[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Looking for FOSS supported PCIe x4 SATA 6 Gb/s HBA with 4 or 8 ports



On Feb 10, 2020, David Christensen wrote:
> On 2020-02-10 13:55, Dan Purgert wrote:
>> On Feb 10, 2020, David Christensen wrote:
>>> [...]
>>> It has the following expansion slots:
>>> 
>>> - One PCI Express 2.0 x16 add-in card connector
>>> - One PCI Express 2.0 x4 add-in card connector
>>> - One PCI Express 2.0 x1 add-in card connector
>>> 
>>>      a.  While migrating backup data, I recently saw a Syba PCIe x1 two
>>>      port SATA II 3 Gb/s HBA model SD-SA2PEX-2IR throttling under
>>>      sustained load -- it ran at 80-100 MB/s for 4-5 minutes, then at
>>>      ~7 MB/s for two hours. Unacceptable.
>> 
>> That controller makes no sense -- I think someone made a typo somewhere
>> on the specs (says it's a 2.5 Gb/sec PCIe x1 interface -- maybe they
>> meant it's PCIe 1.0 compliant, at 2.5 "gigatransfers" per second).
> 
> I assume you are referring to the Syba product page:
> 
> http://www.sybausa.com/index.php?route=product/product&product_id=172&search=SD-SA2PEX-2IR

That, and the SIL PCIe controller chip's datasheet.
> 
> 
> Looking at Wikipedia, PCI Express link performance table:
> 
> https://en.wikipedia.org/wiki/PCI_Express#History_and_revisions
> 
> 
> I believe 2.5 GT/s produces 2.5 Gb/s under PCIe version 1.0.  So, the Syba
> specification is correct in this case.

They apparently conflated "GT/sec" and "Gb/sec" (uhh, "conflated" might
be the wrong word).  2.5 GT / sec with an 8/10 encoding scheme, while
"yes" 2.5 gbit worth of data is being transmitted, results in only 2000
mbit (2.0 gbit) of usable data.  The other 500 mbit being eaten by the
overhead.

> 
> 
>> As I recall, PCIe 1.0 was in the neighborhood of 200MB/sec sustained for
>> a x1 slot;
> 
> Wikipedia indicates 250 MB/s throughput.

I was close :)

> 
> 
>> couple that with a slowish (or damaged) drive, or a
>> RAM-starved system, and a 7MB/sec transfer isn't exactly outside the
>> realm of possibilities.  Granted, age of the card could also be a
>> factor.
> 
> When I connected the drive to a motherboard SATA port, the transfer
> maintained 80-100 MB/s for the entire duration.  That eliminates the
> drive, the cable, and the RAM.  I concluded the HBA was to blame.  In
> hindsight, it may have been a loose connection.  It's not my computer,
> so I will have to wait for an opportunity for further
> troubleshooting.

Yep, certainly does point to either a loose connection, or a failing
SATA controller then.

But yeah, I guess this isn't anything more than sidebar to your initial
question :|.  I'll have to dig up what cards I've used in the past (been
a long time since I've had a desktop ... last two PCs have been
laptops).

-- 
|_|O|_| 
|_|_|O| Github: https://github.com/dpurgert
|O|O|O| PGP: 05CA 9A50 3F2E 1335 4DC5  4AEE 8E11 DDF3 1279 A281

Attachment: signature.asc
Description: PGP signature


Reply to: