Got a new WU of hcdwa_2-kind. I downloaded it half an hour ago, it will need 37 hours for calculating, and the WU will run out on Friday, 2 o'clock PM CET.

I feel a little bit under pressure, now. ;)

I doubt that it will take so long on your machine. But anyway, thanks for letting me know - the wu template is broken, that's why deadline is so short.

---

M4 Project homepage
M4 Project wiki

Got a new WU of hcdwa_2-kind...

This morning I had just seen a "hcdwa68" WU. I think that was done after 15 or 18 hours. Run on a 2.0 GHz Turion. When I'm home, I write the exact duration.

Bye, Grubix.

Got one of those too. However, my deadline is set at 22.11.2009 so the problem is probably solved :)

My deadline was 13.11., finished after 11 hours on a AMD Phenom X4 9750 @ 2.4 GHz. Some "owvha68" needed 10 hours on the same computer.

Bye, Grubix.

Star Date 63161.1, interim report

After watching two Deep Space Nine Episodes and starting the third one ("Second Skin", "The Abandoned" and now "Civil Defense"), my laptop calculated for four or five hours, and I reached the eleven percent. I doesn't seem like it will get any faster, and so I have to face the idea to put my computer in the bathroom overnight. The positive thing is that I made friends with the WU meanwhile. It introduced itself to me as "Gerdy". She is invited for lunch tomorrow.

Furthermore, another WU has entered my BOINC. It's a OWVHA68 and will run out on November 25th. Estimated calculation time: 19 hours.

On this way we'll never finish Enigma@home in the next two days. ;P

My deadline was 13.11., finished after 11 hours on a AMD Phenom X4 9750 @ 2.4 GHz. Some "owvha68" needed 10 hours on the same computer.

Bye, Grubix.

So what?? We have optimized apps don't we???? ;)

Anyhoo, if you had a linux box running the 64-bit optimized app and cranked your Phenom to 3.0GHz (like mine) then you'd have completed the same task in about 6.3 - 7.4 hours depending on the "owvha68" WU. [Enable Gloating Mode]

PS I have had one WU vzfdg48_2_11298803_0 take 33,282.41 secs (9.2 hours) so I'd stop complaining about the owvha68's, as the vzfdg48_2's will kill ya! :)

PPS Hey TJM, how's that CUDA app comin' along?

Mike D

---

Yeah, those vzfdg's are killers. Cuda sounds a good idea, Then I'd need to figure out how to install it, which I need to do anyway...

So what??

Nothing. That was just an information for other users. I wrote this only because it was discussed in this thread. I have no problem with the runtime, it's okay for me.

We have optimized apps don't we???? ;)

Yes, that's right. After this posting, I even got help with the installation.

... so I'd stop complaining about the owvha68's...

I have not complained. If it has sound like so, I apologize for that.

Bye, Grubix.

Star Date 63161.1, interim report

After watching two Deep Space Nine Episodes and starting the third one ("Second Skin", "The Abandoned" and now "Civil Defense"), my laptop calculated for four or five hours, and I reached the eleven percent. I doesn't seem like it will get any faster, and so I have to face the idea to put my computer in the bathroom overnight. The positive thing is that I made friends with the WU meanwhile. It introduced itself to me as "Gerdy". She is invited for lunch tomorrow.

Furthermore, another WU has entered my BOINC. It's a OWVHA68 and will run out on November 25th. Estimated calculation time: 19 hours.

On this way we'll never finish Enigma@home in the next two days. ;P

If it's too long to complete in deadline, just abort it. Before you do, copy everything from the BOINC slot folder and upload somewhere, I'll grant credits for the part already done manually.

---

M4 Project homepage
M4 Project wiki

Gerdy will run out in seventeen hours, and will be completely calculated in two hours. No problem. :)

Her ugly half-step-cousin-in-law was through after twelve hours.

Oh, and "Civil Defense" is a very funny episode.

So, this is it.

Rest in Pieces, Gerdy.

2009/11/10 - 2009/11/13

No one knows your true worth but me. (73 credits are a bit few, aren't they?)

After erratic deadlines (e.g. 66 hours for a 27h WU - 10637038 9941834), it seems to be some credit and duration problems :

A few minutes ago, I've got 5.42 when 15.57 claimed for more than 2 hours of CPU time (10749363 10041827 13 Nov 2009 7:14:23 UTC 13 Nov 2009 10:22:09 UTC Completed and validated 7,927.84 15.57 5.42).

I still have 11 WUs in queue, including a 7Oh WU with 60% completed (10690446 9988296). For the other 10, the expected durations have magically been multiplied by a factor of around 3 this morning before they start, so they are now running in High Priority and completion occurs at 33% of expected.

Enigmatic, isn't it ?

For now I disabled all workunits other than remaining awgly100.
I'll reenable them as soon as I find what's wrong with workunit templates. I suspect that the number of estimated fpops is too high to be stored as integer.

Don't worry about the credits, as soon as I finish other things I'll run a script to check and fix values, it will grant the (correct_credits - already_granted_credits) to users/teams.

mdoerner wrote:

PPS Hey TJM, how's that CUDA app comin' along?

The CUDA app is very far from beeing finished.
I'm still learning basics of GPU programming, the main problem is lack of time, so the progress is very slow.
I also have hardware issues, until I'll find super cheap and low power CUDA-capable card (I'm looking for something like 8400/8500/8600 with broken signal output - people sell them cheap because they're useless, but I haven't seen one in a while) I'm forced to use the emulator, which doesn't speed up things.

Another issue is the server hardware and available bandwidth. Recent test shown that a rate of 1500 workunits/hour basically saturates my upstream (I won't even mention how 'fast' is it). That's because BOINC adds a lot of overhead traffic - the workunit files are extremely small, but the scheduler request/replies are large.
That's why I'm experimenting with different workunit lengths. Fast GPU application would probably kill the server and/or my internet connection with the network traffic, plus I'd have to pay a lot for the traffic on my reverse proxy - when I exceed the monthly limit I have to pay $2/GB, which is definitely not good.

---

M4 Project homepage
M4 Project wiki

Good luck.

I also have hardware issues, until I'll find super cheap and low power CUDA-capable card (I'm looking for something like 8400/8500/8600 with broken signal output - people sell them cheap because they're useless, but I haven't seen one in a while)

I have one of those cards (GeForce 8600 GT). Very cheap and works great, and I've been running SETI's CUDA on it for about a month. It holds down the temperature well. What do you mean "broken signal output" and "useless"?

Are you referring to RFI? (If so, thanks for telling me :)

---

-Dan Q

danq.co

So what??

Nothing. That was just an information for other users. I wrote this only because it was discussed in this thread. I have no problem with the runtime, it's okay for me.

We have optimized apps don't we???? ;)

Yes, that's right. After this posting, I even got help with the installation.

... so I'd stop complaining about the owvha68's...

I have not complained. If it has sound like so, I apologize for that.

Bye, Grubix.

Relax, I was being sarcastic. Sorry that sarcasm doesn't travel well over the internet...;-) Hence, the smiley faces.

Mike D

---

Short stupid question: This project will finished today, correct? Or is there still something else to calculate for Enigma@home?

Sorry, duplicate post, please see original below.

I don't run Enigma constantly and generally download workunits every few days. My last download request was answered with, I believe, 234 work units. It didn't concern me at first because the bulk of the WUs were estimating a very fast run. I ran through about half of them before I was surprised to see that some were running out in 9 hours instead of 16-24 minutes. Running longer files is fine with me, but I can't handle large amounts of them at one time since I'm crunching other projects concurrently. This is creating a scheduling issue. I'm not concerned about credits, but I am concerned that I am having to abort some units. The WUs that are taking nine plus hours to run, originally showed estimated run times of less than 30 minutes. Could this be why I got too many? Everything I have left currently shows completion estimates of 12-26 minutes, but I'm not sure that is correct. Will you be able to generate more accurate completion times? If not, could you limit the number of units generated per download request?

On projects like World Community Grid, no matter how empty the cache, you get a maximum of 15 work units per download request. Most of the World Community Grid projects take 6-7 hours to run. Spinhenge limits are 20 workunits per download request and they won't allow download requests more frequently than every 15 minutes. Each Spinhenge WU takes about 30 minutes to run. Other projects I run seem to have similar limits.

I do have my cache set at 10 days since I only have dial-up internet. A larger cache allows me more leeway in how often I need to connect.

Regards,
Zen

---