[QUOTE=James Heinrich;379315]Perfect timing, I'm just finishing the last 25 of 1000 assignments I picked up, so I can get back to my [url=http://www.mersenne.ca/tf1G.php?available_assignments=1]pet project[/url]. So you'll be losing whatever limited firepower I've given over the last couple months, sorry.[/QUOTE]

No apology needed.

Thanks for your effort and work.

[QUOTE=James Heinrich;379315]Perfect timing, I'm just finishing the last 25 of 1000 assignments I picked up, so I can get back to my [url=http://www.mersenne.ca/tf1G.php?available_assignments=1]pet project[/url]. So you'll be losing whatever limited firepower I've given over the last couple months, sorry.[/QUOTE]

I may be buying my first 580 tonight to dedicate to TF. They're getting really cheap now. My work machine could support three, but I'm not sure I also want to spring for a 1 kW PSU.

Humour is such a subjective thing...

[url]https://www.youtube.com/watch?v=tkAQLLbJbDw[/url]

[QUOTE=chalsall;379324]Humour is such a subjective thing...

[url]https://www.youtube.com/watch?v=tkAQLLbJbDw[/url][/QUOTE]

I found that quite funny :)

[QUOTE=Mark Rose;379325]I found that quite funny :)[/QUOTE]

It cannot be denied that Canadians [URL="https://www.youtube.com/watch?v=aog_zGmWbFI"]are quite weird.[/URL]

[QUOTE=chalsall;379327]It cannot be denied that Canadians [URL="https://www.youtube.com/watch?v=aog_zGmWbFI"]are quite weird.[/URL][/QUOTE]

Dafuq is this

[QUOTE=TheMawn;379330]Dafuq is this[/QUOTE]

A favourite of mine.
[url]https://www.youtube.com/watch?v=uA_3MnVen08[/url]

[QUOTE=chalsall;379314]Just a heads up...

[SNIP]

And, please, let me know if anyone thinks doing this is sub-optimal.[/QUOTE]

Looks good on my end. I just grabbed a couple hundred more of whatever GPU72 wanted me to work on next and with a few exceptions (poor abandoned little things) they're all 73 to 74 in the upper 66M.

It's nice to see we're pushing everything to 74 now as it indicates to me we have a good buffer now. Also it's nice to see that the lowest assignments are in the 66M area. I didn't know we were so far ahead of the wave.

I'm still taking a mix of 68-->69 (45M range) and 69-->70 (34-35M) unless extra firepower is needed to stay ahead of the wave. My AMD GPUs are less efficient in doing 73-->74, so I'll leave that to the people with NVidia cards.

Are we even close to having the amount of firepower needed to do 75 bits? When would the ideal time be to start that?

[QUOTE=TheMawn;379561]Are we even close to having the amount of firepower needed to do 75 bits? When would the ideal time be to start that?[/QUOTE]

It doesn't really make sense to until 74M or so if I read [url=http://www.mersenne.ca/cudalucas.php?model=12]this graph[/url] correctly. Rather than trial factor to 75 bits, it makes more sense to run CUDALucas. AMD cards should only factor to 73 bits as mfakto slows down significantly beyond that from what I understand.

[QUOTE=Mark Rose;379566]It doesn't really make sense to until 74M or so if I read [url=http://www.mersenne.ca/cudalucas.php?model=12]this graph[/url] correctly. Rather than trial factor to 75 bits, it makes more sense to run CUDALucas.[/QUOTE]

Depends on the card... For 580s (GF110 chip), for example, it's "profitable" to go to 75 starting at 65M. For 680s (GK104) not until 70M. And for Titan's (GK110) not until 83M.

[QUOTE=Mark Rose;379566]AMD cards should only factor to 73 bits as mfakto slows down significantly beyond that from what I understand.[/QUOTE]

Correct.

But this is largely moot -- we don't have the fire power at the moment to sustain going to 75 (although those who want to are free to, and some are already).

[QUOTE=TheMawn;379544]I didn't know we were so far ahead of the wave.[/QUOTE]

Actually, we're not really (yet), but we're getting there. Keep in mind that the [URL="http://www.mersenne.org/thresholds/"]Primenet assignment rules[/URL] for LL Category 4 assignments is destined to be offset by 100,000 candidates (it's at 69,560 at the moment, climbing by 120 a day).

If the offset was 100,000 today the assignments would be starting at 68,562,079, and by the time it reaches the 100,000 offset it will probably be in the 70M range or so.

I think we should really think about putting in another option for which GPU is used on the PCs for optimal "Let GPUto72 decide". Some people here are using the GTX 5xx cards where it is good to go to 74 Bits from 50M, some more use the 6xx cards to go to 74 Bits from ~55M, but I wonder how many people are using the very new Cards or the AMD Cards, where they really shouldn't be doing work to 74 Bits there at all.

So I think instead of taking this work into the "Let GPUto72 decide" only into the low-exponent work or even another one, saying for which cards it is profitable.

Another suggestion would be to add a text to the get-assignments-page saying to check on the analysis on where to push the exponents, so they can adjust the limits.

The complication in that idea is when you combine multiple pieces of hardware. If the entire project were running on a single GTX 770, then yeah, by all means, TF to 73 and then LL. However, because a GTX 580 is also available, it might make more sense to TF to 74 bits first. Maybe the GTX 770 does only LL and the GTX 580 does enough TF to keep both cards fed and then LL during the rest of the time?

It gets even worse when you involve a CPU, because TF on a CPU is just bad all around. You CERTAINLY wouldn't be doing TF to 71 and then LL. Not when GPU's exist. Clearly, the optimal way is something along the lines of "GTX 580 does as much TF to 74 as it needs, and LL the rest of the time, then the GTX 770 and the CPU both doing LL 100% of the time".

It boils down to using each piece of hardware to the best of its abilities, keeping in mind the relative amounts of each that you have. For us, we have so much "CPU" that we barely keep up by putting most GPU's to TF. The GTX 770 is an interesting example actually because just yesterday I was looking at the top LL producers and the costs of the cards, and the GTX 770 is actually the sweet spot in terms of LL iterations per dollar. If people are using their GPU's for LL, I would certainly hope they're being efficient with them by picking good LL GPU's.

I don't know if Chris is still doing this but at some point he was offering to trade LL results from his CPU's for TF results from other people's GPU's in order to squeeze more efficiency out of the community. Suppose you really want to do lots of LL tests. You put all your GPU's to LL work, despite the fact that they're WAY better off doing TF. What Chris did was offer to give you the credit for his LL results done on a CPU in exchange for the credit for TF results done on your GPU's instead of LL. In the end, you get the credit for doing the LL's just like you wanted to, but overall, the project's productivity goes way up because both pieces of hardware were doing the optimal thing.


CPU's basically throw a big wrench into the optimization problem.

I think most people that have AMD cards and use GPUto72 know their cards are less efficient doing TF over 73bits. They most likely will set the 'will factor to" to 73 bit. Of course it would be interesting to see who has what kind of cards/firepower.

[QUOTE=VictordeHolland;379623]I think most people that have AMD cards and use GPUto72 know their cards are less efficient doing TF over 73bits. They most likely will set the 'will factor to" to 73 bit. Of course it would be interesting to see who has what kind of cards/firepower.[/QUOTE]

When you Let GPU72 Decide, it ignores your "will factor to" setting.

On the Work Distribution Map (primenet)
34M range Assigned TF: 69

[URL]http://en.gpu72.com/reports/available/[/URL]
34M range Reserved from Primenet 3,121 Assigned TF: 468

That doesn't seem right?

[QUOTE=VictordeHolland;380178]That doesn't seem right?[/QUOTE]

Yeah... I noticed that as well, and it isn't right. I can't explain how it happened though; started a little more an a month ago, and only for that single range...

As no one is being assigned 34M candates by Primenet at the moment I decided to simply let sleeping dragons lie. But since you brought this forward I have released everything not assigned by GPU72 back to Primenet, and am now bringing everything back in.

What are you two exactly talking about?
The GPU72 TF assignments are NOT reserved from PrimeNet as TF assignments. For PrimeNet, which is not "gpu-aware", the TF level is much lower and these ranges are long time ago "done" for TF, from PN's point of view. That is why GPU72 reserves exponents "for LL", we do LLTF to few bits higher than the PN's TF limit, and we then release them back. Therefore, GPU72's TFs (or part of them) do not appear as TF reservations on PN.

Or is this different now? At least this is how we started, if I remember right, and the subject is discussed on the forum many times, when "newbies" complain that their TF reservations appear on PN as reserved by other user for LL ("For Research").

So, if I didn't understand wrong, the "issue" described by Victor is quite normal...

[QUOTE=LaurV;380219]What are you two exactly talking about?
The GPU72 TF assignments are NOT reserved from PrimeNet as TF assignments.For PrimeNet, which is not "gpu-aware", the TF level is much lower and these ranges are long time ago "done" for TF, from PN's point of view. That is why GPU72 reserves exponents "for LL", we do LLTF to few bits higher than the PN's TF limit, and we then release them back. Therefore, GPU72's TFs (or part of them) do not appear as TF reservations on PN.

Or is this different now? At least this is how we started, if I remember right, and the subject is discussed on the forum many times, when "newbies" complain that their TF reservations appear on PN as reserved by other user for LL ("For Research").

So, if I didn't understand wrong, the "issue" described by Victor is quite normal...[/QUOTE]
Not any-more, *most* LLTF and DCTF show up in the Work Distribution map as assigned TF. Don't know when the change took place, but some time ago.

[QUOTE=LaurV;380219]Or is this different now? At least this is how we started, if I remember right, and the subject is discussed on the forum many times, when "newbies" complain that their TF reservations appear on PN as reserved by other user for LL ("For Research").[/QUOTE]

It's been different for quite some time, at least for most candidates.

Some are still being reserved by "For Research" / "wabbit" (recycled captured by one of my spiders), but most are reserved by "GPU72" / "gpufactoring" using a different spider.

Christ, could you load the 46M range in GPU72 for DCTF (the last DC range for 68-69 bit), at the current rate (5000 exponents/week) I expect the 45M range (to 69bit) to be ready in 4-7 days.

Thanks in advance.

[QUOTE=VictordeHolland;380485]Christ, could you load the 46M range in GPU72 for DCTF.[/QUOTE]

Golly! I've been elevated from Christopher to Christ!

[QUOTE=chalsall;380487]Golly! I've been elevated from Christopher to Christ![/QUOTE]

I thought he was just really mad.

[QUOTE=chalsall;380487]Golly! I've been elevated from Christopher to Christ![/QUOTE]
The flying spaghetti monster took over control there for a second!

[QUOTE=VictordeHolland;380503]The flying spaghetti monster took over control there for a second![/QUOTE]
Blessed be the Noodly Appendage!

Need some additional DCTF'ing...
 

Just to let everyone know that we're running a bit low on DC Cat4 candidates -- only about ten days worth or so.

So, for those willing, taking a hundred or so of DCTF "What Makes Sense" or "Let GPU72 Decide" would be much appreciated.

Makes Sense[SUP]TM[/SUP] makes sense to me. I'll grab a few next time I'm visiting the shop.

[QUOTE=chalsall;380790]Just to let everyone know that we're running a bit low on DC Cat4 candidates -- only about ten days worth or so.

So, for those willing, taking a hundred or so of DCTF "What Makes Sense" or "Let GPU72 Decide" would be much appreciated.[/QUOTE]

I've queued 3 days worth on each of my cards, or 224 assignments. How many days will that by us?

After that I've got a bunch of 55M->74 assignments queued.

[QUOTE=TheMawn;380825]Makes Sense[SUP]TM[/SUP] makes sense to me. I'll grab a few next time I'm visiting the shop.[/QUOTE]

Thanks much.

I've already burnt through my $150 USD budget of DCTF on EC2 for this month.

[QUOTE=Mark Rose;380837]I've queued 3 days worth on each of my cards, or 224 assignments. How many days will that by us?[/QUOTE]

The "Cat 4" workers do about 30 DCs a day. So you're buying us about a week.

Thanks! :smile:

[QUOTE=chalsall;380868]The "Cat 4" workers do about 30 DCs a day. So you're buying us about a week.

Thanks! :smile:[/QUOTE]

I've got about 2.5 days of 55M TF to finish and then I'll work on DCTF for a while. I'll put the big guns on clearing up to 35M (about 12 THz-d), which should be done in about 3 weeks (~600 GHz-d). I'll leave the three smaller barrels (~130 GHz-d) on fetching DCTF "what makes sense" without a big queue. I'll be completely only ~14 WMS DCTF a day, so hopefully someone else is also working that area a bit.

Anyone have a spare GTX 580? I'll buy a bigger power supply for my work desktop and plunk it in there :grin:

[QUOTE=chalsall;380790]Just to let everyone know that we're running a bit low on DC Cat4 candidates -- only about ten days worth or so.

So, for those willing, taking a hundred or so of DCTF "What Makes Sense" or "Let GPU72 Decide" would be much appreciated.[/QUOTE]

I just put some resources on it (4x100)

I put 30 WMS on each GPU. In light of the request, I put them at the head of the queue. They should be done before my LLDC work gets too stale.

Thanks guys! Gotta keep the "churners" fed! :smile:

Back to LLTF...
 

[QUOTE=chalsall;380921]Thanks guys! Gotta keep the "churners" fed! :smile:[/QUOTE]

Thanks for the "surge" everyone -- we've now got about seven weeks worth of buffer.

Time to go back to LLTF'ing.

Not to add anything to an already full plate, but I noticed a few things.

First, the assignment preview seems out of whack again (advertised was 72M from 72 to 74, got 68M-69M from 73 to 74).

Also, I was assigned some 75M and even M76,028,009. Are we really there, yet?

Is a fix going in? When I run the submission spider it is getting 500 timeouts on login.

Right at the moment (5-7pm EDT 04-Sep-2014) PrimeNet is [url=http://www.mersenneforum.org/showthread.php?p=382128#post382128]changing servers[/url]. So don't expect to be able to submit anything for the next couple hours.

gpu72.com slow
 

[url]www.gpu72.com[/url] web pages load very slow
also, MISFIT trying to talk to gpu72.com gets
error during ProcessAutoFetch the operation has timed out

Same here, lots of timeouts and unfinished request on gpu72.

[QUOTE=TheJudger;382409]Same here, lots of timeouts and unfinished request on gpu72.[/QUOTE]

How are things now?

I discovered another change on the Primenet server which broke a few of my spiders, and so many instances were running concurrently retrying fetching. The fixes were done yesterday (Sunday); the change on Primenet appears to have occured on the 3rd.

Please let me know if there are still issues -- it might be time to start caching some more of the pages (for example, the Available Candidates page and the Release Level page) -- right now those are rendered when requested.

I like having the Available Candidates page being real-time. It's useful for grabbing oddball work.

One thing that has stopped working recently is the "Fetching preview" when getting assignments in Chrome (probably a Chrome change). It still works in Firefox. I think there may be a race condition in the JavaScript because the request data comes back fine if watching the requests. I haven't been able to fully figure out the issue.

[QUOTE=chalsall;382465]How are things now?

I discovered another change on the Primenet server which broke a few of my spiders, and so many instances were running concurrently retrying fetching. The fixes were done yesterday (Sunday); the change on Primenet appears to have occured on the 3rd.

Please let me know if there are still issues -- it might be time to start caching some more of the pages (for example, the Available Candidates page and the Release Level page) -- right now those are rendered when requested.[/QUOTE]
Maybe this problem is related: [url]http://www.mersenne.info/trial_factored_tabular_data/3/0/[/url] is empty.

[QUOTE=chalsall;382465]
I discovered another change on the Primenet server which broke a few of my spiders[/QUOTE]

Out of curiosity, what PrimeNet pages do you rely on? I only know of the private web page we designed together and that shouldn't have changed.

Do we need to discuss more private pages so that you are more immune to us making layout changes for the general user?

[QUOTE=Prime95;382502]Out of curiosity, what PrimeNet pages do you rely on? I only know of the private web page we designed together and that shouldn't have changed.[/QUOTE]

This was most defintely my fault. My spider was watching for the line after the meaningful data saying "You are visitor number" to check that the reports had fully completed.

[QUOTE=Prime95;382502]Do we need to discuss more private pages so that you are more immune to us making layout changes for the general user?[/QUOTE]

Worth discussing... I will look at what reports my spiders scrape this weekend, and make some suggestions.

Still no counts below 1M....
 

[url]http://www.mersenne.info/trial_factored_tabular_data/3/0/[/url]

How does the DCTF vs LLTF WMS situation look at the moment?

[QUOTE=Mark Rose;382609]How does the DCTF vs LLTF WMS situation look at the moment?[/QUOTE]

Also, can I take it that P-1 is pretty far ahead of the wave? I started a few workers, since I have lots of RAM, but the assignments coming in were 75M-79M. I've already moved two of the four back to DC.

On a tangential note, changing work type preferences seems much more trouble free now. I just stop P95 and change the preferences there. Then I shut the program down and go modify worktodo.txt to reflect the new situation. Finally, I log into PrimeNet>CPU and change the preferences there. I've shuffled things a few times recently with no confusion of work types, and no need to ditch wrong assignments.

I see from the last week report on gpu72 that nucleon is back in business... Nice! Let's kick those freaking exponents, hehe...

I think that is because the winter came there, down under? :razz:

OTOH, my main machine had a [URL="http://www.mersenneforum.org/showthread.php?p=382941"]clogged water pipe[/URL], so i need to stop it for cleaning (2xgtx580 plus a titan). It will be back crunching tonight or tomorrow morning.

Available TF Assignments page blank? NOT!
 

[STRIKE]Actually, only the 332M table is missing outright. The LL and DC tables are empty.[/STRIKE]

Apparently this was a momentary situation. All seems normal now. :smile:

Recently surpassed 10,000,000 saved.
 

That is about 40,000 P1+LL+DC in the current low end LL.

Cat IV assignments
 

Over the last week or so the range 78-79M has had about 2000 LL tests assigned but I have not seen similar output on mersenne.info.

Is GPU72 starting to hand out LL tests with <73 bits TFing? This range doesn't seem to be assigned to GPU72 and assignments seem to be coming directly from GIMPS.

Just curious, not concerned…

Also what has the daily recycle rate of cat IV assignments been like for the last while?

[QUOTE=flagrantflowers;383766]Over the last week or so the range 78-79M has had about 2000 LL tests assigned but I have not seen similar output on mersenne.info.

Is GPU72 starting to hand out LL tests with <73 bits TFing? This range doesn't seem to be assigned to GPU72 and assignments seem to be coming directly from GIMPS.

Just curious, not concerned…

Also what has the daily recycle rate of cat IV assignments been like for the last while?[/QUOTE]
IIRC, only 10-20% of cat4 LL tests are completed, the rest is recycled.

Preview gives me
[CODE]66,123,xxx [/CODE]

but assignment then gives me

[CODE]
78,xxx,xxx
[/CODE]

There have been somewhere in the range of 4000 assignments from 78-79M in the last 2-3 days. A lot of those must be TF'd to 72bits according to mersenne.info.

[QUOTE=flagrantflowers;383818]There have been somewhere in the range of 4000 assignments from 78-79M in the last 2-3 days. A lot of those must be TF'd to 72bits according to mersenne.info.[/QUOTE]
Yes, and that number has continued to grow to 7098 LL assignments in the 78M range in the last couple of days.

Yep looks like GPU72 is not getting enough horsepower. Maybe some stuff ought to be released at 73.

Yep, looking at [url]http://www.mersenne.org/primenet/[/url]

[code]
----------=-----=-- | -----=-----=-----=----- | -----=-----=-----=----- |
Exponent Range | Assigned | Available |
Start Count P | ECM P-1 LL LL-D | ECM P-1 LL LL-D |
----------=-----=-- | -----=-----=-----=----- | -----=-----=-----=----- |
66000000 55644 | 1623 191 2067 11 | 12588 3190 |
67000000 55575 | 156 674 10456 30 | 1 5521 2801 |
68000000 55332 | 4509 243 13846 9 | 1505 |
69000000 55390 | 2 349 16360 17 | 2884 |

70000000 55309 | 14052 530 6141 1 | 179 |
71000000 55285 | 15491 1249 3987 2 | 243 |
72000000 55431 | 16194 156 4253 1 | 226 |
73000000 55165 | 18676 368 2148 1 | 106 |
74000000 55050 | 18170 400 2550 | 8 |
75000000 55307 | 20079 8 1183 | 9 |
76000000 54924 | 20132 4 753 | 2 3 |
77000000 55009 | 18509 1099 1392 1 | 5 4 |
78000000 54900 | 1906 408 7093 | 11560 15 16 |
79000000 54938 | 64 7 | 20797 160 6 |

80000000 55027 | 566 | 20672 70 4 |
81000000 55021 | 1083 1 | 20300 50 |
82000000 54887 | 345 1 | 20787 65 4 |
83000000 54822 | 842 | 20502 42 1 |
84000000 54592 | 2312 | 18862 50 |
85000000 54739 | 2078 1 | 19036 37 1 |
[/code]

would indicate to me that PrimeNet has run out of Cat4 assignments.

It looks to me like the higher ones are being given out for LL because all of the ones in the gap are reserved for TF.

Most 78M LL assignments actually GPU72 TF
 

There is a caveat with the 78M LL assignments, according to [URL]http://www.gpu72.com/reports/available/[/URL] , GPU72 has 7000+ TF assignments in that range. Primenet shows only 1800+ TF assignments, so many of the LL assignments are actually GPU72 TF (but reserved in the 'old' way).

@Mark Rose, luckily we don't run ECM on these exponents, as the header in your table would suggest ;).

Looking in [URL="http://www.mersenne.org/assignments/?exp_lo=78100000&exp_hi=79000000&execm=1&exdchk=1&extf=1&B1=Get+Assignments"]here[/URL] the exponents given out are either TF'd to 74 or given to wabbit aka ForResearch aka GPU72, so everything looks fine.

Except that maybe currently there are too few P-1 workers so all the Cat4 people have to do that by themselves first.

[QUOTE=VictordeHolland;383879]There is a caveat with the 78M LL assignments, according to [URL]http://www.gpu72.com/reports/available/[/URL] , GPU72 has 7000+ TF assignments in that range. Primenet shows only 1800+ TF assignments, so many of the LL assignments are actually GPU72 TF (but reserved in the 'old' way).
[/quote]

Well that is good news!

[quote]
@Mark Rose, luckily we don't run ECM on these exponents, as the header in your table would suggest ;).[/QUOTE]

I only copied the columns from what PrimeNet spat out on that page. :)

[QUOTE=manfred4;383882].....
Except that maybe currently there are too few P-1 workers so all the Cat4 people have to do that by themselves first.[/QUOTE]

I had wondered about the assignments being handed out for P-1 being so high, but had not paid attention to the ranges for the different categories.

Perhaps I will shift my workers from DC to include a couple more P-1s.

We used to have a P-1 surplus did we not?

All I know is when I complete a GPU72 TF->74 assignment it gets handed out by PrimeNet for LL almost immediately.

I think we most desperately need TF fire-power. With the weather getting cooler here I plan to restart my 580. I think when a TF completes to 74 since there is such a small buffer it gets assigned to LL before it has a chance to be assigned for P-1. For instance, there is nothing available for LL P-1 right now. So it is LL-TF we need the most.

[QUOTE=garo;383905]I think we most desperately need TF fire-power. [/QUOTE]
Can someone confirm this? I can put some 590s to work.

[QUOTE=blip;383908]Can someone confirm this? I can put some 590s to work.[/QUOTE]

All the evidence is pointing towards that. I'm tempted to pick up a couple more 580's this weekend. They're getting cheap now.

Cooler weather would make it easier for me to restart a 570, but only if I come up with a large enough PSU. An overclocked 580 + a a stock FX-8350 do fine on a 750 W, but it pulls just under 600 W. It wouldn't carry another 220 W. I'm not sure I could make it work, even by seriously underclocking both cards.

Besides, the 570 has some eroded PCIe power connectors, which concerns me somewhat.

Is that 600 W at the wall? What's the efficiency of your power supply?

If it's 85% efficient, you're only pulling 510 W off the supply side and have room, barely, if the 12 V rail configuration allows it.

I'm pulling 765 W at the wall from an 85% efficient 750 W power supply. I could still get another 100 W out of it before hitting the limit.

I would, however, be concerned about those power connectors.

[QUOTE=Mark Rose;383966]Is that 600 W at the wall? What's the efficiency of your power supply?

If it's 85% efficient, you're only pulling 510 W off the supply side and have room, barely, if the 12 V rail configuration allows it.

I'm pulling 765 W at the wall from an 85% efficient 750 W power supply. I could still get another 100 W out of it before hitting the limit.

I would, however, be concerned about those power connectors.[/QUOTE]

Yes. The 600W is reading from a Kil-a-watt, and it is a Bronze rated supply. It has 4 x 18 Amp, 12 v rails.
[URL]http://www.newegg.com/Product/Product.aspx?Item=N82E16817341041[/URL]

I am surprised to see it still on the market. The NewEgg listing says "Fire Power" for the brand, but the unit shown still has "OCZ" markings. It still lists a 5 year warranty.

I am starting to wonder, again, about the 1 kW OCZ Gold which I had to stop using when its regulation went crazy. I guess I might try to make contact with the latest reincarnation of the company. Gotta look for a receipt, though.

On the 570's issues, I have considered trying to connect it by wire-wrapping and soldering the pins. That, or replace the connector altogether. I know that LaurV talked about tightening the female connectors, but that still makes me nervous after seeing blackened plastic on the cable ends.

Trying to squeak by with the 750 W supply and 2 GPUs would be an interesting experiment, but I have at least a vague memory of BSODs under those circumstances.

EDIT: It might work, though, if I throttled back on both cards. The 580 is rated at a 781 MHz factory OC (+10 MHz from stock), though I'm currently running it at 844 MHz. That adds about 20 W at the wall. The 570 came with an 844 MHz factory OC, but stock is 732. I'm not in a position ATM to check on what that does to the power draw.

[QUOTE=blip;383908]Can someone confirm this? I can put some 590s to work.[/QUOTE]

I can confirm that.

We are a little underpowered in the LLTF region.

We are a little overpowered in the DCTF region.

We're still not doing that terribly badly (read: almost no LLs are being issued below 73 "bits" for "Cat4", and never for Cat1, 2 or 3 below 74) but more TF'ing would be very valuable.

It is really amazing to see, returned TFings are reallocated for LL almost immediately.

I now have two 590s working on LL-TF.

[QUOTE=chalsall;384000]We are a little underpowered in the LLTF region.

We are a little overpowered in the DCTF region.

We're still not doing that terribly badly (read: almost no LLs are being issued below 73 "bits" for "Cat4", and never for Cat1, 2 or 3 below 74) but more TF'ing would be very valuable.[/QUOTE]From watching the classic status page, it looks like the issue is greater through put of LL tests. Last 3 weeks have been the best in a while (about November of 2013). This is a good thing.

I tried to reinstall my GPUs. They both died. :sad:

[QUOTE=ET_;384040]I tried to reinstall my GPUs. They both died. :sad:[/QUOTE]

Wow! That seems a bit much of a coincidence. Could there be some other factor at work than the cards failing?

[QUOTE=kladner;384046]Wow! That seems a bit much of a coincidence. Could there be some other factor at work than the cards failing?[/QUOTE]

I used both 24/7 on the same computer for many months, and then detached the GTX580 when the computer started complaining.

I ran the system for 6 months without Nvidia drivers, using the GTX 275 only for the display. As the 275 had a rattling noise (I suspect the hub of the fan), I exchanged the GTX275 with the 580. The PC freezed after a short while. And kept freezing when I reinstalled back the GTX 275.

[QUOTE=Uncwilly;384039]From watching the classic status page, it looks like the issue is greater through put of LL tests. Last 3 weeks have been the best in a while (about November of 2013).[/QUOTE]

I don't /think/ this is it... Over the last thirty days we've been averaging about 266 LLs completed a day; this is down from about 300 a day a year ago.

I suspect the issue is the Cat 4 "churners", combined with the fact the Cat 4 range increases by about 386 candidates a day (266 + 120). (The extra 120 is because George has Primenet increasing by this amount until it reaches a 100,000 candidate offset for Cat 4 (we're currently at 76,520); note the ~38,000 available Cat 3 candidates -- these are all TFed to at least 74, and P-1'ed.)

[QUOTE=chalsall;384050]note the ~38,000 available Cat 3 candidates -- these are all TFed to at least 74, and P-1'ed.)[/QUOTE]
We have foreseen that. Remember when I said (here, on the forum) that in case Cat4 TF-ed to 74 are not available, we should assign the highest Cat 3 instead? I think nothing was done in this direction, and I still think that is better for the project if a user gets a well TF-ed exponent for LL, than if he gets a higher one improper TF-ed. Of course, the exact "odds" here depend of the churning rate.

[QUOTE=chalsall;384050]I don't /think/ this is it... Over the last thirty days we've been averaging about 266 LLs completed a day; this is down from about 300 a day a year ago.

I suspect the issue is the Cat 4 "churners", combined with the fact the Cat 4 range increases by about 386 candidates a day (266 + 120). (The extra 120 is because George has Primenet increasing by this amount until it reaches a 100,000 candidate offset for Cat 4 (we're currently at 76,520); note the ~38,000 available Cat 3 candidates -- these are all TFed to at least 74, and P-1'ed.)[/QUOTE]

That makes a lot of sense to me.

It looks like only 2100 candidates in the Cat4 range were brought to 74 bits in the last week. That leaves us with a shortfall of 600 on a weekly basis, or about 86 a day. That's about 4 THz-d/day in needed additional trial factoring throughput.

[QUOTE=Mark Rose;384054]That leaves us with a shortfall of 600 on a weekly basis, or about 86 a day. That's about 4 THz-d/day in needed additional trial factoring throughput.[/QUOTE]

Yes. The good news though is only about 7% of all Cat 4's actuallycomplete, so we will get another shot at most of those sub-optimally TFed.

Also, so everyone knows, if "Spidy" seeks that anything is about to be released at below 73 it will release a bunch at 73 (or 74 if we're holding any at that level awaiting P-1'ing). Thus, very very few (if any) should be being assigned for LL'ing (or P-1'ing) at below 73. (And, again, nothing is being assigned at below 74 in the Cat 1 through 3 ranges.)

But, definitely, some more LLTF'ing would be useful and appreciated (even if at the expense of DCTF'ing -- we've got at least a six month buffer there at the moment).

[QUOTE=LaurV;384052]We have foreseen that. Remember when I said (here, on the forum) that in case Cat4 TF-ed to 74 are not available, we should assign the highest Cat 3 instead? I think nothing was done in this direction, and I still think that is better for the project if a user gets a well TF-ed exponent for LL, than if he gets a higher one improper TF-ed. Of course, the exact "odds" here depend of the churning rate.[/QUOTE]

Yes, but... That would involve some additional specialized code by George et al.

Another option would be to simply stop increasing the Cat4 offset by 120 a day, and perhaps drop it back to down to 75,000 (a tiny drop; just a round number). The TF'er could probably (just) sustain the Cat4 demand if that happened.

I understand the desire for the 100,000 offset -- it will make the curves cross for expiry at almost exactly 360 days -- but perhaps for the next few months or so we "hold" at "O minus 25,000"? (Hey, NASA regularly does that kind of thing, why can't we? :wink: :smile:)

If we dropped it to 20 a day from 120 we could still keep up (and slowly build buffer) and still head towards the goal of 100,000 ahead. That sounds preferable to me. We would theoretically save about two tests per month if 7% are completed on first assignment. Not a huge amount, but it would come for free by slowing down the growth until we have more LLTF capacity. It would take 3 years instead of 6 months to reach the 100,000 ahead goal.

How many a day can one card do? Two 590s were added above. That should make a fair difference AFAIK.

and hopefully soon we have the next version of mfakto which process tf74 faster and hopefully supports intel hd4* chips... however fast those may be...

[QUOTE=henryzz;384073]How many a day can one card do? Two 590s were added above. That should make a fair difference AFAIK.[/QUOTE]
Overall they did 48 in the last 24 hours. I had to stop one gpu on one of the cards, as the system froze when using both.

[QUOTE=blip;384075]Overall they did 48 in the last 24 hours. I had to stop one gpu on one of the cards, as the system froze when using both.[/QUOTE]

Hmm, well with my system with the 690 (amd 4800+) I have to keep the cpu busy otherwise it freezes between 1-18hours..

[QUOTE=blip;384075]Overall they did 48 in the last 24 hours. I had to stop one gpu on one of the cards, as the system froze when using both.[/QUOTE]

Even so, 1.5 GTX 590's should give about 1 THz-d/day. That's about 25% of what we're currently short (71 -> 74). Granted, there are a lot of exponents sitting at 73 bits, so in the meantime that will provide almost half the needed 73->74 capacity.

Let's say we release 25,000 CAT4 candidates at 73bits (to get to the 100,000).
With only 7% completing (assuming GPU72 gets them back afterwards), of which ~1/74 would have a factor between 73-74bits.
In that case we'll only have 'missed' about 24 factors.
Looking at it that way I'm inclined to say releasing CAT4 at 73bits for the time being is not a big deal.

We all knew the new assignment rules would come at a cost. A few less candidates saved by TF and more triple checks is that cost to advance milestones faster in the future.

[QUOTE=NickOfTime;384074]and hopefully soon we have the next version of mfakto which process tf74 faster and hopefully supports intel hd4* chips... however fast those may be...[/QUOTE]

It already supports Intel onboard HD4***.

[QUOTE=legendarymudkip;384113]It already supports Intel onboard HD4***.[/QUOTE]

Not yet? There is a -pre version which works... but they are _not_ for actual production.

[QUOTE=kracker;384118]Not yet? There is a -pre version which works... but they are _not_ for actual production.[/QUOTE]

I'd forgotten that... But it still passes the selftest and (as far as I know) no bug has been found that means it finds bad results.

[QUOTE=legendarymudkip;384120]I'd forgotten that... But it still passes the selftest and (as far as I know) no bug has been found that means it finds bad results.[/QUOTE]

Yes and no, Bdot told me that some factors have a rare chance of being missed even if the full selftest had passed. The latest code on github is fine I think, but the last -pre build on the mfakto thread had that bug. (Only on Intel HD I think)

Oh, btw, is the last version of mfakto (the one which speeds up the 74th bit, see mfacto thread) available for windows? Exe, and production ready? In the light of the last discussions in this topic, about Cat4, I could switch a 7970 from "max 73" to "let gpu decide", but the version of mfakto which I use is too slow for this card for bit 74.

The latest code passed all tests except for the general performance comparison. For an unknown reason, some kernels are still slower than before. This makes me hesitate to release it ...