Why do they keep pushing their SSD performance into the extreme?

File: Screen Shot 2016-11-19 at 10.55.18 PM.png (402KB, 700x359px) Image search: [Google]

402KB, 700x359px

>>57632932
caching the ads on facebook requires faster storage

this is apples answer

>>57632932
640k/s ought to be enough for everyone

>>57632932
Computer hard drives have always been the bottleneck ( at least for me ). I am just now starting to get interested in the new NVMe Samsung 960. I have an 840 Pro that I have had for a couple years now, and that 1 part just makes my entire computer run so much smoother. I can't wait for a new build with a PCIx SSD.

>>57632932
System performance is not limited by the processor. It's limited by the storage device.

$$$$$$$

That is all.
Once the price started to tumble following the release of the first consumer based SSD's the powers that be realised that they need to make moar moneyz.

Release something new = more $/happy shareholders.

The numbers look good. That's the real reason.

Plus it means you don't have to code an efficient bootloader.

they have to compensate for all the low-end hardware choices somehow.

>2 cores
>current year

>>57633195
>saying the OSX bootloader isn't efficient
Have you ever used a mac?

>>57633247
No, I'm not a faggot.

>>57633271
Then don't talk about what you don't know.

The OSX bootloader is fast as fuck, even on a mechanical disk.

Consumer grade PcieX4 SSDs will match or exceed those speeds in 2017 (the 960 already does).
This is just the new normal.

>>57632932
compensating for their shit filesystem from the 80s
http://arstechnica.com/apple/2011/07/mac-os-x-10-7/12/#hfs-problems

we should ban SSDs

>>57633338
Please no.

I don't want to go back to my spinning rust overlords. I don't care if people are stupid enough to care about one company's marketing spiel.

>>57633279
why would the drive matter for the bootloader?

>>57633470
Wasn't that the whole point you were making? lol

>>57632932
>They had to make their own custom controller for these SSDs too to achieve those speeds, which is why they may as well solder them simplifying manufacturing in because no one is making anything else compatible anyway
bullshit.
a) not confirmed that those are custom. markings on chips mean shit.
b) NVMe is a superior industry standard now to Apple's custom NVMe lite with missing features
c) soldering makes it harder to manufacture, not easier
d) non-replaceable makes it harder to recover when your mainboard dies

>>57632932
Disk performance relates to networking performance. A slow disk will make a 10 GB line rather irrelevant.

Remember that the system will want to dial home and upload all your hot info to the mothership and that had better be quick to make sure you will not disconnect the line before all has been uploaded.

Insert "Mwahaha" at will.

File: 28967346.png (8KB, 493x402px) Image search: [Google]

8KB, 493x402px

>>57632932
>faster and bigger ssd
>brighter screen with higher contrast
>skylake cpu and faster ram

If Apple would have just updated these specs on last year's macbooks they would have been fantastic.
But no, they had to have their thinner/usb-c/touchbar bullshit.

>>57633584
What ISP offers 10gbit speeds?
Aren't they all trying to cuck their users with data plans?

>>57633115
Yeah that's why they capped ram at 16gb

>>57633328
That's one huge review.

>>57633683
who cares about RAM when your SSD is as fast as DDR2 already?

>>57633700
Latency.

>>57633700
RAM is very slow. SSDs are even slower.

>>57633223
Apple is using top of the line hardware.
Wanting more than a dualcore in a 13" ultrabook is just retarded and thankfully x86 doesn't offer such shit like some ARM chipmakers. One x86 IA core from Intel needs like 10W TDP, so do the math.

>>57633223

The 15 inch machines have always had quad-core HQ i-Series processors. Can't you even read a spec sheet or do you get your information only from /g/?

>>57633683

>that's why they capped ram at 16gb

This is Intel's fault, you're aware of that right?

>>57633878
NVMe can do 2.8 µs.

>>57633990
>his latency is measured in microseconds
stay pleb

>>57633990
Thanks for proving my point

>>57633999
>>57634028
let's show us the use-case where that's not fast enough.

>>57633957
>have to use shit LPDDR3 because your anorexia shrunk the battery
>blame Intel
top kek

using ssd on sata2 and it is blazing trough everything

>>57632932
Bigger numbers sell.

Unfortunately these maximum throughput numbers say little about real world performance.
What really matters are IOPS, but they won't mention those because they probably suck.

>>57634104
IOPS only matter in poorly optimized applications.

>>57634104
>implying HFS+ would scale
top kek

Doesn't matter anyway, SSD's life expectancy is absolute shit.

>>57634123
So that would be 99% of applications then.

They want to mine some burstcoins

>>57633328
>>57634125

>HFS

Apple has already moved to APFS since Sierra. Catch up.

>>57634161
Right now it's okay but if they use smaller transistors then the life expectancy will go down even more. That's why they try to add more layers instead. Unless they find a magical way to double the number of layers NAND based SSDs will stagnate.

>>57634167
fair enough

>>57634035
The use case was to replace DDR2/DDR3 RAM with SSDs.

>>57634194
they didn't.
you are retarded if you trust your data with some clearly marked non-default developer preview fs.

File: latest.jpg (109KB, 1280x532px) Image search: [Google]

109KB, 1280x532px

>>57633124
>>57633115

Unless you're running a program off of a RAMDisk, this is always going to be the case to varying degrees. Sometimes it's negligible, other times it's going to put the rest of your hardware at idle for second(s) at a time, especially with a mechanical drive. Makes perfect sense to make SSD's faster and not be held back by a standard designed with magnetic storage in mind, especially when everything else has become so much faster.

>>57633914

enjoy having no storage and adapters hanging all over the place from your overpriced """pro""" mousepad

>>57633905
>RAM is very slow
K E K
E
K

>>57632932
>nobody makes anything compatible!!!1

Apple does. Apple makes them. Apple could sell replacements for them if they break if they didn't weld the fuckers in.

>>57634247

Compared to something like SDRAM it is.

>>57633668
I have a '15 rMBP and I really like the touchbar on the new one.
Shame they dropped the magsafe and sd reader tho.

>>57633905
>RAM is very slow.
Explain your reasoning here, are you arguing that if all the storage available to the cpu was as fast as its Cache that it would result in significantly higher throughput?

>>57634276
SDRAM is RAM.

>>57634293

I'll be getting one soon. Hows the overall performance, nice and snappy?

>>57633672
no person or enterprise that has a 10gbit line isn't also going to be running in RAID, it doesn't matter if they're using ssd's, a larger number of magnetic beyblades, or a gargantuan number of ZIP drives at that point.

>>57633925
I am clearly talking about the 13. note how it has pro in the name.

>lololo but the 15

>>57634388

>a dual core i7 isn't Pro
>a quad core i7 is Pro

Is this how you think? I know some like-minded people you can talk to. >>>/v/

>>57634123
Or servers.

>>57634306
>>57634247

Yes. Compared to CPU caches RAM is very slow latency wise.

https://youtu.be/Nsf2_Au6KxU?t=125

Watch the first four minutes if you don't want to watch the entire video.

>>57634435
At that point you might as well just buy more RAM.

>>57634354
>no person or enterprise that has a 10gbit line isn't also going to be running in RAID
bullshit.
not everyone is getting it to run some servers.

>>57634440
Apple uses Intel CPUs with eDRAM, which is basically a L4 CPU cache.
thus Apple destroys all those PC laptops that don't use such CPUs.

>>57633542
the two aren't related at all. If you have a shitty bootloader (like osx) on a shitty disk (any mechanical disk in mbp's from pre 2014) you're going to have a shitty machine that boots slower than you grandmother's desktop from 1992

>>57634458
i guess i forgot about buildings that just have a shitload of pcs like schools, offices and whatnot.

but you know what i mean

>>57634243
>enjoy having no storage
2TB is anything but no storage
>adapters hanging all over the place
most ppl have nothing connected to their laptops most of the time except power.
i have one usb-c to usb-a for some occasional usb-a sticks.
for my iphone i got a usb-c to lightning cable.

>>57634664
>for my iphone i got a usb-c to lightning cable.
also that's one charger less to carry now that mbp's usb-pd charger can do the iphone with that cable.

>>57634664
>2TB

in what universe?

>>57633124
>System performance is not limited by the processor. It's limited by the storage device.
Yes but it's not limited by your storage device's sequential transfer speeds but by access times and random access performance.

>>57634194
>Apple has already moved to APFS since Sierra. Catch up.
Bullshit. Beta was announced for 2017 and there is no public timeline for a release. When it's about filesystems this will take long. Well, either long or users might loose their data.

>>57634586
i don't think you know what a bootloader is if you think the disk matters.
your bootloader just reads some few kbs from the disk and starts the OS.
it shouldn't matter for your boot time at all, which is mostly your OS starting up after the bootloader loaded it.

>>57634717
http://www.apple.com/shop/buy-mac/macbook-pro?product=MLH42LL/A&step=config#

I really like the current ssd performance.

Im a musician that plays a lot of virtual instruments in real time, some of these high end software instrument sample libraries are huge, my favorite piano software is a 46gb sample library, and it absolutely gets the lowest latency playback from these super fast ssd's.

>>57633124
That's a stupidly gross oversimplification and you know it

>>57634726
nope, depends on use-case.
just because SSDs improved your shitposting doesn't mean CPUs don't matter for other things like compiling.
just because sequential read/write doesn't matter for shitposting doesn't mean it doesn't matter for other things like video editing.

>>57634850
Get a real instrument

>>57634850
so you are saying you can distinguish a 2015 from a 2016 mbp?

>>57634430
has nothing to do with gaming.

you could slap i7 on some celeron for all I care, it does not have the performance.

>>57634922
>Intel slapping i7 on everything is okay but when Apple does it with pro i am getting butthurt
retard pls

File: streamFLOPs_mm256_load-aligned.png (76KB, 1500x1050px) Image search: [Google]

76KB, 1500x1050px

>>57634306
>>57634440
>>57634517
When you have a program capable of utilizing your CPU good enough (e.g. triad streaming benchmarks) you will notice drastic performance differences even between the different cache levels. "L4" cache doesn't say much as L3 is already considerably slower than L1.

Image related. It shows the achieved FLOP/s for a benchmark (y axis, log scale!) when the data size is increased (x axis, log scale!). You see that whenever the data reaches a certain size the performance suddenly drops and then stays constant again until it drops again at a certain size. This is because the data no longer fits into a certain cache level.
You get the highest performance while everything fits into L1. The first drop occurs when data is also retrieved from L2, next drop when L2 is too small and data is additionally retrieved from L3, last drop is with data in RAM.
As you can see the first drop (L1 -> L2) is already the most dramatic for performance.

For CPU performance a massive amount of a fast cache are important. Massive L1 all the way, some big "shared L3" caches sound nice but don't do much.

>>57634989
you are the retarded one,

it makes no sense in both cases.

>>57634995
except synthetic benchmarks designed to show cache latencies are not the real world.
test your actual use-case.

>>57633271
turns out you are one, congrats

>>57634784
Can't believe it took this long for someone to point this out. /g/ really has gone to shit.

File: 1431284853724.jpg (64KB, 551x550px) Image search: [Google]

64KB, 551x550px

>>57634194
haha you retard, they didn't you're the one who should catchup

>>57634898
>>57634892

I'm not sure what the point of these were. A lot of musicians choose to use software instruments in the studio, they just offer so much more versatility. You can record one take, and turn around and tweak it however you want, as if it's being played in real time, every time.

I, like many other musicians, prefer to practice on the instrument I intend to perform on, so thus I find myself playing these software instruments in real time. Honestly anything more than 10ms latency becomes problematic very quickly. I think any competent musician would notice. 7200rpm hdd were barely fast enough to use these kinds of systems, ssd's are really the answer here.

I dont use macs, but I certainly notice the performance increase when I use the pci ssd in my asus zenbook rather than a 7200rpm HDD in my desktop computer, its still noticebly better than the esata SSD in my desktop too.

its not a placebo, there is a definite improvement in sample fetching speeds. It makes a huge difference in playability

File: screen-shot-2014-02-05-at-11-58-58-pm.png.cf.png (215KB, 399x406px) Image search: [Google]

215KB, 399x406px

Hey nigger faggot who cares my Mac Book Air from 2105 is still the tits

File: happy diamond dogs.jpg (166KB, 1017x1019px) Image search: [Google]

166KB, 1017x1019px

>>57632990
/thread

>>57635044
>pointing out obvious and redundant information for the massive ammount of worthless pajeet shitposters that are on /g now
why would anyone do this other than to troll or shitpost ironically..
/g has become pure cancer since the "/must always be up becuase we're faggot shillfag shitposters who want to advertise out shitty products for free" threads became allowed by mods/janitors. need to spend half an hour to hide this shit just to get to a post that might be worth shitposting in, or conversing with.

>>57634776
I'm not him but the final version is to be expected in 2017 and now you can have experimental one by diskutil format

>>57635096
> 2105

>mfw a macbook hair from 89 years from now still sucks compared to a memepad

>>57633223
Name 1 (one) 13-inch laptop with non-ULV quad-core 45W processor.

>>57634850
Noiserv?