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.
Amfeltec pcie expansion what do you guys think?
- Thread starter William Payne
- Start date
- Sort by reaction score
You're welcome.
For my needs i consider a 6,1 nowadays, but still wait until they get a bit cheaper:
- TB<->PCIe/eGPU cases are cheaper now
- TB2 is good for some time, offering up to x4 2.0
- Somewhat efficient chaining capability, negating PCIe switch needs
- TB2<->3 adapters are one time cost not much higher than eg. active DP converters/dualhead2GO
- 6 native DP outputs and ok GPUs (can add computing GPU capacity via TB anyway)
For my needs i consider a 6,1 nowadays, but still wait until they get a bit cheaper:
- TB<->PCIe/eGPU cases are cheaper now
- TB2 is good for some time, offering up to x4 2.0
- Somewhat efficient chaining capability, negating PCIe switch needs
- TB2<->3 adapters are one time cost not much higher than eg. active DP converters/dualhead2GO
- 6 native DP outputs and ok GPUs (can add computing GPU capacity via TB anyway)
Last edited:
Just FYI, the Mac Pro 2013 will most probably be cut off the eGPU support anyway from 10.13.4, along with all the other TB2 macs unfortunately..
amfeltec Squid PCIe2 4-slot card with new M2s….why crashing?
Hi there
I hope you can help… my apologies for not understanding PCIe lanes and throughputs very well
I have a 2009 Mac Pro 4,1 flashed to 5,1 with upgraded CPU 12 core 3.33ghz, 32 g RAM, Mac OS 10.12.6, SoftRaid 5.6.5
My PCIe slots are:
Upgraded 1) GPU NVidea GTX680, 2) amfeltec Squid 4-slot M2 PCie Gen.2 card for boot disk, 3) USB3/eSATA card, 4) UAD-2 Octo audio DSP card
I have been using the amfeltec card with 2 x 256g Samsung SM951 AHCI chips under SoftRaid Raid 0 as a 500 gig bootup disk. It has been working perfectly for for approx. 18 months… really great
However I have been finding it hard to keep the system drive to within 500 g … I need more space ….and so I recently ordered another 2 units of the Samsung SM951 256g AHCI M2 disks….. I want to increase the size of the bootup disk on the Squid card to 1TB… as Raid 0 the 4 x 256 g disks configured as one become a single 1 TB disk in SoftRaid
I didn’t foresee any problems with this strategy
However I am getting consistent failure…. although what or where the glitch is I don’t know! ) :
In summary it seems that using more than just 2 x M2 disks as a Raid 0 bootup with 64k block size drive results in consistent drive crashes of some sort. Can you help me with this?
In SoftRaid 5.6.5 I have initialised the 4 disks. I have certified them… all disks are good….. I have tried running all 4 of them as Raid 0 bootup disk…..many times …. but no luck…
If I try Carbon Copy Cloner to clone a backup system disk to the new 1TB Raid 0 disk on the amfeltec PCIe card the process only gets about 50 or 70 gig into the clone and then it crashes. Funnily enough I seem to be able to copy large files by drag and drop in the Finder.
I have tried reducing the Raid 0 disk to just 3 of the M2 chips… same problem…. crash of the clone process….crash of SoftRaid, … crash of the computer…. almost always requires Force Restart of the computer ) :
I tried reinstalling OS Yosemite from a Recovery volume of a backup disk onto the 4xM2 bootup Raid disk … again… it just freezes at a point….cannot complete. It really feels like the Squid Board has a major glitch with this config. Carbon Copy Cloner says that: “The task was aborted because a subtask did not complete in a reasonable amount of time”
When creating the raid 0 disk I have also tried using 32k block size (crashed earlier into the clone process), 128 k (crashed later), 256k (completed CCC clone process without crashing but was unusable…like glue… as a bootup disk)
The original, very stable, 2 x M2s are Samsung MZHPV256HDGL-00000 Firmware: BXW2500Q
The 2 new ones are: Samsung MZHPV256HDGL-000H1 Firmware: BXW24H0Q
I can make a new 512G bootup Raid 0 disk using the 2 new M2 disks…. I don’t see any fault with the new disks themselves
I would be very grateful if anyone has a clue on how to help fix …or understand…this problem for me. Is there a mismatch in the M2 disks? It feels like something is running too fast and then short-circuiting…or getting some type of interrupt …. What could it be? I’m just at a complete loss at this stage
Thanks very much for any insight into this you may have
kind regards
Maxxy
Hi there
I hope you can help… my apologies for not understanding PCIe lanes and throughputs very well
I have a 2009 Mac Pro 4,1 flashed to 5,1 with upgraded CPU 12 core 3.33ghz, 32 g RAM, Mac OS 10.12.6, SoftRaid 5.6.5
My PCIe slots are:
Upgraded 1) GPU NVidea GTX680, 2) amfeltec Squid 4-slot M2 PCie Gen.2 card for boot disk, 3) USB3/eSATA card, 4) UAD-2 Octo audio DSP card
I have been using the amfeltec card with 2 x 256g Samsung SM951 AHCI chips under SoftRaid Raid 0 as a 500 gig bootup disk. It has been working perfectly for for approx. 18 months… really great
However I have been finding it hard to keep the system drive to within 500 g … I need more space ….and so I recently ordered another 2 units of the Samsung SM951 256g AHCI M2 disks….. I want to increase the size of the bootup disk on the Squid card to 1TB… as Raid 0 the 4 x 256 g disks configured as one become a single 1 TB disk in SoftRaid
I didn’t foresee any problems with this strategy
However I am getting consistent failure…. although what or where the glitch is I don’t know! ) :
In summary it seems that using more than just 2 x M2 disks as a Raid 0 bootup with 64k block size drive results in consistent drive crashes of some sort. Can you help me with this?
In SoftRaid 5.6.5 I have initialised the 4 disks. I have certified them… all disks are good….. I have tried running all 4 of them as Raid 0 bootup disk…..many times …. but no luck…
If I try Carbon Copy Cloner to clone a backup system disk to the new 1TB Raid 0 disk on the amfeltec PCIe card the process only gets about 50 or 70 gig into the clone and then it crashes. Funnily enough I seem to be able to copy large files by drag and drop in the Finder.
I have tried reducing the Raid 0 disk to just 3 of the M2 chips… same problem…. crash of the clone process….crash of SoftRaid, … crash of the computer…. almost always requires Force Restart of the computer ) :
I tried reinstalling OS Yosemite from a Recovery volume of a backup disk onto the 4xM2 bootup Raid disk … again… it just freezes at a point….cannot complete. It really feels like the Squid Board has a major glitch with this config. Carbon Copy Cloner says that: “The task was aborted because a subtask did not complete in a reasonable amount of time”
When creating the raid 0 disk I have also tried using 32k block size (crashed earlier into the clone process), 128 k (crashed later), 256k (completed CCC clone process without crashing but was unusable…like glue… as a bootup disk)
The original, very stable, 2 x M2s are Samsung MZHPV256HDGL-00000 Firmware: BXW2500Q
The 2 new ones are: Samsung MZHPV256HDGL-000H1 Firmware: BXW24H0Q
I can make a new 512G bootup Raid 0 disk using the 2 new M2 disks…. I don’t see any fault with the new disks themselves
I would be very grateful if anyone has a clue on how to help fix …or understand…this problem for me. Is there a mismatch in the M2 disks? It feels like something is running too fast and then short-circuiting…or getting some type of interrupt …. What could it be? I’m just at a complete loss at this stage
Thanks very much for any insight into this you may have
kind regards
Maxxy
I generally do not have this issue anyway, as long as Apple does not entirely disable root or changes
the kernel to break my kexts
Similar to the MSI system (Z270+TB3 card) i use for testing we could patch the subsystem to reflect the PCIe device directly (this disables chaining and some TB functionality, similar to passing a TB device in Linux via VTd as PCIe into a VM), and at least with a 5,1 definition it assumes there are actual PCIe slots (4 @ x16/x16/x4/x4 2.0 + Airport @ x1 1.1) that i can occupy with the TB device ID. If needed, because a way higher level hack is likely sufficient anyway. Worst case you'd have to hack away into a full 5,1 detection or spoof the GPU IDs (to 79xx, same anyway) if based on eg. D3/5/7xx GPU presence.
It also does not affect me - i do NOT look at using eGPU at all, i need PCIe for other devices - the 6 DP ports are perfectly fine for my usage, even more so if there ever is MST support with AMD GPUs.
Nvidia cards for CUDA are not eGPU anyway and not affected by Apple unless they kill entire PCIe over TB which yields it useless.
Last edited:
My apologies if I sidetracked your thread... it looked like a good place to start on an issue similar to mine...thanks for your understanding...best to you