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

[h9826790]

Is this "dynamic"? If cards are installed in slots 3 & 4, will they only experience that bandwidth penalty if they are both actively doing something simultaneously?

Put another way, if the card in slot 4 is "idling", does the card in slot 3 effectively have a mode of operation indistinguishable from a normal x4 slot? And vice versa.

Yes, even though I don't know if there is any card can be really truly idle (cost zero bandwidth) when installed in slot 4. I think the penalty is virtually not exist when that card is idle.

That's why I said the card in slot 3 "may" run slower, but not "will" run slower.
[doublepost=1468883034][/doublepost]

I just unRARed two 45GB archives simultaneously. I did it first on my Apple SSUBX with M.2 adaptor then with my Crucial M500 on SATA 3.

Apple SSUBX: 6:39
Crucial M500: 12:10

My old Mac Pro 5,1 is packaged up to be sold so I was unable to run any tests on SATA 2, but believe me when I tell you it's a very noticeable difference. The more archives you try to unRAR simultaneously, the more you notice the difference. This is real world usage. Anyone telling you it's just placebo, doesn't know what he's talking about.

Thanks for the test. For some reason, I miss this post, sorry for the late reply.

This make me really tempted to upgrade to PCIe SSD, however, I have 1st to figure out how to downgrade my dual GPU to a single card, and if it's worth to do that. Otherwise, no slot avail

 

[ActionableMango]

I asked long ago, but I didn't get a clear answer so I thought I would ask again.

Is anyone successfully using this setup for a bootable Windows installation? If yes, did you need to do anything special to get it working?

 

[handheldgames]

Hello,

Is there a need for a SUMMER 2106 update on the first post? Or is that info still valid (price and availability)?

Thanks

thanks for the poke. task has been added to my todo list.

 

[Squuiid]

Is this "dynamic"? If cards are installed in slots 3 & 4, will they only experience that bandwidth penalty if they are both actively doing something simultaneously?

Put another way, if the card in slot 4 is "idling", does the card in slot 3 effectively have a mode of operation indistinguishable from a normal x4 slot? And vice versa.

Yes, even though I don't know if there is any card can be really truly idle (cost zero bandwidth) when installed in slot 4. I think the penalty is virtually not exist when that card is idle.

That's why I said the card in slot 3 "may" run slower, but not "will" run slower.

Unfortunately I found the opposite to be the case.
Both the XP941 with UXM6401Q firmware and the SM951 MUST be used in Slot 3 or 4 to prevent sleep issues and run at full speed.

I tried both SSD cards in Slot 3 with a 10GbE card in Slot 4 and both SSDs throttle down to less than half their max speed when the 10GbE card is sitting idle in Slot 4. Even if I disable the card in OS X's Network settings the SSDs will not run at full speed. As soon as I physically remove the 10GbE card the SSD performs as expected.
Given this I have found that I have to leave Slot 4 EMPTY in order for the system to perform as it should.

Unfortunately both these SSDs can ONLY be placed in Slot 3 or 4 and, in my experience at least, are not happy sharing bandwidth with the other slot.

The only ways around this would be to get an XP941 with firmware UXM6501Q (which do not exhibit sleep problems and run at full speed in Slot 2), or disable sleep which I'm not prepared to do.

 

[h9826790]

Unfortunately I found the opposite to be the case.
Both the XP941 with UXM6401Q firmware and the SM951 MUST be used in Slot 3 or 4 to prevent sleep issues and run at full speed.

I tried both SSD cards in Slot 3 with a 10GbE card in Slot 4 and both SSDs throttle down to less than half their max speed when the 10GbE card is sitting idle in Slot 4. Even if I disable the card in OS X's Network settings the SSDs will not run at full speed. As soon as I physically remove the 10GbE card the SSD performs as expected.
Given this I have found that I have to leave Slot 4 EMPTY in order for the system to perform as it should.

Unfortunately both these SSDs can ONLY be placed in Slot 3 or 4 and, in my experience at least, are not happy sharing bandwidth with the other slot.

The only ways around this would be to get an XP941 with firmware UXM6501Q (which do not exhibit sleep problems and run at full speed in Slot 2), or disable sleep which I'm not prepared to do.

Yes, that's a well known bug, I didn't account for that when I said the card should run faster in slot 2.

 

[lozion]

Anyone selling a card and or sm951? Txs.

 

[flowrider]

^^^^I've got a Lycom DT-120 that I'm not using.

Lou

 

[lozion]

Ok pm me the price?

 

[flowrider]

^^^^Just noticed your in Canada. The postage would make it not worthwhile for the both of us

Lou

 

[lozion]

Its pretty light and I'd cover the shipping so..

 

[fhturner]

I bought my x4 Amfeltec because CalDigit's new USB 3.1/6G eSATA card is sitting in slot #2 and for PCIe SSD card I was more interested in reliability rather than speed. I know, however, there is an untapped hyper speed x16 territory out there and I might change my mind in the future.

I would swap those cards, if I were you. There's really no reason to bottleneck your four x4 SSDs into a single x4 slot**. The CalDigit card has no need for the x16 slot and for *most* tasks and scenarios, likely won't make use of much more than a single lane (~500MB/s) anyway. So, an x4 Slot 3 or 4— even if shared w/ another card— will probably not constrict the CalDigit card in any way...

** = Strike all that. At first I thought you were saying that the Amfeltec card was "x4" because it was in an x4 slot. Upon re-reading, I now understand that the Amfeltec card itself is x4, and not the x16 version. I'll leave my answer intact for those w/ the x16 card contemplating what to put where, since my remarks on the CalDigit card still apply...

Fred

 

[flowrider]

Its pretty light and I'd cover the shipping so..

$15 bux - US.

Lou

 

[MacinJosh]

I'm looking to get about 1TB of fast SSD storage. What would be the most cost-effective solution? I don't need the fastest. Would two 512GB Samsung 850 Evos M.2 in two separate NGFF adapters in Raid 0 do the trick? Locally sourced it would be 414 euros for me. Cheaper if I ordered the components elsewhere. That would give me max 1GB/s? Is this recommended? 1TB of SSUBX seems to be way more expensive in eBay. I don't care for the performance. As long as it's better than SATA III.

 

[lozion]

$15 bux - US.

Lou

Ok I'll pm you

 

[patdohere]

This might be of interest to some of you. I just received my Angelbird Wings X2 PCIe card. It supports several RAID modes, speeds up to 950 MB/s and is bootable. The MX2 card is also available for ordering.

View attachment 625648 View attachment 625651 View attachment 625652 View attachment 625653

Hi,

O.N.Y.X I was looking into buying one of these for my Mac Pro 5,1. I was wondering does option boot and everything work fine for you? I know that option boot is a common problem with the Tempo Pro SSD which is known in the FAQ for the 5,1 model.

http://www.sonnettech.com/support/kb/kb.php?cat=457&expand=_a2&action=b760#b760

 

[h9826790]

Hi,

O.N.Y.X I was looking into buying one of these for my Mac Pro 5,1. I was wondering does option boot and everything work fine for you? I know that option boot is a common problem with the Tempo Pro SSD which is known in the FAQ for the 5,1 model.

http://www.sonnettech.com/support/kb/kb.php?cat=457&expand=_a2&action=b760#b760

Option boot breaks 99%, but not 100%. There is another forum member and I able to get the option boot with 5,1 firmware.

He teach me that install the bare card, and do the SMC / PRAM reset, then install back the SSD (with OS) may able to get it back. Which may take quite a few try, but once success, the option boot will stays there until the next reset.

I tried that, it actually works.

However, I later accidentally get a permanently fix. My 4,1 (flashed 5,1) died, I sent it to Apple, they did a lots of test, reset, etc. End up the CPU tray get replaced. I don't know how it works, but with the new CPU tray, even the firmware still 5,1 (which is stored on the logic board), I can get the boot manual no matter what I do now.

 

[patdohere]

Option boot breaks 99%, but not 100%. There is another forum member and I able to get the option boot with 5,1 firmware.

He teach me that install the bare card, and do the SMC / PRAM reset, then install back the SSD (with OS) may able to get it back. Which may take quite a few try, but once success, the option boot will stays there until the next reset.

I tried that, it actually works.

However, I later accidentally get a permanently fix. My 4,1 (flashed 5,1) died, I sent it to Apple, they did a lots of test, reset, etc. End up the CPU tray get replaced. I don't know how it works, but with the new CPU tray, even the firmware still 5,1 (which is stored on the logic board), I can get the boot manual no matter what I do now.

Whats your system configuration is this with the SonnetTech SSD Tempo Pro or Angelbird Wings X2? I have a native Mac Pro 5,1. Also another off topic question. What speeds do you get with RAM Disk. With Blackmagicdesign Disk Speed Test I'm peaking at around 2037 MB/s and 2423 MB/s. Which makes me think that upgrading to the Amfeltec Squid PCIe 4x M.2 Storage option I would not be able to fully utilize the speed.

http://barefeats.com/hard210.html

 

[h9826790]

Whats your system configuration is this with the SonnetTech SSD Tempo Pro or Angelbird Wings X2? I have a native Mac Pro 5,1. Also another off topic question. What speeds do you get with RAM Disk. With Blackmagicdesign Disk Speed Test I'm peaking at around 2037 MB/s and 2423 MB/s. Which makes me think that upgrading to the Amfeltec Squid PCIe 4x M.2 Storage option I would not be able to fully utilize the speed.

http://barefeats.com/hard210.html

Tempo SSD (non pro version, but the same firmware as the pro, and therefore same boot manager problem).

RAM disk speed

In real world, cache avail (benchmark intentionally disable cache, otherwise, it's benchmarking the system, but not a particular hardware). Therefore, you can actually benefit from the Amfeltec PCIe 4x SSD card if you use it correctly.

 

[patdohere]

Thanks h9826790. I will go and try a SMC PRAM reset to see if it resolves the problem. But looking into the Amfeltec as a upgrade path down the road.

 

[Squuiid]

Unfortunately I found the opposite to be the case.
Both the XP941 with UXM6401Q firmware and the SM951 MUST be used in Slot 3 or 4 to prevent sleep issues and run at full speed.

I tried both SSD cards in Slot 3 with a 10GbE card in Slot 4 and both SSDs throttle down to less than half their max speed when the 10GbE card is sitting idle in Slot 4. Even if I disable the card in OS X's Network settings the SSDs will not run at full speed. As soon as I physically remove the 10GbE card the SSD performs as expected.
Given this I have found that I have to leave Slot 4 EMPTY in order for the system to perform as it should.

Unfortunately both these SSDs can ONLY be placed in Slot 3 or 4 and, in my experience at least, are not happy sharing bandwidth with the other slot.

The only ways around this would be to get an XP941 with firmware UXM6501Q (which do not exhibit sleep problems and run at full speed in Slot 2), or disable sleep which I'm not prepared to do.

OK, after much trial and error and swapping around of cards I'd like to retract this statement!
XP941 SSD in slot 3 can happily 'share' with slot 4, with very little, if any impact on performance.

Turns out the SM951 just didn't play nice with the Intel X520 10GbE cards.
I now have my cards in the following arrangement and all works perfectly, at full performance, even when transferring to the SSD in slot 3 via the 10GbE card in slot 4. Very impressed.

Slot 1: EVGA GTX 680ME or NVIDIA GTX 1080FE
Slot 2: Sonnet Allegro Pro USB 3.0
Slot 3: Samsung XP941 512GB SSD AHCI
Slot 4: Intel X520-DA2 10GbE

I've tested both in Windows 10 in UEFI mode and macOS and both systems and cards perform to their full capacity. AJA shows 900MB/s write, 1100MB/s read and 10GbE network transfers hit 1GB/s when pulling from the SM951 which I've now moved to my VMware server.
I also tested latency in Windows 10 using LatencyMon and found it to be very good (low).

 

[buster84]

I'm having a Very strange problem with two of my mac pro 5,1 computers using sm951's. Both have a 256gb sm951 with a lycom adapter. When i first got them everything was good, they both maxed out in the 1500mbs range now when i do a speed test im getting 200mb write and 400mb read and its fluctuating alot.

At first I thought maybe the card was bad so i switched them, nothing changed. So then i swapped ports, again nothing changed. After that i moved the card to my other mac pro and again nothing changed. So then i thought maybe my card went bad. Well then i ran a test on my second sm951 in my other mac pro and it also has very bad read/write that acts the same way as the other one. Either i have two defective sm951s or two defective lycom card, or two defective mac pro motherboards.

These sm951s are have a 3 year warranty so if there bad ill have to figure out how to redeem them, but what i dont understand is that one of my mac pro's had the new drive installed and only ran for a month (then put in storage) and when i decided to use it again the hard drive is showing up real slow in black magic.

Has anyone else noticed this problem? Is there any other reliable hard drive test outside of black magic? The fluctuation is pretty high range, it gets fast and slow but my write never goes above 600mb when letting it test over and over again.

I've included some photos. These were in slot 3 and the same drive. I would post photos of the second drive on the other mac but it acts exactly the same way. I'm wondering if the lycom cards are bad or something else is going on.

 

[h9826790]

I'm having a Very strange problem with two of my mac pro 5,1 computers using sm951's. Both have a 256gb sm951 with a lycom adapter. When i first got them everything was good, they both maxed out in the 1500mbs range now when i do a speed test im getting 200mb write and 400mb read and its fluctuating alot.

At first I thought maybe the card was bad so i switched them, nothing changed. So then i swapped ports, again nothing changed. After that i moved the card to my other mac pro and again nothing changed. So then i thought maybe my card went bad. Well then i ran a test on my second sm951 in my other mac pro and it also has very bad read/write that acts the same way as the other one. Either i have two defective sm951s or two defective lycom card, or two defective mac pro motherboards.

These sm951s are have a 3 year warranty so if there bad ill have to figure out how to redeem them, but what i dont understand is that one of my mac pro's had the new drive installed and only ran for a month (then put in storage) and when i decided to use it again the hard drive is showing up real slow in black magic.

Has anyone else noticed this problem? Is there any other reliable hard drive test outside of black magic? The fluctuation is pretty high range, it gets fast and slow but my write never goes above 600mb when letting it test over and over again.

I've included some photos. These were in slot 3 and the same drive. I would post photos of the second drive on the other mac but it acts exactly the same way. I'm wondering if the lycom cards are bad or something else is going on.

Did you pick the 5GB size for the rest?

 

[buster84]

Did you pick the 5GB size for the rest?

I did the 5gb test I also just got finished running some more tests. I deleted one of the drives then formatted it as ntfs for windows and ran a test with disk mark 1gb and 32gb test. I then reformated that drive as OSX journaled and now im getting full speed again. So im lost as to what the problem is outside of the disc slowing down when osx is installed on the hard drive. I also find it strange since I actually did a fresh install on the drive before i erased it to test it on windows. I'll go ahead and re-add osx and retest.

Based on the new tests its not a hardware problem.

 

[buster84]

I did the 5gb test I also just got finished running some more tests. I deleted one of the drives then formatted it as ntfs for windows and ran a test with disk mark 1gb and 32gb test. I then reformated that drive as OSX journaled and now im getting full speed again. So im lost as to what the problem is outside of the disc slowing down when osx is installed on the hard drive. I also find it strange since I actually did a fresh install on the drive before i erased it to test it on windows. I'll go ahead and re-add osx and retest.

Based on the new tests its not a hardware problem.

Well i reformatted it again and its now working great. Full speed and everything. Im going to reformat my other desktop as well, i guess they slowdown overtime for some reason. Still a little confused as to why it was slow on a fresh install earlier. Maybe formatting it as ntfs fixed it. I guess ill find out when i re-do the other one.

 

[AidenShaw]

Well i reformatted it again and its now working great. Full speed and everything. Im going to reformat my other desktop as well, i guess they slowdown overtime for some reason. Still a little confused as to why it was slow on a fresh install earlier. Maybe formatting it as ntfs fixed it. I guess ill find out when i re-do the other one.

Did you boot Windows - since Windows supports TRIM on any drive that self-identifies as supporting TRIM?