TPU's WCG/BOINC Team

[4x4n]

It might be time to buy bags of ice and throw it in there- save the meat

Or dry ice if there's someplace near you that you can get it. It will last longer. Just don't let it contact the meat directly. A piece of cardboard should work fine. Also, if you use a lot of it, make sure the area is well ventilated so the CO2 can escape.

Power is back, thanks for the advice guys. I did pack it with ice and everything made it just fine. I keep it about -10f so I figured it would be ok as long as we didn,t go too long without power. In 2006 we had another big storm and we lost power for 10 days. I had a freezer full then too, had to give it all away.

 

[Arjai]

OK, I will not be able to have a constant wifi connection until the 10th of September. My pay raise did not clear through the maze of idiots in the Corporate office, yet.

So, @[Ion] can you hold up the Daily's until then? Thanks!

I have been hitting the coffee shop, every couple days, to upload my WU's and to stop in here, once in a while.

Take Care Team!! CRUNCH ON!!

 

[[Ion]]

OK, I will not be able to have a constant wifi connection until the 10th of September. My pay raise did not clear through the maze of idiots in the Corporate office, yet.

So, @[Ion] can you hold up the Daily's until then? Thanks!

I have been hitting the coffee shop, every couple days, to upload my WU's and to stop in here, once in a while.

Take Care Team!! CRUNCH ON!!

Shall be done!

 

[t_ski]

So does anyone notice any improvement in the new 7.6.6 client?

 

[[Ion]]

So does anyone notice any improvement in the new 7.6.6 client?

Yes. 7.0.x on my 3930k and 3770k kept giving "Transient HTTP Error" while attempting to fetch work, and they ran dry a few times. With 7.6.6 all seems well.

 

[Nordic]

@t_ski, I posted my findings on the last page.

Since I now know you can see the numbers from boincs built in benchmark, I decided to test if there was a difference from 7.4.42 to 7.6.6. I did not expect any, but I had an increase on my whetstone score on two systems which makes we wonder if it was not coincidence. Both times I tested, I ran the the benchmark twice and recorded the highest score. Each time tested was after a reboot.

2500k - 4709 Whet - 14747 Dhry - Boinc 7.4.42
2500k - 5591 Whet - 1396 Dhry - Boinc 7.6.6
G3258 - 3298 Whet - 13565 Dhry - Boinc 7.4.42
G3258 - 3975 Whet - 13382 Dhry - Boinc 7.6.6
2217u - 1907 Whet - 5962 Dhry - Boinc 7.4.42
2217u - 2295 Whet - 5423 Dhry - Boinc 7.6.6

That is about a 20% increase in the whetstone score from all three? I don't understand how a different version of boinc could increase this cpu benchmark score. Hopefully it results in more ppd.

 

[xvi]

Just got power back. Crunching shall resume soon.

 

[[Ion]]

According to the Top Hosts page on BOINCStats, the top-two hosts by RAC are at 37k and 35k PPD. Minsk has been doing about 38.5k (albeit only over a couple days). This means that if I can keep it fed with the OET WUs it loves so much, I'm all but guaranteed a top-3 (out of 2.5 million hosts) spot. Throw on a 10% OC (reasonable, I think; the 6128 does 20% easily) and that pretty much guarantees #1

Does anyone know a reasonable max temp for the Opty 6300 CPUs? They're running mid-upper 40s now which I assume must be OK but how high can I go?

 

[m&m's]

Does anyone know a reasonable max temp for the Opty 6300 CPUs? They're running mid-upper 40s now which I assume must be OK but how high can I go?

AMD doesn't like heat, do not exceed 60°C for a long period of time. AMD chips are known to slowly degrade themselves above 62°C. The best would be to keep it below 55°C.

 

[[Ion]]

AMD doesn't like heat, do not exceed 60°C for a long period of time. AMD chips are known to slowly degrade themselves above 62°C. The best would be to keep it below 55°C.

Well, that's much cooler than I thought was permissible. My FX8350 was running very close to that so I'll check when I get home.
But it sounds like I have 8C or so of headroom at the moment which should certainly let me get another 10% or so out of the CPUs.
Thanks
I'd really hate to lose 64 cores worth of CPUs

 

[Norton]

AMD doesn't like heat, do not exceed 60°C for a long period of time. AMD chips are known to slowly degrade themselves above 62°C. The best would be to keep it below 55°C.

+1 I would also recommend sticking to 55C or less.

 

[m&m's]

Now I want to predict average ppd based on the whetstone score. Given time, maybe I will be able to predict it reliably.

It will be very hard based on whetstone only. Lets say a Windows user scores 5000whetstone and 13000dhrystone while a Linux users scores 5000whetstone and 22000dhrystone, what is the effect on the PPD? I'm pretty sure the person with a better dhrystone score will have a higher PPD. So I think that you need to take in consideration both scores.

 

[[Ion]]

+1 I would also recommend sticking to 55C or less.

Alright. Well the basement definitely seems like a good place then b/c it's nice and cool down there.

It will be very hard based on whetstone only. Lets say a Windows user scores 5000whetstone and 13000dhrystone while a Linux users scores 5000whetstone and 22000dhrystone, what is the effect on the PPD? I'm pretty sure the person with a better dhrystone score will have a higher PPD. So I think that you need to take in consideration both scores.

I could do a regression in R to try and figure out something. Busy this week but could be a weekend project

 

[ChristTheGreat]

Hey guys!

I might reduce my main rig and server BOINC time a little bit, just for economy, while it is not alot, as Saturday, I got a news. My wife is pregnant , so any economy I can do so we can continue having a life (cause the house needs some renovation)

 

[Norton]

My wife is pregnant

Congrats to you both!

 

[Peter1986C]

Hey guys!

I might reduce my main rig and server BOINC time a little bit, just for economy, while it is not alot, as Saturday, I got a news. My wife is pregnant , so any economy I can do so we can continue having a life (cause the house needs some renovation)

Congrats!

 

[Arjai]

OK, so, here I am, sweating it out outside the coffee shop, loading up my cache of done WU's. I check my bank balance, rent is due.

Turns out my Wifi provider, somehow got it in their head to auto charge me. Good news, I have wifi. Bad news, my rent is gonna be short! I love my new apartment, for a ton of good reasons. But the rent, controlled through HUD, is a bit more than I believe it should be. The woman, in charge of my VA HUD stuff, will not meet with me. In Fact, she is currently on weeks vaca due to Labor DAY.
Does anyone else see the irony in this?

I am working my ass to the bone, shopping at food selves because I am perpetually broke, I make too much money for EBT, but can't afford to eat enough to maintain my weight!

I have lost over 20 LBS!!

I am hoping, one of two things, three actually, happens soon. 1. I get a raise that is at least $2.00 an hour ( pipe dream ). 2. I manage to convince the HUD lady to sit down with me in a reasonable amount of time. 3. I hit some sort of Lottery/ Million dollar idea, in the next two months.

None of those things are likely to happen. I'm unsure what to cut back. Looks like it may be the girlfriend that goes next!

Then for sure, I will have no life. As I whittle away into a skeleton of myself. Sound's good, no?

If I can manage to make it through January, I am going to enroll in school. Get, at least, a Welding Certificate. If nothing else, I can find work that will pay better than what I am doing now.


30% of $25 an hour is more rent but, 70% of $25 an hour is much better than 70% of $10 an hour!!

Anyway, @[Ion] looks like I can relieve you sooner than I thought. So, go ahead and do the post tonight, I am beat. I will resume on Thursday night, hopefully a bit more rested/ less of a downer!!

Thanks a Bunch for helping out!!

 

[[Ion]]

A stripped-down WCG system really is remarkably efficient on the RAM:

64 WUs running on a CLI environment using slightly under 5GB of RAM. Not bad

 

[manofthem]

Hey guys!

I might reduce my main rig and server BOINC time a little bit, just for economy, while it is not alot, as Saturday, I got a news. My wife is pregnant , so any economy I can do so we can continue having a life (cause the house needs some renovation)

Congratulations buddy, that's awesome! Enjoy the family addition!

 

[Norton]

Beta's are flowing again!

I have about a dozen or so either in progress or finished up so far...

 

[t_ski]

Beta's are flowing again!

I have about a dozen or so either in progress or finished up so far...

You got betas? This is what I got...

Spoiler

 

[Nordic]

It will be very hard based on whetstone only. Lets say a Windows user scores 5000whetstone and 13000dhrystone while a Linux users scores 5000whetstone and 22000dhrystone, what is the effect on the PPD? I'm pretty sure the person with a better dhrystone score will have a higher PPD. So I think that you need to take in consideration both scores.

It actually was not that hard to predict ppd based on the whetstone and dhrystone scores. I got a formula that so far seems to give pretty close ppd predictions based on what I expect from my hardware.

The math I used to do this is pretty simple.
My 2500k got 5934 ppd today and my g1620 celeron got 1978 ppd today according to boinc stats. My 2500k has whet and dhry scores of 4921 and 25461. My g1620 has whet and dhry scores of 3408 and 17779. I then did ppd / (score * threads). I got a two ratios, one for whetstone and one for dhrystone. I then averaged those ratios from the two systems coming to the final ratios of 3.3815 and 17.5698.

Or you can look at this complete formula.
Predicted ppd =((Whetstone score*Threads)/3.3815+(Dhrystone*Threads)/Dhrystone Ratio)/2

The formula is good I think, but the ratios are not perfect. Still though, the predicted ppd is right in the ballpark of what I expect from my systems.

A stripped-down WCG system really is remarkably efficient on the RAM:

64 WUs running on a CLI environment using slightly under 5GB of RAM. Not bad

Did you use anything from the new linux tips thread, or stuff you already knew? Anything to add?

 

[[Ion]]

It actually was not that hard to predict ppd based on the whetstone and dhrystone scores. I got a formula that so far seems to give pretty close ppd predictions based on what I expect from my hardware.

The math I used to do this is pretty simple.
My 2500k got 5934 ppd today and my g1620 celeron got 1978 ppd today according to boinc stats. My 2500k has whet and dhry scores of 4921 and 25461. My g1620 has whet and dhry scores of 3408 and 17779. I then did ppd / (score * threads). I got a two ratios, one for whetstone and one for dhrystone. I then averaged those ratios from the two systems coming to the final ratios of 3.3815 and 17.5698.

Or you can look at this complete formula.
Predicted ppd =((Whetstone score*Threads)/3.3815+(Dhrystone*Threads)/Dhrystone Ratio)/2

The formula is good I think, but the ratios are not perfect. Still though, the predicted ppd is right in the ballpark of what I expect from my systems.

I even took the liberty of making a quick excel spreadsheet for people to play with. I won't leave it up forever, but let me know how close it predicts your ppd.
https://docs.google.com/spreadsheets/d/1pAoTwn7MDZnorMo07-ZW8ME3IDGrQhUIIJsORid6Uhg/edit?usp=sharing



Did you use anything from the new linux tips thread, or stuff you already knew? Anything to add?

Just stuff I already knew from past experience. I'll read the Tips after work tonight to see if I have anything to add.

 

[Nordic]

I made this excel file that will allow you to input your number of threads, whetstone score, and dhrystone score. It will then output a predicted ppd for that system. I made it because I am curious how well my formula predicts, and this makes it easy for you to test. So let me know how well it works.

https://docs.google.com/spreadsheets/d/1qKICdaipq91f3jmrtGvzjZuiRmi6T87vINnhpf4IuRo/edit#gid=0

I won't leave this link up forever. If you saw the link in my previous post, it seemed to have an issue. This one is working fine.

 

[m&m's]

It actually was not that hard to predict ppd based on the whetstone and dhrystone scores.

I said that it would be hard based on whetstone only and that you needed both scores and you did use both scores.

Predicted ppd =((Whetstone score*Threads)/3.3815+(Dhrystone*Threads)/Dhrystone Ratio)/2

Now, you consider that both scores have a value of 50% (you divide the score by 2), but we don't know that. Maybe it's 60/40% or 70/30%. Are the units sent by WCG more floating point or integer dependent or is it 50/50?
Following your equation a computer that scores 5000 Whetstone and 20000 Dhrystone will have a PPD ~25% higher than another computer scoring 5000 Whetstone and 10000 Dhrystone, but if WCG is more floating point dependent then the difference will be less or if WCG is more integer dependent then the difference will be more.
Also the type of units that is run will influence the PPD.

In other words, we need to know what is the influence of each score on the PPD to do a proper math equation. A good way to find out would be to run a computer a month on Windows and another month on Linux and then compare the scores and PPDs.

At the end of the day, what you want to do is predict approximately the PPD, so it might do the job as it is.