Re: Gigabyte BIOS Upgrade F11b BETA solved a lot of diffuse hardware problems

Norm Why wrote:
>>>>> Using that tool, '(05) Reallocated Sector Count' was displayed for the
>>>>> 500GB Samsung EVO SSD, but not for the 500 GB Seagate Barracuda SSD. 
>>>>> It's connected to a different controller, a PCIe (x1) SATAIII card.
>>>> Sorry, I thought that 500GB Seagate Barracuda was a HDD.
>>>>
>>>> The HDTune tool is too old to display SSD tables properly.
>>>>
>>>> You were probably using some Toolkit to get that number,
>>>> and maybe the Toolkit has the SMART table in it somewhere.
>>>>
>>>> My SSD samples here, tend to have rounded numbers,
>>>> like 256GB or 512GB, rather than 250 and 500 values.
>>>>
>>>> The red glow, means SATA commands are coming in. The drive
>>>> can still be writing, as it rearranges the data, but the
>>>> LED does not light while this is happening. Any LED glows,
>>>> indicate your OS is doing something. You would need to
>>>> probe further, to figure out whether the operations are
>>>> read or write. (Task Manager has I/O columns you can turn
>>>> on for this.)
>>>>
>>>> The SSD has a processor inside, and it is running for as
>>>> long as the drive has power.
>>>>
>>>>    Paul
>>> Thanks again Paul,
>>>
>>> I'm still trying to solve some performance issues. Here is:
>>>
>>> -----------------------------------------------------------------------
>>> CrystalDiskMark 3.0.3 Shizuku Edition (C) 2007-2013 hiyohiyo
>>>                            Crystal Dew World : http://crystalmark.info/
>>> -----------------------------------------------------------------------
>>> * MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]
>>>
>>>            Sequential Read :   181.235 MB/s
>>>           Sequential Write :   156.012 MB/s
>>>          Random Read 512KB :     2.018 MB/s
>>>         Random Write 512KB :   150.887 MB/s
>>>     Random Read 4KB (QD=1) :    26.341 MB/s [  6431.0 IOPS]
>>>    Random Write 4KB (QD=1) :    38.755 MB/s [  9461.7 IOPS]
>>>    Random Read 4KB (QD=32) :   127.105 MB/s [ 31031.5 IOPS]
>>>   Random Write 4KB (QD=32) :   106.460 MB/s [ 25991.2 IOPS]
>>>
>>>   Test : 50 MB [D: 21.7% (101.2/465.8 GB)] (x1)
>>>   Date : 2021/04/10 15:58:02
>>>     OS : Windows 8.1 Pro [6.3 Build 9600] (x64)
>>>
>>> Test of Seagate Barracuda. Note anomalous Random Read 512KB :     2.018 
>>> MB/s. Could this low number indicate problem with Seagate Barracuda or 
>>> with cable. I'm using a 7 conductor cable, 2 pairs +/- data, 3 ground. Is 
>>> my Seagate damaged? Is my cable insufficient? What about 'twinax'? Maybe 
>>> my Startech PCIe SATA controller is to blame.
>> I did "not much of a test here", on a Win8.1 setup and
>> crystaldiskinfo 3.0.4 and the 512KB results are just
>> a bit lower than the sequential. There's no huge dive
>> like in your 2.018 MB/s result.
>>
>> It's not PIO mode, because that would affect read and write.
>>
>> It's not error rate, because the other results are
>> too good for that. Why would it just ruin the 512KB stuff ?
>>
>> I'm sure by now, you've compared your exact model number,
>> to graphs already published on the web. And I bet theirs
>> don't dip like that.
>>
>> You do see the occasional ATTO that is out-of-sorts. Where
>> one particular transfer size is not as good as it could be.
>> But I don't know if that would be 512KB - it would usually
>> be some smaller transfer size.
>>
>> See if you can dig up someone elses results for your drive.
>>
>> And if it were AHCI versus non-AHCI, I doubt that would do it
>> either. There is a good chance a 512KB read or 512KB write,
>> would be one "transaction" and multiple packets. Rather than
>> being some crazy series of tagged queue requests to do the
>> job.
>>
>> The cabling is usually pretty good. Only if the cable
>> is kinked, with a permanent pinch mark, would I be
>> concerned. This sometimes happens when importing cables,
>> the cables are wrapped in a bundle, and if some weight
>> falls on the bundle, the cable could get crushed.
>>
>> There is a counter in SMART for CRC errors on the cable.
>> But that counter would be for only one of the two
>> directions. On a CRC error, there is likely a retransmit
>> request. I don't think it's an ECC, and repaired on the
>> spot.
>>
>> I can't think why that would be.
>>
>>    Paul
> 
> Thanks for your thoughts. Here is a picture from Wikipedia. "Cross section 
> of a SATA 3.0 cable, showing the dual Twinax conductors for the differential 
> pairs." 
> https://en.wikipedia.org/wiki/Twinaxial_cabling#/media/File:SATA3-TwinAxCable.jpg
> 
> Better class 6 Ethernet cable is twinax. I would think better SATA 3 cable 
> should be twinax. What do you think? I can't find such cable anywhere, just 
> 7 conductor. 

SATA cabling is nicer electrically than Ethernet. SATA is
a "brute force" technology, in terms of signal processing,
and it's "how fast of a sine wave can I run down a coax".

One thing I like about high speed interconnect, is you
can scope it, and it can be a "blur" and... it still works.
That always freaks me out :-) The trick there is clock
extraction and knowing when to sample the blurry thing.
The 8B10B code gives a bounds on number of edges per
symbol time, that ensures there's something to extract
a clock from.

There's really no excuse for not being able to
recover a signal off that SATA cable.

    Paul