True, although a lot of apps write data to the hard drive, I'm sure the difference really becomes apparent while working with large video files.
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.
1TB Fusion vs. 512GB SSD performance?
- Thread starter iBighouse
- Start date
- Sort by reaction score
True, although a lot of apps write data to the hard drive, I'm sure the difference really becomes apparent while working with large video files.
Ok, look, I don't want to argue and you could be right. But - it sounds weird to me. I think that 300-370Mbps is just too low for a PCIe SSD, SanDisk or not. The difference between 128Gb drive and 256Gb can't be 200Mbps, and I know I measured more than ~550Mbps in Blackmagic disk speed test on my friends 13" rMBP (with a 256Gb drive). My Samsung 512Gb SSD in a 15" rMBP easily hits 700Mbps in Blackmagic - so that would mean (since we both have 2013. models with Samsung drives) that your drive is almost twice as slow just because it is 128Gb? While I don't doubt your honesty, this is something hard to believe (as I mentioned, Samsung Pro and Evo drives come in 128, 256, etc. sizes and the speed difference between 128 and 256 is never more than 50Mbps!). Are you sure that your drive is working properly?
Also, if what you say is true, how do you explain the 650Mbps read speed on my FD?
Look, from what I read - and that could be wrong, but it sounds reasonable to me - the SSD part of the Fusion Drive is a standard Apple SSD (that is either SanDisk, Samsung or Toshiba). They use the same PCIe drives across the line, like the 128Gb drives they use in, say, MacBook Airs or 13" Retina MacBook Pros. So, they are not using some different, "slower" drives. Now, these PCIe drives should be fast, hitting anywhere 500-700Mbps depending on the manufacturer. The write speed is slower on the FD because (and I admit, I don't quite understand this technical stuff, I'm just interpreting what I read) the write process is a bit more complex - OS X writes on the SSD for consistency and at the same time moves this data to the HDD for permanent storage, so the result is a bit slower 350-400Mbps. Again, this is just what I read - but it matches my speeds almost to the letter: I have 650Mbps read speed and around 370Mbps write speed on my FD. If I split the two drives, writing on the pure SSD drive should be higher, while the read speeds should be unaffected.
You make a compelling argument, but if what you say is true - shouldn't I have 400Mbps read speeds on my FD (or even less)?
Either way - whether you're right or I'm right doesn't change the fact that - outside benchmarks - using the FD (when it is using the SSD part, which is, honestly, most of the time, and 100% of the time for most used apps and OS) is perceptibly not different than my pure, 512Gb Samsung PCIe SSD, which the fastest drive that Apple currently offers and, as I said, hits 700-750Mbps speeds in all the tests I tried.
So it comes down to this: claiming that some mythical "slower SSDs" are present in FDs is not true. Because when you say something like "Apple is using slower drives for FD" - everyone would naturally assume that they are not using their standard SSD drives but are, in fact, using some cheaper, slower options. Not true. If all the reviews were happy of the new PCIe speed of MacBook Airs that come with 128Gb, then all I'm saying - that's the speed you're getting from a SSD part of a FD drive. Don't scare people off - it's a very, very fast SSD. It is faster than most SSDs you find in PCs these days, it is far from "slow" and, again, the speed of the SSD portion is NOT a reason to avoid FD. Neither is temperature. Get a pure SSD if you want greater reliability and speed consistency. That's it. While the FD is running from a SSD part (again, most of the time) you will not be able to perceive a difference compared to the pure SSD. Even if benchmarks say otherwise.
My problem is that people come here for some honest advice and that this elitist attitude and phrases like "Apple uses slower SSDs" (which is debatable at best) and "FDs generate more heat" and stuff like that may confuse people and drive them off from solutions that would benefit them the most. SSDs are better, but they are NOT the best solution for everyone, as they have a great, cheaper alternative - the Fusion Drive.
/rant off
Totally and 100% agree with all you just said. I find it so annoying that peoples (like Redheeler) on this forum always advise against Fusion Drives in every threads based on very tiny speed difference in real life usage.
I can corroborate your benchmark on the reading side, which is between 650 MB/s and 700 MB/s on my iMac (late 2013 PCIe 1TB Fusion Drive). Hardly a "slow" SSD by any standards.
I'm not denying that the Fusion Drive 128GB SSD is about 320MB/sec. In that sense you are correct. In that one parameter it is significantly slower than, say, a 512GB SSD.
However -- reads typically outnumber writes by 4-to-1 or more. On some workloads it is 10-to-1. Anybody who is interested can examine this on their own system by looking at Activity Monitor, select Disk button at top, and and comparing number of "Reads in" vs "Writes out" at the bottom.
Using the 4-to-1 case and comparing to >256GB SSD, 75% of the disk activity (the reads) is about 95% as fast, and 25% of the activity (the writes) is about 45% as fast. The overall real world I/O difference is determined by the weighted arithmetic mean of these, which is 600MB/sec: http://en.wikipedia.org/wiki/Weighted_arithmetic_mean
So for that workload the real world I/O difference is about 600 MB/sec (FD 128GB SSD combined reads and writes) vs about 720MB/sec for a larger internal SSD. That's about a 17% difference.
Even *that* difference is only the I/O portion of a task which requires CPU, GPU, and other resources. If the overall task is split 2-to-1 between non-I/O and I/O elements, then the actual aggregate real-world perceptible difference of the 17% I/O difference is much less. If you're waiting on CPU and GPU 66% of the time, a 17% difference in I/O rate may not even be noticed.
Even a lower I/O rate may not make any difference if you're not waiting on it. This is determined by disk queue length. This can be examined using the Terminal command iopending: https://developer.apple.com/library...win/Reference/ManPages/man1/iopending.1m.html
Anyone interested can examine the relative contributions of CPU, GPU and I/O to their workload by using iStat Menus.
As a professional video editor I work with large video files frequently and SSD vs Fusion Drive doesn't make that much difference if the files are on an external Thunderbolt array. Mac internal SSD is way too small for large video files, so the main real world difference is how fast an app loads which is mostly reads, and FD is good at that.
So for playing the "benchmark olympics", internal SSD is definitely faster. In the real world of production work using large video files it often doesn't make much difference for the reasons listed above.
Do you know the brand of your SSD? Mine is a SanDisk (PCIe SD128, in a late 2013 3TB Fusion Drive), I will bench it eventualy.
Last edited:
The issue is that benchmark is going purely to the SSD portion of the fusion arrangement, so the benchmark is only representative of ~128GB of data across your 1.128TB drive, or of a 3.128TB drive if you have the 3TB Fusion (as I do - but no longer use as a boot drive).
That's why every now and then things are just... slow, because some-to-most data you're accessing is NOT on the SSD portion.
Apple's implementation of fusion puts aside ~4GB as a "scratch" portion, so that's why the benchmark ends up going there, and why it appears it's really fast. In reality, it's more complicated. I got so frustrated with the inconsistency of fusion I moved to a pure SSD solution over Thunderbolt, and would never - ever - go back.
FYI, it's also not a PCIe 1TB Fuson drive. It's a PCIe SSD, and a regular old SATA III HDD.
Last edited:
When did you feel the slow HDD? I've been a FD user since November, in my day to day use, for me it is as fast as a SSD. The only moment where I felt the HDD was when I paste my 900GB of datas after Mac OS X installation.
I'm not a videast though, I use my Mac for heavy duty photography on Photoshop. And also scanning and gaming.
I totally understand how Fusion Drive work and how it is a software layer that manage a SSD and a HDD, and that benchmark only show the particular speed for either the SSD or the HDD.
The thing is for me, and my usage, Fusion Drive make a lot of sense.
In fact, If I had a computer with a 128GB SSD and a 1TB HDD without any "fusion drive" implementation, where I could manually decide what goes on which drive, I know I would waste a lot of SSD space simply by keeping OS and Apps and some specific files on the SSD and most of my "long term data" on the HDD. There would always be a considerable amount of SSD space I wouldn't use simply because I don't want to waste my time micromanaging that some files I likely will want to access them fast, some not.
In the end, I would probably end up having a half filled SSD with a half filled HDD.
In fact, using a Fusion Drive, there is probably more situations where I end up accessing a file on the SSD that would otherwise have been stored on the HDD (If I was manually managing different drives) that situations where I end up opening a file on the HDD that I would have manually kept on the SSD.
Personally, having faster access speed is something that is convenient and fun, but not "vital" to what I do. If some times I open an app that I don't use often and it take 2-3 secondes instead of 1/4 of a second, It doesn't really affect my work in the end.
I understand that for some specific usage, access speed and being able to predict it is really important, but for most usage I just don't think it matter that much.
In fact, thinking about it, had I opted for a 256 SSD (with an external HDD) instead of a Fusion Drive, I would probably have wasted more time globally taking time to manually move and manage my files between the two drives than the time Fusion Drive made me wait a little bit more.
I notice it greatly if for example there's an app I don't open super often - like Chrome. Watching that icon bounce in the dock and over while I hear the hard drive grinding quietly in the background - does my head in. I notice it when my iTunes library is slow to refresh, since fusion has decided that since I don't use all that music - clearly a lot of it is on the HDD portion.
I fully understand. I opted for the 3TB fusion, but now exclusively use a 1TB SSD for everything, and the 3TB (well...3.1TB) is now my scratch disk/photos tertiary backup. There's no question in the difference in speed, despite my external Thunderbolt 1TB boot SSD "benchmarking" at only ~400MB/s reads. That number is a load of nonsense. Seek times are where SSDs are at. Fusion simply doesn't compare, and I'm so glad to be rid of it as my primary OS drive. If you're happy, then fine!
Try calling your neighbour "unintelligent". When they get angry, say that they are comparatively unintelligent to Stephen Hawking. See if they understand your reasoning.
Here's the thing - everything is relative. The fastest car is slow compared to a supersonic jet. And the supersonic jet is incredibly slow compared to a light particle. And I could call your SSD slow and hot compared to the new Samsung 2.15Gbps drive that draws only 2 milliwatts of power.
But there is a thing called common sense, and there are comparisons made in practical terms. I'm sure we can all imagine 'fast' computers and 'slow' computers, and their general feel. Labelling a piece of hardware that is many times faster than what is considered an 'average level' as "slow" - is just wrong in every practical sense.
The only sense I can make out of your numerous claims is that a thing becomes slow the moment something faster appears. If one reads yours and likeminded posts on these forums, they would get the impression that i5 CPUs are slow for any production work, that M290X GPUs are inadequate for anything but surfing, that anything below 24Gb RAM is not enough and, oh yeah, that one of the fastest drive solutions in the PC world today is - you've guessed it - slow.
I'm just going to say it - in no universe is FD slow. Even compared to a pure, maxed out SSD, it's not slow - as the differences in speed are, in most normal usage cases, anecdotal and imperceptible. The differences to ordinary HDDs are, on the other hand, quite obvious. The OS and apps launch in seconds.
That's not slow - that's fast.
How can you make this claim? Did you proclaim the iMac 5K as some sort of non-compromise zone? Like, if I decide I want a great screen, than I'm somehow irrational if I don't get the top model?
As for the 'when you pay less, you get less' - I'm not sure what you're getting at. Did you know that when you buy an iMac 5K Retina, you get less performance compared to a 12-core, dual GPU, $10.000 dollar Mac Pro? Just trying to be informative, in case you were, like, thinking that you'll get the same performance - while, in fact, you get less for paying less.
Last edited:
Quite the contrary, 'slow' has meaning on its own in normal human communication. No one calls a McLaren F1 'slow' - although its speed is snail-like compared to, say, a comet.
When people come here for advice, they use common language conventions. They don't speak in relativistic terms - and I'm sure you're well aware of that.
So, I challenge your claims. In normal communication, when talking about modern computers - yes, 'slow' indeed has a meaning on its own. And no - FDs are not slow in any practical sense. And claiming they are 'slow' is just, pure and simple, bad advice and not informative in any usable way. In fact, I would say it's misleading.
In my post I stated the 128 GB SSD is "slow". Normally I'd use the phrase "slower than a 256 GB SSD", but my comparison was implied by the post I quoted, which explained that 128 GB SSDs are slower than 256 GB SSDs.
That quote said the 128GB SSD in Fusion Drive is pathetically slow, and you agreed that's why you don't recommend FD -- because the SSD portion is so slow.
In fact Fusion Drive is pretty fast. I have four Macs, two with SSD and two with FD, plus a Windows PC with a 1TB SSD. There's no question SSD is fast relative to a pure rotating drive. The Samsung 850 in my Windows PC is much faster than the 10k rpm Velociraptor it replaced. My next iMac will definitely be SSD.
However I wouldn't describe Fusion Drive as slow or the SSD portion as pathetically slow. Describing that way is misleading to people who come here looking for advice.
Fusion Drive is a good compromise which often provides much of the real-world performance of SSD combined with larger capacity. If money or storage size are non issues, SSD gives more consistent performance. But if the user needs frequent, high-bandwidth access to over 1TB storage, the Mac SSD is too small. They can't all afford an external Thunderbolt disk array.
Likewise if the user's budget only permits 256GB SSD yet they need more storage, they could end up using a slow bus-powered USB portable drive. This could be slower overall than a 1TB Fusion Drive.
I think most people who have used otherwise similar HDD and FD iMacs can appreciate the benefit of FD.
While this wasnt a scientific experiment by any means, a few weeks ago I was trying out CS6 Premiere Pro on my Base Retina iMac(1 TB fusion, i5, 290m, 24GB Ram) and was surprised how slow it was compared to my 2012 Macbook Air(256 SSD, i7, 8GB RAM) when using the internal drive for everything. It wasnt anything intensive, I just ripped a bunch of small youtube videos, converted to MPEG2 and dragged them to the timeline for editing. My iMac had a really hard time playing off the fly without rendering the timeline. The exact same task on the Air didnt require any rendering.
I really didn't investigate too much since it was a baisc test but the iMac had to keep rendering the timeline with small MPEG2 videos(I think the files were 15BM to 150MB in size)
Benchmarks aside, just based on subjective first-hand experience with both Fusion and SSD drive in three iMac Retina's and a couple of rMBP's, I am taking the SSD every time if I don't need the internal storage capacity of a Fusion drive.
The SSD offers relatively more speed, silent operation, and lesser probability of failure than a Fusion setup. Subjectively, it feels faster for the kind of web production work that I do, even though none of it comes close to pushing the computer to its limits.
The SSD offers relatively more speed, silent operation, and lesser probability of failure than a Fusion setup. Subjectively, it feels faster for the kind of web production work that I do, even though none of it comes close to pushing the computer to its limits.
I can't explain that but I doubt it's related to SSD vs FD disk I/O. Total I/O rate for a single stream of 1080p/30 MPEG2 is about 10 megabytes/sec, which even the slowest USB 2 portable hard drive could handle.
The i7 in the MBA does have 2x the virtual cores via hyperthreading, so under an ideal workload it is theoretically 40% faster at the same clock rate but the RiMac has a faster clock. If anything the iMac should have more CPU resources.
Timeline rendering of CS6 effects heavily leverages the GPU via the CUDA API, which is nVidia specific. Your iMac doesn't have an nVidia GPU, but neither does the MacBook -- it uses the integrated Intel GPU.
I think the Mercury rendering engine in CC uses OpenCL more effectively but I thought in CC it was more CUDA oriented.
This assumes both MBA and RiMac are running CS6, inc'l all patches. If they are running different versions, anything is possible.
I have some trouble believing that Retina iMac is in any way slower than a 2012 MacBook Air - as it is (and feels) faster than my 2013 top of the line 15" Retina MacBook Pro. The 2012 MacBook Air has a SATA SSD (I'm guessing Redheeler would call that pathetically slow
and the iMac has a PCIe SSD based Fusion Drive. Also, your iMac has 3x the RAM and double the number of physical cores, running at much higher speed (the mobile i7 used in Airs has only 2 cores with HT and has a much lower clock - it's incomparable) Especially for the tasks you mentioned (small Youtube videos). I'm sorry, there's just no way the 2012. Air can be faster at anything. There must be another reason.As I mentioned several times before, my experiences are very positive with the iMac 5K and Fusion Drive, and I'm comparing it to the 15" Retina MBP (which has a true quad core i7, a 700Mbps 512Gb PCIe SSD and 16Gb RAM) which feels slower in general. And disk-based tasks (like launching apps, rebooting, etc) don't seem much faster on the rMBP. My FD experience has been great so far.
I really read strange things here. I trust you guys are honest with your claims, but I just find them hard to believe.
I have studied and read lots of articles on this subject and I think which way you jump re: FD SSD/HDD (dual drive) or SSD only has to be a personal decision based on your usage.
For my own personal situation it would have to be an SSD only or at a push the Fusion Drive. I cannot see the logic in installing an SSD alongside your existing HDD.
I run a NAS, so with the exception of iTunes (which simply doesn't play well if removed from the iMac) most of my data is on the NAS. That way I can access and work on my files wherever I go. I simply prefer to have my own personal 'Cloud' than rely on one provided by other businesses.
If I installed an SSD alongside my existing HDD then logic would dictate I should bring back my data first onto the SSD, copy it to the HDD and provide lots of symbolic links from the SSD to the HDD before deleting the material on the SSD. You can simply use preferences then 'accounts' to achieve similar but it works better if the cache is on the SSD.
However, given I would then not be able to access the material remotely I see little point. Therefore for me it would make more sense to simply replace my existing 500gig HDD with a similar capacity SDD.
I could also argue is there any point installing an SSD at all? I have recently removed the bottleneck from my HDD (Parallels .PVM file) with the result the machine is now like a spring chicken again. Apps in the dock bounce just once and they are open. The majority of the machines use is 'Office' type tasks and I find clicking on a 'new document' to be almost instantaneous.
The other thing to consider is cost - if I add up the cost of the drive, the tools to install it, plus I think I would need the 'thermal sensor' too the costs rack up. For what I use the machine for would I see that much difference? After all, I rarely shut the machine down completely so simply hitting any key on the keyboard brings it out of sleep almost immediately, I don't have any boot time issues.
These are all things worth considering before you go ahead.
For my own personal situation it would have to be an SSD only or at a push the Fusion Drive. I cannot see the logic in installing an SSD alongside your existing HDD.
I run a NAS, so with the exception of iTunes (which simply doesn't play well if removed from the iMac) most of my data is on the NAS. That way I can access and work on my files wherever I go. I simply prefer to have my own personal 'Cloud' than rely on one provided by other businesses.
If I installed an SSD alongside my existing HDD then logic would dictate I should bring back my data first onto the SSD, copy it to the HDD and provide lots of symbolic links from the SSD to the HDD before deleting the material on the SSD. You can simply use preferences then 'accounts' to achieve similar but it works better if the cache is on the SSD.
However, given I would then not be able to access the material remotely I see little point. Therefore for me it would make more sense to simply replace my existing 500gig HDD with a similar capacity SDD.
I could also argue is there any point installing an SSD at all? I have recently removed the bottleneck from my HDD (Parallels .PVM file) with the result the machine is now like a spring chicken again. Apps in the dock bounce just once and they are open. The majority of the machines use is 'Office' type tasks and I find clicking on a 'new document' to be almost instantaneous.
The other thing to consider is cost - if I add up the cost of the drive, the tools to install it, plus I think I would need the 'thermal sensor' too the costs rack up. For what I use the machine for would I see that much difference? After all, I rarely shut the machine down completely so simply hitting any key on the keyboard brings it out of sleep almost immediately, I don't have any boot time issues.
These are all things worth considering before you go ahead.
Spoken like someone who's never experienced the sheer joy of using an SSD as your main Mac drive
Your Mac may seem like a 'spring chicken' but with an SSD it would run like an Ostrich
I have some trouble believing that Retina iMac is in any way slower than a 2012 MacBook Air - as it is (and feels) faster than my 2013 top of the line 15" Retina MacBook Pro. The 2012 MacBook Air has a SATA SSD (I'm guessing Redheeler would call that pathetically slow
At a high level the iMac is much more powerful computer no doubt say in gaming but the difference i've seen is the some lag at times in preview. Now we are not talking about preview but Premiere Pro obviously but I just wanted to point that out that the iMac is in no way as consistently fast as my 2012 MacBook Air for browsing files. Sometimes its quick sometimes it hangs.
If I had to in particular nail down where I saw a major difference in Premiere Pro it was two areas. First, when I added clips to the project I had to render them on the iMac and not the Air. Second, it was adding a clip to the timeline and then dragging the clip smaller directly on the timeline. On the iMac you needed to render the clip in order to play it. On the Mac Air you didn't. I didn't have to render anything on the iMac if i can recall when I edited the traditional way trimming the clip down in the preview window and adding it to the timeline(or whatever it's called).
I'm going to run Blackmagic Disk Speed Test on both tonight if I can remember to do so. Also I'd like to give Final Cut X a whirl on both machines.
This could all be user error on my part but i'm curious what the difference was exactly. I was assuming there was some different setting in the render playback but maybe it's something with the fusion drive...I dunno. But these are small YouTube clips so im surprised that even if there was some custom setting on the Air that I enabled a while back the iMac's Premiere was definitely using the defaulted software configurations.
One possible reason could be different playback resolutions. In CS6 if you right-click on the playback window, it is adjustable. If the iMac is playing back at a higher resolution than the MBA that could slow it down.