Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
what is the big deal about SSD performance when it is paired with 7W CPU? and mostly base model carry 128GB/256GB

SSD performance in MBA is not a factor in 2019 anymore, capacity and overall specs, yes...

it is not aimed at Pro (or whoever apple feels Pros are) ... Absolute performance of SSD just does not matter.

We did seriously get off on a tangent didn't we!

you're absolutely right. when we're talking about these sorts of machines, top end numbers are essentially bragging rights as most use cases in an ultrabook are not going to take nearly full advantage of the full performance capabilities of these drives.

especially when most of the data that they handle is internet and wireless based. the liklihood of saturating the drive's performance is fairly low.

What is more benefiicial than raw throughput is queue length and time to retrieve the data and how responsive a program is to load. the difference between the benchmarks I provided is going to be completely and utterly negligble t the vast majority of use cases especially given the CPU offered
 
Please continue reading...

I will agree, based on the benchmarks, the Surface itself is defnitely using a sub par pissant of a drive. inexcusible.

but NVME drives are much faster than what Microsoft is apparently using. Good NVME drives are faster than what Appe put in the Air.

As I said, continue reading instead of jumping defensively as I've provided benchmarks

Have you ever compared ssds in use? The topic gets discussed quite a lot. Outside of extremely io bound applications, the fastest NVMe drives don't produce a lot of tangible benefit. Data that's in any kind of active use is still loaded into ram, so this becomes a matter of improving a small fraction of the total execution time by a typical factor of 10 - 30x. I'm not saying that drives aren't slow, just that in most workflows they aren't your bottleneck. If you're doing something highly IO-bound, it likely isn't loading from your boot drive anyway.

yeah. I'm bitter. I wanted a new MBA to replace my old dead one and Apple dropped the ball.

I too gave up. Notebooks have actually increased in price across the board over the past couple years, but Apple has increased quite sharply. Aside from the other aspects, I don't care for the risk of transporting these things, given their replacement cost.
 
Have you ever compared ssds in use? The topic gets discussed quite a lot. Outside of extremely io bound applications, the fastest NVMe drives don't produce a lot of tangible benefit. Data that's in any kind of active use is still loaded into ram, so this becomes a matter of improving a small fraction of the total execution time by a typical factor of 10 - 30x. I'm not saying that drives aren't slow, just that in most workflows they aren't your bottleneck. If you're doing something highly IO-bound, it likely isn't loading from your boot drive anyway.
absollutely correct!

I've done some work aroun benchmarking performance of drives for specific database engines and applications. it's quite impressive what sort of behaviours drives, memory, and how the data is read/ written makes a difference on the overall devices.

what sort of work your drive is forced to do is really work load dependent. And even Application dependent.

Don't know if you have experience with Progress OpenEdge as a DB engine, but I've had the (dis)pleasure of having to benchmark it, and provide recommendations and configurations to major financial institutions on using this engine for full scale banking application back ends.

this is one of those examples where pure raw throughput is fairly meaningless as the bulk of the transactions are via 4k read/writes and completely non-sequential. having a set of drives capable of 5,000MB/s is irrelevant IF the queue depth of those transactions skyrocket due to overwhelming pressure by too many potential disk access at the same time.

Nevermin what is in RAM vs read of storage. this gets things even trickier as well, as depending on program and data set, the actual Disk itself may be in almost no use (like the DB engine above, it loads data direct into memory, where it's processed before being served to the user, further rendering raw disk speed irrelevant)

So while we're worthlessly pissing about max theoretical bandwidth, I've definitely fallen prey to being goaded into a meaningless debate.
 
Meh. Sounds like you're in a very intimate relationship with your operating system.

Psst. OS's are supposed to get out of the way. You might be holding it wrong. Lol

Yeah, I spent 20 years trying to get Windows out of the way. Then I finally gave up and switched to Mac.
 
I got a 2015 Macbook Air, 11 inch, 4/128, for $500 from Groupon.

Using it right now.

Upgraded to Mojave with no problems.
 
Yeah, I spent 20 years trying to get Windows out of the way. Then I finally gave up and switched to Mac.

Me too (actually almost 30 years). Then I switched back after the past 6 or so.

I'm still open to whatever device or OS gets as far out of my way as possible. That's the prize.. my money.
 
  • Like
Reactions: PC_tech and LordVic
View the video. It does what you are wanting.
[doublepost=1551220250][/doublepost]
I think the base Air should be $899. I would have bought one at that price. As is, I agree with you that it is overpriced.
But the old Air *is* $899 at Costco all day everyday... (build-to-order i7 model, no less, at least at US Costcos)
 
Big fan of these comparisons as it shows the current crop of hardware, but ultimately it all boils down the OS. At the end of the day, if you prefer windows, there are dozens of options. I you want macOS, (unless you are a hackintosh type) you are limited to what apple sells.
 
  • Like
Reactions: OdT22
absollutely correct!

I've done some work aroun benchmarking performance of drives for specific database engines and applications. it's quite impressive what sort of behaviours drives, memory, and how the data is read/ written makes a difference on the overall devices.

what sort of work your drive is forced to do is really work load dependent. And even Application dependent.

Don't know if you have experience with Progress OpenEdge as a DB engine, but I've had the (dis)pleasure of having to benchmark it, and provide recommendations and configurations to major financial institutions on using this engine for full scale banking application back ends.

That sounds like a neat job. I haven't spent much time with databases.


this is one of those examples where pure raw throughput is fairly meaningless as the bulk of the transactions are via 4k read/writes and completely non-sequential. having a set of drives capable of 5,000MB/s is irrelevant IF the queue depth of those transactions skyrocket due to overwhelming pressure by too many potential disk access at the same time.

Nevermin what is in RAM vs read of storage. this gets things even trickier as well, as depending on program and data set, the actual Disk itself may be in almost no use (like the DB engine above, it loads data direct into memory, where it's processed before being served to the user, further rendering raw disk speed irrelevant)

I could see this. They're just requesting required pages, although I'm wondering whether the type of data being represented makes a difference here. If it's something that can be easily retrieved from some kind of B-tree structure rather than say something GIS related, I would expect the requests to conform to this. That is somewhat speculative on my part.

I was thinking more along the lines of video processing. There you may have multiple streams corresponding to different channels (much more efficient as the underlying engine can use simd arithmetic types without extra data movement), where deinterlaced raw video or a complicated comp of some kind could actually make use of that bandwidth, particularly if memory mapping of files on disk is allowed and used.

So while we're worthlessly pissing about max theoretical bandwidth, I've definitely fallen prey to being goaded into a meaningless debate.

Yeah I noticed that. An ssd has a lot of advantages over the older spinning drives for a notebook, but it falls off at some point when used as a system drive. I don't doubt that Apple uses expensive ones, but I suspect that at fixed price points, a lot of people would benefit more from higher capacity ssds than faster ones.
 
  • Like
Reactions: LordVic
"When it comes to key feel, the MacBook Air wins out" - Geez Louise. The Macbook Keys are beyond awful. How much worse can the Surface be than that?
 
  • Like
Reactions: WatchFromAfar
That sounds like a neat job. I haven't spent much time with databases.




I could see this. They're just requesting required pages, although I'm wondering whether the type of data being represented makes a difference here. If it's something that can be easily retrieved from some kind of B-tree structure rather than say something GIS related, I would expect the requests to conform to this. That is somewhat speculative on my part.

I was thinking more along the lines of video processing. There you may have multiple streams corresponding to different channels (much more efficient as the underlying engine can use simd arithmetic types without extra data movement), where deinterlaced raw video or a complicated comp of some kind could actually make use of that bandwidth, particularly if memory mapping of files on disk is allowed and used.



Yeah I noticed that. An ssd has a lot of advantages over the older spinning drives for a notebook, but it falls off at some point when used as a system drive. I don't doubt that Apple uses expensive ones, but I suspect that at fixed price points, a lot of people would benefit more from higher capacity ssds than faster ones.

I know very very little about video processing so I can't really speak to the performance requirements of that activity.

the DB I work in is only b-tree for searches. So it has a heavy disk IO requirement especially on index hits. but as said, fast IO here is responsive ness and ability to get to the data quickly. Not transfer the data in bulk.

really this emphasizes the point though that workload, software efficiency all around tend to make a larger difference on performance these days than raw throughput (when you get into SSD territory).

overall what it comes down to is bottleneck. where is it for the application specific requirement. in the case of the MBA, the bulk of the bottleneck in all performance is likely the CPU. you can throw a 10000000MB/s SSD all you want at it for boot/application purposes, but if the CPU is going to calculate slower than the data can be accessed, than the speed is ultimately irrelevant for most tasks.

It's frustrating because while the new MBA did bring about upgrades in a few specs, like resolution of the display, many other parts were essentially downgraded. the CPU, keyboard, Soldered Storage, the displays colour accuracy (and that's something considering the bad tN panel of the previous gen)

Simpy put, while the device is a decent little laptop. it should NOT have warranted nearly 20% price increase.

you know what the 2018 MBA feels like? The 2014 Mini.
 
I like your bookshelf and all the stuff in it. Very tasteful. :)

These reviews are very useful at separating marketing hype from reality, so keep them coming.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.