[QUOTE=flashjh;361884]I just went into both my systems and had them export the completed work so far today, anything else will post at the normal time.[/QUOTE]

Thanks Jerry. :smile:

I should have another 3 or 4 TF to 74 and submitted in the next 9 hours.

[QUOTE=Mark Rose;361889]I should have another 3 or 4 TF to 74 and submitted in the next 9 hours.[/QUOTE]

Coolness, thanks.

If I may share, I find this to be an excellent example of "just in time" management.

And, if I may further share, a team I helped assemble may win a big contract because we were the only ones who were able to "get the atoms in" on time (read: we understood the temporal shipping limitations in relation to the deadlines).

Timing is often everything.... :smile:

Wow suddenly there are hundreds.

[QUOTE=petrw1;361893]Wow suddenly there are hundreds.[/QUOTE]

There were always hundreds.... :smile:

He likes to keep a few hidden up his sleeve, just in case ;)

I just dumped 7 to 74. :smile:

[QUOTE=chalsall;361882]Don't Panic. [SUP](TM)[/SUP]

Everything is under control. Nothing not TFed to at least 74 and P-1ed is being released for LLing.

With his permission, I sometimes have Jerry's cycles dedicated to taking 68M's up to 74, which are then released back to Primenet for P-1 assignments. I've "taught" spidy how to have Primenet release high candidates which are already fully and appropriately "cooked".

I'm doing this now, in preparation for a batch of candidates TFed to 74, but not yet P-1'ed, to be released this evening.

Please rest assured I never do this without being "at console".[/QUOTE]

You may use my GPU cycles like this if you wish.

[QUOTE=Chuck;361944]You may use my GPU cycles like this if you wish.[/QUOTE]

Thank you.

From Dec 24 to Jan 31, my PC is going to be offline. A GTX 670 and GTX 660 Ti, both producing ~300 GHz-Days each will be down for a week. Just letting you know ahead of time in case there's something that can be done. It seems like we might be nearer the knife's edge than we would like.

[QUOTE=TheMawn;361953]It seems like we might be nearer the knife's edge than we would like.[/QUOTE]

Thanks for the heads up.

During that period I will be able to have the machine hosting my 580 (thanks again Jerry! :smile:) in Linux 24/7. That will cover half of it.

And, now that Jerry is back (mostly) on-line, with LaurV's contributions, and with owftheevil doing TFing for any P-1 assignments he gets not yet TFed to 74, plus many others who have brought additional firepower to bear, we should be OK. In fact, we're pulling ahead again. :chalsall:

Thanks everyone. And hopefully Santa will be generous with "kit" this year.... :wink:

[QUOTE=garo;361876]I am getting a new (used) 580 today![/QUOTE]

The bad news is it doesn't fit in the Antec Solo case that houses my main rig. The good news is my second box is in a full-tower and it does fit.

@Chris-
Just extended my GPU72 reservations. I was attempting to make a video in order to ask for advice on how to stop my desktop from crashing seconds after startup, but it is now working, so...
Yea, mfaktc is back up and running on it now (and it's stayed alive long enough for me to post this) :)

[QUOTE=c10ck3r;362228]@Chris-
Just extended my GPU72 reservations. I was attempting to make a video in order to ask for advice on how to stop my desktop from crashing seconds after startup, but it is now working, so...
Yea, mfaktc is back up and running on it now (and it's stayed alive long enough for me to post this) :)[/QUOTE]

BSOD? If so, have you looked up the code?

With mfaktc actually running (has it found a factor?) I'll assume the GPU is not the culprit (you could run a graphics benchmark to be sure)

Run a CPU stress test (assuming you're not currently running Prime) and a RAM stress test.

Never ignore the PSU as well, if it is so it will be disguised.

[QUOTE=TheMawn;362246]BSOD? If so, have you looked up the code?

With mfaktc actually running (has it found a factor?) I'll assume the GPU is not the culprit (you could run a graphics benchmark to be sure)

Run a CPU stress test (assuming you're not currently running Prime) and a RAM stress test.[/QUOTE]

It crashed again less than an hour later. 'Bout to check the PSU.

If you don't find it, try a teardown. Clean out all parts with a compressor. Check the mb for swelled capacitors.

I can't access GPU72.com. I've got less than a day's assignments left. I'll check again in a bit.

Just letting you know in case something died.

[QUOTE=TheMawn;362331]Just letting you know in case something died.[/QUOTE]

Seems OK from our perspective.

Is there a particular URL you can't access?

This /could/ be a problem on your end.

I cant get through either. Trying to connect to [url]www.gpu72.com/[/url] gives a server not found error.

[QUOTE=owftheevil;362333]I cant get through either. Trying to connect to [url]www.gpu72.com/[/url] gives a server not found error.[/QUOTE]

Is that an error you are seeing from my server, or from your browser? I suspect the latter...

It appears we're under a big DDOS attack.

Are we having fun yet? :wink:

[CODE][/CODE]DNS timeout from UUNET, Level3, Google public.
Doesn't give name not found.
[CODE]C:\Users\Scott>nslookup gpu72.com 8.8.8.8
Server: google-public-dns-a.google.com
Address: 8.8.8.8

DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
*** Request to google-public-dns-a.google.com timed-out

C:\Users\Scott>nslookup gpu72.com 4.2.2.2
Server: b.resolvers.Level3.net
Address: 4.2.2.2

DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
*** Request to b.resolvers.Level3.net timed-out

C:\Users\Scott>nslookup garbageentrytotest.com 4.2.2.2
Server: b.resolvers.Level3.net
Address: 4.2.2.2

*** b.resolvers.Level3.net can't find garbageentrytotest.com: Non-existent domai
n
[/CODE]

Yup. Thanks.

Something big's going down!

Don't yet fully understand it.

[QUOTE=sdbardwick;362336][CODE][/CODE]DNS timeout from UUNET, Level3, Google public.
Doesn't give name not found.
[CODE]C:\Users\Scott>nslookup gpu72.com 8.8.8.8
Server: google-public-dns-a.google.com
Address: 8.8.8.8

DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
*** Request to google-public-dns-a.google.com timed-out

C:\Users\Scott>nslookup gpu72.com 4.2.2.2
Server: b.resolvers.Level3.net
Address: 4.2.2.2

DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
DNS request timed out.
timeout was 2 seconds.
*** Request to b.resolvers.Level3.net timed-out

C:\Users\Scott>nslookup garbageentrytotest.com 4.2.2.2
Server: b.resolvers.Level3.net
Address: 4.2.2.2

*** b.resolvers.Level3.net can't find garbageentrytotest.com: Non-existent domai
n
[/CODE][/QUOTE]

[QUOTE=chalsall;362332]Is there a particular URL you can't access?
This /could/ be a problem on your end.[/QUOTE]The problem's here too. DNS fails to resolve (|[url]www.|gimps.)gpu72.com[/url]
On the other hand, 50.21.180.9 pings just fine.

[QUOTE=James Heinrich;362338]The problem's here too. DNS fails to resolve.[/QUOTE]

Thanks James.

So, then, what this most likely means is someone has found a way to multiply their DNS queries to bother legitimate users.

I would argue that this is actually the fault of the "founders" of the internet.

Why would the the founders of the internet assume that every participant was benign?

OK here as well.

No issue here. Perhaps an international routing error?

If it is DNS then you could have no visible issue if you (or someone nearby) have recently accessed GPU72.com and the lookup is cached.
If it's DNS-DDOS then the issue could be intermittent, and after one successful lookup you also won't see a problem until the DNS cache expires.

Ok from Thailand. (or it was fixed meantime?)

I'm okay now. Not sure what it was. I had used downforeveryoneorjustme.com and it said the site was running fine which had me worried. I'm "glad" to see others had the issue and am also glad it's better now.

For me, anyway.


It happens to the best. A couple of weeks ago or so YouTube was down via DDOS. I don't know if it's more of a thing now than before.

[QUOTE=TheMawn;362360]I don't know if it's more of a thing now than before.[/QUOTE]

It's definitely more of a thing now - because you get all these kids who are 15 years old and too silly to know better and think it's funny to attack random websites...

Thanks for your observations guys.

I'm not quite sure why, but this forum and GPU72 have been seeing some very interesting (hostile) activities recently... :sad:

It's always interesting, in so many ways, having a "presence" on the Internet.... :wink:

What's kind of fucked is if they took any time at all to read up about the project, they'd find much, much more damaging ways to cause harm than preventing access to the server.

[QUOTE=TheMawn;362429]What's kind of fucked is if they took any time at all to read up about the project, they'd find much, much more damaging ways to cause harm than preventing access to the server.[/QUOTE]

That would take effort, though!

[QUOTE=blahpy;362441]That would take effort, though![/QUOTE]

Thank [blip] for that!



P.S. I've been searching for an atheist equivalent for the turn of phrase "Thank God" to no avail for years. Does anyone know of one?

I like "Thank the almighty Atheismo" but it doesn't have the same ring as "Thank God"

Thank [URL="https://en.wikipedia.org/wiki/Flying_Spaghetti_Monster"]FSM[/URL]! (Not exactly atheistic, but generally gets the point across.)

[QUOTE=TheMawn;362452]P.S. I've been searching for an atheist equivalent for the turn of phrase "Thank God" to no avail for years. Does anyone know of one?[/QUOTE]

Although not atheist, I sometimes use "Thank the Universe".

[QUOTE=TheMawn;362452]P.S. I've been searching for an atheist equivalent for the turn of phrase "Thank God" to no avail for years. Does anyone know of one?[/QUOTE]

I'm not atheist, but "thank goodness" sounds like a decent alternative. Or just throw in a cuss word to make it sound like you don't believe in God. :smile:

[QUOTE=Mini-Geek;362471]I'm not atheist, but "thank goodness" sounds like a decent alternative. Or just throw in a cuss word to make it sound like you don't believe in God. :smile:[/QUOTE]
I'm an atheist, but still object to blasphemy (as avid readers of my posts will have noticed).
I particularly object to OMG.

David

PS I see that WTF is considered perfectly acceptable and common here, and I don't mean "What Trial Factoring" bit level.

Submitted my last little batch of TF. Signing off for a week.

[QUOTE=TheMawn;362811]Submitted my last little batch of TF. Signing off for a week.[/QUOTE]
Merry Christmas!

I hope Santa can get all that firepower you asked for down the chimney:smile:

Although of course it isn't really needed for TF:davieddy:

[QUOTE=TheMawn;362452]P.S. I've been searching for an atheist equivalent for the turn of phrase "Thank God" to no avail for years. Does anyone know of one?[/QUOTE]I find that "Thank $DEITY" works well. Readers can assign whatever value they like to the variable, including null (or nil), and everyone is happy.

[QUOTE=xilman;362818]I find that "Thank $DEITY" works well. Readers can assign whatever value they like to the variable, including null (or nil), and everyone is happy.[/QUOTE]

Brilliant!!! :bow:

But how do you implement that in an audio stream? :wink:

[QUOTE=chalsall;362821]Brilliant!!! :bow:

But how do you implement that in an audio stream? :wink:[/QUOTE]

Taking a cue from Portal, "Thank *insert deity here*." :smile:

Not sure if this is to be posted here or elsewhere, last night I got worried (again) of the fact that I am not finding (too many) factors in my TF quest and I took all the 336M range from 65 to 66 bits. That was 22302 exponents (twenty-two thousands, three hundred and two), all unreserved. This was about 9-10 hours on 4 GPU cards, gave me about 500 GHzD and found about 350 factors on the way (that is, ~350 exponents cleared).

So, it is ok, we can find factors. Normal work resumed.

[QUOTE=LaurV;363015]So, it is ok, we can find factors. Normal work resumed.[/QUOTE]

Well, according to [URL="https://www.gpu72.com/reports/worker/2423ae6e8f696d5e7d1447de91ca35a6/"]your stats[/URL], you're actually slightly ahead of the predicted find rate. So you know, the predicted rate is based on a linear regression of the overall GPU72's empirical find rate for each bit level.

[QUOTE=LaurV;363015] Normal work resumed.[/QUOTE]

Normal???

[QUOTE=petrw1;363019]Normal???[/QUOTE]
i.e. TF to 74 at LL front on two 580's and TF to 73 at LL front on a 7970, but in fact, this "normality" is going to be broken soon, following the last discussions I switched the preference to DC-TF range. Old work still queued, but it will be finished soon, like end of the year.

@chalsall: not sure what you what to show me, but the most of the saved work in the last time, which can be seen on that graphic, is due to your and kracker's "aliens" who did P-1 work for me, and also few of my CPUs who did P-1. Only about 15-20% is due to factors found by TF ("exotic" ranges, not assigned by gpu72, where I found the most of the factors are not shown on those tables). That is why I always have this feeling (it can be subjective) that I am finding too less factors by TF to 74. However, I will continue to TF to 74 for a while... still in debt to you and kracker :blush: (well, excepting the fact that you let me "keep" the TF credit, hehe, I have already a big "surplus" of TF since the third card is in the game, and you stopped your aliens, and kracker stopped his - and he did not say anything about few of my cores which did P-1 for him :razz: in this time, that was just for fun and I wanted to see his face when he saw it, but maybe he didn't see it yet, he said he is not interested in "credit" and maybe he does not check the stats often... well, xmas, everybody busy... hehe)

[QUOTE=chalsall;363016]the predicted rate is based on a linear regression of the overall GPU72's empirical find rate for each bit level.[/QUOTE]
It escaped me first time, and I don't want to mix it with the previous gibberish in my post.

If you do that, then it is wrong. The cheaters will lower the rate in time. You have to use the "theoretical" 1/b, where b is the bitlevel. For example: if I do TF to 74 in a "fresh" range, i.e. where no P-1 was done, then my chances to find a factor is about 1 in 74. James' site has a nice calculus somewhere.

[QUOTE=LaurV;363023]@chalsall: not sure what you what to show me, but the most of the saved work in the last time, which can be seen on that graphic, is due to your and kracker's "aliens" who did P-1 work for me, and also few of my CPUs who did P-1.[/QUOTE]

Not the graphs -- I'm referring to the "Predicted" vs. "Found" columns in the table at the top of the page. In the "LL TF" range you've found 211 factors, instead of the 206.036 predicted based on the overall "kill" rate.

Edit: Cross posted... WRT "cheaters", we actually haven't had this problem. This was actually one of the reasons why this feature was implemented -- to be able to determine if anyone was "gaming" the system.

And, WRT James' calculus, yes. But I'm more of an empirical kind of guy....

I just did notice. Thanks for the [SIZE="1"]two [/SIZE]P-1. :smile:

[QUOTE=LaurV;363023]i.e. TF to 74 at LL front on two 580's and TF to 73 at LL front on a 7970, but in fact, this "normality" is going to be broken soon, following the last discussions I switched the preference to DC-TF range. Old work still queued, but it will be finished soon, like end of the year.[/QUOTE]

Thank you for helping out during Jerry's "blow-out". Craig has also rejoined our little "party", so we're looking good.

And, since you brought up DCTFing... For anyone working down there, TFing the 33M range to 71 would be best for the project. Now that several of us have moved their CPUs from P-1'ing to DCing, we're losing our lead even faster than before.

Not critical yet -- we've still got several months -- but WMS for DCTFing is 33M to 71, not the higher ranges to 69.

[QUOTE=kracker;363028]I just did notice. Thanks for the [SIZE=1]two [/SIZE]P-1. :smile:[/QUOTE]
:w00t: mwaaahaha, this was just a very unlucky setup! I didn't check that, but you are right, I only produced two P-1 for you...

Nevertheless, I still want to see your face :razz:

(midnight here, guests are sleeping, me is a bit drunk, going to bed before succeeding to make an even more fool of myself)

:pancakebunny:

EDIT: Going back to my classic, DC again.... I just might try LL, only have like 2 or 3 LL lifetime done...wish I had more GPU power :razz:

[QUOTE=chalsall;363030]... For anyone working down there, TFing the 33M range to 71 would be best for the project....

...WMS for DCTFing is 33M to 71, not the higher ranges to 69.[/QUOTE]
That's not what "Let GPU72 decide" says.
My 7970 is doing 42M to 69, because that is where it gives 150% efficiency compared with 33M to 71, but the 580's are set to "Let GPU72 decide" and the decision is "37M to 70"... (?!?!?)

(now I am for sure going to bed!)

[QUOTE=LaurV;363036]My 7970 is doing 42M to 69, because that is where it gives 150% efficiency compared with 33M to 71, but the 580's are set to "Let GPU72 decide" and the decision is "37M to 70"... (?!?!?)[/QUOTE]

:doh!: Thanks for pointing that out. SPE....

[QUOTE=kracker;363034]Going back to my classic, DC again....
...wish I had more GPU power[/QUOTE]
Thanks for the P-1 cycles. Wish you find a (missed) prime with those DCs. That would be [U]really[/U] cool!

Related to the second part, we still can arrange... just say the words...

[QUOTE=kracker;363034]:pancakebunny:

EDIT: Going back to my classic, DC again.... I just might try LL, only have like 2 or 3 LL lifetime done...wish I had more GPU power :razz:[/QUOTE]

Atm, I'm not GPU limited, but power and cooling limited. :)

It looks like I've fried a GTX Titan. :(

Don't talk to me, I'm in mourning. :)

The long story - the machine with 2x Titans, had an external 'box fan' blowing on it. We had a wind storm here, curtains were blown into the vent of the fan, pretty much stopping air flow. Machine doesn't boot up with the suspect Titan. remove Titan, machine boots. When I get time - going to try some troubleshooting. But I smell burning from it - I think it's toast.

-- Craig

Basically what happened to two of my 580s. Water system has a leak and the system didn't shutdown due to the lockup fun the liquid. Cards kept going and two fried from heat.

Who makes the titan?

[QUOTE=flashjh;363224]Basically what happened to two of my 580s. Water system has a leak and the system didn't shutdown due to the lockup fun the liquid. Cards kept going and two fried from heat.

Who makes the titan?[/QUOTE]

Nvidia.

It's similar to GTX780 but without the DP modules disabled.

-- Craig

I meant which manufacturer, they may RMA it for you?

[QUOTE=flashjh;363226]I meant which manufacturer, they may RMA it for you?[/QUOTE]

Probably EVGA or ASUS, they are the only ones who "distribute" titan's. If there is no sign of physical damage, you should be good with both.

We mourn with you!

[QUOTE=flashjh;363228]We mourn with you![/QUOTE]

Indeed. Very sorry if our suggestion you rejoin our party cost you. :sad:

[QUOTE=chalsall;363233]Indeed. Very sorry if our suggestion you rejoin our party cost you. :sad:[/QUOTE]

It's cool :)

It was the wind storm blowing my curtains into my fan that killed it :)

It was already crunching on P-1. Titans are pretty good at P-1. 70GHz-days/day (ish).

Here's the breakdown with mersenne primes
2x AMD HD7990s : TF 1300GHz-days/day
1x Titan : P-1 70GHz-days/day
2x GTX580s : TF 850GHz-days/day

-- Craig

[QUOTE=nucleon;363251]2x AMD HD7990s : TF 1300GHz-days/day
[/QUOTE]
Do you mean for one? :shock: (for two would be somehow too less, but for one is bloody high! I could get ~900G from mine, close to maximum 1.02T if I only do like 40M-45M expos to 69 bits - the most profitable variant; also, the 7970 Ghz edition gets 630G in average, doing 42M to 69 bits).

2x 7990

They run too hot, so they're clocked down a bit.

Currently running 65M@74bits.

-- Craig

I just wanted to make a note that I had thought I'd burned out my 580 from heat at one point (I run a 480 and a 580) because when I unplugeed the 580 the computer ran just fine. Turned out though that the heat had actually stressed the (1050 watt) power supply, and so it only had the juice to run one card after that. I replaced the power supply some time later, and poof, the 580 started running again. I was happy that the 580 wasn't dead, but somewhat upset that I lost almost a month of computing time on it.

[QUOTE=nucleon;363262]2x 7990

They run too hot, so they're clocked down a bit.

Currently running 65M@74bits.

-- Craig[/QUOTE]

Yeah, I heard that was a issue with the heatsink (dual 7990's). Are they really close together?

[QUOTE=kracker;363276]Yeah, I heard that was a issue with the heatsink (dual 7990's). Are they really close together?[/QUOTE]

Yeah, after I wrote my post, I checked the cards. Yeah they are pretty close.

I can unplug one and put it in the pci-e x4 slot to give it more space.

-- Craig

Tom's Hardware has a article about that here:

[URL="http://www.tomshardware.com/reviews/radeon-hd-7990-crossfire-overheat,3539.html"]http://www.tomshardware.com/reviews/radeon-hd-7990-crossfire-overheat,3539.html[/URL]

Putting more space should help but probably won't completely remove the problem.

[QUOTE=kracker;363317]Tom's Hardware has a article about that here:[/QUOTE]

This is an example of why I /so/ love "the empirical"! Theory is great and helpful, but it's only a theory.... :smile:

It seems the other pci-e slot isn't suitable.

It doesn't work quite well - speed per gpu drops to 150Ghz-d/day. (Typical is 325-340GHz-d/d)

I've installed msi afterburner to get better details on speed.

It seems it's running at defaults 950core, 1500mem.

Here is the ini file. I'm open to suggestions.

[CODE]Verbosity=2
SievePrimes=105000
SievePrimesAdjust=1
SievePrimesMin=5000
SievePrimesMax=200000
SieveSizeLimit=64
NumStreams=3
VectorSize=2
GridSize=3
WorkFile=worktodo.txt
ResultsFile=results.txt
Checkpoints=1
CheckpointDelay=300
Stages=0
StopAfterFactor=2
PrintMode=1
TimeStampInResults=1
ProgressHeader=Date Time | class Pct | time ETA | GHz-d/day Sieve Wait
ProgressFormat=%d %T | %C %p%% | %t %e | %g %s %W%%
AllowSleep=1
GPUType=AUTO
SieveCPUMask=0
SmallExp=0
SieveOnGPU=1
GPUSievePrimes=75000
GPUSieveSize=126
GPUSieveProcessSize=24
FlushInterval=3
TestSieveSizes=1,2,3,4,5,6,7,8,9,10,11,13,16,19,20,21,22,25,30,36,43,50,58,67,77,88,100,120,170
TestSievePrimes=254,310,1846,21814,67894,82000,111157,222222,444444,777777,1000000[/CODE]


-- Craig

From former discussions I remember that GPUProcessSize of 24 has a chance in 6 to miss a factor, I think you should use 16 or 32. I use 16. The only parameter which seems to influence the speed is the VectorSize (correctly set to 2 for GCN cards, the default of 4 is much worse). Also, primes, I go by the default 84486, but there is not much speed difference between 60, 80, 100k. I go by GPUType=GCN, but it does not really matter for speed, assuming it is AUTO and correctly detected as GCN.

What really makes a difference: go to the "performance" tag in the CCT and set the Overdrive/Power Control to about +15%, otherwise the card will try to keep the power consumption "in check" and throttle after few seconds (it shows that the GPU is used 99%, but in reality is only used 60-70%, the rest is throttling). For me, with 0% (or even +5%) it starts with almost 1T, but back off to ~600G in the first ~5 minutes. When I set it to +10%, still backs off after a longer time. With +15% it continues to run max speed. Of course, this means you need to properly cool the cards, but my AMD cards are air cooled too (I mean, they are not water cooled, only for the gtx stuff I use water cooling).

About the second slot, the card works on in x8 or even x4 slots, as mfakto does not have big traffic. You can check the bios, maybe the slot has strange settings there (different features disabled?).

[QUOTE=LaurV;363363]From former discussions I remember that GPUProcessSize of 24 has a chance in 6 to miss a factor, I think you should use 16 or 32. I use 16. The only parameter which seems to influence the speed is the VectorSize (correctly set to 2 for GCN cards, the default of 4 is much worse). Also, primes, I go by the default 84486, but there is not much speed difference between 60, 80, 100k. I go by GPUType=GCN, but it does not really matter for speed, assuming it is AUTO and correctly detected as GCN.

What really makes a difference: go to the "performance" tag in the CCT and set the Overdrive/Power Control to about +15%, otherwise the card will try to keep the power consumption "in check" and throttle after few seconds (it shows that the GPU is used 99%, but in reality is only used 60-70%, the rest is throttling). For me, with 0% (or even +5%) it starts with almost 1T, but back off to ~600G in the first ~5 minutes. When I set it to +10%, still backs off after a longer time. With +15% it continues to run max speed. Of course, this means you need to properly cool the cards, but my AMD cards are air cooled too (I mean, they are not water cooled, only for the gtx stuff I use water cooling).

About the second slot, the card works on in x8 or even x4 slots, as mfakto does not have big traffic. You can check the bios, maybe the slot has strange settings there (different features disabled?).[/QUOTE]


Thanks for the reply.

I've adjusted power setting to +15% but so far I can't see any improvement.

I've also adjusted GPUProcessSize to 16. It's auto detecting GCN aok.

Yeah I have a feeling there's something funky in the motherboard. I'm sure I've read somewhere where these slots had issues with GPUs.

-- Craig

[QUOTE=nucleon;363360]It seems the other pci-e slot isn't suitable.

It doesn't work quite well - speed per gpu drops to 150Ghz-d/day. (Typical is 325-340GHz-d/d)

I've installed msi afterburner to get better details on speed.

It seems it's running at defaults 950core, 1500mem.

Here is the ini file. I'm open to suggestions.

-- Craig[/QUOTE]

Do what I did. Use a ribbon cable to connect the second GPU to the MOBO and hang the GPU outside the case.

Craig: It could be that PCI-E 2.0 x4 is just not enough to feed two HD 7900 chips. People have shown that PCI-E 3.0 x16 or x8 has had extremely little impact in most applications, but 3.0 x8 is four times the bandwidth (and you have two chips going, to boot)

I haven't noticed any decrease in performance either since I added a GPU, which caused a switch to x16 to x8/x8

mfaktc/o does not care about PCI bandwidth. LL tests may, I'm not sure.

Hmm... possible GPU mismatch for [URL="http://mersenne.org/report_exponent/?exp_lo=37631929&exp_hi=&B1=Get+status"]37631929[/URL]. Anyone want to TC?


M( 37631929 )C, 0x149e74d21e55____, n = 2097152, clLucas v1.02

[QUOTE=kracker;363890]Hmm... possible GPU mismatch for [URL="http://mersenne.org/report_exponent/?exp_lo=37631929&exp_hi=&B1=Get+status"]37631929[/URL]. Anyone want to TC?[/QUOTE]

Submit your result, and I'll TC with a CPU.

[QUOTE=chalsall;363891]Submit your result, and I'll TC with a CPU.[/QUOTE]

Done. :smile:

[QUOTE=kracker;363892]Done. :smile:[/QUOTE]

Thank you. :smile:

This candidate is now on one of my machines.

Edit: ETA ~24 hours.

[QUOTE=chalsall;363895]Edit: ETA ~24 hours.[/QUOTE]
:whistle: This guy has half of the computing power in the country... even if that country is so small as Barbados, how can we compete with him? :shock:

Just remind yourself that it's a cooperative venture not a competitive one, and then focus on the lines of people you can beat. :-D

That was a trap post, I expected him to jump and say "no, I only have 6 computers" (which he said somewhere already, talking about those servers), and then I would come and say "yes, that is what I said, half of that country's computing power" :razz:

[QUOTE=LaurV;363932]That was a trap post, I expected him to jump and say "no, I only have 6 computers" (which he said somewhere already, talking about those servers), and then I would come and say "yes, that is what I said, half of that country's computing power" :razz:[/QUOTE]

LOL...

Sorry for not taking the bait. :razz:

[QUOTE=kracker;363890]Anyone want to TC?[/QUOTE]

You'll be pleased to know that your DC matches my TC.

[QUOTE=flashjh;361218]Just lost my watercooled system due to a leak. Have to wait until Saturday to see if it can be salvaged. If anything is broken, it will take time to get fixed, so -~1400 GHzDays/Day until it comes back up (if ever) :no:[/QUOTE]
Alright, it's back up and running, finally. After several weeks of mailing parts, troubleshooting and an awesome friend who did the waterblock swaps, all 4 cards came back online.

BTW - [URL="http://www.gigabyte.com/products/product-page.aspx?pid=4334#ov"]This board[/URL] is awesome! Discovered that one of the waterblocks was leaking all over the board and then it received a gush from the break. After cleaning the corrosion off, it's still able to handle 4 cards under full load with the CPU maxed out as well. Highly recommend if you're in the market.

[QUOTE=chalsall;363996]You'll be pleased to know that your DC matches my TC.[/QUOTE]
Thank you! :smile:

[QUOTE=flashjh;363998]Alright, it's back up and running, finally. After several weeks of mailing parts, troubleshooting and an awesome friend who did the waterblock swaps, all 4 cards came back online.[/QUOTE]

Coolness!!! (Joke intended... :wink:)

Thanks for the news Jerry! :smile:

MIA
 

We seem to have lost TheJudger? No assignments in GPU72 suddenly....maybe everywhere?

[QUOTE=petrw1;364004]We seem to have lost TheJudger? No assignments in GPU72 suddenly....maybe everywhere?[/QUOTE]

Not lost (at least, I hope).

Oliver let me know that he wouldn't be doing any work for at least two weeks over the holidays.

[QUOTE=LaurV;363932]That was a trap post, I expected him to jump and say "no, I only have 6 computers" (which he said somewhere already, talking about those servers), and then I would come and say "yes, that is what I said, half of that country's computing power" :razz:[/QUOTE]

And since his processors are Xeons running at 2 Ghz, that got me convinced that we are really not lacking computing power, but just data moving capacity. So instead of focussing on processors our comparing charts should be about motherboards, chipsets and memory technologies. Anyone an interest in implementing that in the GIMPS clients?

[QUOTE=tha;364084]And since his processors are Xeons running at 2 Ghz, that got me convinced that we are really not lacking computing power, but just data moving capacity. So instead of focussing on processors our comparing charts should be about motherboards, chipsets and memory technologies. Anyone an interest in implementing that in the GIMPS clients?[/QUOTE]

Sorry, I don't understand. Care to expand? :smile:

[QUOTE=kracker;364094]Sorry, I don't understand. Care to expand? :smile:[/QUOTE]

The bits are being reordered like in a multiplication, and they are being moved like from a register to a memory with all bits remaining in the same order. Basically every computer program does just these two things. George will be the one who can best tell which operations are the most critical and what the bottlenecks are inside a computer running Prime95. But reordering the bits is not the problem, getting the bits to the places where they are being reordered is what limits work being done.

I'm not understanding you either. I don't think the chipset or motherboard is of serious importance, unless that component happens to outright malfunction.

Up to a point, the CPU is the biggest consideration. At a point, memory and CPU become equally important as the memory can become too slow to keep the CPU fed. As far as I know, the "speed" at which bits are moved between the components is basically negligible.

ADDENDUM: I would be utterly shocked if you could tweak Prime95 to make a 2GHz Xeon run as fast as a 4 GHz 4th gen i7.

ADDENDUM: If I understood you correctly, you figure that because Chris is a big player with Xeons, Xeons must be really good hardware. 6 "okay" computers can do a lot of work, not counting any GPUs he may be running also.

[QUOTE=TheMawn;364127]As far as I know, the "speed" at which bits are moved between the components is basically negligible.[/QUOTE]

Incorrect. Memory bandwidth can be very important.

As I discovered during this exercise, tuning your CPUs to your caches, and your memory, can increase your throughput significantly.