ASICs and BOINC

[AMDave]

I agree. I had the same thoughts when the ASICS kicked in. I pulled my machines from the project.
No scientific benefit that I can see and no contest. It just turned into a BOINC credit-fest for ASICs.
Note that this project certainly has a place in providing funding to start-ups, just not a place in the points for distributed computing for scientific advancement.
We had the a similar quandry with the AMD GPU contributions to Donate@home in the hope that further client development would follow, which resulted in naught. Ripped off.
That's not to say there's a level playing field in hardware support in any group of projects. Far from it.
But I don't see any other application that these ASICS can perform that can contribute scientifically.
That would require a different ASIC to be developed.
And that raises another overdue development change ... why are there no ASICs for the science applications that can actually make a difference?

[Dirk Broer]

For example, steVe did a quarter of my entire BOINC career's worth of cobblestones in 1 day, that's right In 4 days he generated as many BOINC points as I have accumulated running up to 30 machines in some years and about a dozen GPUs in recent years over at least the last decade. Sort of makes the whole "game" pointless.

It's worse than that: on the 5th of July he managed 1,068,285,000 credits on BU (FreeDC)
On the bright side: most BU crunchers do not crunch at 100-odd other projects, so we can still outshine them in (Mega) Milestones, Badges, Trophies (did I hear SkyNet POGS?)...

[Dirk Broer]

And that raises another overdue development change ... why are there no ASICs for the science applications that can actually make a difference?

A question I have asked before in different words. We should start with porting BOINC to a FPGA (field-programmable gate array).
Once you get a project running through a FPGA it can be miniaturized into an ASIC -provided the chip maker sees profit in it, or if you have enough cash/(bitcoins?).

[AMDave]

Peaked your interest, eh?
Yes FPGAs are the first step (and re-useable)
Here's a blog note with a couple of relevant links: http://www.fpgaathome.org/
There are other FPGAs templates have been deployed for some other science apps
FPGA has also been discussed for WCG and FaH.
AFAIK some of the universities have deployed FPGA arrays in the past to supplement the progress of their projects but it hasn't become a "thing"
I haven't looked into that solution as much as I could have, so there may be some hurdles and plateaus encountered in that method, but surely the logical next step is a hardware specific ASIC.
Now that there are numerous ASIC manufacturing paths (albeit for BTC and DGC) they could be leveraged for the well established DC clients.
There is a short window of opportunity on that and it "should" be tested.
I had heard a rumour, in the past, of one project purchasing ASICs but lacking the upgrade path of FPGAs they became obsolete.
More than several years down the track, having established that GPUs are not the 'panacea' of scientific DC, it's worth looking at the FPGA & ASIC solution paths again.
I feel that the FPGA solution path has been under utilized and I should know more about it than I do.

[Nflight]

Nice Link AMDave, Looks like utilization of the unknown opportunities that exist but are not being felt in the DC world could use a boost.

[Dirk Broer]

Not being felt in the DC world?

[vaughan]

Well I'm maxed out at around 10M a day. One quad core machine just refuses to run Bitcoin Utopia so its assigned to Asteroids and its GPU is on PG PPS sieving duty.
Unsure whether to get some of these ASICS USB stick thingies or to just give up on BOINC totally, the cobblestones race has slanted so far to Bitcoin Utopia its not a level playing field anymore. I'm sure I could do some more NPLB tasks. I'll think about it.

[Dirk Broer]

There seem to be USB stick thingies (Neo Fury) that can get 11 GH/s. There's even an Australian ebay seller for it

• No. of chips: 6 Bitfury chips,
• Hashrate: ~11GH/s out of the box, up to 15GH/s overclocked,
• Power usage: ~0.86W per GH,
• Dimensions: 26x60mm without usb plug,
• Voltage: 5V
• Power requirement: 2-2.5A

The daily score of one of these thingies will be about as great as the total of your combined CPU/GPU fleet.
The same output can be reached by using 2 to 3 Bi-Fury sticks, which are now dumped by the miners -at least over here.
If you have 10 PCs and stick one bi-fury in each you may reach 50M a day on BU alone -and that's beside the 10M for your original projects that you can still run on them.

[Nflight]

Thanks for the links there Dirk, If i had some cash maybe I would. But right now the Bitcoin Utopia realm os over for me. Seems the sgminer.exe that runs in the background is now considered a Trojan horse and my anti-virus won't let me exclude this from the permit-table operating equipment. Rats

[alez [ TSBT Pirate ]]

I have found a link from Berkley which will not open at this time, may be too many people reading the site! Who knows...

The Link is here: But it was here that I was looking for aid from App_config.xml files from the internet.

I went digging to optimize the GPU once everyone finds the niche to get those pesky GPu's working. The LINK is here; Seems the person at the Scottish BOINC team found a niche accelerator in his attempt to secure quicker and more efficient means to push the output of the GPU he owns. This looks promising will take a look after i return from my meeting this morning and reply with my findings.

Nflight, sorry this optimisation only works on Einstein due to the unique way it works. Bitcoin has extremely low cpu usage to feed the GPU so you would simply be wasting cores. Further due to a bug / not full development of the GPU app it works on time not Flops so effectively every GPU regardless of power gets payed the same credits.