Topic: [BitCentury] Metabank 120Gh 65nm Pre-Order Proxy [CLOSED] - page 4. (Read 79010 times)

Good news folks,

I met with Timur and Legkodymov face to face today at Metabank. Long story short, Metabank will start delivering large portions of our batches starting late tomorrow, and hopefully by Saturday or Sunday we will have received all units. This also means that shipping is likely to start by Friday at the latest. The first few units will take a lot more time, until we get the packing process assembly line well established.  I will be sure to provide updates on which order #'s have shipped, and will be emailing tracking numbers etc. The other good news is that Timur confirmed that they will provide sufficient single units to cover all individuals that ordered only 1 unit, as well as 1 or 2 individuals that did group buys through us that also required some single units. For others, we will be shipping double or triple units depending on how many were ordered.  Timur said he would be there tomorrow for the start of the deliveries, but also I made sure that we were all on the same page as far as what would be delivered to us tomorrow so that no more flip/flops (pardon the FPGA pun) of information occurred.

Further to this Legkodymov  also started talking about that they are working on a way to do away with the motherboards and just interconnect the hashing units with IDE-like cables via various board pin-outs, and that a design should eventually be published. The goal of this is to increase the physical space between hashing cards such that double fans can be mounted on every heat sink to better cool the chips and overclock them closer to 3GH/s. Your units won't ship like this, this is just a work in progress that you may choose to implement yourself down the road once the design is published, so keep an eye out for that.

As to failure rates of equipment. The explained that they test all gear for a few hours before handing it over to their customers such as ourselves. In their experience they have aprox a 10% failure rate which we, the customers, should never see.  From all the units they have shipped out so far (not sure the number, but easily several dozen by now), only 3 hardware problems have occurred (including mine, which perhaps I triggered myself). I asked what most of the 10% of the failure rates were, and they said it was PCBs or something gone wrong in the assembly process. The bitfury ASICs themselves appear to be quite reliable, and no problems with power supplies, fans or raspberry pi's (which BTW have a 4GB SD card). If we do encounter any faulty parts during our testing process, Timur has confirmed that we will be given priority to have hardware either fixed or swapped out such that everything is dealt with before I have to leave Moscow.

In regards to the power supply voltage, even they weren't 100% sure if they were dual voltage or not as nearly all their customers are from Russia and they have no need for 110V, but they believe highly likely they are not dual voltage and only 230V. So customers in the USA, unless you have specifically modified your power panels to support 230V outlets (not something that is ever done by default), do NOT plug in and use the shipped power supplies as it may cause major problems or even be dangerous, you will have to source your own 110V power supplies that should be readily available at most computer retail or online stores. Estimated power wattage use at the current un-overclocked chips for a single unit is believed to be around 130 Watts, with spikes up to 165W. This is second hand information, and unfortunately I don't have my KillaWatt here to inform you ahead of time, but I hope as a community we can share some info with each other.

The sample box will be swapped out tomorrow, I decided not to bring it in today for a few reasons.

That's it for now folks, look for another update after we collect a bunch of units less than 24 hours from this post.

Cheers
Luis

Oh and PS., we also did a stop over at DHL to confirm a few details, and guess what... "In Soviet Russia..."  - http://tinyurl.com/insovietrussiahaha - ... The customs people work for DHL ;-)  No conflict of interest there huh?! But hey, this should really work to our advantage and hopeful minimize any outbound issues. We actually showed the DHL lady the box and the insides, and she said "hand made computers" should be no problem to ship.

Ok, I just called Timur to see if he was in the office. He is out in meetings with his suppliers, but he has acknowledged the unit that is not working and requested I bring it in. Very likely it is a short somewhere. More importantly, he stated that he is able to get enough single units for us from his supplier(s) and will try to deliver as many as possible tomorrow and Thursday.

Given this is a complete 180 from what I got from their lead engineer just a few hours ago, at this point I'll believe them when I see them delivering all the units. Until then, just need to stay on top of the situation. Hopefully, I will manage to see Timur later tonight and confirm things when I drop off the unit that is not working. We now have several of their cell phones and they haven't ignored calls, so that's something.

Thanks everyone for your patience and sorry for the ongoing drama.
Luis

@fpgaminer, I don't have a multimeter with me unfortunately. What I did learn from the lead engineer is that one of the things that causes the power supply not to start is if there is no hashing unit seated in the very first slot - that is the slot closest to the power supply on Motherboard Type A. If this happens, you have to wait at least 15 minutes or so for the power supply to clear the short (excuse the non-technical terms), and then try again with a card seated in the first slot. I've tried this and just heard a little "poof" sound from the power supply, whereas before just turning it on and off with different slot configurations didn't make any sound, but having said that I haven't been able to replicate the "poof" either. LOL@Purple dinasaur. The case actually looks nice from a distance.

@trinsic - You are correct, units shipped to North America would require a power supply swap out for 120V; however, please don't rush to buy one just yet, I will try to remember to double check this detail, that it is indeed only 230V power supply, with metabank tonight. I can confirm that I don't see any voltage button switch on the power supply and it only says 230V on a sticker, so it is highly likely it does not support 120V. I'm hesitant to remove the power supply and look under it or even inside. The entire system appears quite finiky (sp?) so I'd rather not touch it too much. And yes, we will definitely be individually wrapping each card, likely with bubble wrap after they are inside anti-static bags.

So yeah the Pi is powered by the IDE cable and as you pointed out plugging that IDE cable in incorrectly will kill the Pi. The Pi is powered by 5v and GRND wires on the IDE cable.

Bad news all around really isn't it Some of us wont have problems with the PSU but others may have to buy a PSU if you're sending to anyone in the US or other 120v countries. Delays on triple units and more delays for double and single units.

Hopefully there is some good news from Timur. Sounds like they're extremely busy.

Units will have to be disassembled before shipping or we risk getting them destroyed and in small pieces

spiccioli

Have I mentioned that these frequent updates are great?

By the way, have you seen http://bitfury.metabank.ru/?p=37?  Shows a couple of points to check for shorts, assuming you have a multimeter handy.

I don't care if it looks like a purple dinosaur, as long as it hashes when it gets here

Hi guys,

Here's the latest update:

- We went to Metabank late yesterday evening to pick up our sample unit (a single).
- After 1 hour wait in the reception, legkodymov came down with our unit.
- We opened it up, 1 motherboard type A , 5 hashing cards, 1 raspberry pi, 1 power cable, and (first) surprise... a 230V-only power supply (no 110V/120V switch on it).
- Over the course of an hour we waited there, another 7 or 8 people showed up, and we saw at least 4 or 5 other units go out the door. So they are definitely starting to ship.
- I asked legkodymov how many units worth of components I could pick up tomorrow (Tuesday) and he said, we can't pick anything up until the 20th (surprise #2) and they can only ship triple units to us (surprise#3), and that any spare parts would only be available after the next weekend or probably october (surprise#4) - some of these are obviously a deal breakers.
- Obviously this is not what we had agreed to on Friday, and so I immediately asked to speak to Timur. He wandered off, and never came back.
- After waiting a long while, we tried to call Timur directly (we have his cell), and nothing, no response.
- Another Metabank employee was also there helping to deliver units and get signatures of confirmation from customers that they received units. When I asked him where was Timur, he said he did not come in today and that at this hour (around 9Pm or so) he is sleeping as he was supposedly programming all day and working long hours. I also learned that he only comes into the office maybe twice a week, which they neglected to tell us before, but that he expected him to be at the office tomorrow (Tues).
- Once we got to our place, I immediately tried to power up the box and (surprise #5), the box was DOA. The power supply would not even power on.
- Thinking it was a short with one of the cards, I disconnected one at a time from the motherboard, then after that still didnt work, the fans, then tried different wall outlets, power cables etc. Needless to say after 1 hour of troubleshooting, I gave up and called Metabank.
- Surprisingly, I got a call back from legkodymov. I asked him why he wandered off and didn't tell me TImur was not even in the office that day, he appologised and that he misunderstood what I had said. In regards to the DOA unit, he was surprised it didn't work as he explained they had just finished testing it. He requested I take a picture of the unit, which I've done, and am now waiting for his reply (after 11AM Moscow time).
- We have also in the mean time found a store with anti-static bags, and are going to buy a few this morning to confirm they fit properly.

All in all not a great experience (understatement), most troubling was the change of direction in what they said from Friday; however, I am going to talk to them again ASAP, specifically Timur if I can to sort this out, as obviously this is not a good situation for any of us.

Some other things we learned:
- Weight of a single unit is 4.61 Kg (without packing material).
- Width = 45cm by 44.5 cm length and 10 cm height - this is the same as was posted a couple weeks back in the PDF file provided by Metabank in this thread.
- How the raspberry Pi is powered isn't confirmed since the box was DOA, but only the IDE cable was provided, so I am assuming it is through that.
- All the hashing cards are very loose in the box, any tilting of the box and you can hear things drag around, especially the raspberry Pi.
- The hashing cards don't have to be installed in sequence. There can be gaps in between. Below in pics you can see how metabank delivered it to us. They set it up so each card was in front of a fan.
- Another miner at Metabank reception mentioned to us that the triple units come with more powerful fans. Not sure on the power supply. They said ours was 500W, which is way overkill for a single, but there was no easy way to verify this as there was no sticker indicating this. We didn't remove the power supply out to see if there were stickers on the bottom side.
- Here's 3 additional pictures we took of our DOA sample box:

1. Top View: http://i.imgur.com/7tJTo6J.jpg

2. Raspberry Pi & power view: http://i.imgur.com/AXuyuBF.jpg
- Note the carrugated cardboard under the motherboard. The motherboard is not screwed down either, but doesn't move too much due to the weight of the hashing cards unless you tilt the box.
-  Legkodymov at one point explained that if anyone unplugs the raspberry Pi IDE cable and puts it back backwards (observe the red line on one side of the cable), it *will* kill the Raspberry Pi and possibly kill the motherboard. So users beware!

3. Top view of hash cards: http://i.imgur.com/GLCfWr1.jpg
- Note that on the flip side of the PCB, just under each capacitor, the heat sink for that card has been drilled or sanded down forming 3 concave grooves. In this picture it is very evident on the left most card, but this has been done to every heat sink on each card. I imagine this has been manually done to prevent some kind of a short with a connector on the PCB. Stuff like (drilling/sanding holes, not to mention cutting out pieces of cardboard to put under the motherboard) is probably adding to the labour required to prepare each unit. Also note that the cards sitting right beside each other have the heat sinks touch the capacitors. This happens at the top (seen in the picture), but also to the capacitors below the heatsink (not visible in picture). I dont imagine this is a problem unless the heatsink gets excessively hot.

Overall, this definitely looks like something that came out of a workshop, rather than any kind of professional fabrication.

Regards,
Luis

Good questions. Like I said, we only got a couple of quick minutes to look at the box, so after Monday/early Tuesday, once we have one in hand, I can confirm these and other details.


We have sourced some types of packing material such as bubble wrap, certain sizes of boxes and we are going to confirm on airbags as soon as possible. Nearly all the packing companies are closed on weekends. We still have some details to finalize which are specific to the size of the various components we are packing. Thus we can only provide more information late next week after we get the sample box in hand and can take precise measurements and then confirm packing material with packing companies. I will provide details and pictures as soon as I am able.

Cheers.
Luis

Thanks for the pictures Luis!

May I ask what the PCBs are powered by? Looks like Type A motherboard by the ATX and P4 4-pin and Type B by a 6-pin PCI-e? What about the Pi?

Also do you an an idea on the packaging yet? To protect the motherboards and modules you should probably put each into anti-static bags and carefully bubble wrapped possibly spaced with airbags or more bubble wrap and all fit into the case. Any remaining space in the case filled with more bubble wrap to prevent anything moving about.

Hi guys,
As promised here's some pictures of the Metabank box we saw when we met them on Friday. We only had about 2 minutes to look at the box, so didn't go into many technical details, but picture number 2 explains a few specifics. I forgot to mention two details yesterday: 1) the plastic box indeed has its various walls glued together and the plastic may not survive international shipping if the box is dropped. The plastic is thin and I can bend it with my fingers a bit. We will of course do our best to pack it well. In image #2 you can see all the silver colored capacitors. Not all hashing units are like this, as seen in previous pictures, Metabank has confirmed that we may get random hashing cards with the bigger capacitors mixed with the silver ones.

Image 1 - Overview of box (excuse my fuzzy picture). I believe this is what a double unit would look like with 10 hashing cards.


Image 2 - Main picture explaining details. Behold my awesome Photoshop skills!


Image 3 - showing power connectors of Motherboard Type A and B, as well as Raspberri Pi IDE-like connection.


Image 4 - Zoom in of inter connection of motherboard Type A and B


Image 5 - Picture of the Metabank building. There are many business' in there, and in fact Metabank isn't even a business recognized by the reception desk, they are registered under a different name. You can make out the word "bank" in the blue/white Cyrillic sign outside, but that's actually for a different bank inside the building.


Image 6- Proof that capitalism thrives in Moscow   Although the guy on the left doesn't exactly look like he's "Lovin' It"

Great news to hear!  I hope you have time to also enjoy the trip; I'll bet Moscow is very beautiful place.

Thank you for taking the time to write out these detailed updates.

Привет из Москвы!  

- Today we went to the Metabank offices and met with Timur (owner) , Anatoly (aka. legkodymov - main hardware engineer) and Vasily.
- When we asked when our units would be ready for pickup they explained not until late Thursday but more likely Friday, the 20th of Sept, as per one of their original posts.
- I asked then, why they kept saying 15th on the phone, and what they meant was they would receive all the PCBs and various other components by the 15th ready for shipping. So obviously a miscommunication/lost-in-translation there somewhere. My apologies for this.
- We requested to receive as many units early as we could (before the 20th), but they agreed only on the basis if we would assemble them ourselves.
- We agreed to this, as their main obstacle to deliver all units by 20th (assembled) is the man-power required to do so.
- By late Monday or Tuesday morning, we will have a fully functional "sample" Metabank box, which we will use as the blueprint to assemble all the other boxes.
- To be clear, assembly does not mean soldering components - I made sure we were clear on that. We will just receive motherboards, hashing units with heat sinks already attached, power supplies, raspberry pi's  and the exterior cases as well as all necessary cables and screws.
- Between Tuesday and Friday we are expecting to receive many of these components, hopefully all of them; however exactly how many we will receive on what day will be further discussed when we meet with Metabank again first thing next week. They needed more time to review their inventory; however in theory if they get all components by 15th, this should not be a problem - but I will provide another update after our next meeting.
- They also confirmed that if we found any components that didn't work, we could just come back and exchange it. They are keeping spares for this purpose.
- In regards to the issue with the single units, they explained they would be able to give us some (less than 20). This should cover all the customers that ordered only 1 unit, but is not enough for those that bought multiple units through us for their own group buys.
- We really pushed to get more single units, but they said that their contract with the external box manufacturer has ended and he is not going to make more.
- I conceited that, ok, we (Bitcentury) will use some of the external boxes from the units we bought for ourselves and give those to our customers, but we still need more single motherboards.
- They explained that, this too could be problematic because their order of PCBs was completed a long time ago, and they are almost all delivered already. To make a new order would take many more weeks, not to mention cost. They said, possibly some time in October, perhaps they could do it.
- So I requested for them to just take some of the double or triple units (which have 2 motherboards) and split them into single units.
- They said it's not that simple because overall there are 2 types of motherboards. I'll call them type A and B. Type A is used for the single devices and has 8 slots for hashing boards and an onboard connector for the Raspberry pi. Type B has also 8 slots, although only 7 may be used for hashing boards, and it has no connection for the raspberry pi easily accessible.
- To build a double or triple unit, you need a little small IDE-like cable (excuse my ignorance, I will post pictures tomorrow) that interconnects motherboards type A and B together.
- A single unit is filled with 5 hash boards on type A motherboard, a double with 10 (using TypeA+B motherboard), and a triple has a total of 15 hashing boards (8+7).
- They mentioned that one possible solution is if we split a double into a single, the Type A will accept raspberry Pi, but to do the same for the Type B requires some minor soldering and a small component. I will post some pictures tomorrow to explain this more clearly, and so after the pictures, feel free to also ask any technical questions you may have so I can write them down and ask them for when we meet Metabank again first thing next week.

Thanks everyone for your patience and especially your civility, considering all of what goes on around these threads
Have a great weekend,
Luis

I would strongly recommend not to rely neither on Sep 15 nor on overclocking.

It seems that Metabank will be done with shipping not earlier than the end of September (or even later).  This is a total BS about 40 units per day. Or there are some hidden customers noone knows about. The whole Metabank preorder was about 400 pcs of single units. Ten days to ship at specified rate. Why does it take longer then?

As for overclocking - Metabank runs its chips at 2.5-2.7 Gh/s. It means that overclocking to 3 Gh/s will raise power consumption might be one and a half or even two times. Besides, each device is "missing" an additional extension card which would drive device up to 120 Gh/s at 2.5Gh/s chip speed and regular power consumption. E.g. Metabank's devices don't really have any hoarded resources to rely on.

I've personally made a refund. I recommend that you reestimate profits from October shipped units - will they cover network hashrate growth?  I don't really think so.

The former Metabank's customer.

Hi guys,
So we made yet another call tonight, as the guy rushed us off yesterday. Here's some additional info:

1. They (Metabank) are having a meeting tonight as other people and other group buys also ordered multiple individual units, so they will have to deal with individual units one way or another. Right now, they are preparing most of the units as either doubles or even triples, but the reality is they have already shipped individual units, as many of the first units were individual units. I have a feeling this has more to do with their costs and time/effort to have to assemble & test more units, than if they combined many of them - but that's just my speculation.

2. When we asked for an ETA, the guy really would not commit to any date at all. All he said is that we (Metabank) really have to stick to the order sequence and if we have any problems and have delays we can not have people show up because we told them to come on X date based on an estimate. When we have units ready, we call the people in the sequence to come pick up the units that have been assembled and tested.

3. We asked if they were still targeting the 15th of Sept to have all units assembled, tested and delivered. He said, yes that is still their target date, but this is only an estimate and goal and cannot be guaranteed in any way. Also given they are already about 1 week late on their "40 boxes per day" initial target, I seriously question all units shipped by the 15th, but we'll certainly be ready if that's the case. The 15th is only a few days away now, so it's not like they are saying "two weeks" ;-)

We also confirmed a bunch of other details, like pickup address, procedures, time of day they are open etc not so much relevant to our customers.

As far as the units not being 120 GH, we totally hear you guys on that, but there isn't a whole lot we can do about that, believe me, at least 75 pages (!) out of the 350+ pages of the Russian 65nm thread are people complaining they didn't get the 120 GH and only 96-105 or so. Hopefully like the avalons and BFL units at some point somebody will figure out how to overclock them and share with everyone.

Regards,
Luis

Thanks for the update; it sounds like an interesting time for all involved.

Not taking a shot at you Luis, and while I actually think compared to other vendors Metabank has done better than average so far I'm nevertheless intrigued that on one hand Metabank are using the justification that they sold GHashes, not boxes, to ship fewer physical units, but at the same time they are apparently underdelivering on the GHashes they promised.

Those of us like me who only ordered 1 x 120 GHash unit will be hoping that we're not unfairly punished with late delivery because we didn't order 2 or more units. I know, nothing you can do about it.

I am sure you will have enough stories to write a book by the time this is all done and dusted!

Keep them updates comin'!  It helps distract me from watching the Oculus shipping thread(s)...

No problems here.  Heck, you can round up my order if you're lazy

Hi guys,
We managed to get another call into Metabank yesterday after many hours of re-dialing due to busy signals.  They didn't want to talk very long due to the call volume, but here's the latest updates we managed to get from them:

1. By default Metabank is assembling double hashing units. This is to save on assembly time, but obviously cost (to them). If you've ordered a single unit from us (BitCentury) we'll obviously need to get them to provide an individual unit. If for example you ordered multiple units but for a total of an odd number (i.e. 3), expect to receive a double hashing unit + a single unit.  Whether this is good news or bad news, depends on one's perspective I suppose. If you have a major objection to getting double units, please email me ASAP ([email protected]) so I can take note and see what I can do about it, but given we're a proxy service, I make no promises - we get, what we get. In short their explanation/justification was that: " We sold GigaHashes, not boxes. So, so long as we deliver the gigahashes sold, it's not our problem if it all fits in 1 box or 3".

2. All units will be tested and assembled by Metabank prior to handing  over to BitCentury. This is contrary to what a few select customers were reporting earlier in the 65nm Russian thread, where they picked up all the pieces individually, then got home and realized something didn't work or fit together properly.  On this note, we will still test the units ourselves prior to shipping to you - just in case!

3. On that note, if we do encounter faulty parts, Metabank says they will offer replacements, so this should be relatively easy for us to do, given we will be doing the testing in Moscow.

4. The software, contrary to some suggestions in the 65nm Russian thread (one guy reported having to reboot every 2 days), is not buggy, but expect updates for "optimization" - whatever that means.

5. When we asked for ETA of when we can start picking up devices, the guy on the phone said he didn't have the list in front of him, but he said assembly/testing is rapidly ramping up this week and they are still targeting to deliver everything by 15th. This is somewhat noticeable in the Russian thread, as more and more people are claiming they received their units, but actual official numbers are not available it seems. As of the time of this posting, we still have no email from Metabank requesting we come pick up the units.

That's it for now.
Cheers,
Luis

Ok found the source of the "10% of all units shipped so far" (As of yesterday). It's directly from Metabank, so that's good.
https://bitcointalksearch.org/topic/m.3096944

Cheers
Luis

Hi fpgaminer,
You're welcome. No, I had read 2 estimates. One That they had shipped a total of 5 units which I think possibly was just the number of people that came forward on the thread talking about their units, and then a few pages later, somebody posted that according to their estimates a total of 29 have shipped (over the period of several days - so not 40 units a day as Metabank had initially estimated).

The last thing I read yesterday was that one guy believed (not sure where he is getting his info) that they have now shipped aprox 10% of all units, although I have no way of knowing if that's accurate or not.  (sorry I lost URL)

Another guy just yesterday...
https://bitcointalksearch.org/topic/m.3095676

...Thinks that perhaps this week they finished May 31st orders. Since our first order is from June 2nd, this doesn't sound promising unless they quickly ramp up production volumes, which they say they are doing.

I don't think it's public knowledge exactly how many  all units add up to, but if the public spreadsheet is any indication, there's at least 419. So with the more optimistic 10% estimate, that's 41 units shipped would still put us well below our initial order (aprox #113 best case scenario); however, there's also rumblings in the forum now that metabank may be delivering at least some of the units out of order, which is concerning (look at green highlighted sections of google docs spreadsheet I posted before), especially since we still haven't gotten any communication from them to come pick up any of our units. fsb4000 also just wrote yesterday "Well finally the first person who received the device and that does not look like a friend MetaBank. This gives hope that soon I'll call"
https://bitcointalksearch.org/topic/m.3095766

I'm going to try to arrange for my translator to make another call early next week, and see if we can get some realistic ETAs, even though Metabank has asked people not to call - but what else can they do if they don't offer info up front!? Lastly, sometime this coming week, I will be away on my way to Moscow. My visa has officially been approved, so now just waiting for the courier gods to send me my passport back before I have to leave.

Have a great weekend everyone,
Luis

Thank you for the update, digitalmagus.  I'm fine with the no-refund policy; expected no different.  I'm still excited the get the units.

As in, they're shipping 29 units per day so far?  That's not bad at all.  Any estimate, based on that figure, of when our units might be ready for pick-up?

By the way, there has been a few technical updates on their blog here.  Nothing significant from what I saw.

And here's a few other useful posts I was able to gather by skimming through over 100 pages of the Russian 65nm threads....

Tech related posts:
How to edit cgminer.conf through ssh/WinSCP to point to another pool other than ghash.io
https://bitcointalksearch.org/topic/m.3067979
https://bitcointalksearch.org/topic/m.3068380

Sept 3rd bullet point summary of past ~100 PAGEs of 65nm Russian thread:
https://bitcointalksearch.org/topic/m.3069643

NEW Russian tech thread related to configuring and optimizing Metabank units. This one will hopefuly have way less spam and more useful content. We'll definitely be keeping an eye on it:
https://bitcointalksearch.org/topic/bitfury-asic-287147

I've stopped reading (for the time being) the main 65nm Russian thread after page 310 (Already at 330). If anybody else has other useful info they extracted, by all means share it.

Cheers,
Luis