The New 2019 MacBook Air Features a Slower SSD Than 2018 Model (imore.com) 121
The new 2019 MacBook Air with a True Tone display, upgraded keyboard and a price cut has been out for a week already, but we're finding out more about. The latest bit of information from Consomac confirms an unfortunate drawback: the SSD is slower than the previous 2018 model. From a report: The French site conducted some tests on the new 2019 MacBook Air using Blackmagic Disk Speed Test and it achieved speeds of 1.3 GB/s read and 1 GB/s write. Compare it to the 2018 MacBook Air, which achieved 2 GB/s read and 0.9 GB/s write. Apple's newer laptop improved slightly on the writing side, but its performance downgraded by 35% on the reading side. That can be attributed to a slower SSD Apple included in the new MacBook Air.
that is truly courageous (Score:5, Funny)
At a time when SSD prices are plummeting and speeds are increasing exponentially, to ship an inferior drive to what they had in the last revision, that is courage!
Re:that is truly courageous (Score:4, Interesting)
Re: (Score:3, Insightful)
Write speeds that are slightly faster and read speeds that are drastically slower when reads are far more common than writes, I'd say yes.
Re:that is truly courageous (Score:4, Insightful)
Most reads (90%) actually come from RAM. Writes have to happen and are the biggest thing that cause people to be frustrated with speed because you cannot delay it, many applications will hang the UI while doing it.
Although at 1GB/s I highly doubt anyone will notice. These are office-work machines and still superior to the SATA SSD most computers come with these days.
Re: (Score:2)
Actually, the GP is at least somewhat correct. For most workloads, you can predictively prefetch data into RAM in ways that largely mask read speeds. That's how they made hard drives, often with rotational latency of up to about 11 milliseconds, actually be usable.
By contrast, data that is being written to disk cannot usefully be pre-written, because you don't know the data to write until you're ready to write it. You can sometimes write behind (by pretending that the write has completed, and not blocki
Re: (Score:2)
You can't necessarily draw conclusions about an entire system based on a single piece of that system. One of the most I/O-intensive things that the average user does with a computer is launching apps. Elsewhere, I did the math, and calculated that the increased disk read speed in the previous MBA mo
Re: (Score:1)
Man you talk like an ass for someone making sweeping generalizations.
In literal terms the only thing effected by slower reads are load times. Load is a word which means "the amount of time it takes for you to read data into main memory".
Once you've loaded everything you need then the disk is out of the picture. So "slower performance all around" is incorrect. The only thing that slower drive reads effect is loading time.
Not defending the slower disk. Considering how they market themselves as a premium produ
Re: (Score:3)
Actually, the GP is at least somewhat correct. For most workloads, you can predictively prefetch data into RAM in ways that largely mask read speeds.
So why get an SSD?
That's how they made hard drives, often with rotational latency of up to about 11 milliseconds, actually be usable.
So you are saying that we should settle for "usable" when it comes to SSD's now?
Re: (Score:3)
Actually, the GP is at least somewhat correct. For most workloads, you can predictively prefetch data into RAM in ways that largely mask read speeds.
So why get an SSD?
Because they don't mask read speeds fully. But you'd be amazed how much they do mask it. I remember talking with engineers about the sorts of prefetching that macOS does to speed up boot times by creating a boot cache that can be read as a single blob, and that made the difference between a boot taking seconds and taking several minutes on slower drives.
Either way, no matter how much you cache things, order-of-magnitude differences still matter. The difference between 1.2 GB/s and 2 GB/s (a factor of les
Re: (Score:3)
Irrelevant. Users don't perceive read speeds. What they perceive is delays in getting things done. Slower read speeds can cause, but do not necessarily cause meaningful differences in overall performance. It all depends on a fairly large number of other factors, including but not limited to caching, prefetching, and even random lu
Re: (Score:2)
Also, this ignores the fact that in tests based on synthetic workloads, the measured disk performance often doesn't match reality. For example, HDD manufacturers might claim 30 MB/s random access time on a 5400 RPM drive, but in the worst case, you might get only a tenth that speed, depending on how badly the elevator algorithm or whatever handles the read reordering of a given set of requests.
For an SSD example, suppose you have two drives, one of which has a larger tagged command queueing buffer, allowi
Re: (Score:2)
Irrelevant.
It's absolutely not irrelevant, in fact you directly followed this up with your admission that you agree that slow read speeds can cause meaningful differences in overall performance. You just directly contradicted yourself.
I didn't say that they couldn't. What I said was that in practice, it doesn't usually matter.
It all depends on a fairly large number of other factors, including but not limited to caching, prefetching, and even random luck.
Yes there are many things that can be done to improve performance, hacks to compensate for poor speeds but you seem to be apologizing for the decrease in read performance by suggesting all these things mitigate even the need for improved read performance which is just wrong.
The onus is on you to prove that the difference in read performance actually affects the end-user experience. You have not done that. Without doing so, the claim that there is even a need for improved read performance is nothing but speculation.
The big thing you seem to be missing is the fact that very little of an application's launch time is actually spent reading the actual bits from disk. At some point in the past — somewhere between 7200 RPM disks and the most dog-slow SSD ever built — app launch time stopped being dominated by disk I/O.
Citation?
Sure. Let's start with the estimate from FAST: Quick Application Launch on Solid-State Drives (Usenix) [usenix.org]. On an iPhone 4 (101.2 MB/s), for a few randomly chosen apps, warm
Re: (Score:2)
It matters, but not nearly as much as you seem to think it does. This is a MacBook Air we're talking about.
Re: (Score:2)
You'd figure paying premium dollar would get you premium parts...
It's 10% cheaper than the previous year's model. Paying less usually gets you less. :-)
Re: (Score:2)
1GB/sec sequential writes. They will notice when it's random I/O.
I have a machine with an older Samsung SSD. 550MB/sec sequential write speed, verified. The system gets extremely slow and unresponsive when downloading via BitTorrent at over about 15MB/sec, freezing for several seconds at a time with 20MB/sec.
The random writes of BitTorrent really hammer performance. It doesn't help that the machine has only 4GB of RAM so switching tabs in Firefox causes a few megs to load from the disk cache. Tab switches t
Re: (Score:2)
I actually think the Chinese tariffs may be the cause.
SSD tests are misleading (Score:2)
There's so many issues in testing an SSD. There's a lot of cache and fast access memory management that goes on. There's serial versus random reads and writes. There's the command Queiing. there's temperature issues on sustained writes. And there's access under conditions where multiple tasks are trying to access rater than just one. All of which means your test harness has a lot to balance. There's no single number for performance either.
I can't say if this SSD is slower or faster, I'm just saying t
Re: (Score:1)
Re: (Score:2)
"exponentially"
lol, no.
Re: (Score:2)
Re: that is truly courageous (Score:2)
Itâ(TM)s possible the downgraded drive doesnâ(TM)t actually impact performance of the system... the MacBook Air is not a performance device, and by using a lower-spec without impacting performance frees up money to improve other components and maintain the old price-point for the laptop.
Re: Most won’t notice (Score:2)
Macs have much faster ssd drives over the majority of windows systems.
Macs ship in small quantities, and at much higher prices than âoethe majority of windows systems.â Yes, the $1,000 MacBook Air has better build components than a $249 Windows laptop - thank you Captain Obvious.
It is an Air... (Score:2)
If this were a Macbook Pro I would say that was a strange change, but since it's the Air I don't see as great a need for really speedy SSD... for lots of uses of the Air the improved write speed would be slightly more useful if anything.
Re: (Score:3)
Except for the fact that a newer model, you should expect at worse performance to stay the same or be better.
Sure an Air you don't expect a super computer, however due to Apples long lifecycle on their products, you should expect a better new version.
Re: It is an Air... (Score:1)
Long lifecycle? The whole last batch of Macbooks have the ticking time bomb keyboard. Better hope your keyboard fails while Applecare is current. (And even then they will just install another time bomb keyboard) Resale value after Applecare? Why not buy lottery tickets instead?
it is though (Score:1)
you should expect at worse performance to stay the same or be better.
A) Why? Products change all the time. The speed they offer is still much better than spinning disk.
B) It is for write speeds.
Re: (Score:2)
Agree with you for once. Nobody should expect the performance to stay the same or be better because this is Apple.
Re: it is though (Score:1)
Re: (Score:2)
He's not entirely wrong though. Read speeds aren't irrelevant... they can certainly add to the speedy feel of a system. But in reality read speeds are effectively masked by caching. Any modern system with a decent amount of RAM should have a read cache hit ratio of at least 50%. Sure, probably not on initial boot (unless you're doing fancy stuff with preloading caches from previous sessions) but after running for a little bit the relative speed of the underlying drive isn't as noticeable. Write speeds ARE m
Re: it is though (Score:1)
In fact, it would be entirely expected for both to increase, and when measured over a timeframe denoted in years, increases would even be considered failures were those increases not large enough.
There is no legitimate need for a consumer in 2019 to justify a speed decrease in either reads nor writes. Frankly, I canâ(TM)t fathom why they would
Re: (Score:2)
The speeds they offer us is much faster then Real to Real tape too. However for most computing needs we need to read data. While Ram Caching is a useful tool. I myself tend to run with some heavy read sessions causing the disk to go beyond the cache.
Now this is a mobile system. So we are not going crazy with Logging, or DB activity, however it will want to load data from disk and dump it into RAM.
Re: (Score:1)
If this were a Macbook Pro I would say that was a strange change, but since it's the Air I don't see as great a need for really speedy SSD... for lots of uses of the Air the improved write speed would be slightly more useful if anything.
A ~40% performance deficit in read performance certainly is going to hurt when you're loading up applications and games or loading datasets into those applications. Write performance (at only 10% difference) is unlikely to have that much of an impact for a device like this.
Re: (Score:2)
Apple introduces the Mackbook Toy lineup. Mackbook Toy costs twice as much as a Mackbook Pro because it's twice as much fun. It's twice as much fun because it does half as much work. Apple logic.
its a status symbol (Score:1)
Apple has some really great tech but they are more interested in building status symbols than work horse computers. The air is for execs to feel good about themselves when they open it in front of other execs in the fancy conference room. If you actually want to get work done, they want you to fork out for a Pro.
Re:its a status symbol (Score:5, Insightful)
I would say it was a status symbol 10 years ago. Today someone who has a Mac Laptop is mostly stuck on it, because they didn't want to port over their software to a Windows version.
pre-iPhone days, you were Mr. Fancy pants if you walked into starbucks with a thin Mac Laptop. Today, people look at you funny for having a laptop in general.
For business folks lately you will see them with the always classy Think Pads.
Re: (Score:3)
The big advantage of the Mac is: my shel scripts run the same like on linux.
Well, Windows 10 changed that, but why would I use that?
Linux has no useable mail program and is not scriptable like Mac OS X or macOS is.
However I like to use GitBash on Windows, works most of the time perfectly (can not do a MySQL log in though, somehow the invisible password or some input thing goes wrong).
Re: (Score:2)
The big advantage of the Mac is: my shel scripts run the same like on linux.
Well, Windows 10 changed that, but why would I use that?
There's long been a variety of ways to run your shell scripts on Windows or even DOS, although running shell scripts on DOS often required massaging. I had both sh and csh for DOS. And even before Cygwin, there was MKS Toolkit, though that was commercial software.
Re: (Score:3)
No usable mail program on Linux?? I've been using Thunderbird for over a decade, works great. You must have some silly criteria for "usable"
I've horror stories about MS Outlook on Mac
Re: (Score:2)
No usable mail program on Linux?? I've been using Thunderbird for over a decade, works great. You must have some silly criteria for "usable"
I consider it unuseable.
Worst thing I remember: define a smpt connection, make a mistake, so it does not work. Delete it. Make a new one with the same "name" does not work: the name is already in use. Wow: I DELETED the old one ... and it is hidden somewhere in a config file that you don't find it with grep to kill it the hard way.
Deinstall Thunderbird. Install it again
Re: (Score:2)
Re: (Score:2)
and that all this time, you didnt know it.
How do you come to that retarded idea?
Writing a bash script on macOs, Linux, BSD has ZERO overhead in making it run on each of those systems.
This indicates that you dont know that there is more than one shell available on the platforms you listed, that you have literally always been able to make all those platforms shell compatible with each other
Actually I always knew that. Idiot. When I started on Unix around 1987, MK Systems already existed.
And you are wrong. On
Re: (Score:2)
Not a status symbol, my employer gives us a choice and Mac OSX is far more useful than Windows. Plenty of business software targeted for it, my scripts and favorite Linux wares run on it. It has longer battery life and more power than the Dell alternative they offer us too.
Of course, I wouldn't buy one for home, my laptop and desktop run Linux Mint. then sure, Thinkpad is great way to run Linux.
That Ubuntu crap Microsoft has with windows is like strapping an outboard motor on a goat and calling it a fish
Re: (Score:2)
Today someone who has a Mac Laptop is mostly stuck on it, because they didn't want to port over their software to a Windows version.
If they have a MacBook then they don't have to port their Windows apps over.
I have a MacBook that I run Windows in a virtual machine, to switch back and forth is just a keystroke. I just helped my brother set up a dual boot on a MacBook so that he could work from home using Windows during the day and then reboot into macOS for his wife to use in the evenings. He's got it set up with a wide format screen at his desk so he's got plenty of screen space, a Windows style keyboard, and the pointing device of hi
Re: (Score:2)
I would say it was a status symbol 10 years ago.
Today Apple products mark you as a shallow fad follower, lacking in the fiscal planning department.
They've been using slower SSDs for a few years (Score:5, Interesting)
e.g. Say you're comparing a NVMe SSD with 2 GB/s sequential speeds and 35 MB/s 4k speeds, with a SATA SSD with 500 MB/s sequential speeds and 50 MB/s 4k speeds. Say the operating you're using to compare involves 1 GB of sequential data, and 200 MB of 4k data. Which disk will finish the task faster?
Obviously the NVMe drive, right? It has 4x faster sequential speeds, while only being 1.4x slower at 4k speeds. And there's 5x more sequential data than 4k data. Everything favors the NVMe drive, right?
NVMe: (1 GB) / (2 GB/s) + (200 MB) / (35 MB/s) = 0.5 sec + 5.7 sec = 6.2 sec
SATA: (1 GB) / (500 MB/s) + (200 MB) / (50 MB/s) = 2 sec + 4 sec = 6.0 sec
Surprise! It's the slowest drive speeds which matter the most. If you want a fast SSD, get one whose slowest operation is as fast as you can get. That means it needs to have fast 4k speeds. In typical tasks, a SATA SSD with good 4k speeds will outperform a NVMe SSD with poor 4k speeds.
Unfortunately, Apple started putting terrible NAND into their Macbook SSDs a couple years back. I suspect this is fallout from ditching Samsung as their supplier, and switching to Toshiba NAND. If you scroll down to the disk benchmark section [notebookcheck.net], the 2018 MBPs have the worst 4k speeds of any laptop SSD I've seen. 11 MB/s reads, 20 MB/s writes for the 15". The 13" is even worse [notebookcheck.net] Typical 4k scores for other laptops are 30-70 MB/s reads, 100-150 MB/s writes.
Apple managed to dodge criticism for this significant downgrade to the Macbook SSDs because a bunch of clueless reviewers didn't understand that OS X's new filesystem simply created links when copying files, instead of actually copying the file. (This was probably done to minimize the space occupied by files on the disk, since Apple regularly puts SSDs with insufficient capacity in their Macbooks, and charges an arm and a leg for capacity upgrades which can only be made at the time of purchase. Once you buy it, there's no way to upgrade it. The actual copy isn't made unless/until you do something to make the copy different from the original.) And they erroneously declared the SSD in the new Macbooks were the fastest they've ever seen, when in actually they're the slowest ever and their benchmark was erroneously reporting the time to create a link as the time to make a copy.
Re: (Score:1)
You generally don’t buy an Air for video editing. Email, Word processing and browsing is more in line for those devices.
Re: (Score:2)
https://9to5mac.com/2019/07/12... [9to5mac.com]
Re: (Score:3)
Why people think "performace" means "throughput" is something I'll never understand. Throughput is _always_ secondary to latency, and really only becomes interesting when it becomes a latency number (ie "I need higher throughput in order to process these jobs in 4 hours instead of 8" - notice how the real issue was again about _latency_).
-- Linus Torvalds
Re: (Score:1)
> Did Linus misspell performance, or are you a dumbass?
They're not mutually exclusive, but to answer your question:
https://lkml.org/lkml/2005/5/16/152 [lkml.org]
Comment removed (Score:5, Funny)
A long lasting Joke (Score:3)
Macbooks. A long lasting Joke since 2014.
What SSD? (Score:3)
Re: (Score:2)
you mean NAND flash chips
But even there, they aren't addressed individually. They go through a controller. The NAND flash chips and the controller are what we call an SSD, whether it's replaceable or not. And yes, they should be replaceable.
Re: (Score:2)
An SSD is a modular drive and replaceable.
Please point me to a definition in literature where modularity and replacability are a requirement.
I'm gonna call it. (Score:2)
I'm gonna call it. Innovation at Apple is officially dead.
It's now just a company run by accountants selling the same tired old designs dating back to the Jobs era over and over at ridiculous prices to braindead sheep.
Re: (Score:3)
Nice of you to call it 8 years after the fact.
cooling? (Score:4, Interesting)
Re: (Score:2)
--Skip the high-capacity thumbdrives - go with an external SSD. Larger form factor, but much more reliable.
Re: (Score:2)