I don't have access to the marketplace either. Let me know if you're still selling it
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.
Highpoint 7101A - PCIe 3.0 SSD performance for the cMP
- Thread starter handheldgames
- WikiPost WikiPost
- Start date
- Sort by reaction score
- Status
I don't have access to the marketplace either. Let me know if you're still selling it
Who is gonna be the first one to buy HighPoint SSD7102 and dump the new firmware? I bet that they changed the SPI flash memory to a bigger one, but it's not a problem for me to replace it.
Datasheet: HighPoint SSD7102
Broadcom has a DOS utility to read and write the EEPROM: https://www.broadcom.com/products/pcie-switches-bridges/pcie-switches/pex8747#downloads
View attachment 798002
[doublepost=1540407201][/doublepost]Btw, I bought a SSD7101A yesterday, so today they announce SSD7102 - I'm always lucky like that
I ordered mine yesterday. I will PM you when I get it installed.
Planning on doing this myself. Taking the lid off the 7101A shows that the stock thermal solution is just awful. The heatsink is okay, but majority of the cooling for the PEX 8747 chip is coming from the forced air onto the chip itself and the two thick posts connecting it to the rest of the sink. Those thick thermal pads really have terrible conductivity.
I tried running the card with the power to the fan disconnected and it resulted in instability, though the heatsink itself barely scraped 29degC. That shows how poor the thermal transfer is between chip and the majority of the sink and explains why your card runs fine with such a tiny heatsink on the PEX 8747.
Are you using a thermal compound glue or just a paste?
[doublepost=1541553577][/doublepost]
The only Mojave APFS RAID boot success I've had was with the Angelbird Wings X2. That pretends to be a single Apple SSD when in RAID mode, meaning you can merrily boot an APFS formatted volume stored on a RAID array. Unfortunately, the performance of a single NVMe drive eclipses it entirely, making the whole endeavour somewhat pointless.
The copper heatsinks I used come with a sticky thermal compound pad already attached to them.
Be gentle when removing the blue plastic cover off the sticky pad.
Ah, I know the stuff. I have various ridiculous heatsinks in a box somewhere, I'll try to make some kind of Frankenstein's monster out of the card.
I did similar modifications to my Highpoint card. First I used this on the PEX:
It's made of aluminium and no copper.... but with this it was running very(!) warm - too warm to touch. I was concerned, so I improved the passive cooling:
https://forums.macrumors.com/threads/which-m-2-pcie-x16-adapter-card.2120993/page-2#post-26588513
...and now I modified it even further: removed the fan and replaced the big cube-like cooler with a flat 100x60x10 cooler:
It's running stable for quite a while now. Performance (970 Evo Raid) looks like this:
Idle thought re : M.2 cooling . . . . I wonder if the heat dissipation clips used on RAM modules could be used on M.2 blades . . . . or . . . a dedicated hi-tech ( silent ) fan run off an unused SATA power socket.
At least in winter ( in japan ) heat is not a real problem for almost 4 months of the year.
Now that I have only one SATA HDD installed in the main internal case area .. in socket 4 ( rear of the case ) my M.2 adapter gets a lot more airflow.
I also have two ( slim ) SSD's in the DVD drive area so both they AND the PSU now get a lot more cooling due to the DVD drive having been removed.
My Japanese brother-in-law runs a CNC milling machine factory and I'm familiar with CNC programming . . I might try to come up with a pure copper M.2 heatsink connected ( top & bottom ) with non-conductive thermal paste in the relevant, hottest areas + clips . . or . . . connected directly with longer screws to the resident adapter install hole.
At least in winter ( in japan ) heat is not a real problem for almost 4 months of the year.
Now that I have only one SATA HDD installed in the main internal case area .. in socket 4 ( rear of the case ) my M.2 adapter gets a lot more airflow.
I also have two ( slim ) SSD's in the DVD drive area so both they AND the PSU now get a lot more cooling due to the DVD drive having been removed.
My Japanese brother-in-law runs a CNC milling machine factory and I'm familiar with CNC programming . . I might try to come up with a pure copper M.2 heatsink connected ( top & bottom ) with non-conductive thermal paste in the relevant, hottest areas + clips . . or . . . connected directly with longer screws to the resident adapter install hole.
Nice. This looks much better than the postage stamp sized approach that was posted earlier. These chips get HOT and if one is not careful, removing active cooling can easily decrease the longevity of PEX 8487 PLX's operation.
They do get hot without any cooling and are rated to operate between 0 and 70C.Nice. This looks much better than the postage stamp sized approach that was posted earlier. These chips get HOT and if one is not careful, removing active cooling can easily decrease the longevity of PEX 8487 PLX's operation.
With the copper heatsink, average ambient temp of 26C and PCIe fans spinning at 1300 min, the heatsink's temp is anywhere between 55C and 62C.
While the switch is operating within specs, I found a similar spec (dimensions/form) fan that is quieter by 7dB and will be ordering it soon. While the replacement fan should bring the temps down by quite a bit and also be quieter, I also plan on upgrading the heatsink to Enzotech's pure copper clip-on heatsinks with fins as high as 12mm vs the current 2mm height.
These are my own results and will vary in different environments.
Right, this is what I came up with.Ah, I know the stuff. I have various ridiculous heatsinks in a box somewhere, I'll try to make some kind of Frankenstein's monster out of the card.
Stop laughing at the back!
This actually works really well; the heatsink is a 60x10mm block of copper, readily available on Amazon with the fan and retaining clip removed. I originally intended to use thermal adhesive and just glue it on there (straight) for good, but as luck would have it two opposing holes in the sink aligned with holes on the board from the original fan and heatsink arrangement.
I used a couple of PCB standoff nuts I had in a box, a couple of bolts with matching threads, drop of Arctic Silver and it works perfectly. There was no way I could cope with this card with that ridiculous little fan running all the time.
Only thing to note is this:
One side of the sink has a rebated edge which was just sufficient to rest gently on this inductor. Also, of course, there's nothing stopping you from cranking down the pressure on the PEX 8747 chip until you crushed it, so go easy.
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Since these are essentially differently-shaped PCIe x4 slots, has anyone tried an m.2 to PCIe adapter for Bitcoin mining to host aPCIe USB 3.1 Gen2 ASM1142 card? Would be a way to free up a PCIe slot and also alleviate the bottleneck that exists with slots 3 and 4 sharing bandwidth.
These are PCIe 3.0 slots, so an x1 link will will give full speed that ASM1142 is capable of. It would go m.2 to USB plug(carrying PCIe signal) to bitcoin-style PCIe riser to ASM1142 card. Could mount riser and USB card in open 5.25" bay. Might even fit neatly in one or two 3.5" drive sleds.
This: https://www.ebay.com/itm/273548447313
And This: https://www.ebay.com/itm/273541594119
These are PCIe 3.0 slots, so an x1 link will will give full speed that ASM1142 is capable of. It would go m.2 to USB plug(carrying PCIe signal) to bitcoin-style PCIe riser to ASM1142 card. Could mount riser and USB card in open 5.25" bay. Might even fit neatly in one or two 3.5" drive sleds.
This: https://www.ebay.com/itm/273548447313
And This: https://www.ebay.com/itm/273541594119
Who is gonna be the first one to buy HighPoint SSD7102 and dump the new firmware? I bet that they changed the SPI flash memory to a bigger one, but it's not a problem for me to replace it.
Datasheet: HighPoint SSD7102
Broadcom has a DOS utility to read and write the EEPROM: https://www.broadcom.com/products/pcie-switches-bridges/pcie-switches/pex8747#downloads
View attachment 798002
[doublepost=1540407201][/doublepost]Btw, I bought a SSD7101A yesterday, so today they announce SSD7102 - I'm always lucky like that
I talked to the Highpoint sales rep asking them what is the status of my order and they asked me what platform I will be using it on. After I replied they answered:
" Thanks for the promptly reply.
unfortunately, current release only support windows platform, end of December and Mac support is target on next year."
Yes it should - google BFB0512MA for the 28dB-A or BFB0512HA for the 32.5 dB-A models.
The one that came with the card is BFB0512HHA and its noise level is 37.5 dB-A
Edit: corrected the original fan's model number
Last edited:
amfeltec card has a screaming fan too! noctua low-noise adapter has a resister in it. looks like 82 ohms. http://www.silentpcreview.com/forums/viewtopic.php?t=57475
Newly installed 7101A. My PCI fan speed has increased from the baseline of 800 to 910. No other changes to machine. Does this card installation normally result in a higher PCI fan speed?
Popped a second 1 TB 970 PRO onto this card today, decided to try software RAID0 with Disk Utility for a giggle before I finished setting it all up properly. This happened.
Regarding the Amfeltec Squid Gen 3 x16 PCIe quad slot card or the Highpoint 7101a quad slot card:
Does anyone know if I can use a single 970 EVO NVMe in either card to boot and stripe {0} two other 970 EVO NVMe blades for storage on the same card??
Will I get approx. 3000 for all? (the separate 970 boot and the other two striped 970’s on the same card)
I thought tsialex might have discussed this on one of the threads I have been reading??
I am a newbe reading like crazy!!
2009 cMP flashed 4,1 to 5,1 upgraded with as much hardware as I can understand!
Thanks for this forum!!
[doublepost=1543043042][/doublepost](I have updated boot ROM to 140.0.0.0 thanks to tsialex)
Just got a reply from Bare Feats:
Unless your boot ROM version is 140.0.0.0, you can’t boot from an NVMe flash like the 970 on the Mac Pro tower.
I have not tried reading/writing to two different sets of flash blades simultaneously on the Squid3. It has a PCIe switcher that acts as a traffic cop. I’ll have to experiment to get you a solid answer.
Does anyone know if I can use a single 970 EVO NVMe in either card to boot and stripe {0} two other 970 EVO NVMe blades for storage on the same card??
Will I get approx. 3000 for all? (the separate 970 boot and the other two striped 970’s on the same card)
I thought tsialex might have discussed this on one of the threads I have been reading??
I am a newbe reading like crazy!!
2009 cMP flashed 4,1 to 5,1 upgraded with as much hardware as I can understand!
Thanks for this forum!!
[doublepost=1543043042][/doublepost](I have updated boot ROM to 140.0.0.0 thanks to tsialex)
Just got a reply from Bare Feats:
Unless your boot ROM version is 140.0.0.0, you can’t boot from an NVMe flash like the 970 on the Mac Pro tower.
I have not tried reading/writing to two different sets of flash blades simultaneously on the Squid3. It has a PCIe switcher that acts as a traffic cop. I’ll have to experiment to get you a solid answer.
Last edited:
I now you are newbie, but you already posted this 3 times and two times in threads off-topic. Don't hijack threads, it's a bad way to get help and no one who have the knowledge will help you if you continue with this attitude.
Your questions are already answered multiple times into this thread:
- You can install up to 4 blades into a Squid PCIe 3.0 or Highpoint SSD7101A, the 4 blades share the total PCie 2.0 x16 bandwidth, a little over 6000MB/s after the overhead.
- If you want to boot from the PCIe adaptor, you can use one for boot and up to three to create a RAID-0.
- If you install one or two, you will get the maximum throughput possible from the 970 EVO.
- Since 4 970EVOs have more throughput than the PCIe 2.0 x16 slot, you will get around 6000MB/s with 4 blades installed.
- You can't get past the limitation of the 16x PCIe 2.0 slot total bandwidth, the 4 blades will share around 6000MB/s.
- With a RAID-0 of two 970EVO you will get around 5000MB/s from the RAID, maybe a little less, since benchmarks some posts above were made with 970PRO 1TB.
- The PCIe switch shares the total bandwidth, if you are using one for boot and actively using the RAID, the bandwidth will go for the RAID, it's not statically divided but shared between all blades.
Last edited:
No. Not since 10.12.
Can't boot to RAID in 10.13 High Sierra or 10.14 Mojave.
This isn't NVMe limitation. 10.13 and 10.14 will not boot from any RAID volume.
Unless it's a hardware RAID which disguises the array from the OS, such as on the Angelbird Wings X2. But then you're not getting NVMe performance out of that card.
Is this a direct result of the APFS implementation..? Can't we still choose HFS+ on HS..? I know it's nearly impossible to workaround on MOJAVE.. RAID configurations are important to me - I not willing to 'update' to a recent version of OSX until I know any/all constraints or restrictions that might affect my software RAID setups