Don't get me wrong, but you are using wrong terms here. Don't use bifurcation term to cards with PCIe switches, bifurcation only applies to PCIe lane partitioning and not PCIe switching.
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.
Which M.2 PCIe X16 adapter card?
- Thread starter MisterAndrew
- Start date
- Sort by reaction score
Thanks for letting me know, is appreciated! I'm learning new things, because comments like this.
Marketing and sales people literally don't understand the tech, even with terms that are as old and used since railways exist.
I found the HighPoint SSD7101A-1 pretty good. get this sort of speed with 2x 970 Pro in raid0. Due to the MLC on the 970 Pro the speed is fairly consistent
Hello
I'm trying to understand bifurcation switching bridging and partitioning.
I read your posts and the document explaining how to flash the roms.
I want to achieve the same thing as you : use the speed of the x16 2.0 slot to access a x4 3.0 nvme ssd (or 2 or 4 in raid).
The squid and the highpoint cards can do that apparently.
If I use only one nvme ssd 970 pro in one of these cards, that gives me 3300/2800 on PC, will I get the same speed?
Or on my Mac 5,1 will it be limited to 1500/1500 for one?
If it is limited to the x4 2.0 per ssd, then I can use two cheaper ssd that give 1500 each and then 3000 once combined by the card.
I think this has already been discussed and answered in this thread but I cannot be sure of the final answer.
Thanks a lot for your help
Quote from another post.
Top performance / high cost [tops at 3200 MB/s with one blade, 6200 MB/s (10000~120000 MB/s for MP7,1) with two to four], all tested in the thread:
Amfeltec Squid: Amfeltec x16 PCIe with 4 SSDs: 5900+ MB/s. Supports four double side M.2 blades, up to 110mm. The PCI Express Gen 3 version uses PLX PEX8732 switch. Just a heatsink and fan for the PCIe switch, no heatsink for the blades. The Gen 2 version doesn't have a fan and doesn't allow full speed from individual Gen 3 SSDs.
That tells me the the first gen can give 3200 with ONE ssd. Is that correct?
I'm trying to understand bifurcation switching bridging and partitioning.
I read your posts and the document explaining how to flash the roms.
I want to achieve the same thing as you : use the speed of the x16 2.0 slot to access a x4 3.0 nvme ssd (or 2 or 4 in raid).
The squid and the highpoint cards can do that apparently.
If I use only one nvme ssd 970 pro in one of these cards, that gives me 3300/2800 on PC, will I get the same speed?
Or on my Mac 5,1 will it be limited to 1500/1500 for one?
If it is limited to the x4 2.0 per ssd, then I can use two cheaper ssd that give 1500 each and then 3000 once combined by the card.
I think this has already been discussed and answered in this thread but I cannot be sure of the final answer.
Thanks a lot for your help
Quote from another post.
Top performance / high cost [tops at 3200 MB/s with one blade, 6200 MB/s (10000~120000 MB/s for MP7,1) with two to four], all tested in the thread:
Amfeltec Squid: Amfeltec x16 PCIe with 4 SSDs: 5900+ MB/s. Supports four double side M.2 blades, up to 110mm. The PCI Express Gen 3 version uses PLX PEX8732 switch. Just a heatsink and fan for the PCIe switch, no heatsink for the blades. The Gen 2 version doesn't have a fan and doesn't allow full speed from individual Gen 3 SSDs.
That tells me the the first gen can give 3200 with ONE ssd. Is that correct?
Last edited:
Do you live in Bunbury, Western Australia?
Bifurcation (or more accurately, host cards containing 'active' PCIe switch* technology) will give you maximum speeds in SLOT 1 or 2 - both PCIe 2.0 (x16) slots (electrically, mechanically, physically and in terms of *memory interface bandwidth* through to the QPI [for the sake of simplicity - we might just call this the CPU])
The reason so many members, contributors, enthusiasts (and experts - though to a MUCH LESSER degree) refer to the "1500/1500" data transfer speed limitation is the "memory interface bandwidth" constraint of a PCIe 2.0 (x4) slot.
SLOT 3 and SLOT 4 are clearly marked "PCIe 2.0 (x4)". The data throughput ceiling on S3 and S4 means users inserting PCIe Add-In Cards capable of 'negotiating an 8 lane link' OR 'negotiating an 16 lane link' will never see speeds above "1500/1500".
Whilst there is a bit more to understanding "dynamic lane negotiation" in a PCIe slot. I will answer at least one part of your query with categorical precision:
(A) Where the AIC - in this case - a PCIe m.2 NVMe Host Adapter card WITH active PCIe switch* technology is installed in S1 or S2 you will get a near identical result on your cMP 2010 (i.e. YES - you will get the same "3300/2800" on your Mac Pro or marginally better).
* Known GOOD (most frequently cited, tested and verified working) PCIe switch chipsets are identified by model number and contained on the following list of cards (new entrants means this list is not exhaustive):
- Highpoint SSD7101 with PLX switch (PLX PEX8747)
- AMFELTEC SQUID with PLX switch (PLX PEX8732)
- IOCREST (aka Syba) SI-PEX40129 Dual M.2 NVMe Adapter Card with ASMedia switch (ASM2824)
- Lycom DT-130 with ASMedia switch (ASM2824)
Bifurcation (or more accurately, host cards containing 'active' PCIe switch* technology) will give you maximum speeds in SLOT 1 or 2 - both PCIe 2.0 (x16) slots (electrically, mechanically, physically and in terms of *memory interface bandwidth* through to the QPI [for the sake of simplicity - we might just call this the CPU])
The reason so many members, contributors, enthusiasts (and experts - though to a MUCH LESSER degree) refer to the "1500/1500" data transfer speed limitation is the "memory interface bandwidth" constraint of a PCIe 2.0 (x4) slot.
SLOT 3 and SLOT 4 are clearly marked "PCIe 2.0 (x4)". The data throughput ceiling on S3 and S4 means users inserting PCIe Add-In Cards capable of 'negotiating an 8 lane link' OR 'negotiating an 16 lane link' will never see speeds above "1500/1500".
Whilst there is a bit more to understanding "dynamic lane negotiation" in a PCIe slot. I will answer at least one part of your query with categorical precision:
(A) Where the AIC - in this case - a PCIe m.2 NVMe Host Adapter card WITH active PCIe switch* technology is installed in S1 or S2 you will get a near identical result on your cMP 2010 (i.e. YES - you will get the same "3300/2800" on your Mac Pro or marginally better).
* Known GOOD (most frequently cited, tested and verified working) PCIe switch chipsets are identified by model number and contained on the following list of cards (new entrants means this list is not exhaustive):
- Highpoint SSD7101 with PLX switch (PLX PEX8747)
- AMFELTEC SQUID with PLX switch (PLX PEX8732)
- IOCREST (aka Syba) SI-PEX40129 Dual M.2 NVMe Adapter Card with ASMedia switch (ASM2824)
- Lycom DT-130 with ASMedia switch (ASM2824)
Thanks for this detailed answer.
So if I put one of these cards in slot 1 or 2 and I patch everything correctly,
I will use a nvme ssd at full speed. 3200/2800 in my example.
And if put 2 in raid 0 (if the card allows it) , i will approximately double this.
Any other card in slot 1 or 2 OR these cards in slot 3 or 4 will give me 2.0 x4 which means 1500/1500.
Am I correct?
Ps: not at all i live in France mate
So if I put one of these cards in slot 1 or 2 and I patch everything correctly,
I will use a nvme ssd at full speed. 3200/2800 in my example.
And if put 2 in raid 0 (if the card allows it) , i will approximately double this.
Any other card in slot 1 or 2 OR these cards in slot 3 or 4 will give me 2.0 x4 which means 1500/1500.
Am I correct?
Ps: not at all i live in France mate
D'accord - Bonjour!
You got it.. this is a near perfect interpretation (but for the secret* information I've not revealed yet ).
*there are some funny anomolies reported by forum members and in some cases experienced by me personally.
One example: your Mac completely freezes (music may keep playing - but hard reset required) when you only have one NVMe SSD in a DUAL m.2 port host board with an ASMedia (ASM2824) switch.
Others have reported this and I have experienced it personally. There's still speculation on what causes it BUT having analysed a lot of these reports where..
(1) members complain that they have a faulty NVMe SSD (regardless of manufacturer)
(2) some members complained that they had been sent a 'bad' SYBA / IOCrest Dual Port m.2 adapter
(3) I've experienced it when using the DT-130 and the IOCREST but I never experienced it when using the HIGHPOINT SSD7101
The only commonality here is the ASM2824. So.. although I think they are excellent and far more affordable host cards - I sold my Highpoint SSD7101 to buy two ASM2824 switch-based cards (1x LYCOM - DT-130 and 1x IOCREST (aka Syba) SI-PEX40129 Dual M.2 NVMe Adapter Card) - I now consider this intermittent complaint about the NVMe hardware across the spectrum to be a KNOWN ISSUE relating to the ASM2824 chipset.
I have never experienced a problem EVER when I have 2 NVMe SSDs installed on a host board with the ASM2824.
I may come back to report on some other detail-centric PCIe AIC secrets a little bit later if I can find the time a little bit later today.
You got it.. this is a near perfect interpretation (but for the secret* information I've not revealed yet
).*there are some funny anomolies reported by forum members and in some cases experienced by me personally.
One example: your Mac completely freezes (music may keep playing - but hard reset required) when you only have one NVMe SSD in a DUAL m.2 port host board with an ASMedia (ASM2824) switch.
Others have reported this and I have experienced it personally. There's still speculation on what causes it BUT having analysed a lot of these reports where..
(1) members complain that they have a faulty NVMe SSD (regardless of manufacturer)
(2) some members complained that they had been sent a 'bad' SYBA / IOCrest Dual Port m.2 adapter
(3) I've experienced it when using the DT-130 and the IOCREST but I never experienced it when using the HIGHPOINT SSD7101
The only commonality here is the ASM2824. So.. although I think they are excellent and far more affordable host cards - I sold my Highpoint SSD7101 to buy two ASM2824 switch-based cards (1x LYCOM - DT-130 and 1x IOCREST (aka Syba) SI-PEX40129 Dual M.2 NVMe Adapter Card) - I now consider this intermittent complaint about the NVMe hardware across the spectrum to be a KNOWN ISSUE relating to the ASM2824 chipset.
I have never experienced a problem EVER when I have 2 NVMe SSDs installed on a host board with the ASM2824.
I may come back to report on some other detail-centric PCIe AIC secrets a little bit later if I can find the time a little bit later today.
Thanks. That is very helpful.
The only card I can find right know is the highpoint.
Do I need to check the model number or any model is ok?
The only card I can find right know is the highpoint.
Do I need to check the model number or any model is ok?
Best HighPoint card for a MP5,1 is SSD7101A-1 v2.00, SSD7102 and SSD7103 have additional resources that only work for PCs. You can use any of the three models, but the best cost/benefit ratio is SSD7101A-1 v2.00. You don't need to pay for things that won't work with your Mac, but again, any of the three will do fine.
The older v1.01 release of SSD7101A-1 is a good one to buy too, it just don't have the software controlled fan.
Hey !
I bought a new Lycom DT-130 (ASM2824) to speed up my cMP 4.1 -> 5.1, running Mojave 10.14.6 with a Samsung 970 Evo 1TB. It just arrived, but after installing it I experienced some strange behaviour. I searched the forum and I couldn't find anyone else yet that had these problems:
I used to run the 970 Evo as a startup disk with a Lycom DT-120 (PCIe slot-2). So far so good. No problem at all and never any failure. Today the DT-130 came in. I mounted the SSD and placed the card in slot-2 again. The Mac started up and all seemed to work. Until I tried Blackmagic Speed Test and selected the drive for a test. The Mac crashed completely and rebooted by itself. I did a PRAM zap and tried again: same result. I am sometimes able to copy small files to the drive, but as soon as I tried to copy a bigger file -> instant crash. And sometimes the whole system seems to crash randomly. I tried the 970 in both slots, but no difference.
I removed the card (felt already pretty hot after a short time) and re-inserted the DT-120+ 970 Evo. All works fine again, including running Blackmagic Speed Test. So it looks like the problem is not with the SSD drive itself. Does anyone have a clue what could cause the problems between my mac and the DT-130? Or maybe it's just a faulty card?
I have a cMP 4.1 -> 5.1 , dual X5690, 64 GB RAM, running Mojave 10.14.6 (clean install), ROM 144.0.0.0.0.
Thanks in advance!
Last edited:
Indeed, cMP does not support bifurcation. This is why we need a card with switch or controller. I am talking about a card with ASM2824 switch, which is reported working in a cMP by other users.
Thanks @zedex for the info! I completely missed your post from feb 22 for some reason. I will add a 2nd SSD blade and report back.
Bifurcation is not supported with Macs at all.
Intel PCIe Lane Partition depends on UEFI/BIOS Setup configuration of the lane arrangement before boot and Apple never implemented it or something equivalent.
Has anyone heard anything about the Highpoint SSD7140 8x M.2 NVME PCIe card?
Eight NVMEs on a single PCIe card would be super-useful for super-fast mass storage. I can’t find any info about it outside of Highpoint’s website though.
Eight NVMEs on a single PCIe card would be super-useful for super-fast mass storage. I can’t find any info about it outside of Highpoint’s website though.
Hello, Martys1,
I'm looking forward for Your experience in terms of speed test.
Thank You for sharing
Andrea
Last weekend I installed a 2nd blade on the Lycom DT-130 card. All problems solved! Here are my speed results with a Samsung 970 evo 1TB:
Only thing I experience is that when I test with large 4gb files the writing speed suddenly drops to about 50% after the first seconds. But reading speed stays high. When I select a 1gb file for speedtest then the writing speed stays high as well.
Only thing I experience is that when I test with large 4gb files the writing speed suddenly drops to about 50% after the first seconds. But reading speed stays high. When I select a 1gb file for speedtest then the writing speed stays high as well.
Attachments
thank You, then it is worth thinking about. Just a last question, sorry for asking: MacPro 5,1? PCI Slot 2? What if PCI Slot 3 or 4 (my slot 2 is blocked by a graphic card)?
I m using this card with a Mac Pro 4.1-> 5.1, running Mojave 10.14.6 with rom 144.0.0.0.0. I would not recommend using this card in slot 3 or 4. It is an 8-lane card and slot 3 and 4 share 4 lanes if I remember correctly. So in order to run this card in full speed you would need to install it in slot 1 or 2.
You are right. Thank You. Maybe I can move the graphic card to slot 2... or find a riser adapter...
Thank You.
Thank You.