Topic: Cointerra running at 80% ???? (Read 2213 times)

I was under the impression the only reason they were shipping with these specs is because all their existing customers expressed a preference for earlier and under-powered, rather than later and to spec.  Reality being what it is, what is your preference?  Maybe you can make arrangements with them.

It looks like 80% is the best the first group of Cointerra customers can expect.


I wonder what sort of discounts they'll give since they're delivering late and under powered? I'm not happy waiting knowing my new machine will be slower than expected.

bad news for whom?

They posted today that they expect to begin shipping the initial run of units as the end of this week...good news for the first batch buyers but bad news for low powered miners.

gras - i think they made a deal with all their december batch customers to move them into january, with compensation.  so i don't think there are any december orders expecting hardware in december!

Its hard to say from the vid even if they had the rig running on 'all four cylinders'.

If I had a December order, I'd probably be freaking out a bit, because I'd want it now and figuring out any performance issues could take a while.  From an outside position, I think this is a great starting point.

I think that if they can't get it to goal quickly, then they should let the customers with outstanding orders pick if they want 'right' or 'right now' for their unit.

The way the difficulty keeps going up, I know which way I'd lean.

theyve been making updates and announcements weekly, usually on tuesdays so maybe this tuesday we will have more news...?

Does anyone know how big the delays are? and if the February batch is delayed as well?
I can't find any information about this on the internet 

Well said. But I'd be a lot more impressed if there was a real 'flow' of information rather than a snippet with nothing to back it up. These are digital chips, not mixed signal where there's a lot more issues. If the chips had a built in test structure it would be very easy to run specific test patterns to see what the problems might be. This isn't a new CPU core - it's a bog standard fixed integer arithmetic function that can be realised in less than 5k gates per stage (or any awful lot less if you're clever enough) then simply replicated many times to build the pipelines. So where's the problem?

None of my business, of course, simply professional curiousity. Contrary to what some of you may think, I do wish Cointerra well and hope that they live up to their customer's expectations.

Title could just have easily been, "Cointerra running at 80% !!!!" with a smiley face.   We design and build signal processing hardware at my job and 80% of goal at initial turn on would not be bad.   In some cases we would not be allowed to test run new hardware at full capacity until it was better characterized at lower speeds.  Also, even if I have indications or real test data that showed I could run at higher speeds, I would not be allowed to report that to a broader audience even inside the company until other cognizant staff and managers were in agreement with the findings.

So no matter what Cointerra reported between 80% and 120% you could spin the interpretation based on how honest you think they are and how you believe they run their tests and reporting on new hardware.

Bronto had a pessimistic interpretation that may or may not turn out to be valid.

Even if it does end up hashing at or above their 2TH design goal, will it ROI?   1.6 TH delivered to day is going to be worth 2.0 TH delivered 2 weeks later.   I would take the 1.6TH at the earlier date. 

If I were Cointerra I would want to avoid major failures of equipment delivered to customers.  That would be a huge headache for any company in the near term and for long term business opportunities in their technology space.

I'm going to have to agree with the 'don't worry about it yet' sentiment.  You're right; engineers generally know if they've got a brick as they bring it up...  but there's frequently a lot of room for tweaks and improvements.  I don't know if they'll reach 100% of their projections, but they're within arms reach of a reasonable noise margin only a few days after a major holiday season.

As for their communication... I haven't been following them closely, but it looked like they admitted their delays, volunteered their current under-performance, and have been drastically more transparent and professional than any manufacturer to date.

I can understand a little anxiety, but I don't think this company and product warrant the typical fidgeting we see on these forums.  Let's wait it out and see how they handle themselves.  

I'm only concerned that paid up customers get told the truth about what's going on. If a company is ahead of it's delivery date, then it's nice to be kept informed. When you're late, you need to communicate any potential delay and/or problem right away, preferably with a fix.

Ouch... guess they truly are out of the woods. Brace yourselves and make sure you remember the location of your nearest emergency exit in case of survival!

Agreed,

80% for a first boot is pretty damned good. I wouldn't be concerned yet

To be honest, there's a ton of room for improvement on these things. Take KnCMiner for example: even on the October shipments, there are (not KNC supported) methods of getting those to hash nearly at the November Jupiter rates. I agree with Joe here, getting 80% of the specified performance on a first run is pretty good.

If you can only reach 80% of your promised target you just give 20% off the next purchase to compensate. 

Appreciated, amigo. I've been 'there' before many times in my professional life and this situation is unfortunately following a classic path, engineers tend to know within the first few hours if they've got a dog. Don't get me wrong, I want their project to succeed and their customers to get what they ordered, but if this was a commercial company they were developing a system for, the customer would be on their doorstep and in their labs wanting answers.

I think you are jumping the gun, I have no skin in this game (no pre-orders) but its clear to me that slapping the chip on the board and firing it up with version 0.01 of the firmware and getting 80% of the specified performance does not suck at all.

Remember raising the performance is about firmware timing, voltage adjustments, miner tweaks and cooling (to name just a few) how much of that do you think they were able to do so far? Not much is my guess.

Wait another week before you panic.

Just my two pesos.

Just had a visit to Cointerra site where they posted, on January 14th,  that they had a fully built up box working at 1.62TH, or " a little over 80% of (their) performance target.

They also said that their engineers were "making continuous tweaks and refinements each day to improve the power efficiencies on the board to increase performance". Well, they said the engineers had one die on the package running at 132Gh/sec on the 7th of January, so four of them should meet the target spec. They would have known within hours if there was any kind of problem, which there clearly is.

Maybe I'm just cynical (and I am) but 20% seems an awfully large gap to close with 'tweaks' unless someone has seriously screwed up on a major part of the design. I haven't bought one of their rigs, but I'm sure that every one of their anxious customers would gladly sacrifice a bit more power consumption just to get their paws on their purchases. Remember that Cointerra have $1.5 million of Private Equity funding - which they obtained before the public got in on the act - and I'd hate to think that any delay had been introduced into the supply of rigs to the public customers as a consequence of the presence of this funding.

Remember that these guys are already running late and if there's still 'tweaks' to be done at this stage, the December buyers will be lucky to see their rigs before the end of January; small 'tweaks' might mean changes to printed circuit boards or different components to be sourced, either of which might be $2-3 changes but may take weeks to source and/or change.

Perhaps one or two of those customers might like to ask Cointerra some very pertinent questions about what the problems are? Don't want another Hashfast, do we?