Nobody should take laptopmag seriously.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
2018 MacBook Pro Features 'Fastest SSD Ever' in a Laptop According to Benchmarks
- Thread starter MacRumors
- Start date
- Sort by reaction score
Nobody should take laptopmag seriously.
For about the billionth time, this benchmark does not copy a file. It creates a files, reads it then deletes it. My Mac benchmarked around 760Mb/s on HFS and the same on AFPS. The guy from laptopmag had/has no clue what he’s talking about.
I might be missing something here but don't we just need a copy test between two 2018 MBPs connected by an Ethernet cable to get the true copy speed?
[doublepost=1531599961][/doublepost]
Or this. That's a great idea too.
[doublepost=1531599961][/doublepost]
Or this. That's a great idea too.
The gaming performance difference is likely due to the courageous metal API in OSX.
Also did they really try to say RX580 is super fast? Beyond hilarious. Almost as ridiculous as an eGPU...where you can't swap out the GPU. Oh and did I mention that GPU is already a 3 year old mid range?
Also did they really try to say RX580 is super fast? Beyond hilarious. Almost as ridiculous as an eGPU...where you can't swap out the GPU. Oh and did I mention that GPU is already a 3 year old mid range?
That would be limited by gigabit Ethernet. Once again, BlackMagic does not copy a file. It has a 5, 4, 3, 2 and 1Gb test. That would make the app 15Gb plus change if it worked by simply copying an existing file.
[doublepost=1531600087][/doublepost]
That’s a relevant, on-topic post in a thread about SSD performance.
Yeah you would be limited to like 114 MB/s or so in real world speeds with Ethernet. Theoretical limit being 125 MB/s. Even with a 10Gb USB-C adapter 1.25 GB/s theoretical with real world speeds somewhere around 1.15 GB/s or less. That is if such USB-C to 10Gb Ethernet adapters exist.
Another example, my Inspiron has a 9260ac WiFi card that hits 1.7 Gbps link rate in HT160 mode with my R7800 router but I only get 114 MB/s (~910 Mbps) in transfers to my NAS since Ethernet to my NAS is the limiting factor.
Maybe a USB-C to C would work as even with overhead you have enough bandwidth.
Another example, my Inspiron has a 9260ac WiFi card that hits 1.7 Gbps link rate in HT160 mode with my R7800 router but I only get 114 MB/s (~910 Mbps) in transfers to my NAS since Ethernet to my NAS is the limiting factor.
Maybe a USB-C to C would work as even with overhead you have enough bandwidth.
Last edited:
That is disingenuous at best. Sell customers a pretty old, mid-tier gpu for top dollar. Get a gtx 1080TI for that price instead.
Good point about gigabit Ethernet. How should it be done then? Could two MBPs be linked with a USB-C cable?
I wasn't referring to Blackmagic. Just an old fashioned large file transfer.
[doublepost=1531600450][/doublepost]
True. What's going on with this GPU banter, mods?
I don’t see any way to do a copy without being a victim of some other subsystem. You can’t have a second internal disk afaik, so that’s out.
If you copied from/to a fast external disk array, even if USB-C/thunderbolt doesn’t actually cap out, you’d still be seeing overhead from that work.
Even if you managed to do an internal copy on the same drive without copy on write, by formatting HFS+ for example, you’d be doing both the read and write at the same time (reading from source file, writing to target file) and that isn’t what’s attempting to be measured here.
The purest way to test the write and read speed of the drive is to create a file and read it.
Yup. These results are sketchy. You can definitely get non-Apple laptops with SSD speeds in the gigabytes per second.
There's also too little information. What GPU did the Dell model they tested have? Could it be that the MateBook and the ZenBook have virtually the same GPU results because they have the same GPU? Why is the multicore benchmark best by a long shot on the MacBook — did they not test against other Coffee Lake CPUs?
Something newer coming out doesn’t make it any slower. Your point about not being able to upgrade is undeniably true though.
That's always the case. The aim is to find the subsystem that will give the result closest to the true SSD speed.
No idea but these should give you a reasonable(ish) idea.
For write (please ensure you have 5 Gb free first)
dd if=/dev/urandom of=5gigfile.bin bs=1m count=5120 oflag=direct
I see about 144 Mb/s from a spinner on a 3 year old Dell running Mint
For read, it’s not a great test but run
hdparm -Tt —direct /dev/sda1 (You may need to use root or sudo and you may need to change disk device)
And check the O_DIRECT disk reads. I see 427Mb/s cached and 114Mb/s non-cached.
They aren’t perfect tests but they should be ballpark correct.
Note: I edited the above to /dev/urandom from /dev/zero as /dev/zero return unrealistic results.
"Fast" and "slow" are relative terms in the computer world. But "all our stuff is disposable" is an absolute term in Apple 2018's world.
They do not because, as you noted, it’s a different platform. They are a best effort I put together quickly to ascertain realistic disk speeds in Linux, because I assumed when you asked if BlackMagic has a Linux version you were actually asking in good faith. Seems I may have been wrong there.
IOW you put these parameters together to achieve a particular end result. Hmm, doesn't that speak volumes to the validity of a benchmark when the parameters chosen can significantly influence the outcome?
Do you think maybe this influence can be applied to any other benchmarks? Say one which shows the new MBP significantly outperforming other systems?
Yes, that result being an accurate picture of the performance of the disk. You know, a benchmark. BlackMagic and Crystal are long-established, credible benchmarks that are used by credible publications (though laptopmag does not count as a credible publication) All the mental hoops you’re jumping through to deny their results because you don’t like them are really quite remarkable. If you have nothing factual to add, we have nothing to discuss. Have a good one.
Razer allegedly makes good, slim laptops these days. ASUS too, from what little I've seen. There might well be more good choices, I really don't know; personally I wouldn't buy a windows laptop unless there was no alternative. From my perspective, a windows laptop can't beat Apple's attention to detail and tight integration between hardware and software. Win10 doesn't handle touch gestures or high-DPI displays anywhere near as smoothly as MacOS, and you likely won't get a whole system as tightly engineered as a macbook anyway (*cough*keygate...*ahem* lol); what quality is the screen, what color gamut, viewing angles and so on? Cooling system, is it quiet, efficient? And so on.
I have a fat gaming desktop PC as my main rig, that's what it is really good at. Windows has both hardware and software support, PCs have expandability and lots of choice of components from tons of vendors. Only Mac with even remotely decent graphics performance is the iMac Pro which is glued shut and can't be user upgraded in any way and is very expensive due to proprietary components and xeon CPU/ECC RAM, starting off at €5000 and rocketing away from there. Also, few games run under MacOS anyway.
But for a laptop, expandability isn't important - to me anyway. I'll buy a good one and then use it until it's worn out/obsolete; it's more important here that each part of the laptop is tailored specifically for its task, like in a Mac, resulting in a package that is slim, light and still quite powerful.
I'm seriously looking at a quad-core 13" MBP now. It 'only' has intel integrated graphics, but compared to my 2011 macbook with its sandy bridge iGPU, that graphics is absolutely lightning fast. It's good enough for my needs, and at 1.3kg it's so tiny and light. That it doesn't have a dGPU which guzzles battery and belches heat is OK; I'd rather have longer battery life and a lighter, slimmer laptop instead. Usability #1, portability #2; performance #3! Hence a Macbook.
The test is accurate as designed - it doesn't necessarily represent any real-world usage scenario though, but if all the test did was create a virtual copy of a test file then the test would be over basically as soon as it started, with nearly infinite disk performance as a result.
You'd need a software driver that lets the thunderbolt interface act as an ethernet network adaptor, but as you wouldn't have any hardware acceleration for packet/header generation, CRC calculations and so on your CPU might well bottleneck badly at thunderbolt 3 interface speed.
Seriously, the disk test is fine as-is. "Fine", in that it's a synthetic best-case scenario that doesn't even try to mirror a real-world situation. If you want an accurate disk benchmark you'd need something like a recorded disk I/O stream captured from genuine applications that actually stresses the disk subsystem. Database or fileserver apps, and so on. Anandtech and other serious review sites run tests like that.
This performance figure (and Apple's stated spec) is more like pressing down the clutch and gas pedals of a car as far as they'll go and see how high the rev indicator needle goes before it maxes out. It doesn't mirror any actual work getting done.