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.
MP 1,1-5,1 Syba I/O Crest SI-PEX40129 usage?
- Thread starter TECK
- Start date
- Sort by reaction score
Does anyone know of a card like the SI-PEX40129 that is as fast without needing to be installed in the first pci slot, yet still as fast in slot #2? The SI-PEX40129 has to be installed in slot one to achieve the maximum transfer speeds as far as I know and if you have a 2-slot gpu like me, that kills a slot.
MacPro5,1 have four slots:
Slots 1 and 2 are 16x PCIe lines each, are connected to the CPU and have the same identical throughput. Both slots 1 and 2 have a real life throughput of ~6200MB/s. Don't matter if you use slot 1 or slot 2 for your PCIe card, same specs, same limitations besides the physical disposition on the backplane.
Slots 3 and 4 are 4x PCIe lines each and both are connected to a IDT PCIe switch and share 4x PCIe lines from the south bridge. Both slots 3 and 4 have a real life throughput of ~1450MB/s.
Slots 1 and 2 are 16x PCIe lines each, are connected to the CPU and have the same identical throughput. Both slots 1 and 2 have a real life throughput of ~6200MB/s. Don't matter if you use slot 1 or slot 2 for your PCIe card, same specs, same limitations besides the physical disposition on the backplane.
Slots 3 and 4 are 4x PCIe lines each and both are connected to a IDT PCIe switch and share 4x PCIe lines from the south bridge. Both slots 3 and 4 have a real life throughput of ~1450MB/s.
I wish I knew why Teck says this below, but I've tested it with every operating system I've installed and the card is slower in slot #2 than #1 for sure. What else could be going on if it is slower? I get about 1,100 MB/s write and 1,700 MB/s read now. When I move it to slot #2 I get about half that from what I recall.
@TECK
If you want the max speed out of your I/O Crest, you will have to put the card in PCIe slot #1 and move your GPU card
to slot 2&3, - loosing one slot in the process...
@TECK
If you want the max speed out of your I/O Crest, you will have to put the card in PCIe slot #1 and move your GPU card
to slot 2&3, - loosing one slot in the process...
Same exact logical and electrical specs for both slots 1 and 2, if your card is slower with slot-2 and performs as expected with slot-1, something is wrong with the slot-2 of your Mac Pro.
Btw, how much slower, you didn't posted about the performance difference between both slots.
Maybe he is referring to the size limitations when you have a wider GPU like Sapphire Pulse, when you have to make adjustments to get everything to fit.
There are no speed/throughput differences between slots 1 and 2.
Okay, switched the slots and ran tests before and after. Seems to be roughly half the speed, but perhaps the numbers will make more sense to you or others. Now I am wondering about the damaged surface mount (resister?) and missing large capacitor. I thought if they were a problem it might be more obvious, but perhaps this is the result. My GPU seems to work the same in both slots, but I didn't do a benchmark. I have a use for the extra pci slot, but I am not sure I would like to live with half the speed on my startup drive.
Slot 1 - SI-PEX40129:
Slot 2 - SI-PEX40129:
Slot 1 - SI-PEX40129:
Slot 2 - SI-PEX40129:
I can't say what is/are your problem(s) but, slots 1 and 2 from a MacPro5,1 have the exact same specs/performance and when benchmarking a correctly working MacPro5,1 and PCIe card, the throughput will be identical.
MacPro5,1 Slots 1 and 2 are 16x lanes, 5GT/s and PCIe v2.0.
If System Information is detecting a x8 card as x4, something is wrong with the slot or the card, since the card seems to be working as expected with slot 1, slot 2 (or the circuit that provide lanes for the slot) is defective.
The link speed detected as 8GT/s is because of System Information is looking at the downstream of the PCIe v3.0 switch instead of the real 5GT/s of the upstream/Mac Pro PCIe v2.0 slot.
Btw, 16GT/s is the throughput for one lane of PCIe v4.0, not possible with a MacPro5,1 (since the PCIe is v2.0) and also not possible with the SI-PEX40129 ASM2824 PCIe switch (since it's a 8x PCIe v3.0 switch and not a v4.0), you are mistaken. See below the Wikipedia PCIe comparison table:
I should have said, that it was what was being reported, I've heard the mac was limited before because of what you said. I reread more of that quote from the other thread. They really do seem to be saying what I have experienced—that the card's transfer is faster when it is in slot #1 and by the amount I am indicating. I just wish I could find any technical explanation for it. I seem to have read one somewhere when I first came upon this issue, but I don't know where. AlexMaximus was the one who was indicating the card's increased performance btw, not TECK. There are also photos of the cards in that oddball orientation as well. I really don't like to think of my mac being broken in some way, but it is hard to overlook electronic components missing or broken on the PCB as have no effect at all.
The speed improvement goal with those AHCI SSD blades was to bring it up from around 1150MB/sec up to around 1890MB/sec (which was the 2015/16 speed average in newer Macs). But I have two different blades with different speeds. However, this was not the main concern. It was more about to give the Vega 7 better cooling headroom. But with a theoretical NVME 970 Pro Blade upgrade, this would become more significant.
Look, believe you or not, there are no measurable difference between the slots 1 and 2, anyone that contradict the specs is mistaking something else unrelated to the 16x 5GT/s of the Mac Pro PCIe v2.0 slots 1 and 2.
You have a problem with your slot-2 circuit.
Btw, the post you tried to quote is incorrect from the start.
2015 and 2016 MacBook Pros are PCIe v3.0 and a blade that runs the maximum throughput 3200MB/s with a 2015 or 2016 MacBook Pro like the 1TB factory installed SSPolaris will run at ~1450MB/s when installed to a MacPro5,1 with a dumb adapter and 2900 to 2950MB/s with a good PCIe switched card with a 8X switch or 3200MB/s with a 16x switch. Obviously you need a 12+16 to M.2 adapter for this even to be possible.
Anyway, this became circular and I'm repeating myself here.
I have a MP 5,1 with two of these cards. One in slot one and one in slot two. They benchmark identically, no difference in speed at all. I moved the gpu to slot 3.
Like Alex said, it seems like there is something wrong with your second PCIe slot.
Last edited:
Oh drat, thanks everyone. I guess I am going to have to live with things or get a new motherboard or get it repaired. I am going to do a benchmark on my GPU to see if it is being slowed down at all in slot number two, because that is more important as I use it for VR. Especially annoying because I just installed the bluetooth upgrade with the minipcie card.
Last edited:
I did a test video card benchmarks are close 150981 in slot 1 vs 150434 in slot 2 (but at one point it was 15300 in slot 2). Update: Right now I the GPU cards in slot one. I have knocked off, smushed or otherwise broken so many of those tiny resistors(?) off things I am surprised the computer and gpu still works at all. My NVME drive is very slow at this point, it seems not much better than SSD connected to sata (about 400 MB/s read). Probably due to another three resisters that were knocked off or damaged on the Syba I/O Crest. I'm really heartsick over my mac. I had no idea I could damage these resistors so easily (I couldn't see them, so I didn't even know they were there in some places).
The cost of a used Syba I/O Crest stinks, but probably less than the cost of getting it repaired, especially where I live. So I will get another one that doesn't have resistors broken. I will probably start replacing things like the motherboard. When it comes to the video card, those four resistors are still at least half connected so maybe I will take that somewhere to get resoldered.
The cost of a used Syba I/O Crest stinks, but probably less than the cost of getting it repaired, especially where I live. So I will get another one that doesn't have resistors broken. I will probably start replacing things like the motherboard. When it comes to the video card, those four resistors are still at least half connected so maybe I will take that somewhere to get resoldered.
Last edited:
I had a similar situation and destroyed resistors in the past on that card by fumbling in the dark and missing the slot.
If you can't find someone at your place that repairs that, I know an excellent repair place that is specialised at that. But you would need to send that card to Germany.
So, I got a used Syba Crest I/O card for $50 + Shipping, that takes care of that issue. I have no idea when I destroyed all those resistors, but somehow I don't think it was the card resistors that was causing the issue of speed, since the card worked fine in one slot and slower in another. It is probably the resistor, capacitor or something else missing on my mobo. I will see when I replace the card when the 'new' one comes. Then maybe I will deal with my video card issues, since all the resistors are there and it appears that just the solder was broken, I am hoping someone local can do that for me. Unless hot air is easier than a metal soldering tip, I think I am done with trying to solder small things.
Great news. I got the new (used) Syba Crest I/O card and it is working at top speeds in slot #2 as everyone has said. I must have knocked off some of those resistors a very long time ago. Thanks for being so knowledgeable everyone, especially tsialexx for all the time you spend helping others.
I finally had a chance to test out wireless VR on the mac 4,1 and it is working exceedingly well (there's a wigig pci card involved).