Amfeltec x16 PCIe with 4 SSDs: 5900+ MB/s

[tsialex]

Anyone checked out the Kickstarter M.2 SSD Carrier For 4 M.2 NVME PCIe GEN3 x16 80Gpbs? It doesn't look like he's going to make his goal but I was wondering if this would work.

Why they don't talk about the PCIe switch? Why test results below 12000MB/s? Results are comparable to a PCIe 8x switch, like a 8x 4-Port card.

Seems very sketchy to me…
[doublepost=1551913970][/doublepost]

 

[handheldgames]

Why they don't talk about the PCIe switch? Why test results below 12000MB/s? Results are comparable to a PCIe 8x switch, like a 8x 4-Port card.

Seems very sketchy to me…
[doublepost=1551913970][/doublepost]View attachment 825069

VERY SKETCHY... and my .02 cents / comments from the peanut gallery....

I think the results are based on assumptions, rather than actual boards.

When it comes to M.2 PLX, there are 2 chipset manufacturers dominating the market:

Broadcom has 16x and 8x versions of the PLX used in the Highpoint 7101/2-A, amfeltec squid, qnap m.2, and countless motherboards. 16x versions max transfer is just over 6000 MB/s in the cMP. 8x versions max out at 3000mb/s in the cmp.

Asmedia has a low-cost 8x PLX that's used in the 2 m.2 slot IOCrest/Syba board and other 4 slot offerings with a max transfer of about 3000MB/s, about 1 SSD at a time in the cmp.

Developing a custom board design is futile and a waste of cash. There are plenty of ODM manufacturers available with existing designs for the aforementioned chip-sets.

The best solution would be to leverage a group purchase for a set number of boards, with a cascading discount based on the order volume, presenting it as such.

Working directly with the manufacturer can greatly lower the cost per unit, that can be passed on directly to the user.

 

[tsialex]

VERY SKETCHY... and my .02 cents / comments from the peanut gallery....

I think the results are based on assumptions, rather than actual boards.

When it comes to M.2 PLX, there are 2 chipset manufacturers dominating the market:

Broadcom has 16x and 8x versions of the PLX used in the Highpoint 7101/2-A, amfeltec squid, qnap m.2, and countless motherboards. 16x versions max transfer is just over 6000 MB/s in the cMP. 8x versions max out at 3000mb/s in the cmp.

Asmedia has a low-cost 8x PLX that's used in the 2 m.2 slot IOCrest/Syba board and other 4 slot offerings with a max transfer of about 3000MB/s, about 1 SSD at a time in the cmp.

Developing a custom board design is futile and a waste of cash. There are plenty of ODM manufacturers available with existing designs for the aforementioned chip-sets.

The best solution would be to leverage a group purchase for a set number of boards, with a cascading discount based on the order volume, presenting it as such.

Working directly with the manufacturer can greatly lower the cost per unit, that can be passed on directly to the user.

Double and add a little more for PCIe 3.0, you are using PCIe 2.0 values. That card is sold as a PCIe 3.0 - probably has an Asmedia 8X PCIe 3.0 switch.

The board has PEX8 written, btw.

 

[handheldgames]

Double and add a little more for PCIe 3.0, you are using PCIe 2.0 values. That card is sold as a PCIe 3.0 - probably has an Asmedia 8X PCIe 3.0 switch.

The board has PEX8 written, btw.

Yepp Ya caught me... I was stating max bandwidth on a cm @ pcie 2.x , not a pcie 3.0 slot.

The kickstarter pics and the writing offers little to trust.

 

[joevt]

Anyone checked out the Kickstarter M.2 SSD Carrier For 4 M.2 NVME PCIe GEN3 x16 80Gpbs? It doesn't look like he's going to make his goal but I was wondering if this would work.

The picture shows only two M.2 connectors. The board is not offset like the amfeltec, so putting two M.2 connectors on the backside will interfere with the adjacent slot.

Why does it say only 80 Gbps? That's only 10 GB/s. A PCIe slot can do 8 GT/s * 16 lanes * 128b/130b = 126 Gbps. A single NVMe drive can do around 3500 MB/s. Four of that means 14 GB/s. If you have NVMe drives that can do 3000 MB/s, then that would give you 12 GB/s. I got 11 GB/s on my Z170X-Gaming 7 PC with the Amfeltec.

How are they going to make it less expensive than the amfeltec or the highpoint? I don't see any useful differences.

 

[KingCachapa]

gen 2 is for recommended for Mac because the Mac pros are pci-e 2.0 and you will get better performance using Squid PCI Express Gen 2 Carrier Board on a Mac pro but if you have a hackintosh that use pci-e 3.0 then you get the maximum speed using Squid PCI Express Gen 3 Carrier Board

is not that Squid PCI Express Gen 2 is faster than Squid PCI Express Gen 3
is just that version 2 runs better on a Mac Pro
version 3 is definitely faster but it won't work as good as version 2 on a mac pro because the pci-e 2.0 bus speed

but on a hackintosh a version 3 will smoke and blow a version 2 out of the water
as long as that hackintosh has a 16X pci-e 3.0 slot

this is the new card but it is for Mac Pro 2013
http://amfeltec.com/products/mac-pro-late-2013-carrier-board-for-m-2-pcie-ssd-modules/

does this still hold true? (Gen 2 being better for our cMPs than Gen 3?)

 

[tsialex]

does this still hold true? (Gen 2 being better for our cMPs than Gen 3?)

Never was.

Back in the day you had to change the PCIe slot speed manually, default was PCI 1.0 speeds, but since 138.0.0.0.0, the firmware does it automatically.

 

[KingCachapa]

Never was.

Back in the day you had to change the PCIe slot speed manually, default was PCI 1.0 speeds, but since 138.0.0.0.0, the firmware does it automatically.

Good to know thanks!

 

[tsialex]

Good to know thanks!

This idea originated by one more non-sense article from Barefeats, if he checked the PCIe configuration, he would clearly see that was just a matter of initialisating the slot correctly, since MP51 firmwares configured PCIe 3.0 cards as PCIe 1.0, this behaviour ended with 138.0.0.0.0.

Barefeats has very good comparatives, said that, some things are out of his deep and he never corrects past articles that are wrong.

 

[KingCachapa]

This idea originated by one more non-sense article from Barefeats, if he checked the PCIe configuration, he would clearly see that was just a matter of initialisating the slot correctly, since MP51 firmwares configured PCIe 3.0 cards as PCIe 1.0, this behaviour ended with 138.0.0.0.0.

Barefeats has very good comparatives, said that, some things are out of his deep and he never corrects past articles that are wrong.

I can see how that happened haha... it's an adventure following these threads and finding what you think is new info that is so quickly irrelevant within a year

 

[ezylstra]

This idea originated by one more non-sense article from Barefeats, if he checked the PCIe configuration, he would clearly see that was just a matter of initialisating the slot correctly, since MP51 firmwares configured PCIe 3.0 cards as PCIe 1.0, this behaviour ended with 138.0.0.0.0.

Barefeats has very good comparatives, said that, some things are out of his deep and he never corrects past articles that are wrong.

I had looked that that article and even thought I'd ruled out PCIv1 mis-negotiation. Now I see it saturates with two m.2 cards, which is exactly what we'd expect with PCIv1.

Always good to have reasoned replies to help correct these bits of misinformation!