2014 MacBook Air: Performance benchmarks
macworld.com
macworld.com
http://blog.macsales.com/19008-performance-testing-not-all-2...
You only need one sample to make the point that it may be slower, and that's very much contrary to expectations.
2013 models had drives from different manufactures and some of those were slower than others.
Here's an example: Say 2013-A-SSD > 2013-B-SSD and we have the same two A and B SSDs in 2014 machines.
Then 2013-A-SSD > 2014-B-SSD.
With just one machine (and without knowing the specifics) you can deduce from the above that "2014 drives are slower than 2013s" when in fact they are the exact same 2 models that 2013 Air's had.
I think what you mean is "You might get a machine that's slower than _a_ 2014 model". Assuming every 2014 model performs at the baseline, unless that baseline is lower than the previous generation's you're still getting better performance than the _official_ 2013 model.
If I have a 2013 MBA I expect the SSD when I upgrade to a 2014 one to be at least as fast as the old one.
It just tells you you'll get a speedy SSD of X capacity (true), and you're not supposed to obsess over manufacturer and a couple percentage points of performance. Else, you wouldn't have bought the Air, but some custom rig.
That's a perfectly reasonable expectation to hold, as well. One really shouldn't end up worse off (for whatever reason) merely by buying a more recent model, especially when it's a computer that's being purchased.
Ah. Why would they do that? Someone could spend $3000 on MacBook Pro 16GB + SSD highest spec and risk the chance of getting bad parts? I never knew this was a practice. Is this even common in the industry otherwise Apple?
Apple has done this various times in the past, like shipping a 16x DVD drive when they advertised 8x. More problematic was the rMBP where LG LCDs were noticeably lower quality than Samsungs. http://9to5mac.com/2013/03/20/macbook-pro-with-retina-displa...
http://www.marco.org/rmbp-irtest.html
I am not sure which one of two manufacturers display is faulty, but per discussions on Apple's support Communities (https://discussions.apple.com/message/18669644#18669644), rMBPs carrying display panel from LG tends to suffers from this problem more commonly compared to those from Samsung.
EDIT: Just ran this test on a retinaMacBookPro to determine if I have a LG or Samsung Panel and found it to be working.
http://geeksblogs.com/lg-or-samsung-heres-a-little-guide-to-...
EDIT: I mean about the replacement part. Sure, they'll probably replace it after a several back and forth, plus the replacement will take serveral weeks.
If your friends aren't seeing noticeable problems though, there's nothing to warn them about. It's not that one display is inferior to the other in terms of color accuracy or brightness or quality. One display (I think it's the LG) has a noticeable ghosting and bright pixel problem in some cases.
It also lets them have a stronger negotiating position to drop a crap supplier. They can only maintain that negotiating position, though, if they buy drives from all of them; if they only used drives from whoever was the at-the-time "best" supplier, then if that supplier started to suck, all the others would demand a king's ransom for the "privilege" of switching to them.
All the component manufacturers that supply Apple are also supplying other contracts.
An example is the high-definition screens; 'Retina' Apple laptops use interchangeable screens from Sharp, LG and Samsung as availability permits. There was a run of returns last year due to some issues with the LG screens.
You may see the worst-performing variants of a given model as defining the baseline, with some being above the baseline.
But it's perfectly legitimate to define the baseline at the best-performing variants of a given model, with some unfortunately being below the baseline.
Personally, I prefer to set expectations high, and define the baseline based on the best that has been achieved so far. The worst-performing variant from 2014 should still be expected to exceed, or at the very worst be equal to, the best-performing variant from 2013.
Is the generation of the processor always the dominant factor nowadays? Is there a convenient way to estimate performance without poring over benchmarks? What are people's favourite sites for finding out about these things?
No.
> Is there a convenient way to estimate performance without poring over benchmarks?
Unless you are willing to quite a lot of time learning, not really. And unless you've been willing to do that, you haven't been able to conveniently estimate performance for about 14 years now.
> What are people's favourite sites for finding out about these things?
http://www.anandtech.com/bench/ is pretty good, as they standardize all results into a convenient database with easy head-to-head comparisons.
Lenovo has a similar problem in their new models. X240 has only one dimm slot and currently supports max 8GB of memory. While investigating the issue I found a discussion[1] where somebody mentioned that Intel would need to add support for these new dimms in the "memory reference code"[2].
[1] http://forums.lenovo.com/t5/X-Series-ThinkPad-Laptops/16GB-m... [2] http://en.wikipedia.org/wiki/Memory_Reference_Code
I had 6GB of RAM in my main development box until very recently. Even ran a few VMs on it. 8GB is a good amount for most power users, especially considering that the limited CPU will probably be your bottleneck first.
The takeaway is not everyone has the same use case. Also IMHO bloated DOMs and careless coding† are the real pain point.
† It doesn't have to be optimised to the metal but seriously, quite a few people just develop as if we all have ideal Turing machines.
Flash is trending towards fewer higher capacity chips. In some cases this is resulting in less parallelism and slower response times for individual reads and writes.
Not saying anything about this article and set of hardware, just in general that this has been my observation with new models of SSDs as they are released.
Unfortunately the way most applications are written the response time of SSDs is the primary benefit because they present low concurrency and don't parallelize the procesing of data as it comes off of disk. Still way better than seek times on spinning disk, but far from what could be.
I still miss the days when SLC was the norm, and you didn't need complex error-correcting-codes and bad block management. Now SLC is considered "high-end", even in enterprise applications, and becoming rarer to find, while 2-bit MLC, which used to be considered the inferior, consumer-level grade, has also become more difficult to find and a "professional" feature. Considering that endurance and retention decrease exponentially with each additional bit-per-cell while capacity only increases multiplicatively, the tradeoffs don't seem quite so great.
The extreme endurance torture tests that get reported often don't tell the whole story either - the more flash cells are cycled, the "leakier" they become and retention goes down significantly. Figures I've seen for SLC are 10 years retention after 100K P/E cycles, earlier-generation MLC 5 years after 10K P/E cycles, newer MLC is 5 years @1.5~3K, TLC is 1 year @ <1K. Of course retention tests don't make for as interesting news articles as endurance ones since they're almost like watching paint dry, but IMHO they are just as if not more important, and manufacturers should provide warnings that flash-based devices are intrinsically unstable and their retention ability is measured in years. "Bit rot" is a reality with NAND flash. I only hope that people who think they've "backed up" data onto used SSDs, memory cards, and other forms of high-capacity flash don't find that much of it has literally self-erased and disappeared after only a few years.
One of the blessings of an education in hardware; you learn to think in a distributed, parallel model.
People are really bad at that until they learn/train otherwise. Our brains are parallel at low levels, but conscious thought seems to be mostly serial.
Whatever you do, get the 3 year care is all I can say.
Considering the terrible customersupport that I've received multiple times from Apple, I bought my latest Macs at one of those stores.
I bought 2 Macbook airs together, both have the same issue.
The 2014 drives were faster than the 2013 in many of the test cases, specially 2nd page.
For example my Sony Vaio Pro is significantly faster than any of the numbers mentioned in the article. It gets nearly 1GB/sec in read speeds: http://i.imgur.com/24GdPiz.png
The drive is partitioned so ignore the odd size.
(yes, I know Gates' quote is apocryphal)
http://www.cnet.com/news/sony-agrees-to-sell-pc-business-rev...