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.
LaCie Thunderbolt Little BigDisk Disassembled - Pics SSD XFer Rates Added
- Thread starter ender21
- Start date
- Sort by reaction score
There is a crazy amount of silicon on that PCB.
I count some 10 micro-controllers on that board. I do wish we could make out the markings on the chips. Any chance even a camera phone close up of them would be possible, ender21?
Also, I couldn't quite make out the configuration of the daughter-card to which the drives connect. It almost looks like it's using a port multiplier configuration rather than two dedicated SATA controller connections. That might explain the throughput bottleneck... 6.0 Gbps / 2 = 3.0 Gbps, or 300 MB/s without the 8b/10b encoding overhead.
I count some 10 micro-controllers on that board. I do wish we could make out the markings on the chips. Any chance even a camera phone close up of them would be possible, ender21?
Also, I couldn't quite make out the configuration of the daughter-card to which the drives connect. It almost looks like it's using a port multiplier configuration rather than two dedicated SATA controller connections. That might explain the throughput bottleneck... 6.0 Gbps / 2 = 3.0 Gbps, or 300 MB/s without the 8b/10b encoding overhead.
Since I still don't have my SLR back from Canon, I can't take any better macro shots. The pics as they are were taken from the iphone 4 and if I got any closer it wouldn't focus.
I'll take new pics of the daughter-card because I think your assessment may be spot-on.
No, it's from Marvell and the traces go straight to the SATA connectors, so I'm guessing it's the SATA controller, I just can't tell what flavor.
I bit the bullet yesterday and cancelled my LBD order and ordered a Pegasus R4 instead. It should be here tomorrow, and my two 128GB M4 SSDs by Friday. I'll be sure to post some benchmarks for comparison. My plan is to have one RAID 0 set with the two SSDs, and then either a RAID1E or 0 set with two of the included HDDs.
Ah ha! I forgot I have an old Rebel SLR (autofocus broken, but manual works), so I threw my macro rig on that and voila.
Agreed. Though I would have expected whatever chipset they're using on the HDD version to be the same for the SSD version to cut down on costs.
But now that I think about it further, things are starting to point to at least the possibility the SSD version will be different.
It makes sense that the Sata interface on-board is 1 channel of 6Gbps, because the daughter-card is splitting the bandwidth, and because the HDDs they're using are only 3Gbps anyway.
It'll be interesting to see a teardown of the SSD version when it comes out, though I doubt I'll be plunking down the money for it!
Nice!
So it looks like a Marvell 88SE9182, SATA 6 Gbit/s controller.
You also got a good shot of the Parade PS8321 DisplayPort 2:1 multiplexer, which starts to answer my question about all the silicon on this board. I guess Thunderbolt licensing brings with it all types of extra work at this point for OEM's designing a two port device.
It kinda sucks that it appears the backplane is limiting this device, and that LaCie is forcing people to buy the SSD version to get two full 6 Gbit/s SATA ports, when the controller is more than capable of providing them.
Last edited:
That may very well be the case. It'll be a little bulkier than it needed to be, but it sure would cost less to only put one really fast SSD in it.
It would be great to have a bus-powered, single 2.5-inch drive Thunderbolt storage solution, but if you really need all those freakin' chips on there to make a 2-port device, I'm not sure if 10W will be enough.
Many thanks for your photos!!!
The Marvell 88se9182 is a dual port SATA 6Gbps controller. The daughter card doesn't split the bandwidth because is only a passive card. There are 4 couple of balanced traces (8 signals) coming out from the controller. Each sata channel has 4 signals (tx+,tx-,rx+,rx-), so both channels are routed to the daughter card.
LaCie uses 3 equal connectors: 2 for drives and 1 to connect the daughter card to the main board. The daughter card's bottom is also used for the box front panel's led and switch.
It's also important that there is a Winbond W25X10BV serial flash chip with the firmware close to the marvell controller.
At this point the question is: why are you obtaining such low transfer rate with SSDs? Is it a locked firmware problem?
I have just ordered a Pegasus drive also. I want this drive to be future-proof and upgradable. It really think it is essential that any thunderbolt drive uses the 6Gb/s standard, don't think I could justify the price otherwise.
However, I became a little concerned when reading the user manual for the Pegasus that says it has 'SATA specification of 3 Gb/s transfers with CRC error-checking', but elsewhere I have read 6 Gb/s?
We really shouldn't have to do so much extra digging to figure out basic specs of these drives. Hopefully Lacie haven't imposed firmware limitations on this drive.
Thanks for all your work here ender21!
Interesting. Well, we know in general the Pegasus can be much faster than the LaCie, since Anandtech got 1GB/s using 4 SSDs. I'll be happy if I can get 600-700MB/s with my 2 SSDs, which should theoretically be possible even if each drive only gets 3Gbps.
Hah! I saw those traces, but stopped paying attention to them when 2 pair went to the 15 pin SATA power connector. LaCie used 4 pins normally reserved for the 3.3V power rail to route the data lines to the second drive connector. And people bitched when Apple did some unusual appropriation of SATA power conductors...
Actually it looks more like a W25X40BVNIG, higher capacity, but same difference.
ender21, you've been great with all the photos thus far, but if you get a chance, I'm very curious about the 4 Parade and 4 NXP chips which seem to be all DisplayPort related. I can make out the 2 x NXP LPC1114F's, the 1 Parade PS8321, and an NXP marked with what looks to be ZSD114 although I have no clue what it does. I couldn't quite decipher the other NXP or the 3 remaining Parade chips. Any chance of getting another Rebel shot of the rest of the PCB?
You are right. Moreover LaCie uses the same connector to route traces also for the front blue orb (led+switch)
My guess for the future is that LaCie will change only the color of paint dot on the flash chips (now it's yellow) on the future models
You are right. Moreover LaCie uses the same connector to route traces also for the front blue orb (led+switch)
Yeah, on closer inspection, there ain't nothin' normal about what they've done with that power connector. It sort of begs the question of why they didn't use a card edge connector for the backplane. Can you imagine if you just plugged a single drive into that board directly using normal SATA cables?
I think that might be the hard way to go about it. The Marvell controller should just show up as a normally connected PCIe device, so I would think you could just boot into Windows or Ubuntu or something and make an image of or flash the firmware from there. Which also makes me realize that while I had ender21 running around taking photos to figure out which SATA controller was being used, he probably could have just found out by looking in System Profiler.
As far as I'm aware though, the 88SE9182 doesn't have any available firmware updates or PCIe lane limitations like the 88SE9128 does. The only thing people have complained about with this chip are Marvell's Windows drivers. Which also points out that this drive didn't ship with any drivers, thus it must be going off of whatever is built into Mac OS X. Since I doubt Apple has ever used one in a Mac, I wonder how good Lion's support of the Marvell 88SE9182 SATA 6 Gbit/s controller is?
Thank you so much!
So let's see, in order to get DisplayPort out on either of the two Thunderbolt ports, since LaCie can't predict which one will be used, it looks like we have:
An NXP 6D212 (I'm guessing this is a DisplayPort splitter.)
A Parade PS161HDM DisplayPort to HDMI/DVI Converter.
A Parade PS8321 Dual Mode DisplayPort Source 2:1 Multiplexer (to rejoin the DP and HDMI/DVI signals.)
Another NXP 6D212 (I reckon to split the signal once again.)
2 Parade PS8301's which I believe are DisplayPort re-drivers (we also see these just before the Thunderbolt ports on most new Macs.)
Then 2 NXP LPC1114F's (I reckon that the bi-directional DP AUX channel can't survive the split/MUX/split cycle, and may be wrapped in PCIe for transport by Thunderbolt anyway, so these are for adding that back in.)
And finally 2 PI3VeDP's (I'm guessing these are eDP drivers which combine the AUX signals from the LPC1114's with the output from the PS8301's. There is also one of these on the MBA's logic board right next to a Parade PS8301.)
10 chips just to light up a DP display should you choose to connect one... Until someone comes out with a SoC to deal with all this, Thunderbolt is clearly going to command much more of a premium for 2 port devices than just the cost of the TB controller.
Edit: This is clearly ancient history at this point, but I was linking to this thread and took another look at those pictures and came up with a more detailed BOM.
IC1, IC4: TI SN74AHC1G14 Single Schmitt-trigger inverter gate
IC2: Intel CV82524EFL 4-channel Light Ridge Thunderbolt controller
IC3, IC8: NXP CBTL06DP212 High performance DisplayPort 1.2 2:1 multiplexer
IC5: Parade PS8321 Dual-Mode DisplayPort source 2:1 multiplexer
IC6: Parade PS181 DisplayPort to Dual-Mode DisplayPort converter; HDMI 1.4b compatible with 300 MHz TMDS clock support
IC7: Hosonic HCX-5SB 27.000 MHz SMD crystal unit
IC9, IC10: Parade PS8301 Dual-Mode DisplayPort repeater/redriver with I2C control
IC11, IC12: Pericom PI3VeDP212 2-lane DisplayPort switch/mux for DP driven panels with triple control pins
IC13, IC14: NXP LPC1114FHN33/301 32-bit ARM Cortex-M0 microcontroller with 32 kB flash and 8 kB SRAM
IC15: Atmel AT25080B SPI Serial EEPROM 8 kb (1024x8)
IC23: Winbond W25X40BV 4 Mb serial flash memory with 4 kb sectors and dual I/O SPI
IC30: Marvell 88SE9182 2-lane PCIe 2.0 to 2-port SATA 6Gb/s storage controller
Well, that's good. At least there was no magic smoke released from any components. Most SATA power connectors gang quite a few individual pins together. Using them individually as LaCie did could lead to unexpected results if they didn't protect the circuits properly.
Last edited:
You are right.
A software tool to program and/or configure the 88se9182 controller is absolutely needed. When it's used on a pc motherboard the job is simple because there are bios extensions and you can do that during boot. You can even configure the 88SE9182 for hardware raid which is much better than Little Big Disk software raid...
There is no need for a software tool for windows because PCs rely on bios setup... on a Macintosh (efi) is another thing...
LaCie should develop this software as Promise has done for it's Pegasus drive.
I don't know if there is a tool for Linux and if could be usable on a virtual machine (parallels desktop or vmware...)
repoman27,
What do you think?
Well I didn't use power off the LaCie. I don't have the proper cable for that. I ran power from an old PC I happened to have open at the time, then used a simple internal sata cable.
Just unpacked my Pegasus R4 (4TB), switched the array to RAID0, and ran a quick test:
Sequential 313.37
Uncached Write 1367.51 839.63 MB/sec [4K blocks]
Uncached Write 741.08 419.30 MB/sec [256K blocks]
Uncached Read 107.11 31.35 MB/sec [4K blocks]
Uncached Read 741.94 372.89 MB/sec [256K blocks]
Random 1589.53
Uncached Write 1033.49 109.41 MB/sec [4K blocks]
Uncached Write 1140.98 365.27 MB/sec [256K blocks]
Uncached Read 4296.95 30.45 MB/sec [4K blocks]
Uncached Read 2274.27 422.01 MB/sec [256K blocks]
Not too shabby. SSDs will be here tomorrow. I'm also pleased with how relatively quiet it is for having 4 drives and 2 fans. It's definitely humming but not unpleasantly so.
Sequential 313.37
Uncached Write 1367.51 839.63 MB/sec [4K blocks]
Uncached Write 741.08 419.30 MB/sec [256K blocks]
Uncached Read 107.11 31.35 MB/sec [4K blocks]
Uncached Read 741.94 372.89 MB/sec [256K blocks]
Random 1589.53
Uncached Write 1033.49 109.41 MB/sec [4K blocks]
Uncached Write 1140.98 365.27 MB/sec [256K blocks]
Uncached Read 4296.95 30.45 MB/sec [4K blocks]
Uncached Read 2274.27 422.01 MB/sec [256K blocks]
Not too shabby. SSDs will be here tomorrow. I'm also pleased with how relatively quiet it is for having 4 drives and 2 fans. It's definitely humming but not unpleasantly so.