they could... JawnV6 posted:This is pretty much happening. Your phrasing acts like it isn't, which is sorta confusing? Intel's doing their best to jam it into the mobile segment but everyone else is counting down the days until ARM scales up to servers. I mean fair enough and everything, but why care about AMD64 at all then? It seems pretty drat irrelevant since Intel currently has defacto control over x86-64 in the server market, and is kind of irrelevant in the mobile market. So Intel grabs AMD64, no one cares as ARM marches on and Intel fiddles with an increasingly dead technology, woop, or moves to ARM as well.
|
|
#
?
Apr 19, 2015 21:17
|
|
|
FaustianQ posted:I mean fair enough and everything, but why care about AMD64 at all then? It seems pretty drat irrelevant since Intel currently has defacto control over x86-64 in the server market, and is kind of irrelevant in the mobile market. So Intel grabs AMD64, no one cares as ARM marches on and Intel fiddles with an increasingly dead technology, woop, or moves to ARM as well.
|
|
#
?
Apr 19, 2015 21:24
|
|
|
Oh come on, this market is saturated with out of work desperate coders, I'm sure we can put them on commission, and start them on recompiling legacy code 
|
|
#
?
Apr 19, 2015 21:28
|
|
|
|
| # ? May 21, 2024 16:07 |
|
|
If our financial institutions still run COBOL I doubt anyone bothers to port some Windows 98 greeting card software
|
|
#
?
Apr 19, 2015 22:09
|
|