SATA Express meets the '09 Mac Pro - Bootable NGFF PCIE SSD

[MacVidCards]

My 1Tb SSUBX card showed UP AT x2 in slot 3 so I moved to slot 4 and it worked at 4 lane speed.

And sometimes when used with our new 970/980 cards they play the "no bootable disc found" gag and get stuck on a black screen. So there is some unknown wonkiness going on. But typically a PRAM reset or changing slot of SSD fixes it.

But I have definitely seen some of same issues as others. 3,1 seems more prone to trouble than 4,1

 

[handheldgames]

My 1Tb SSUBX card showed UP AT x2 in slot 3 so I moved to slot 4 and it worked at 4 lane speed.

And sometimes when used with our new 970/980 cards they play the "no bootable disc found" gag and get stuck on a black screen. So there is some unknown wonkiness going on. But typically a PRAM reset or changing slot of SSD fixes it.

But I have definitely seen some of same issues as others. 3,1 seems more prone to trouble than 4,1

Here's a snapshot from the System Information. The report shows the drives as a G revision, ex. 512G whereas 2013 technology was F revision, 1024F.

 

[MacVidCards]

Here's a snapshot from the System Information. The report shows the drives as a G revision, ex. 512G whereas 2013 technology was F revision, 1024F.

yeah, got a couple of them

 

[flehman]

I realize there aren't many data points to go on right now, but any sense if the SSUBX is a more unstable iteration than the SSUAX iteration when it comes to drive recognition, booting from PCIE, recognition of lanes and full PCIE speeds, etc.? I'm considering one of these, but given the choice between improved reliability and the extra 100-200 r/w speed that SSUBX offers over SSUAX it might be that the SSUAX is preferable.

 

[OCDMacGeek]

SATA Express meets the '09 MacPro - Bootable NGFF PCIE SSD

That is a good question.

The 1 TB blade that I have been posting about above is an SSUBX (MZ-KPV1T00/0A2, same one as Synchro). I have never been able to figure out why once the blade fills up so that there is only about 350 GB left, speeds tank and only test at 100 to 200 MB a second. I appreciated all the advice I got, but nothing yielded any answers or results. So, I now have to leave the blade largely unused to retain my high speeds.

Despite having the exact same model, Synchro doesn't have this issue. However, his blade is in a cMP and mine is in a nMP (directly in the Mac, not in an enclosure).

So, there's really no way to know whether the SSUBX has instabilities. It could be that some of them have issues with over provisioning or garbage collection. But, that is really just speculation. From what I can tell, I'm the only one who has had any of these issues.

 

[handheldgames]

I realize there aren't many data points to go on right now, but any sense if the SSUBX is a more unstable iteration than the SSUAX iteration when it comes to drive recognition, booting from PCIE, recognition of lanes and full PCIE speeds, etc.? I'm considering one of these, but given the choice between improved reliability and the extra 100-200 r/w speed that SSUBX offers over SSUAX it might be that the SSUAX is preferable.

The speed is more than a 100-200 MB/s difference between the SSUAX vs SSUBX revisions. The increase is 46% faster write and 72% faster Reads. I believe you are referencing the difference between PCIe 2.0 and PCIe 3.0.

Is SSUBX viable in the cMP? Absolutely. Does one have to reset the SMC and Flash the PRAM to ensure good operation. Unfortunately, it appears so. Is it worth it? Absolutely. 1500MB/S R/W is hard to pass up for a bit of house cleaning. Is the platform stable after the reset? Yep.

Hopefully this answers your questions. Happy hunting!

----------

That is a good question.

The 1 TB blade that I have been posting about above is an SSUBX (MZ-KPV1T00/0A2, same one as Synchro). I have never been able to figure out why once the blade fills up so that there is only about 350 GB left, speeds tank and only test at 100 to 200 MB a second. I appreciated all the advice I got, but nothing yielded any answers or results. So, I now have to leave the blade largely unused to retain my high speeds.

Despite having the exact same model, Synchro doesn't have this issue. However, his blade is in a cMP and mine is in a nMP (directly in the Mac, not in an enclosure).

So, there's really no way to know whether the SSUBX has instabilities. It could be that some of them have issues with over provisioning or garbage collection. But, that is really just speculation. From what I can tell, I'm the only one who has had any of these issues.

This is new territory for the cMP and nMP that's for sure. My 512G is still empty and hasn't been thrashed with data yet. It will be interesting to see if this occurs on this silicon or not and whether it occurs on the SM951.

After you copied the data, did you give the system ample time to go through necessary trim optimization and for rebuilding spotlight? That can throw any benchmark into the bin.

 

[handheldgames]

Thunderbolt PCIe x4 != PCI ex4

Quick takes from the Lab:

Syntech PCIe adapter - Why is it only $10?
As I expected, First generation Syntech PCIe SSD adapters appear to have the PCIe bracket holes in the wrong locations. When properly assembled and inserted into the cMP, the orientation of the PCIe pins may be slightly skewed - resulting in improper SSD detection and initialization. Installing the Syntech PCIe SSD adapter into an external Akitio adapter on the test-bench, the misalignment is more than evident. This may be why the SSD is having issues in different slots. It's unclear if later revisions of the adapter that include a heat spreader correct this issue. Beyond the mishap with the hole placement, this should be an installation consideration - as the board design and socket quality surpass ANY NGFF/M.2 solution I've seen to date.

Thunderbolt2 is no match for internal PCIe. Period.

• Apple/Samsung PCIe F revision SSUAX SSD's from Late 2013 run at full speed in the Akitio Thunderbolt 2 PCIe expansion.
• G Revision SSUBX SSD's from late 2014 max out around 1275 MB/s on rMBP TB2 vs 1500 MB/s on cMP PCIe. About 15% +/-slower than the cMP @ PCIe2.
• Although the Akitio Thunderbolt2 PCIe Expansion advertises a maximum speed of 1375 MB/s it cannot be achieved with more than capable hardware. DOH!

Attached screenshots:
Images from the test bench
644w/685r: Apple Samsung 256f % 32% free - internal SSD
675w/729r: Apple Samsung 512f, 98% free - Akitio
928w/860r: Apple Samsung 1024f, 98% free - Akitio
1202w/1273r: Apple Samsung 512g 99% free - Akitio

 

[TheStork]

Quick takes from the Lab:

Syntech PCIe adapter - Why is it only $10?
As I expected, First generation Syntech PCIe SSD adapters appear to have the PCIe bracket holes in the wrong locations. When properly assembled and inserted into the cMP, the orientation of the PCIe pins may be slightly skewed - resulting in improper SSD detection and initialization. Installing the Syntech PCIe SSD adapter into an external Akitio adapter on the test-bench, the misalignment is more than evident. This may be why the SSD is having issues in different slots. It's unclear if later revisions of the adapter that include a heat spreader correct this issue...

Best to update the OP in this thread, then. Unfortunately, I bought two of them last week.

 

[flehman]

Anyone have a link or photo of the revised (?) Syntech adapter with the heat spreader?

 

[handheldgames]

Best to update the OP in this thread, then. Unfortunately, I bought two of them last week.

Thanks for the reminder. The Spring 2015 update is past due.

Some more thoughts:

• With or without the backplane, the SSD's behave in the same manner in the OS.
• Reconfiguration of PCIe setup may cause the revision G SSD to be problematic. A 2nd reboot after the move - or - a PRAM reset fixes the issue.
• in my cMP, Revision G/SSUBX initializes to 2.5GT/s in PCIe slot 2 and 5.0GT/s in slot 3/4. Revision F/SSUAX SSD's initialize to 5.0 GT/s in slot 2,3 and 4.

 

[Keter]

I have both versions of the Sintech PCI-e adapters and both work exactly the same for me. I did find the adapter with the heatspreader alignment is a bit better, but the mounts for the blade are not perfect so the blade position is slightly off from parallel with the adapter card.

Neither of these card give me issues when plugged into any of the PCI-e slots and both register at 4 lanes and 5 GT/s. I am on a Mac Pro 5,1 that isn't flashed.

 

[ianj1972]

and don't forget you might have daisy chained a monitor through your thunderbolt/pcie adapter and spend an evening debugging why you can't even hit that! :roll eyes:

 

[DougTheImpaler]

OK dumb question time. Which slot is best for the drive in a flashed cMP 4,1? I have my video card, a USB 3.0 card, and the PCIe adapter to worry about. The video card is in slot 1, the PCIe adapter with the SSD is in slot 2, and my USB 3.0 card is in slot 4. The drive is a APPLE SSD SM0512F and it shows up as x2, but I was thinking it should show as x4. Or is that only the 1TB ones?

Believe me, I have no complaints about performance. BlackMagic shows it at around 720MB/s for both reads and writes. I just want to make sure I put it in the right slot to get best performance, and I'm a little confused. Searching the thread for my particular model yields 0 results.

 

[flehman]

OP has stated that the cards are recognizable at 5.0 speeds (4 lanes) in slots 2, 3, or 4 depending on the pedigree of your SSD. Best advice is trial and error from there - try in slot 2, reset SMC and NVRAM, repeat for slots 3 and 4 and compare speeds to see what works best.

 

[DougTheImpaler]

I'd seen that in the OP, that the models mentioned would show x4, but since my particular model isn't mentioned in the thread I just thought I'd ask. Resetting PRAM/SMC isn't something I thought to do, but I will, and I'll swap it around tonight and see where it's happiest.

What I do think I understand is that slot 3 and 4 share 4 lanes, so I have a feeling that if my USB 3.0 card and the SSD are both in 3/4 that it would definitely be limited.

 

[flehman]

True enough. There are other folks around here who know which slots share lanes on a 2009 and can rattle that off for you, or you can find the service manual for the 2009 which may also state one way or the other. I don't have that problem with my 2008 since I know that anything I plug into slots 3 or 4 are splashing around in the PCIE 1.0 kiddy pool.

 

[IowaLynn]

True enough. There are other folks around here who know which slots share lanes on a 2009 and can rattle that off for you, or you can find the service manual for the 2009 which may also state one way or the other. I don't have that problem with my 2008 since I know that anything I plug into slots 3 or 4 are splashing around in the PCIE 1.0 kiddy pool.

Ff you move the blade adapter to slot #2 in a 3,1, you have PCIe 2.0 16x which will help get maximum thruput than your slot 3/4 that are 4x and PCIe 1.0 (likely 400MB/sec writes and 700MB/sec reads, which is still good deal).

 

[Pressure]

Indeed, currently getting around 730MB/s read and 690MB/s write out of my MZ-JPU256T/A06 with a negotiated link speed of 5 GT/s 2-lanes in slot 2 (Mac Pro 3,1).

Sinntech adaptor works without any issues. Will try slot 1 in the weekend.

 

[Keter]

I am wondering if those with flashed 4.1 MP's have some kind of stability problem. I have a SSUAX and SSUBX in my system and both register 4 lane link and 5 GT/s

The SSUAX is attached to Sintech w/o the heat spreader and the SSUBX is connected to the Sintech w/ the heat spreader.

SSUAX is on PCI-E slot 2 and SSUBX is on slot 3

 

[flehman]

Ff you move the blade adapter to slot #2 in a 3,1, you have PCIe 2.0 16x which will help get maximum thruput than your slot 3/4 that are 4x and PCIe 1.0 (likely 400MB/sec writes and 700MB/sec reads, which is still good deal).

I don't have a PCIE SSD at the moment - just researching and fact-gathering for now. My only option for PCI 2.0 performance is slot 2 since my GPU is in slot 1.

I'm not entirely thrilled about seating an expensive SSD directly above my heat-emitting GPU although the 6870 is pretty cool-running as they go. I guess these drives last pretty well in notebooks that get very hot. Should I be concerned about heat from the GPU affecting a PCIE SSD over time when considering this upgrade, or am I being needlessly nervous? This has not been an issue in the 4 drive bays since they are sort of "above the fray" of the heat-generating CPUs, GPU, RAM, etc.

 

[OCDMacGeek]

I have an OWC Helios 2 and a PCIE adaptor that fits my nMP's original 256 gb blade -- I don't know if it's a Sintech because I got it off EBay from a US based seller who wasn't sure. (I replaced my original 256 gb blade with a 1 TB, which is why I repurposed it). I'm only getting about 700 MB/s read and write in that set up, even though I recall the speeds being a bit higher when it was the primary drive in my nMP. This could be the fault of my adaptor. I need to get a Sintech one to be sure.

Sounds like the card is a 2X, as that's the speed limit for 2X cards. You need a 4X card.

Loa

An update on my SSUAX 256 GB blade in the OWC Helios 2. I had tested it using a PCIE adaptor that might not be a Sintech and it got about 700 MB/s read and write. I finally got my Sintech adaptor (got the one with the heatsink).

After installation in this PCIE adaptor in the Helios 2 (via Thunderbolt, naturally) I am still getting the same 700 MB/s read and write. I guess this is a limitation of the OWC Helios 2, but its fine enough for my purposes, for this particular blade. My main disk is a 1 TB SSUBX that gets 1200 MB/s write and 1360 MB/s read (as long as I leave it no more than 60% full, which isn't ideal, but I'm surviving.)

Regarding the Sintech adaptor with heatsink -- my SSUAX blade was bonded to a metal heatsink of its own (see picture) and the Sintech heatsink would not fit over it.

I didn't want to pry it off the blade, so I took off the PCIE card's heatsink. If anyone wants to try fitting that heatsink onto your original PCIE card (not sure if it has the screws for it) then you're welcome to it.

 

[Keter]

I tested a different configuration with the SSUAX and SSUBX.

If I swap them so the SSUBX is on slot 2 instead of slot 3, I am limited to 4x link speed and 2.5 GT/s, 750 MB/s read/write vs when at 5 GT/s, 1.43 GB/s read/write.

I didn't bother to clear PRAM before turning on my system.

 

[Synchro3]

My 'old' SSUBX (MZ-KPV1T00/0A2) has no issues on slot 2, full 4x link speed at 5 GT/s. Only the new SSUBX SSD's seem to be affected.

My main disk is a 1 TB SSUBX that gets 1200 MB/s write and 1360 MB/s read (as long as I leave it no more than 60% full, which isn't ideal, but I'm surviving.)

However, this is a strange problem. I filled 905 GByte and still had the full performance. Did you try to reformat the SSD?

I am wondering if those with flashed 4.1 MP's have some kind of stability problem.

No stability issues on my flashed Mac Pro 4,1 -> 5,1 with my SSUBX SSD since October.

 

[DougTheImpaler]

No stability issues on my flashed Mac Pro 4,1 -> 5,1 with my SSUBX SSD since October.

Now that I've had mine for a few days, I agree. I have not had any stability issues, and it wakes from overnight sleeps just fine.

 

[handheldgames]

I don't have a PCIE SSD at the moment - just researching and fact-gathering for now. My only option for PCI 2.0 performance is slot 2 since my GPU is in slot 1.

I'm not entirely thrilled about seating an expensive SSD directly above my heat-emitting GPU although the 6870 is pretty cool-running as they go. I guess these drives last pretty well in notebooks that get very hot. Should I be concerned about heat from the GPU affecting a PCIE SSD over time when considering this upgrade, or am I being needlessly nervous? This has not been an issue in the 4 drive bays since they are sort of "above the fray" of the heat-generating CPUs, GPU, RAM, etc.

Good question. Placing a SSD next to HDD's is no issue and PCIe Slot 3 & 4 work without issues in all reports. Regarding PCIe slot #2, the cMP video card should not present any issues relative to heat and only has reported issues with:

• Earlier 128GB Samsung XP941's disconnect on deep sleep. Yes 128GB, it was the only version available for purchase in December of 2013.
• Apple/Samsung SSUBX PCIe SSD's may connect at 2.5 GT/s vs 5.0 GT/s, limiting speed to the mid 800MB/s.