GPU to 72 status...

[storm5510]

I do not know what the current thinking is on end-bits for wavefront TF's, and I am not going back through hundreds of pages looking. So, for now, I have it set to 75 bits. If this goes smoothly, I may increase it to 76. I am wondering if I should disable stages for mfaktc? Right now, stages are enabled and stopping after a factor is set to 0. I have had difficulties in the past with multiple stages and stopping if a factor was found. It would simply proceed to the next stage. With larger end-bit, this could make for a lot of wasted time.

[James Heinrich]

Originally Posted by storm5510

I do not know what the current thinking is on end-bits for wavefront TF's

It is actively managed with "Let GPU72 decide", then you don't need to think about it.

[chalsall]

Originally Posted by storm5510

So, for now, I have it set to 75 bits. If this goes smoothly, I may increase it to 76.

I'll let others speak to your stages question.

With regards to TF'ing depth, going to 76 is pretty much optional right now.

We're still going to 77 in 9xM, to keep that range "neat" (less than 3,000 to go). And also in front of Cats 2 and 3 because most clients there haven't upgraded to 30.3b3.

Cat 4 is really not much more than slow P-1'ers for the lower Cats, and are getting 76 already.

Edit: Oh, also, the P-1 wavefront, in 100M, is also getting 76.

[storm5510]

Originally Posted by chalsall

I'll let others speak to your stages question.

With regards to TF'ing depth, going to 76 is pretty much optional right now.

We're still going to 77 in 9xM, to keep that range "neat" (less than 3,000 to go). And also in front of Cats 2 and 3 because most clients there haven't upgraded to 30.3b3.

Cat 4 is really not much more than slow P-1'ers for the lower Cats, and are getting 76 already.

Edit: Oh, also, the P-1 wavefront, in 100M, is also getting 76.

30.3 B3. I am one behind, again. Many do not like to run CF's so that is what I do.

Every TF I have ran for the past several weeks has been in the 109-million area. The 74 - 75 combinations take 70 minutes on this 1080. i have not tried the 1650 yet, but I will. It has done really well with the 74's. I will stay with 75, for now. If things get into a "crunch" situation, then I can go to 76.

I look at the recent results on mersenne.org regularly. What I do not see is a lot of P-1 results. I do not understand how it is keeping up, especially now that the shift to all PRP's has started. I think a lot of the members have become fascinated with it because it is different now than it was; certifications.

Thank you for the reply!

[chalsall]

Originally Posted by storm5510

Every TF I have ran for the past several weeks has been in the 109-million area.

OK. That suggests you've been requesting the minimal amount of work, which is 109M 73 to 74.

Originally Posted by storm5510

Thank you for the reply!

Thank you (and many) for the work. But, seriously...

Think about the planet while you do this kinda thing.

Unless you're heating your place using a GPU instead of a space heater (or outsourcing to a hosting provider which runs RE), please don't do compute without thinking about it.

Barbados is the "Head Pin" in the Hurricane bowling lane in the Atlantic.

[LaurV]

Originally Posted by chalsall

Think about the planet while you do this kinda thing.

Man, are you ok? All cats ok in the house? Everything fine there? Hihi...

[James Heinrich]

Not sure if this is a GPU72-proxy or Prime95 thing. I just tried v30.3 for the first time and got this:

Mersenne number primality test program version 30.3
Optimizing for CPU architecture: Core i3/i5/i7, L2 cache size: 6x256 KB, L3 cache size: 12 MB
Updating computer information on the server
Starting worker.
PrimeNet success code with additional info:
GPU72 assigned P-1 factoring work.
Got assignment e9a3bce92b515f9f50f3bb4185e4xxxx: P-1 M105928607
Received unknown work type: 4.
Visit http://mersenneforum.org for help.
Will try contacting server again in 38 minutes.

[storm5510]

Originally Posted by chalsall

Think about the planet while you do this kinda thing.

I do every day, but probably not in the way you are thinking.

Originally Posted by chalsall

Unless you're heating your place using a GPU instead of a space heater (or outsourcing to a hosting provider which runs RE), please don't do compute without thinking about it.

Are you saying I need to stop running everything? I do think about what I am doing, every minute of the day and night.

[chalsall]

Originally Posted by James Heinrich

Not sure if this is a GPU72-proxy or Prime95 thing. I just tried v30.3 for the first time and got this:

It is highly advisable to *not* use the GPU72 Proxy with 30.x versions of Prime95/mprime. There's something in the new comms stream the Proxy doesn't understand, and it causes errors.

I haven't had the cycles to drill down on what's wrong, but for FC/DC work it doesn't really add anything. It's only really useful for P-1 nowadays.

[chalsall]

Originally Posted by storm5510

Are you saying I need to stop running everything? I do think about what I am doing, every minute of the day and night.

No... Just try not to burn /too/ many long-dead animals in our pursuits of the next MP... Even better if the waste heat is actually used (for example, to keep humans warm during the winter).

[storm5510]

Originally Posted by chalsall

No... Just try not to burn /too/ many long-dead animals in our pursuits of the next MP... Even better if the waste heat is actually used (for example, to keep humans warm during the winter).

I understand what you mean. I rarely ever run my GPU's at full capacity. 80% is typical. The impact on most GPU programs I run is nearly nothing. The only exception is mfaktc. Then, the performance drop-off is only around 10%. I tried it at 50%, the lowest I can take it, and mfaktc still ran above 800 GHz-d/day. These machines add about $12 USD to my utility cost each month. The "waste heat" does very little. An oil-lamp would do more.

I am seriously considering building a new rig around a 12-core Ryzen 9. The only thing stopping me is my own personal hesitation. Some of the things I run now with the GPU I could run with Prime95. It is needless to say the sticker-shock value for all this is a bit remarkable. I want this to be the last one I ever build. This is not something I enjoy doing now. It used to be. I am rapidly closing in on age 65. In a few years, or maybe a bit more, I may lose interest in participating completely, and stop. I have been participating since 2005. This forum, 2009. It has always kept my interest.