GPU to 72 status... - Page 35

chalsall · 2012-01-26, 17:21

Hey all.

I've made a small change to the individual worker's graphs, and would appreciate feedback.

For those of us who have both GPUs and CPUs, the work being done by the CPUs was effectively hidden on the graphs because of the incredible speed of the GPUs. An example is our current leader in LL TF, Craig.

So I've now made the graphs "dual axis". GPU class work (LL and DC Trial Factoring (the green and red lines on the graphs)) are related to the left-hand scale. CPU class work (P-1, LL and DC) is now related to the right-hand scale.

And, yes, I know that some are using their GPUs to do LL and DC work -- the graphs do not take this into account.

Do people like this, or hate it?

And, yes, I also know I need to remove those classes of work for those who are not doing said work so the unneeded scale doesn't show up. That's a bit of work, though, so I wanted feedback before I keep this modification. (And, yes, I also know I need to start adjusting the X axis so the dates don't overlap for long-term users.)

KyleAskine · 2012-01-26, 17:53

Quote:

Originally Posted by chalsall

Hey all.

I've made a small change to the individual worker's graphs, and would appreciate feedback.

For those of us who have both GPUs and CPUs, the work being done by the CPUs was effectively hidden on the graphs because of the incredible speed of the GPUs. An example is our current leader in LL TF, Craig.

So I've now made the graphs "dual axis". GPU class work (LL and DC Trial Factoring (the green and red lines on the graphs)) are related to the left-hand scale. CPU class work (P-1, LL and DC) is now related to the right-hand scale.

And, yes, I know that some are using their GPUs to do LL and DC work -- the graphs do not take this into account.

Do people like this, or hate it?

And, yes, I also know I need to remove those classes of work for those who are not doing said work so the unneeded scale doesn't show up. That's a bit of work, though, so I wanted feedback before I keep this modification. (And, yes, I also know I need to start adjusting the X axis so the dates don't overlap for long-term users.)

I love it!

chalsall · 2012-01-26, 18:06

Quote:

Originally Posted by KyleAskine

I love it!

Thanks for the feedback.

And you're one of the best examples, as you've completed work in all five assignment types.

Anyone against?

It took me two minutes to do this, but removing an axis for those of us not blessed with a GPU (myself included) is about an hour of work.

Dubslow · 2012-01-26, 18:26

Jeesis, who took a 45M expo to 76 bits?

Well, at least the P-1 will be easy

My only comment is that especially for the LL/DC work, the graphs are particularly granular due to the size of the individual work unit. Also, perhaps have a set ratio of the right hand scale to the left hand scale, so that it's easy to compare my ratio of CPU/GPU work to someone else's ratio of CPU/GPU work. Does that make sense?

Note: I missed there was a next page. I've also done all 5 worktypes -- but not concurrently. See my graph to see how this makes everything blocky.

James Heinrich · 2012-01-26, 18:39

Quote:

Originally Posted by chalsall

Do people like this, or hate it?

This particular people loves it!

My P-1 (CPU) and TF (GPU) graph lines are now perfectly aligned so it's very easy to see how both are doing relative to my normal output.

chalsall · 2012-01-26, 18:51

Quote:

Originally Posted by Dubslow

My only comment is that especially for the LL/DC work, the graphs are particularly granular due to the size of the individual work unit.

Yes. That is the nature of the beast -- results are only turned in every few days (unless you have a huge farm) for those work types. The graphs are showing exactly what is true.

Quote:

Originally Posted by Dubslow

Also, perhaps have a set ratio of the right hand scale to the left hand scale, so that it's easy to compare my ratio of CPU/GPU work to someone else's ratio of CPU/GPU work. Does that make sense?

Yes, it makes sense. But no, (IMHO) it doesn't make sense. Does that make sense?

When comparing one worker against another, the comparative graphs for each work type should be used. If you're comparing two worker's individual graphs, map the Y axis in your head.

kladner · 2012-01-26, 19:14

Another thumbs up here.

Dubslow · 2012-01-26, 20:11

I meant that if one line is worth 10,000 GPU GD on a graph, then the equivalent CPU GD value should be the same for everybody's graph.

E.g., compare me and James' graph. They superficially look the same, unless you notice that 7500 GPU GD is 1000 CPU GD for him, whereas for me 7500 GPU GD is 250 CPU GD. Then you realize that proportionally, he has a higher CPU/GPU ratio than I do, but you only notice this from looking at the scale, not at the graph. Thus I propose that 7500 GPU GD should have the same CPU GD value for everybody. (Also note that figuring out what ratios work best for everybody is tantamount to figuring out that 'scaling' factor we discussed a month ago, except that this time, we have a criterion, namely that all the graphs should look reasonable.)

...Does that make any more sense?

chalsall · 2012-01-26, 20:45

Quote:

Originally Posted by Dubslow

Thus I propose that 7500 GPU GD should have the same CPU GD value for everybody.

But not everyone has the same scaling.

Some have only GPU class work; some only CPU class work.

For each worker, the scales are their own.

I say again, when comparing individual workers' graphs scale the Y axis in your head.

What do you think we have to do when comparing Barbados GDP against (for example) the US of A's GDP?

Quote:

Originally Posted by Dubslow

...Does that make any more sense?

No.

garo · 2012-01-26, 20:56

I took a hiatus from GPUto72 and did some LL TF on exponents I collected from Primenet. Now when I want more work it says: Computer says no!

I can wait a few days and it will probably allow me to get more work then but just wondering how you make the hoarding judgment? Thanks for all the good work. Boy am I glad I talked you into it

chalsall · 2012-01-26, 21:22

Quote:

Originally Posted by garo

...but just wondering how you make the hoarding judgment?

I don't; George does.

There's lots available to TF if you're interested....

2012-01-26, 17:21	#375
chalsall If I May "Chris Halsall" Sep 2002 Barbados 9,767 Posts	Change to the individual Worker's graphs... Hey all. I've made a small change to the individual worker's graphs, and would appreciate feedback. For those of us who have both GPUs and CPUs, the work being done by the CPUs was effectively hidden on the graphs because of the incredible speed of the GPUs. An example is our current leader in LL TF, Craig. So I've now made the graphs "dual axis". GPU class work (LL and DC Trial Factoring (the green and red lines on the graphs)) are related to the left-hand scale. CPU class work (P-1, LL and DC) is now related to the right-hand scale. And, yes, I know that some are using their GPUs to do LL and DC work -- the graphs do not take this into account. Do people like this, or hate it? And, yes, I also know I need to remove those classes of work for those who are not doing said work so the unneeded scale doesn't show up. That's a bit of work, though, so I wanted feedback before I keep this modification. (And, yes, I also know I need to start adjusting the X axis so the dates don't overlap for long-term users.)

2012-01-26, 18:26	#378
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 1110000110101₂ Posts	Jeesis, who took a 45M expo to 76 bits? Well, at least the P-1 will be easy My only comment is that especially for the LL/DC work, the graphs are particularly granular due to the size of the individual work unit. Also, perhaps have a set ratio of the right hand scale to the left hand scale, so that it's easy to compare my ratio of CPU/GPU work to someone else's ratio of CPU/GPU work. Does that make sense? Note: I missed there was a next page. I've also done all 5 worktypes -- but not concurrently. See my graph to see how this makes everything blocky. Last fiddled with by Dubslow on 2012-01-26 at 18:44

2012-01-26, 20:11	#382
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3·29·83 Posts	I meant that if one line is worth 10,000 GPU GD on a graph, then the equivalent CPU GD value should be the same for everybody's graph. E.g., compare me and James' graph. They superficially look the same, unless you notice that 7500 GPU GD is 1000 CPU GD for him, whereas for me 7500 GPU GD is 250 CPU GD. Then you realize that proportionally, he has a higher CPU/GPU ratio than I do, but you only notice this from looking at the scale, not at the graph. Thus I propose that 7500 GPU GD should have the same CPU GD value for everybody. (Also note that figuring out what ratios work best for everybody is tantamount to figuring out that 'scaling' factor we discussed a month ago, except that this time, we have a criterion, namely that all the graphs should look reasonable.) ...Does that make any more sense? Last fiddled with by Dubslow on 2012-01-26 at 20:11

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

2012-01-26, 19:14	#381
kladner "Kieren" Jul 2011 In My Own Galaxy! 2×3×1,693 Posts	Another thumbs up here.

2012-01-26, 20:56	#384
garo Aug 2002 Termonfeckin, IE 2764₁₀ Posts	I took a hiatus from GPUto72 and did some LL TF on exponents I collected from Primenet. Now when I want more work it says: Computer says no! I can wait a few days and it will probably allow me to get more work then but just wondering how you make the hoarding judgment? Thanks for all the good work. Boy am I glad I talked you into it