Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
*Pegasus2 R8 32TB model is populated with 5900 RPM SATA drives. All other Pegasus2 models are populated with 7200RPM SATA drives.

Pretty slow on the 4TB drives but that's what you would expect on the type or rather size of that drive. I'd imagine when 6TB drives come out it will be no better. Which is why having 8 of them should make up (if in RAID 0) for its slow rpm with raw data thru put. Should be fast enough to handle what it says it does.

I have 4 x 5,900 RPM drives from Seagate, and they are actually essentially as fast as 7,200 TPM drives of lower capacities (for sequential transfers -- and let's face it, if you're doing a lot of random IO, you should be using an SSD), due to the higher platter density.

5,900 RPM drives shouldn't be a problem. If speed is that important to you, you should be working with your files on a scratch SSD, or better yet, RAM disk.

For sequential transfers, with a RAID10 setup on the R8, you shouldn't have any problems copying files over to the scratchdisk. With 32TB in RAID10, you're looking at 1.16 GB/s real world (my drives get 145 MB/s on average, quite consistently). At that rate, even completely filling a 1TB scratch drive would only take ~16 minutes.

5,900 RPM isn't a problem here.
 
Someone please correct my ignorance: what's the point of Thunderbolt 2 being used with hard drives if the drives have read/write speeds significantly slower than the throughput capable of T2?

Some drive configurations could saturate it.

Passthrough.
 
RAID 0 on the r8 would probably get you about 1200 MB/s on paper. Plus 4K passthrough for displays.
Yes, but then again, Thunderbolt 10Gbps (1250Mbps theoretical throughput, minus protocol overhead) would probably be sufficient for that job. But 20Gbps (2500Mbps) is just stupid with these configurations.
 
You could quadruple the speed of the fastest HDDs and they'd not only be beaten by a single 6Gb/s SSD but would, at most, only saturate a single SATA 6Gb/s channel worth of bandwidth because they only hit over 100Mb/s using sustained transfers and as these systems will likely be used in RAID 5, RAID 50 or some proprietary hybrid RAID configuration, they'll be even less impressive.

If these were half the price, they might be worth it but like Thunderbolt RAID systems, they're a waste of money compared with their USB 3.0 equivalent and for some uses, their Gigiabit equivalent too.

The fastest hard drives available sustain around 240 MB/s.
 
You could quadruple the speed of the fastest HDDs and they'd not only be beaten by a single 6Gb/s SSD but would, at most, only saturate a single SATA 6Gb/s channel worth of bandwidth because they only hit over 100Mb/s using sustained transfers and as these systems will likely be used in RAID 5, RAID 50 or some proprietary hybrid RAID configuration, they'll be even less impressive.

If these were half the price, they might be worth it but like Thunderbolt RAID systems, they're a waste of money compared with their USB 3.0 equivalent and for some uses, their Gigiabit equivalent too.

The R8 won't max out TB2 with HDDs, but it does max out TB1 (and USB 3.0 by a long shot). In a RAID10 setup, the read speeds scales linearly with the number of drives in the system. Thus, with 8 drives, you can reach real world speeds around 1.2 GB/s, which would saturate both USB 3.0 and TB1 (just barely there.)
 
Too Noisy

I'd love to see a smaller, fan-less version that houses a bunch of SSDs.
 
Yes, but then again, Thunderbolt 10Gbps (1250Mbps theoretical throughput, minus protocol overhead) would probably be sufficient for that job. But 20Gbps (2500Mbps) is just stupid with these configurations.

I don't think so.

1. If you are daisy chaining, it all has to go through the same channel, so you'll be glad of that extra bandwidth.

2. 1250MB/s is theoretical, whereas 1.2 GB/s is actually quite possible in real world scenarios, so I'd bet that TB1 would actually slow you down a bit.
 
1. If you are daisy chaining, it all has to go through the same channel, so you'll be glad of that extra bandwidth.
Yes, that may be some corner case where this would be needed.

2. 1250MB/s is theoretical, whereas 1.2 GB/s is actually quite possible in real world scenarios, so I'd bet that TB1 would actually slow you down a bit.
Like I said, minus protocol overhead. You just wrote it differently. I have not tested TB1 to it´s limits, so I´m not sure how efficient it actually is.

The 15k seagates are rated at 258.
Good god, can you be more cryptic, please? :rolleyes:

What drive are you speaking of, because the Cheetah drives I found are rated at up to 204MB/s sustained.
 
Cheers Promise for releasing a diskless version! Will be saturating Thunderbolt 2 with 4 SSD's in no time.
 
For those who do not know thunderbolt 2 is 20 Gb/s (gigabit per second not GB/s or gigabyte per second).

1 byte is 8 bits, so 20 Gb is 2.5 GB.
 
Yes, but then again, Thunderbolt 10Gbps (1250Mbps theoretical throughput, minus protocol overhead) would probably be sufficient for that job. But 20Gbps (2500Mbps) is just stupid with these configurations.

Thunderbolt 2 with 8b/10b encoding overhead is 2000MB/s
 
Like I said, minus protocol overhead. You just wrote it differently. I have not tested TB1 to it´s limits, so I´m not sure how efficient it actually is.

The fastest TB1 speed I've read about was 900 MB/sec. Not even close to 1.2 GB/sec.
 
why not all diskless?

I would like to buy the 8HDD version w/o HDD´s.
Don´t need all that storage but the speed of 8 hdd in raid0 would be nice.
 
Yes, that may be some corner case where this would be needed.

Is it that hard to envision such a scenario? MBP hooked up to an external monitor with a storage array?

Not all "work" computers need to be desktop rendering machines.

Like I said, minus protocol overhead. You just wrote it differently. I have not tested TB1 to it´s limits, so I´m not sure how efficient it actually is.

All I'm saying is you're comparing a theoretical number to a real-world number, so TB1 might not actually be able to handle what you said it could.
 
Thunderbolt 2 with 8b/10b encoding overhead is 2000MB/s
So it´s about 20% overhead. That´s quite a lot. Didn´t expect that.

All I'm saying is you're comparing a theoretical number to a real-world number, so TB1 might not actually be able to handle what you said it could.
That´s why I said "minus protocol overhead". Why would I write that if I didn´t mean theoretical vs. real-world? :confused:
 
Pretty slow on the 4TB drives but that's what you would expect on the type or rather size of that drive.

They gain in bit density what they lose in rotational speed. In an 8 disk RAID, particularly if it's a RAID0, RAID5 or RAID15, you'll saturate the pipe rather quickly with that many disks.
 
So it´s about 20% overhead. That´s quite a lot. Didn´t expect that.


That´s why I said "minus protocol overhead". Why would I write that if I didn´t mean theoretical vs. real-world? :confused:

You said that "TB1 would probably be sufficient for that job."

I was pointing out that due to the theoretical-realworld mismatch, it would probably NOT be sufficient.
 
I have 4 x 5,900 RPM drives from Seagate, and they are actually essentially as fast as 7,200 TPM drives of lower capacities (for sequential transfers -- and let's face it, if you're doing a lot of random IO, you should be using an SSD), due to the higher platter density.

5,900 RPM drives shouldn't be a problem. If speed is that important to you, you should be working with your files on a scratch SSD, or better yet, RAM disk.

For sequential transfers, with a RAID10 setup on the R8, you shouldn't have any problems copying files over to the scratchdisk. With 32TB in RAID10, you're looking at 1.16 GB/s real world (my drives get 145 MB/s on average, quite consistently). At that rate, even completely filling a 1TB scratch drive would only take ~16 minutes.

5,900 RPM isn't a problem here.

Then there must be some more cache on these drives (64mb+?). In either event I've only noticed (at least easily noticed) a jump from 5400/5900 to 10k+.
I'm sure it can do it or at the least merits TB2 over TB1. Not to mention its made to support the new Mac Pro. So you will need the TB2 anyway (in case you wanted to daisy chain a few of these things).

I also remember hearing Phill talk about teaming these TB2 connections... :D
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.