Best work choice for these machines

garo · 2020-04-29, 15:07

So I have a few boxen that are either quite slow - think Core 2 6600 - or are on only intermittently - laptops etc. What is the best work choice for these?

I currently run a combination of DC/P-1 and TF. I find with P-1 the older boxes really struggle as they have very limited memory - hard to spare even 1GB for P-1 without interfering with the other things. With DC I am not finding it hard to complete work before expiry as the exponents get larger some of these boxes can take months to complete work and the exponent can become Cat 1 or even Cat 0 by then.
Any suggestions?

Uncwilly · 2020-04-29, 15:16

How about PRP on cofactors? That is not cutting edge mainline items. But it is an available work type. Memory is not much of an issue. No milestones to have to worry about.

garo · 2020-04-30, 09:34

Thanks. I tried that. But on a two core Core2 6600 when I try with the latest Prime95 version it insists that it needs two cores for the job. Is that correct? I want to let P-1 run on the other core and I keep getting a warning about wanting to use more cores than exist.

kruoli · 2020-04-30, 13:45

Quote:

Originally Posted by garo

Thanks. I tried that. But on a two core Core2 6600 when I try with the latest Prime95 version it insists that it needs two cores for the job. Is that correct? I want to let P-1 run on the other core and I keep getting a warning about wanting to use more cores than exist.

I guess that's a bug in the GUI. I have not yet found a solid way to reproduce it regularly, but sometimes I have the issue in the worker window, when changing the amount of threads per worker and leaving the text box, then it will revert to the old value even if it is nonsense.

My suggestion: Accept the warning and edit local.txt manually, that should work.

Mark Rose · 2020-05-01, 22:16

If you'll continue doing DC, I'd say make sure both cores are assigned to a single worker. You could also manually grab high DC, which won't be a lower cat for many years.

garo · 2020-05-03, 20:59

Thanks both for your answers. I will try editing local.txt

@Mark Rose: That's a good suggestion. These machines are somewhat bandwidth limited with only DDR2-666/800. I assume using two cores for a single worker will more or less halve the iteration time?

garo · 2020-05-06, 10:23

Following up on this. I ran benchmarks on two of the newer laptops we have in the house. I haven't run Prime95 on them but it is amazing how much faster they are both absolutely and relative to clock-speed. I compared i5-8365U and i7-7500U laptops to the Haswell 4770 and they are both faster at P95 despite significantly slower clock speeds. I usually don't run P95 on laptops due to thermal issues but I'm very tempted to run something on these boxen.

paulunderwood · 2020-05-06, 10:53

I haven't benchmarked it, but the cheap i3 9100f (no graphics, 65w tdp, lga 1151 v2) I purchased recently really flies. Reminder to self: set up the XMP.

garo · 2020-05-06, 15:12

Nice. I don't really want to buy new hardware but I will keep this in mind. THough I think if I had to spend money on hardware I would probably get a Radeon card.

paulunderwood · 2020-05-06, 18:52

Quote:

Originally Posted by garo

Nice. I don't really want to buy new hardware but I will keep this in mind. THough I think if I had to spend money on hardware I would probably get a Radeon card.

You can't beat the investment into a Radeon, but for LLR the 9100f gives quite a bang for the buck. LGA 115A v2 boards are fairly cheap -- I have a MSI. When you take into account electricity costs of running older hardware it seems worth buying new hardware for crunching.

kruoli · 2020-05-06, 19:38

Quote:

Originally Posted by paulunderwood

...bang for the buck...

If you would be going solely after that, you should fix your preferenced work type beforehand and select the processor after that. Yes, not what garo asked for.

2020-04-29, 15:07	#1
garo Aug 2002 Termonfeckin, IE 2⁴×173 Posts	Best work choice for these machines So I have a few boxen that are either quite slow - think Core 2 6600 - or are on only intermittently - laptops etc. What is the best work choice for these? I currently run a combination of DC/P-1 and TF. I find with P-1 the older boxes really struggle as they have very limited memory - hard to spare even 1GB for P-1 without interfering with the other things. With DC I am not finding it hard to complete work before expiry as the exponents get larger some of these boxes can take months to complete work and the exponent can become Cat 1 or even Cat 0 by then. Any suggestions?

2020-04-30, 09:34	#3
garo Aug 2002 Termonfeckin, IE 2⁴×173 Posts	Thanks. I tried that. But on a two core Core2 6600 when I try with the latest Prime95 version it insists that it needs two cores for the job. Is that correct? I want to let P-1 run on the other core and I keep getting a warning about wanting to use more cores than exist. Last fiddled with by garo on 2020-04-30 at 10:11 Reason: Edit: Insert missing article just in case I am accused of being a Russian trollbot.

2020-05-06, 10:53	#8
paulunderwood Sep 2002 Database er0rr 5·937 Posts	I haven't benchmarked it, but the cheap i3 9100f (no graphics, 65w tdp, lga 1151 v2) I purchased recently really flies. Reminder to self: set up the XMP. Last fiddled with by paulunderwood on 2020-05-06 at 11:20

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Titan's Best Choice	Brain	GPU Computing	30	2019-10-19 19:19
Too much choice!	fivemack	Factoring	1	2018-06-19 20:15
Redoing factoring work done by unreliable machines	tha	Lone Mersenne Hunters	23	2016-11-02 08:51
Require P-1 and other factoring work to be done on reliable machines	tha	PrimeNet	66	2015-05-20 15:45
Work transfer between 32 and 64 bit machines	tichy	Software	11	2011-01-07 22:57

2020-04-29, 15:16	#2
Uncwilly 6809 > 6502 """"""""""""""""""" Aug 2003 101×103 Posts 2²×2,767 Posts	How about PRP on cofactors? That is not cutting edge mainline items. But it is an available work type. Memory is not much of an issue. No milestones to have to worry about.

2020-05-01, 22:16	#5
Mark Rose "/X\(‘-‘)/X\" Jan 2013 https://pedan.tech/ 2⁴·199 Posts	If you'll continue doing DC, I'd say make sure both cores are assigned to a single worker. You could also manually grab high DC, which won't be a lower cat for many years.

2020-05-03, 20:59	#6
garo Aug 2002 Termonfeckin, IE 5320₈ Posts	Thanks both for your answers. I will try editing local.txt @Mark Rose: That's a good suggestion. These machines are somewhat bandwidth limited with only DDR2-666/800. I assume using two cores for a single worker will more or less halve the iteration time?

2020-05-06, 10:23	#7
garo Aug 2002 Termonfeckin, IE 2⁴·173 Posts	Following up on this. I ran benchmarks on two of the newer laptops we have in the house. I haven't run Prime95 on them but it is amazing how much faster they are both absolutely and relative to clock-speed. I compared i5-8365U and i7-7500U laptops to the Haswell 4770 and they are both faster at P95 despite significantly slower clock speeds. I usually don't run P95 on laptops due to thermal issues but I'm very tempted to run something on these boxen.

2020-05-06, 15:12	#9
garo Aug 2002 Termonfeckin, IE 2⁴·173 Posts	Nice. I don't really want to buy new hardware but I will keep this in mind. THough I think if I had to spend money on hardware I would probably get a Radeon card.