So it's not about the money, or about being open source, or about shipping more than 2 units. So what is Avalon about?

Still, I hope that batch 2 ships on time. ASIC competition is important.

Does anyone know if it's even possible for hard copies of FPGAs to get that sort of performance? Wouldn't Avalon have done that if it would have worked?

When they finally release sources, this should be pretty easy to fix.

They've only had test units for a few weeks. They have no idea if they will still be working in a month; if they fail on a large scale then warranties would kill their profits.

An "Avalon unit" includes every component in the box with fans. The contents of that box are irrelevant. If they say it uses 450W, and it uses 620W, then they are fucking liars. Plain and simple. "It's the power supply" is no more relevant than "it's the ASICs". The simple fact is that the shipped unit does not work as described.

If the chips themselves use 450W, then that is of academic interest only, since the "Avalon unit" sucks far more than that from the wall.

Well, I will agree that fixing this mess is more important right now, but they did have many days before the sale to get things right...

Your website still shows 450W when it has been measured at 620W+ at the socket. Sorry, but you really fucked up the specs, or at least lied by omission. You're selling complete units here, not individual ASICs. You need to give the specs of the complete units or you're a lying piece of shit. Attacking BFL for possibly not meeting specs when you already failed to meet specs and still advertise the wrong specs is just retarded.

Fun fact: BFL's order form still works perfectly and:

• Doesn't mess up the price
• Doesn't mix up the shopping carts
• Doesn't hand out others' personal information

It's just ridiculous to take a swipe at them at every opportunity when you're fucking up even worse than they are in everything but shipping dates.

An OpenWRT image was meant to be released (and it would need sources) a month ago. It never happened. Avalon is actually less open than BFL right now (since BFL released a draft protocol spec). Pretty disappointing.

Looks like it's $500. So $2000 for 90GH/s. Same cost and H/s as a BFL single + little single + raspberry pi to run them, and BFL's two units still take up less space and use less power than the non-upgraded Avalon.

If BFL ships, Avalon will be just barely keeping up with BFL in H/$ and still way behind in H/J.


Yes, I'm still thinking along the same lines. (If it wasn't obvious, I meant "increase mining without making the power usage even worse".) They've effectively allowed you to buy 1/3 of a second unit. This is useful for people who have $2000 but not $3000, space for 4U but not 8U, and free power. "Serious" miners will be getting multiple units anyway (and 4 normal units is as good as 3 upgraded units).

The only undeniable advantage is that Avalon has shipped and BFL has not (and that's a pretty fucking big one), but this has nothing to do with the magic feature.

You could even re-use the power supply and the control board!

Puerto, you're right. I was wrong. The Avalon is modular enough that you can throw out or sell those Avalon chips and put BFL cards in them! That is a killer feature.

Kinda looks like Tom or comproTom (aka drunk Tom, the supposedly compromised account that stirred up some drama recently).

Agreed, but it was just too tempting to turn Puerto's statement around.

It depends on how Avalon's batch 2 goes, how BFL goes and how ASICMINER goes. I'd rather not.


Of course they can do that, if they've paid off their initial NRE and have $500k? to spare for moving to a smaller process. The result will be a product similar in efficiency to BFL products, but most likely months after BFL has already met most of the demand.

I thought the mini rig was already customisable/upgradeable, meaning that Avalon copied BFL.

Um, did I say they were in a good position for the race if they're losing the race? That would be stupid.

Edit: I should clarify. The race is obviously over (Avalon won), which means ASICs are now in the ramp-up phase of the curve (like GPUs right after the first OpenCL miner), but when the curve flattens off and 60GH/s units are normal, it will no longer make sense to use units that are not power efficient. BFL has obviously been targeting the flat-curve phase since they went with 65nm. Avalons will not sell well in that phase.

I think they put them in exactly the spot where they expected to be - a pretty good one if their power usage will be as low as expected.


Agreed.

I had a feeling that you would defend whatever the magic feature was when released, then fabricate some numbers to make it sound amazing. Adding extra modules to bring an Avalon to 90GH/s is roughly the same as buying a little single to add to your 60GH/s single to get to 90GH/s...Except that the Avalon will cost $1500 to start with, we don't know what the add on modules will cost, it still takes up more space, and the power usage is STILL going to be much higher.

I predicted before that the feature would not be enough to overcome the difference in power usage or make people suddenly cancel their BFL orders. Was I wrong?

Except that "the job" is dissimilar. Avalon's goal was to produce a small number of units with an old, low-risk process. BFL has been trying to produce much more efficient units that will remain profitable for a long time even after there is competition, and in a huge quantity.

The time frames are also hard to compare, unless Avalon funded all development with pre orders (and putting all customer pre order money at risk like that would be ridiculous).

Your argument is effectively "company B should be ashamed because company A did something different in a different time frame".

Why did you make this thread? Why are you so obsessed with complaining about BFL?

To a casual observer, many of your posts appear to be more than a little bit misleading. If your statement is true (considering that, statistically, liars say it more often), then it is so only by technicality.

You can get regulators (even single-chip) that go much higher than that. I suppose they wanted to try and cut costs by rolling their own, once they realised that powering their 300W monstrosity would not be cheap. Then they decided to not buy the parts months in advance.