I have 2 GTX295 on an i7 machine. It's crunching around 800 WUs per day. Quota is 100 WUs per cpu and per day, so 800 WUs per day for an i7. It's too small, especially if there is a lot of large angle range WUs during a day.
Please, increase !

My machine

Thank you,
ThierryH

---

I congratulate you on being able to crunch so much in a short period of time. Seti and other projects intentionally set the limit to 100/core. If you were to have a hard drive failure or some other catastrophe. you'd put 900+ WU's on hold which start to eat up database space when you consider the number of people that get large caches and either quit the project or have other failures. There are other good reason that someone would point out.

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

I would say that 2 GTX295s should at LEAST be counted as 4 cores and allow for 100WU/day as well as the CPU cores.

/Anton

Maybe a better way would be to let the user process all of the units until the quoter is reached and then let the user download 10 or so at a time until it is reset at the end of the day.

The above would depend on reporting successful work done and not having the original quoter used up by errors or aborts.

To clarify:

Good
1. Process and report 100 units successfully.
2. Download additional work in small chunks.

Bad
1. Report lots of errors and abort work you done like the look of.
2. No more work for you until tomorrow.

I also recall reading that Seti recommends running 1 or more of the other projects available through BOINC. I don't think any suggestion to expand WU quotas would be taken seriously. I would bet that S@H would prefer to increase the data being crunched per WU than change the daily quota.

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

Maybe a better way would be to let the user process all of the units until the quoter is reached and then let the user download 10 or so at a time until it is reset at the end of the day.

The above would depend on reporting successful work done and not having the original quoter used up by errors or aborts.

To clarify:

Good
1. Process and report 100 units successfully.
2. Download additional work in small chunks.

Bad
1. Report lots of errors and abort work you done like the look of.
2. No more work for you until tomorrow.

This does in no way contradict having a WU dl-quota for GPU cores. Let the work flow. The more the merrier! :)

In ThierryH's case though, 10 units at a time would probably not be enough 'extra ration'...since he computes 12 in parallell...

As stated before Seti@home encourages people to have additional projects to work on. I'd be happy as a clam to get that much done in one day. However its been their policy to not increase the # of Wu's being sent out but to increase the detail/work processed in each WU. I don't see Berkeley changing their way of doing things for relatively few individuals.
I'd bet that they appreciate all your work but instead of complaining about something they aren't changing just for you try working another project in your spare time.

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

In ThierryH's case though, 10 units at a time would probably not be enough 'extra ration'...since he computes 12 in parallell...

Indeed, maybe 1 or 2 extra units per processing unit would be more logical.

I'd bet that they appreciate all your work but instead of complaining about something they aren't changing just for you try working another project in your spare time.

Its probably even more productive to talk about possible solutions and join other projects to take up the slack.

For example: Maybe someone can raise this with the BOINC people for it be be implemented and maybe even submit a patch if the development is open. (And project administrators can use it or not).

Well just letting the manager see the GPUs as a core would work, with 12 cores you could get up to 1200 WUs a day, a lot more than the machine would be able to crunch, I think....

---

Even my gtx 260 core 216 can chew through over 400 wu's a day for the 13 credit wu's... Plus then there is the aborted VLAR wu's... (sorry I wont run wu's that take over 2 hours when they shouldn't and make my PC sluggish)
It took me over 5 days to get 1000 wu's in que (not even a 5 day buffer)
But a long buffer is currently needed because if I came across a bunch of shorties and VLAR then I would chrun through over half of my que in a single day...

---

Do you Good Search for Seti@Home? http://www.goodsearch.com/?charityid=888957
Or Good Shop? http://www.goodshop.com/?charityid=888957

From the boinc_dev mailing list:

From: David Anderson <davea@ssl.berkeley.edu>
Subject: Re: [boinc_dev] Work-fetch and gpu quotas
Date sent: Mon, 19 Jan 2009 17:03:50 -0800
Copies to: BOINC dev <boinc_dev@ssl.berkeley.edu>

I checked in a change to deal with this problem.
I added a scheduler config parameter "cuda_multiplier".
If the basic quota is K,
the actual quota for a host with N CPUs and M CUDA devices will be
K * (N + M*cuda_multiplier)
(i.e. cuda_multiplier reflects how much faster GPUs
are than CPUs for this project).

I'll deploy this on SETI@home (with a value of 5 or so) soon.

-- David

                                                                 Joe

From the boinc_dev mailing list:

From: David Anderson <davea@ssl.berkeley.edu>
Subject: Re: [boinc_dev] Work-fetch and gpu quotas
Date sent: Mon, 19 Jan 2009 17:03:50 -0800
Copies to: BOINC dev <boinc_dev@ssl.berkeley.edu>

I checked in a change to deal with this problem.
I added a scheduler config parameter "cuda_multiplier".
If the basic quota is K,
the actual quota for a host with N CPUs and M CUDA devices will be
K * (N + M*cuda_multiplier)
(i.e. cuda_multiplier reflects how much faster GPUs
are than CPUs for this project).

I'll deploy this on SETI@home (with a value of 5 or so) soon.

-- David

                                                                 Joe

Thank you for the answer.
ThierryH.

---

Today, there was a lot of very large angle range. My box took only 12 hours to crunch daily 800 WUs. Now waiting tomorrow 800 next on GPUGrid.

ThierryH

---

If my calcs are right, when the new rule is implemented you should have an allocation of 1800/day. (If multiplier value = 5)

That should keep you going 24/7 most if not all of the time.

If my calcs are right, when the new rule is implemented you should have an allocation of 1800/day. (If multiplier value = 5)

That should keep you going 24/7 most if not all of the time.

2800 ;)
GTX295 have two GPUs on board. So 100 * ( 8 + 5 * 4 ).
It's effectively enougth. Probably too much for the moment.

---

Say Thierry,
Of what makes are your GTX295s? I'm curious because several users of different projects report that the Nvidia drivers only let them use one core of each card. The EVGA cards are mentioned in particular.

Do you have them working 2 WU/each at one time (for a total of 4)?
Did you have to configure them in any way special? (sli/physix)

I have a pair on the way as well and I am interested in knowing if WinXP x64 works...WinXP x86 apparantly works, Vista (x64 or x86) apparently does not.

/Anton

Seems like it is effective already, seen a 700 units limit being reached on my client (2 CPUs + 1 GPUx5)

Unfortunately I think the client still grossly underestimates the productivity of CUDA and fails to compensate by increasing the queue proportionately.

---

Unfortunately I think the client still grossly underestimates the productivity of CUDA and fails to compensate by increasing the queue proportionately.

Which client?

---

Say Thierry,
Of what makes are your GTX295s? I'm curious because several users of different projects report that the Nvidia drivers only let them use one core of each card. The EVGA cards are mentioned in particular.

Do you have them working 2 WU/each at one time (for a total of 4)?
Did you have to configure them in any way special? (sli/physix)

I have a pair on the way as well and I am interested in knowing if WinXP x64 works...WinXP x86 apparantly works, Vista (x64 or x86) apparently does not.

/Anton

My GTX295s are from Gainward. Usage of one or two GPUs is only a setup. You have to disable SLI to have two GPUs on one card, able to crunch two WUs same time.
Like you could see on my machine, I'm working under XP64. Just be sure to take nVidia driver 181.20 or more to avoid all troubles there was by the past with SETI app.

/Thierry

---

@ [AF>HFR>RR] ThierryH

Now you are down to:
Maximum daily WU quota per CPU 1/day

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<message>
app_version download error: couldn't get input files:
<file_xfer_error>
  <file_name>setiathome_6.08_windows_intelx86__cuda.exe</file_name>
  <error_code>-120</error_code>
  <error_message>signature verification failed</error_message>
</file_xfer_error>

</message>
]]>
 
Validate state Invalid 
Claimed credit 0 
Granted credit 0 
application version 6.08

Hope the Berkeley-crew will fix this soon..


EDIT:
BTW.
Nice rig.. :-)

---