Kinda upset at the lack of response to GBe issues

[MarkHaysHarris777]

And then the killer answer you've somehow been begging for for weeks ... NFS ..... in fact NFS will give you blazing speeds over GbE on my odroid.
Its faster than eMMC/SD/USB storage, my NAS has an SSD especially for all my embedded devices and their NFS shares.
Most of them just get their entire root partition over NFS, so yes GbE is quite usefull (8-9 MBps vs +/- 100 MBps)
Hope that satisfies your question

I get what you've been saying all along, I just don't believe it... and empirically, I have not seen it either;  not on my Pine board !  ---on my T61 ThinkPad, yes;  but not on my Pine board.  It doesn't matter how blazing fast the port is... nor how fast NFS is,... ( and this isn't an Odroid ) the data still has to go through the Pine board, and it isn't fast enough for GbE; not in a practical sense.  Let me put this another way, the Pine board that I have that works at GbE speeds in terms of iperf and iperf3, does not work at GbE speeds when transferring data through the board... not so that it makes much difference one way or the other. And that is true whether I'm running it at 1000 /full,  or 100 /full.

And, I'm not at all arguing with you, nor am I even emotional about it... I was just wondering in the final analysis how really important is this actually ?  

But even if Pine Inc were to acknowledge the boards as hardware defective ( they barely acknowledge emails ) what makes you think they're going to replace all these non GbE working boards with working GbE tested boards... they barely got the shipment of the original boards out !  I'm just saying.

[amc2012]

I was just wondering in the final analysis how really important is this actually ?  

But even if Pine Inc were to acknowledge the boards as hardware defective ( they barely acknowledge emails ) what makes you think they're going to replace all these non GbE working boards with working GbE tested boards... they barely got the shipment of the original boards out !  I'm just saying.

I think many of us have stated how really important it is. 14 pages just in this thread alone, and more posts in others. 

As for what makes me hope they could replace all these non working GbE boards with working tested ones, here's the deal with that - you infer that there's little difference between Fast Ethernet and GbE. So I'd wager most people don't notice they have non-functional boards. So even if the PINE folks said, "hey guys, we know there's an issue - if you have a non working board, send it back to us and we'll test it and send you out a working one if we see the issue," how many people would take them up on it? According to your inference, maybe .005%?

[MarkHaysHarris777]

So even if the PINE folks said, "hey guys, we know there's an issue - if you have a non working board, send it back to us and we'll test it and send you out a working one if we see the issue," how many people would take them up on it? According to your inference, maybe .005%?

That was just one of the purposes of the poll; frankly.  So far the numbers are actually:

19 / 38000+  =   .0005

So, we're talking about .05%  ( not a whopping margin ).  So, it may be that Pine Inc, would be willing to replace a handful of boards for the people severely impacted ( knowing that the number isn't going to be 2,000 );  I mean we're talking about less than 2 dozen .  I want to make clear something that waldo and I straightened out on the irc tonight... I am not minimizing this problem, I'm merely trying to place this problem in perspective while we are trying to fix it. We are going to get to the bottom of this, sooner or later.

To add to the discussion a bit,  I was looking at the schematic again and noticed ( although out-of-date ) that there are several mods that are possible (resistors & caps) depending on which PHY was installed. There was a parts shortage and another vendor was used ( another PHY ) ; what if the mods on the board are for 8211D, but the PHY is an RTL8211E ?  Would the mismatch cause this difficulty ?  The schematic notes that there are five (5) chips that were employed:

  RTL8211CN  RTL8211D  RTL8211E  RTL8201FN  RTL8201EN

Hopefully the poll data will help to narrow this a bit;  and over time , as impacted users file in , we'll know more than we do today.  We have made progress in documenting the problem, isolating the impact cases, logging the hardware, and experimenting with possibilities; we have made progress in knowing what it IS NOT... and we have made some significant progress in understanding the impact scenarios.  We are going to get to the bottom of this, sooner or later.

[Boring]

Dear Mark,

Unfortunately I have to go off topic and please do not see below as an attack on you. I'm really happy you are on the forums.

19 / 38000+ = 0.0005

However this number mumbo jumbo you use here and say only 0.05% has a GBe issue is NOT statistics.

If you want to use your poll for statics first get enough responses. there aren't enough respondents at this moment for statics. secondly first use the numbers from the poll to calculate a statistical value and translate this to the complete population.

So here is what you should do (if there are enough respondents) 'no of defective gbe boards' / 'number of respondents' = 'percentage defective gbe'.

Now the number of respondents is too low to use statistics. We are in danger that the people who have gbe issues are over represented in your poll.

knowing that the number isn't going to be 2,000

I don't know how get this number, but you didn't use statistics to get it. Getting a reliable estimate for the number off defective gbe boards' might be difficult. An estimate on how many people want a replacement is easier to obtain. That can be done by sending out a survey to all the people who have a pine with gbe.

Mark I hope you keep on sharing your expertise, knowledge and positive spunk on these forums. But don't try to use your statistics when I'm around. I will not let you fool people so easily.

[MarkHaysHarris777]

19 / 38000+ = 0.0005

this number you use here and say only 0.05% has a GBe issue is NOT statistics.

 knowing that the number isn't going to be 2,000  

I don't know how get this number, but you didn't use statistics to get it. 

Did I ever mention statistics.  I am an expert statistician, and the above it NOT statistics, nor did I say it was. All the numbers above represent is an order of magnitude set aside as extremes for the purpose of discussion (one being small, and the other being large).  In no sense did I ever suggest nor even imply that I was employing statistics.

The Pine Inc is probably not going to want to replace a 'large' number of boards.  I made up 2000 !  ... but it could be 10,000 the point was that if the number were large there would be no hope for across the board replacement.

On the other hand this is real, not statistics, real:   19 / 38000  =  .0005

The point of that real number (not statistics) is that so far nineteen 'real' people have indicated that their GbE board did not work.  That does not mean that only 19 boards have failed , nor that the failing ratio is .0005 / far from it because its not meant to be statistics.  The number (19) is the actual real number of people TL Lim might be dealing with (an order of magnitude) right now ,  who want their boards replaced.  I think that number is significantly low that he might actually consider replacing them , just to get them to shut up !

---

I think its impossible to determine the actual number of defective boards, even if I were to use statistics, which I did not.  But I suspect the number is quite low because I suspect that the number has to do with a mismatch of mods 'specs' for the board and the actual PHY used... and this did happen for a short time.  

If I ever decided to use statistics, I would state that I am using statistics and I would explain which method, and why I think its valid.  To claim I am using statistics (when in fact I am not, nor even hinted that I was) is a red herring a straw man argument, and an ad hominem all at the same time !

[pfeerick]

Did I ever mention statistics.

Yes you did. You did interpret and analyse numbers. That is beside the point, and not even considering your base number would be wrong, because you based it on the number of backers, not on the number of boards that actually went out, which will be significantly higher (e.g. the backers who ordered 10). To analyse the poll, the number to analyse the poll against would be more the number of forum members, which is approximately 4800 (20 members per page on the forum listing, 240 pages).

Now, before this turns into yet another argument about who is right, and who is wrong. I have something to add with respect to another earlier statement on what the GbE on the pine can actually do, and have video evidence to prove, which I will be uploading later this evening. With a debian linux distro, from the pine64.pro site, with minor modifications not related to ethernet or performance tuning, the pine64 can only achieve read and write speeds of approximately 9MB/s (which when you multiply by 8 to change from MB to Mb) gets you to 72Mb/s... which is near enough to the theoretical 100Mb/s that I wouldn't complain... if this were at 100T, that is. But this was at 1000T (GbE). However, I didn't stop there. I then changed the network speed to 100T... and guess what... negligible difference... maybe a fraction slower in the download phase of data transfer. However, when I run longsleeps networking tuning script... I get a remarkably different result... an upload to the pine64 that was previously around 9MB/s was now approximately 29MB/s... that is one whopping big jump in performance! Now, it was about the same speed as before when it was downloading the file back... but the pine64 processor was pretty toasty by then... so with cooling, it *might* have performed better... I should have had the health monitor open also that time, and may also trying another run with some cooling on it to see what effect that has.

I'll also be trying the armbian distro out on this, from what tk was telling me before, none of this is necessary out of the box... Armbian is configured for performance by default.

[MarkHaysHarris777]

I have something to add with respect to another earlier statement on what the GbE on the pine can actually do, and have video evidence to prove, which I will be uploading later this evening. 

Very interesting;  looking forward to reading your post.

[pfeerick]

This is probably going to be the fourth video in a set of five... the first four have been done, but are taking too long to upload, so I'll post this one first, and I'll post a link to the playlist in the next day or so once all the over videos are uploaded.

As indicated in the video, a USB3 64GB drive is used (PNY Turbo) with benchmarked speeds of read 191.45 MB/s and write speeds of 136.2 MB/s, so it should certainly not be a bottleneck, and pushing the USB interface as well. I could have used a memory drive as a test, but that would only be raw throughput, not real life usage.

https://youtu.be/ZN-rLuC5HlU

[MarkHaysHarris777]

That's fun pfeerick ! ... and great science; thank you !

[amc2012]

OK guys, not sure why tkaiser was banned, but between that and responses I seem to get in my email and then when I come to the forum they're not here (leading me to believe that they've been deleted by a mod), I'm out. Nothing personal, and thanks to everyone for the responses (even MHH777, with whom I had some disagreements), but I'm out. Moving on.

Good luck, folks.