Topic: Acorn M.2 FPGA based GPU Accelerator - page 20. (Read 73507 times)

and the mining industry dies or goes to where it was going in the first place larger farms that serve as power plant load stabilizers .

because the bottom foundation to mining the gamer/miner is killed off in that scenario above that the two of you believe in.

i think the future may be along the line of if a new algo comes along it goes like this:

1. most every gpu on the planet jumps in. because easy to code, so quick to make for public miners

2. then, very fast, fpga assisted kick a lot of stand alone gpus out. slightly harder to code vs gpu alone, but could provide decent boost and power savings

3. then, probably fast again, larger more capable fpgas kick most or all gpu/gpu assisted.  because probably harder to find suitable fpga boards and code for them vs gpu assist alone

4. then asic kicks out fpga. but takes a while to design/manufacture.

possibly skip #2 depending on algo

rinse and repeat.

These numbers have not changed since June

they have boost estimates here (this is 215+ but i believe they have a similar chart for the others)

http://squirrelsresearch.com/acorn-cle-215-plus/ i dunno if they have changed since july.

if they are having problems with these at this point i would think they would have to admit it soon ^tm

that stuff from gpuhoarder is self admitted armchair discussion of where the mining scene is now. basically he is saying (my take, maybe wrong) on some coins/algos acorns may be better stand alone vs gpu assist. lotta variables.   

Here is what GPUHoarder said yesterday:

"GPUHoarderToday at 6:57 PM
So here’s how Acorns went - we built the first Hand Built prototype boards, of which 3 actually functioned (1 was later broken before it got much life) early in the summer. A combination of the two remaining prototypes and AC701 boards were used to prove out functionality - including ethash, CN, Lyra2re2 and X16r algo support bitstream.

GPUHoarderToday at 6:59 PM
We proved out and metered exactly how much state bandwidth we can get on/off acorn in full duplex, and the performance and power of all the bitstreams, as well as area on chip.
That was used to generate test harnesses for miners,
For x16r, we can’t actually test every algo combination - so we timed exact performance on all the GPU algorithm pairs vs acorn - math is applied to get what the average will be, and specific blocks of course tested


"

It was in a sense a response to questions about how they got the numbers (simulation only?) in June but they cannot come up with working SQRL Miner now.

So if they made them up  and believed they could write software to achieve those numbers it has now created an interesting problem for them.

Have to say gear looks nice.


Seems to be they should be able to show us  those  estimates  were real right now  and they don't.

Frankly that =  "bad news"

I forgot to add the information I talk about X16R ... not ETH. Sorry for the missing info. Pill don't have an effect on X16R because of the low RAM dependency.
All cards (1050 Ti also for example) have seen a very large hashrate improvement since the promises about increased X16R hashrate with acorns was released.
The question is, how is the improvement today?

ETH different story.

Exactly. Back in june for example a 1080 Ti was able to reach 15-18 MH/s, depending on PL. Now we are at 24-28 MH/s.
No information if the Acorn FPGA is able to generate the same percentage of improvement or if the software miner were able to close the gap "a little".

Among other things, he said:

"
GPUHoarderToday at 6:52 PM
The thing with GPUs was that for GPUs the ratio of power to revenue -> profit is much higher, so as revenue squeezed profit squeezed to nothing much faster than standalone(edited)
GPU Assist squeezed less, but still much more than standalone
GPU Assist is awesome, and a support nightmare is what I’m anticipating

GPUHoarderToday at 7:08 PM
So I.e. the bitstream we used for testing that can only be flashed over jtag, and can’t be updated in field isn’t very helpful. we also initially early on used three different open source miners with hacked in acorn support for testing and then refining our kernels, also not releasable easily.

GPUHoarderToday at 7:09 PM
Coin X makes $1.00 a day, and costs $0.50 in power on a GPU. Coin Y makes $0.30 a day on Acorn stand-alone
Acorn makes coin X power drop down to $0.20/day - profit goes from $0.50 to 0.80.(edited)
Stand alone same profit as GPU assist
Now half revenue(edited)
Coin X is now break even on GPU
With assist, makes $0.30 a day,
Coin Y still makes $0.15 day
Half it again

Now now coin X makes $0.05/day with assist, and Coin Y makes $0.075


GPUHoarderToday at 7:14 PM
That’s the scenario we are in now - power is such a large component of GPU coin and such a small component of FPGA coin

.................


GPUHoarderToday at 7:41 PM
As soon as it is profitable it gets driven to edge because so many GPU operations are desperate for profit
GPUHoarderToday at 7:41 PM
And there are now a large number of FPGA ops just scanning the horizon constantly for new opportunities



"

Acorn was to be released in August. Whenever there were questions about SQRL Miner (to test it even without Acorn) the answer was that it would be released with Acorns. It is already December 5. Hardware without software is useless. By the way, the numbers illustrating the advantage of using Acorn, were released in June, I believe.

I think the above comments are a little overheated:

1) There are many complex engineering projects that end up behind schedule because of the technical challenge, software/hardware integration issues, suppliers and or money. Examples of projects having no shortage of money and people are the F-35, the large hadron collider and the James Webb space telescope. They were/(are) overbudget and overschedule. This is just the tip of the iceberg.

2) In terms of selling hardware to the public before the there was much software, I believe this was an IBM specialty.

3) There is, in the short history of crypto, a track record of some software whiz squeezing out another 25% - 50% in performance out of the hardware. The 1080ti being the latest example.

Right now is a good time to prepare.

What did he say?

You don't rush selling hardware to the public without the applicable software; especially when there were delays that should have given additional time to develop software lol...

Well, this is a forum for GPUHoarder's Acorn (not any FPGA) so what he plans to do is relevant.

[/quote]

Perhaps you should look at what GPUHoarder was saying on sqrl-acorn Discord starting at 6:52 pm on Dec 4 (today)
[/quote]

.....
yeah this shit again. ctrl c / ctrl v

Perhaps you should look at what GPUHoarder was saying on sqrl-acorn Discord starting at 6:52 pm on Dec 4 (today)

It's pretty clear from the descriptions of the product that its primary use is to boost gpus, not standalone mining. I'm sure standalone mining will be one application, but it was designed as an adjunct to gpus in order to handle compute-heavy work at much lower power usage. So the overall effect is generally an improved hash/watt. I do wish we had real benchmarks.

Your vast experience aside, what is the main benefit of Acorn, in your mind? I mean gains of course. Is there even a concept of ROI for Acorn or ROI is only for newbies?
I do not think that any one who is mining at home can even remotely consider paying the prices, you mentioned when current high cost of electricity and pessimism of the market are taken into account. Keep in mind the price of Z9 mini when it was release - this, together with its gains in Summer, is a reference point for many (regardless if you consider it to be unreasonable or not)

Acorn does not "have a low price". It is currently at least 50% more expensive than RX580 and will not earn more (even with the low power usage) than this GPU. GPUs can be sold and I have my doubts about Acorn. I am done with GPU mining and mentally out of mining all together. I do not believe in this illusive big rebound of crypto. There will be more and more POS coins and centralization will creep in.

You said word for word what I did and thought/think