Folding Forum

Yep, F@H just crossed 40 PetaFLOPS (viewtopic.php?f=16&t=26054) making it faster than the fastest Supercomputer

Adam A. Wanderer wrote:Yes, but F@H seems to work only on a single type of problem, whereas a supercomputer can be reprogramed to work on several different types of problems.

And conversely, the Chinese computer wouldn't do a very good job running FAH without a lot of software changes -- though it woud be interesting to try it.

It's iteresting that the article mentions Hadoop and not FAH. They're both specialized alternatives to "standard" supercomputers (if there is such a thing).

As per here: http://en.wikipedia.org/wiki/TOP500
Lots of intel and Linux, should be relative easy to fold on some CPU-slots ... I would offer my passkey ...

Google has given us access to pretty dramatic resources. We can't publicly comment just how much, but it's definitely big. There's a reason they call it "Exacycle."
https://folding.stanford.edu/home/key-n ... chemistry/

PS I should mention that the Google resources don't show up on the FAH osstats page.

Dr. Pande, if the Google contributions were included, what would be your estimate for the additional PetaFLOPS Google contributes above the 40 PetaFlOPS recently achieved? I always appreciate it when you weigh in on topics on the forum. Thank you for taking the time.

NookieBandit wrote:Dr. Pande, if the Google contributions were included, what would be your estimate for the additional PetaFLOPS Google contributes above the 40 PetaFlOPS recently achieved? I always appreciate it when you weigh in on topics on the forum. Thank you for taking the time.

Sorry, but he already said can't answer your question.

VijayPande wrote:Google has given us access to pretty dramatic resources. We can't publicly comment just how much, but it's definitely big. There's a reason they call it "Exacycle."

If he can't disclose X, then it doesn't make sense for him to be able to disclose (40+X).

Actually, Bruce, I think you're trying to be helpful but, first, my question was to Dr. Pande, and second, your answer did not address my question.

The question to Dr. Pande specifically referenced the increase in incremental aggregated power contributed, not categorized by osstats. His "p.s." clarified his position relative to a restriction on any public announcement that such osstat breakdown would not be forthcoming. However, his clarification did not suggest commenting on an incremental aggregated PetaFLOP power increase was confidential/proprietary information.

If Dr. Pande chooses not to answer, that's his decision. The question can hang out there unanswered forever, which in itself is the answer.

It's entirely possible that Google could be collaborating in such a way that they've got their own assignment/work servers which are not hooked up to the stats system. This would be relatively simple to implement logistically and would keep things isolated and protected, which is likely favorable to Google's sysadmins and security crew.

I should also mention that this is not the first time that they have collaborated with Google. In the early years of F@h, Google co-founder Sergey Brin launched an add-on for the Google Toolbar called Google Compute, which according to one source tripled the number of machines running F@h. Google Compute ended in late 2005, but I am excited to see them Google contributing again nine years later.

In that (commercial) context I also see the NaCl version ... At least partially. Still good to have for part-time folder and as easy "entry drug"

Thanks Google! First they bring us chrome , gmail, etc, then more processing power for F@H. Google has no end to how cool it is. "Exacycle" Cool name too. Why the exa? Is it capable of many Exaflops Coming from Google I would not really be that surprised. Plus, why name it Exacycle if its only a PetaCycle? Google did just buy Motorola recently, which is one of the largest IC producing companies on earth....

"Scratches head in wonder"