GPU to 72 status... - Page 443

chalsall · 2020-04-19, 19:25

Just saw the funniest thing in one of my log files...

Code:

[19/Apr/2020:19:22:34 +0000] gpu72.com 192.71.42.108 - - "GET /robots.txt HTTP/1.1" 200 176 "-" "Go-http-client/1.1"
[19/Apr/2020:19:22:34 +0000] gpu72.com 192.36.53.165 - - "GET /humans.txt HTTP/1.1" 301 240 "-" "Go-http-client/1.1"

Great sense of humor. I wonder if there's an RFC for this. The fetcher has quite a bit of IP space...

James Heinrich · 2020-04-19, 19:46

https://www.robotstxt.org/
http://humanstxt.org/

chalsall · 2020-04-19, 20:02

Quote:

Originally Posted by James Heinrich

http://humanstxt.org/

Cool. Thanks for the knowledge. I'd never heard about this initiative, but I like it. Logical, and potentially useful meta-data. Good place for a copyright et al notice...

Chuck · 2020-04-24, 03:48

Is SRBase and BOINC now out of the picture?

Uncwilly · 2020-04-24, 04:02

Quote:

Originally Posted by Chuck

Is SRBase and BOINC now out of the picture?

SRBase and their BOINC folks are now getting exponents directly from PrimeNet. I recently had a PM exchange with KEP. They are still turning in exponents (you can see this in the recently cleared list). There are other developments that are in the works. I don't want to speak out of school at the moment.

ixfd64 · 2020-04-24, 23:17

I'm disappointed to learn that GPU to 72 no longer supports BOINC due to a service mark dispute. Hopefully this won't mean the end of BOINC integration for GIMPS as a whole.

chalsall · 2020-04-24, 23:39

Quote:

Originally Posted by ixfd64

I'm disappointed to learn that GPU to 72 no longer supports BOINC due to a service mark dispute.

To be clear, I'm more than happy to continue to support Reb's BOINC efforts. However, GPU72 doesn't have the type of work available which they want (72 to 73 work in the higher ranges, which won't be useful for years...). The specialized API I built for them still exists, if they ever decide they'd like to do deeper work again.

Quote:

Originally Posted by ixfd64

Hopefully this won't mean the end of BOINC integration for GIMPS as a whole.

They're now getting work directly from Primenet. And, the REAL GPU72 BOINC system is in early alpha...

Mark Rose · 2020-04-25, 00:45

Exciting times in the TF world!

(I've read the locked thread)

chalsall · 2020-05-09, 01:01

So, it looks like we're going to be OK with regards to TF'ing firepower. So...

By the request of some (but particularly Uncwilly), I have imported some candidates in the 332M range (AKA the 100M digit candidates) for some quality TF'ing time. Thanks to George for expiring a bunch of /really/ old assignments.

For those who want to work there, the LLTF Assignment Form again has the LMH Bit-first and Depth-first options. The former goes to 76 bits, while the later goes to 78.

I'm currently testing to ensure that the Colab code-paths will handle this OK. So, currently, a couple of my instances, and all of Uncwilly's, are doing 332M work. Tomorrow I'll update the Instance Assignment form to let people opt in to do this work type.

Which brings up a question: just how deep to people want to take these? 78 isn't "optimal", but a 77 to 78 run will take about two hours (on a P100). Should I have a Breath first (75), Nominal (78), and Deep (81?).

Feedback appreciated.

kladner · 2020-05-09, 01:47

The idea of ripping through TF is attractive, but I'm not really drawn to the upper reaches. On the other hand, I am mostly into diddling Colab for all the GHz-d I can get.

I don't care what GPU72 deems necessary as long as it advances the main project. What makes Sense, or Let GPU Decide. Just say.
I will note the saying, "If it's free it's not the product. You are."
But, I just signed off my paid account, with 4 P100s, called up a free account and got 2 T4s. This is approximately equal to 3 P100s. I usually stop this kind of run after 5-6 hours, and line up the paid account to run overnight.
3 free accounts and one paid let's me keep as much running as I have time to set up. I can always find at least one that will run with GPUs.

Uncwilly · 2020-05-09, 04:55

Quote:

Originally Posted by chalsall

I'm currently testing to ensure that the Colab code-paths will handle this OK. So, currently, a couple of my instances, and all of Uncwilly's, are doing 332M work.

I requested a manual one for my integrated GPU. And I have set MISFIT to target that range (I hope I got it right.) With the T4's I have at the moment, 75->76 is taking ~21 minutes per. Hoping to find a factor or 2 (still running ~9% behind expected).

2020-04-19, 19:25	#4863
chalsall If I May "Chris Halsall" Sep 2002 Barbados 23047₈ Posts	Just saw the funniest thing in one of my log files... Code: [19/Apr/2020:19:22:34 +0000] gpu72.com 192.71.42.108 - - "GET /robots.txt HTTP/1.1" 200 176 "-" "Go-http-client/1.1" [19/Apr/2020:19:22:34 +0000] gpu72.com 192.36.53.165 - - "GET /humans.txt HTTP/1.1" 301 240 "-" "Go-http-client/1.1" Great sense of humor. I wonder if there's an RFC for this. The fetcher has quite a bit of IP space...

2020-04-24, 23:17	#4868
ixfd64 Bemusing Prompter "Danny" Dec 2002 California 2395₁₀ Posts	I'm disappointed to learn that GPU to 72 no longer supports BOINC due to a service mark dispute. Hopefully this won't mean the end of BOINC integration for GIMPS as a whole. Last fiddled with by ixfd64 on 2020-04-24 at 23:53

2020-05-09, 01:01	#4871
chalsall If I May "Chris Halsall" Sep 2002 Barbados 9,767 Posts	LMH TF -- Back by popular demand! So, it looks like we're going to be OK with regards to TF'ing firepower. So... By the request of some (but particularly Uncwilly), I have imported some candidates in the 332M range (AKA the 100M digit candidates) for some quality TF'ing time. Thanks to George for expiring a bunch of /really/ old assignments. For those who want to work there, the LLTF Assignment Form again has the LMH Bit-first and Depth-first options. The former goes to 76 bits, while the later goes to 78. I'm currently testing to ensure that the Colab code-paths will handle this OK. So, currently, a couple of my instances, and all of Uncwilly's, are doing 332M work. Tomorrow I'll update the Instance Assignment form to let people opt in to do this work type. Which brings up a question: just how deep to people want to take these? 78 isn't "optimal", but a 77 to 78 run will take about two hours (on a P100). Should I have a Breath first (75), Nominal (78), and Deep (81?). Feedback appreciated.

2020-05-09, 01:47	#4872
kladner "Kieren" Jul 2011 In My Own Galaxy! 2×3×1,693 Posts	The idea of ripping through TF is attractive, but I'm not really drawn to the upper reaches. On the other hand, I am mostly into diddling Colab for all the GHz-d I can get. I don't care what GPU72 deems necessary as long as it advances the main project. What makes Sense, or Let GPU Decide. Just say. I will note the saying, "If it's free it's not the product. You are." But, I just signed off my paid account, with 4 P100s, called up a free account and got 2 T4s. This is approximately equal to 3 P100s. I usually stop this kind of run after 5-6 hours, and line up the paid account to run overnight. 3 free accounts and one paid let's me keep as much running as I have time to set up. I can always find at least one that will run with GPUs. Last fiddled with by kladner on 2020-05-09 at 02:09

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Status	Primeinator	Operation Billion Digits	5	2011-12-06 02:35
62 bit status	1997rj7	Lone Mersenne Hunters	27	2008-09-29 13:52
OBD Status	Uncwilly	Operation Billion Digits	22	2005-10-25 14:05
1-2M LLR status	paulunderwood	3*2^n-1 Search	2	2005-03-13 17:03
Status of 26.0M - 26.5M	1997rj7	Lone Mersenne Hunters	25	2004-06-18 16:46

2020-04-19, 19:46	#4864
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 11·311 Posts	https://www.robotstxt.org/ http://humanstxt.org/

2020-04-24, 03:48	#4866
Chuck May 2011 Orange Park, FL 3·5·59 Posts	Is SRBase and BOINC now out of the picture?

2020-04-25, 00:45	#4870
Mark Rose "/X\(‘-‘)/X\" Jan 2013 5562₈ Posts	Exciting times in the TF world! (I've read the locked thread)