Full SSD on iMac is criminal

[bonedaddio]

Hi all, aside from the bashing (fun to follow, but not useful), this is a great thread! Apple does what Apple does, and Apple Charges What Apple Charges... but we as buyers can decide whether to "drink the coolaid" or even just "some of the coolaid".
I'm a long time PC builder and user, and Windows 8 has finally convinced me to move Adobe Master Collection and all my stuff to Mac. Since 2006 I've made most of my living on a MacPro Quad 1,1 running either XP or Win7 bootcamped. I have some specialized programs for 3D work (and QuickBooks Pro) that I'll need to keep a Win7 bootcamp for, but the rest is all going Mac.
After some research at order time, like WilliamG and AshleyPenney, I went with a 1 TB disk drive in my BTO iMac (i7; 680mx 2gb). I already have the WD TB desktop adapter sitting here (works; used it with TB on my MBP, does NOT provide full TB bandwidth according to testing posted elsewhere), and I'm ordering in a Samsung 840 (256gb) and a Pegasus J4 (with two striped SSD's will go insanely high) TB enclosure today. You pays your money and you takes your choice.

I like the shelves... I have a friend who can make me a couple.

If anyone's interested, there's a fascinating "how-to" build your own Fusion Drive:
http://www.petralli.net/2012/10/anal...nal-hard-disk/
Apparently it just works (the fusion drive file management stuff is built into ML) and works well. So... build your own fusion drive with an External SSD and HDD, and use a 256 SSD while you're at it!! I agree, the 768gb SSD option just seemed crazy... especially after some research. Basically, if you use the right external enclosure, and a fast SSD, you'll get all that speed and storage for less money and a hell of lot more flexibility!

I like ThunderBolt, I like the idea of being able to basically move the system bus outside the box as needed. I like the all in one concept... and I'm tired of crawling around on the floor hooking things up. I'll have one TB cable and one USB 3.0 cable running over to a cabinet with all the whatevers inside it; I'll add quiet cooling if needed to it. My desk should remain pretty uncluttered... and I'll finally be able to vacuum under my desk!!
I find this all fascinating, and a great way to while away the time before my 27" gets here.
Best of luck to us all in the New Year!

[WilliamG]

Originally Posted by DisMyMac

Is it something like this?

http://www.bhphotovideo.com/c/produc...t_Adapter.html

Add any SSD you want, make that the boot drive, and that's it?

That's the one. No issues here with it at all.

*EDIT*

Sorry that's NOT the one. I have the Backup Plus, not the GoFlex.

[flynz4]

Originally Posted by cloudyo

Thats not true either. Most of the data on the HDD is not in sequential order. To even read large files the HDDs head has to move multiple times to get from the beginning to the end of a single file.

I did not mean my statement literally... but I acknowledge that it can be interpreted that way.

On a relative basis... large file I/O on HDDs has a minimum amount of mechanical overhead per data transferred. This is why SSDs offer the lowest performance gains over HDDs during large read or write transfers.

They are absolutely stellar on small transfers vs HDDs.

/Jim

[cloudyo]

Originally Posted by flynz4

On a relative basis... large file I/O on HDDs has a minimum amount of mechanical overhead per data transferred. This is why SSDs offer the lowest performance gains over HDDs during large read or write transfers.

Fragmentation is something that gets worse over time. As the HDD fills up there is less continuous free space to place large files in. Meaning: The mechanical overhead will increase the more you use the drive.

But even if we are only looking at sequential read performance, the fastest consumer HDDs barely reach 200MB/s. Compare that to the 400+ MB/s that a typical SSD can deliver.

[WilliamG]

Originally Posted by DisMyMac

Is it something like this?

http://www.bhphotovideo.com/c/produc...t_Adapter.html

Add any SSD you want, make that the boot drive, and that's it?

Quoting this again. You want the Backup Plus, not the GoFlex. They may be the same device, but I have no experience with the GoFlex, only the Backup Plus.

[flynz4]

Originally Posted by cloudyo

Fragmentation is something that gets worse over time. As the HDD fills up there is less continuous free space to place large files in. Meaning: The mechanical overhead will increase the more you use the drive.

But even if we are only looking at sequential read performance, the fastest consumer HDDs barely reach 200MB/s. Compare that to the 400+ MB/s that a typical SSD can deliver.

As stated many times... large file transfer rate is probably the worst metric to use for SSDs. IOPs is much better.

The 2X performance difference you mention is insignificant compared to the increase in IOPs.

HDDs deliver somewhere between 200 - 400 IOPs. By contrast, a consumer SSD will deliver in the 40,000 IOPs range... an increase of 100X. An enterprise SSD might offer 250,000 - 500,000 IOPs... or 1000X.

It is a sin that SSDs are being compared based on MB/sec. That is not where they shine. It is not what makes your computer feel (and be) fast when switching to an SSD. The performance increase is predominately about IOPs.

/Jim

[qamaro]

Originally Posted by flynz4

As stated many times... large file transfer rate is probably the worst metric to use for SSDs. IOPs is much better.

The 2X performance difference you mention is insignificant compared to the increase in IOPs.

HDDs deliver somewhere between 200 - 400 IOPs. By contrast, a consumer SSD will deliver in the 40,000 IOPs range... an increase of 100X. An enterprise SSD might offer 250,000 - 500,000 IOPs... or 1000X.

It is a sin that SSDs are being compared based on MB/sec. That is not where they shine. It is not what makes your computer feel (and be) fast when switching to an SSD. The performance increase is predominately about IOPs.

/Jim

Jim,

I agree. Yes IOPS should be used for comparisons as it will show the greater differences for single drives or versus RAID's (HDD or SSD). The issue is HDD manufactures love to talk about Mbps (not IOPS) and SSD's ones like to push numbers in IOPS. Yes it disadvantages the true benefit of the SSD's but, it's a measure nonetheless. The tools being used on the forums are just for relative relational purposes. While not in IOPS one could do the conversion manually. The real issue is outside of DBA's and IT Architects, most end users don't know how to look or measure, the transactional nature of these drives outside of specific media tasks (Media access times, frame-rates, file conversions, etc.). Which is why Blackmagic is used so much as it shows how drives comparatively handle a known task (e.g. video throughput).

For those that need a primer on how to take the IOPS data you see in SSD ads and convert it to MBps yourself here is the link (see below). Or if you like convert HDD MBps data into IOPS to compare at that level instead. So, as an example a 100K IOPS @ 4K is equivalent to 390.62MBps.

SSD Freaks Converting MBps to IOPS

Script to Measure IOPS performance on Linux / OSX

IOPS Testing in OSX

Until we have better "user-friendly" tools in OSX to compare both IOPS and MBps, for variable tasks (small and large transactions) folks will use whatever is easiest to understand and compare.... IMHO

Here's hoping that Drobo (or others) release this tool and they allow it to work with all drives as it measures both MBps and IOPS for differing transactions. Which they used to show off their latest Drobo's true results.

Drobo Performance Test

(sorry for the long post )

[flynz4]

Originally Posted by qamaro

Jim,

I agree. Yes IOPS should be used for comparisons as it will show the greater differences for single drives or versus RAID's (HDD or SSD). The issue is HDD manufactures love to talk about Mbps (not IOPS) and SSD's ones like to push numbers in IOPS. Yes it disadvantages the true benefit of the SSD's but, it's a measure nonetheless. The tools being used on the forums are just for relative relational purposes. While not in IOPS one could do the conversion manually. The real issue is outside of DBA's and IT Architects, most end users don't know how to look or measure, the transactional nature of these drives outside of specific media tasks (Media access times, frame-rates, file conversions, etc.). Which is why Blackmagic is used so much as it shows how drives comparatively handle a known task (e.g. video throughput).

For those that need a primer on how to take the IOPS data you see in SSD ads and convert it to MBps yourself here is the link (see below). Or if you like convert HDD MBps data into IOPS to compare at that level instead. So, as an example a 100K IOPS @ 4K is equivalent to 390.62MBps.

SSD Freaks Converting MBps to IOPS

Script to Measure IOPS performance on Linux / OSX

IOPS Testing in OSX

Until we have better "user-friendly" tools in OSX to compare both IOPS and MBps, for variable tasks (small and large transactions) folks will use whatever is easiest to understand and compare.... IMHO

Here's hoping that Drobo (or others) release this tool and they allow it to work with all drives as it measures both MBps and IOPS for differing transactions. Which they used to show off their latest Drobo's true results.

Drobo Performance Test

(sorry for the long post )

Thanks for the note. In reality... getting a good IOPs number is much more difficult than the conversions listed in the links. I have given myself the personal goal of turning the industry around on this metric... and hope to make significant progress in 2013. I have also assigned some good people to the task to make it happen.

/Jim

[qamaro]

Originally Posted by flynz4

Thanks for the note. In reality... getting a good IOPs number is much more difficult than the conversions listed in the links. I have given myself the personal goal of turning the industry around on this metric... and hope to make significant progress in 2013. I have also assigned some good people to the task to make it happen.

/Jim

Agreed.. consumer understanding has to start on some basic level though hence the links. But, its good to know someone is on the case on the industry side of things. Catch you on the flip side of 2013.

[njean777]

Originally Posted by 7enderbender

That's correct. Which is why I so far have never bought Macs. I came close several times and there was always something missing that I wanted or needed at the time. Without the Win 8 disaster and the horrid Windows laptops available these days I still wouldn't go near it for this exact reason.
If Apple wants to continue to grow market share in their computer segment beyond loyalists who are willing to justify ANYTHING that Apple comes up with and beyond the very basic consumer segment who buy iMacs because of the design without much need for anything else then they need to offer more professional choices between the low end consumer machines and the very expensive tricked out machines.

And even with a $2000 budget and medium difficult applications in mind (photo editing and home studio recording) it is very very difficult to find a Mac that is not too much of a compromise. I've pretty much ruled out iMacs at this point still and will purchase a Mini with NEC screen and put in new drives and memory myself. iMacs are too difficult to upgrade and there is nothing out of the box that would work for me without going WAY over budget.

And yes, I'd also want a 256 SSD (system and program files only) and no fusion drive.

TBH if you really "need" a specific feature you should just build your own pc. Buying prebuilt machines, that are not apple, are a complete waste of money unless it is a laptop.

[cloudyo]

Originally Posted by flynz4

As stated many times... large file transfer rate is probably the worst metric to use for SSDs. IOPs is much better.

I completely agree. In your initial post you were making it sound like HDDs aren't that bad compared to SSDs when we look at the reading speed of large files. I simply tried to make it clear that SSDs are vastly superior even in that category.

[DisMyMac]

Originally Posted by WilliamG

Quoting this again. You want the Backup Plus, not the GoFlex. They may be the same device, but I have no experience with the GoFlex, only the Backup Plus.

Good tip, thanks. The 128 is a year newer than the 121, but someone said it was all the same (right).

[jraske]

So, I purchased the 27" 3.4 GHz i7, GTX 680MX 2GB, and upgraded the RAM myself to 32 GB . . . I never considered the full SSD, and went with the 3 TB Fusion Drive.

My main reason for going with the Fusion Drive is that I dislike external HDs (I hate cables and extra things on the desk) and prefer as much internal storage as possible.

Also, my gut told me that on a practical, daily basis I would not find the Fusion Drive to be "slow". I use the iMac primarily, although it is stuffed with my family's music and movies, which are typically streamed to Apple TVs and iOS devices.

My bottom line (and if yours most likely if have a similar use case) is that the Fusion Drive is the way to go. This thing flies because the applications, OS, and most used files are on the SSD, and there is no lag when streaming media stored on the traditional HD.

[natesymer]

It used to be the case that you could have three SSDs in an iMac 2011 (One in HDD bay with 3.5" adaptor, and two where the optical bay would rest). I could buy three 512GB Crucial M4s for around $1100, with all of the toolkits and other stuff, around $1200. Correct me if I'm mistaken, but that's ~$100 cheaper than the 768GB SSD from apple. Plus, you have 2x the storage space.

Plus, correct me if I'm wrong, it costs only $500 for the 768GB SSD in the rMBP...

In addition, the fusion drive has the possibility to corrupt files, AND it will destroy the SSD. Why apple went with a 128GB model is beyond me... At least use a 256GB model or an SLC SSD...

Lastly, you can split up a fusion drive into two logical disks. That way, you would have a 3TB HDD and a 128GB SSD. I don't know about you guys, but I really only have around 50GB of data (well, I have 120GB on my second SSD, but... That can go on the HDD).

[smr]

Originally Posted by natesymer

Plus, correct me if I'm wrong, it costs only $500 for the 768GB SSD in the rMBP...

The $500 cost for the 768GB SSD in the rMBP is cost to upgrade from the 512GB SSD.

In the iMac, the 768GB SSD cost is upgrading from a 1TB hard drive.

[gnasher729]

Originally Posted by cloudyo

Thats not true either. Most of the data on the HDD is not in sequential order. To even read large files the HDDs head has to move multiple times to get from the beginning to the end of a single file.

So what if a gigabyte file is in ten pieces? 10 times repositioning the head = 10 times 10 milliseconds = 100 milliseconds; nothing compared to the time to read a gigabyte.

[xgman]

Originally Posted by WilliamG

Do what I did, and get a Samsung 830 or 840 Pro, and a Seagate Backup Plus Thunderbolt 2.5", and boot off that. I'm running the 830 256GB here.

I also got a samsung 840 pro, but . . it doesn't bench quite as fast as the internal ssd. (controller adds a bit of latency) Just saying. If only the 840 was inside. . . .

[excommie]

Originally Posted by flynz4

This is absolutely false. SSD's shine on small files... not large files. Just think about it for a second... they do not have to reposition a head... or wait for the data to rotate to the correct position. They are also faster on reads rather than writes.

For large files on a HDD... you position the heads once... and then stream the data from the HDD. Hence... the long time to move the heads is amortized against a lot of data... hence they have reasonably high "large read" or "large write" performance.

SSDs give the smallest performance advantage during "large writes". They give spectacular performance on "small read".

Luckily... (as you imply)... the vast majority of access on client systems are "small reads". Estimates are over 90%... with the vast majority of the rest being "small writes".

SSDs are a HUGE advancement for client systems.

/Jim

I see your point. But I also have some experience using SSD over FW800 bus and then using the same SSD over the SATA bus. In both cases I was loading the same OS (Mountain Lion). The difference in boot time was negligent. Both times, there was no noticeable difference when opening common applications (Safari, Word, Handbrake, etc...). OS is a collection of small files - using FW800 I was not able to fully saturate the bus - FW800 bus wasn't really a bottleneck when loading OS.

However, there was a noticeable speed increase when opening the iPhoto Library over SATA bus as opposed to FW800 bus.

[WilliamG]

Originally Posted by xgman

I also got a samsung 840 pro, but . . it doesn't bench quite as fast as the internal ssd. (controller adds a bit of latency) Just saying. If only the 840 was inside. . . .

Agreed. But the benchmarks aren't affecting my usage.