I got my two Supertalent 64 GB SSDs as boot drive installed in RAID0 with 128K stripes. Vista64 Bootcamp is running of a 250 GB HD in AHCI mode as are my 1 TB Movie disk and the Blu-Ray.
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.
SSD Drives for Mac Pro Boot Drive?
- Thread starter netdog
- Start date
- Sort by reaction score
I got my two Supertalent 64 GB SSDs as boot drive installed in RAID0 with 128K stripes. Vista64 Bootcamp is running of a 250 GB HD in AHCI mode as are my 1 TB Movie disk and the Blu-Ray.
Unbelievably fast when opening apps. I was copying some Winclone backup files and getting 75-90 MB/s. Still I would probably like to run a benchmark that is familiar to this board to get something objective. What would be a good thing to test with?
![]()
![]()
OS X is 32-Bit with SSD
Vista is 64-Bit with a 7200 rpm HD with AHCI driver
so none of the tests is giving the full potential. that will come with SL.
OS X is 32-Bit with SSD
Vista is 64-Bit with a 7200 rpm HD with AHCI driver
so none of the tests is giving the full potential. that will come with SL.
How about an Xbench disk test to start.
Doesn't it just feel like you doubled the speed of your system though?
Doesn't it just feel like you doubled the speed of your system though?
How about an Xbench disk test to start.
Doesn't it just feel like you doubled the speed of your system though?
yeah, I'm impressed when I compare it with my new Unibody MBP
The old banger basically beats the MBP by a factor of 4.
Xbench http://db.xbench.com/merge.xhtml?doc1=383098&doc2=1&setCookie=true
Here's my results... (2 Intel SSD's in RAID0)
As expected your drives beat Intel on write speeds (except for small random writes) and the Intel's beat your SSD's on reading but not by a huge margin.
Cool!
Code:
Results 438.45
System Info
Xbench Version 1.3
System Version 10.5.7 (9J61)
Physical RAM 6144 MB
Model MacPro4,1
Drive Type Intel SSD RAID0
Disk Test 438.45
Sequential 286.82
Uncached Write 276.71 169.90 MB/sec [4K blocks]
Uncached Write 253.02 143.16 MB/sec [256K blocks]
Uncached Read 189.87 55.57 MB/sec [4K blocks]
Uncached Read 898.12 451.39 MB/sec [256K blocks]
Random 930.26
Uncached Write 1095.54 115.98 MB/sec [4K blocks]
Uncached Write 417.89 133.78 MB/sec [256K blocks]
Uncached Read 2346.21 16.63 MB/sec [4K blocks]
Uncached Read 1760.86 326.74 MB/sec [256K blocks]As expected your drives beat Intel on write speeds (except for small random writes) and the Intel's beat your SSD's on reading but not by a huge margin.
Cool!
I want to configure my MacPro as follows:
1x INTEL X-25 Postville 160GB as boot drive
3x WD 640GB Raid0 for data
1x 2TB WD Caviar Green for Time Machine
So Boot and Data will be two separate partitions... will the Time Machine backup work properly? How will I be able to restore the backup to these two partitions in case of a crash?
1x INTEL X-25 Postville 160GB as boot drive
3x WD 640GB Raid0 for data
1x 2TB WD Caviar Green for Time Machine
So Boot and Data will be two separate partitions... will the Time Machine backup work properly? How will I be able to restore the backup to these two partitions in case of a crash?
If at all possible, you might want to swap the 640GB drives for 1TB. You'd not only get additional capacity, but the outer tracks are faster. So by keeping the array at 50% or less, you will end up with higher throughputs.
Not sure on Time Machine, but I'd think it can do what you want. The rest of the drives look good. It's about the max you can go throughput wise if your system is an '09 though, as the ICH10R is limited to 660MB/s.
Good drives (I'm using 8x RE3's on a card). The Caviar Blacks aren't bad either, so I'm using them for backup.
They're attached to an Areca ARC-1231ML (setup in RAID5). Throughputs have hit 1.39TB/s, though that's due to the cache (2GB worth). If I shut it off, or use very large files (>2GB), it drops down to ~800MB/s (reads).
Nah... Crazy fast is 4TB/s or better.
Think parallel cards (either SAS or FC) to racks of drives.
Too rich for me though. 
Yup I get the same thing. I get between 1.2 GB/s and 2 GB/s if I stay within the total size of: (number_of_drives × cache_size). Mine cuts down to about 550 MB/s instead of 800 MB/s like you though. Ooo, geek-superority! Grrrr...
Hhehe, The OS is on a small partition (short stroke), so it's on the fastest tracks. The inner tracks ruin things as you well know.Yup I get the same thing. I get between 1.2 GB/s and 2 GB/s if I stay within the total size of: (number_of_drives × cache_size). Mine cuts down to about 550 MB/s instead of 800 MB/s like you though. Ooo, geek-superority! Grrrr...
The worst case, all capacity = single partition, it produces 550 - 600MB/s range IIRC, and the stripe size was set at 64K. Currently the stripe's been increased to 128K as well. Also helps.
IIRC, you're also running a 128K stripe size, aren't you?
The OS is on a small partition (short stroke), so it's on the fastest tracks. The inner tracks ruin things as you well know.
The worst case, all capacity = single partition, it produces 550 - 600MB/s range IIRC, and the stripe size was set at 64K. Currently the stripe's been increased to 128K as well. Also helps.
IIRC, you're also running a 128K stripe size, aren't you?
Yeah, right now I'm running a 4.5 TB 3-drive stripe at 128K and a 1 TB 3-drive stripe at 256K.
I'm also trying a USB/eSTAT drive cradle. Man in USB mode what a total waste of a USB port.
I didn't know/recall you had a second array. What's it for? Backups?
I'll presume this is the second array (1.5TB set). USB would be a waste of a port or super s_l_o_w, depending on which take you prefer.I'm also trying a USB/eSTAT drive cradle. Man in USB mode what a total waste of a USB port.
What did you do for an eSATA port? Just a bracket, or did you go with a card?
Sorry - but: NEVER NEVER NERVER never never (...) put a BackUp on a RAID-0 ! NEVER!!!
I don't really consider a stripe set a proper RAID level, as it has 0 redundancy. Just speed. It's possible to use, and safer than you might think though, as the power saving features should keep the drives spun down most of the time (until a backup process is started), as the drives aren't being used. But I still wouldn't trust it myself, and wouldn't recommend it.
I prefer 10/5/6 (an array type that has redundancy). Otherwise, it's best to use the backup drives individually, or as a JBOD for low risk + low cost solutions.
What about a Stardom Prodrive, http://www.stardom.com.tw/pro_drive_feature.html
You could using two 2.5" SATA HDD and it would be much cheaper than an SSD, but I'd imagine there would be nothing stopping you using two 2.5 SSD in one of these enclosures. Boy that would fly.
You could using two 2.5" SATA HDD and it would be much cheaper than an SSD, but I'd imagine there would be nothing stopping you using two 2.5 SSD in one of these enclosures. Boy that would fly.
They're fine for mechanical drives, but not SSD's, especially in a stripe set. The reason is, they're only capable of ~200MB/s max throughput (it has a Port Multiplier chip in it), and SSD's can exceed that alone on some drives, let alone 2x.
You don't want to use SSD's with PM chips right now. They just aren't fast enough. Even a PM chip that can utilize SATA 6.0Gb/s devices, would only be good for ~500MB/s at best, so say 2x of the Intel Postville's (G2's) would saturate it.
Will the bottleneck still happen when the Pro Drive is in RAID 1? Would it make any sense to put two Intel X25-V 40 GB SSDs in a Pro Drive as a mirror pair for a boot drive?
It would depend on the specifics (what else is running simultaneously with the SSD's). This also assumes each disk has it's own SATA port, not shared using the Pro Drive device. If that device is used, there will be a bit of a bottleneck if the reads exceed 200MB/s). In the case of the X25-V, IIRC, reads are ~180MB/s (single drive mode).
As per a RAID 1, you can get a slight performance increase, so long as the OS is multi-threaded and supports split seeks. There's a price though, and that's the writes are a little slower. I presume OS X is capable of this, but not 100% sure.
Also, if the Pro Drive device is used, the PM chip will add latency to the operations, so the slight performance gain may disappear.
Hope this helps, as I'm not sure if you're wanting to use the Pro Drive or not.