Just wanted to say a big thanks to SciManStev for pointing me in the right direction in my quest to get my 480's crunching for seti.
It's a steep learning curve but all you volunteer testers deserve more than a high five.
I think with a little more reading and some free time i am ready to try for it.
Big thanks especially to Todd - for posting your boinc directory and the correct files i will need, Vyper and Raistmer your a credit to the cause.

No problem..

Make sure that you join GPU Users Team instead for expressing your gratitude :P ..

Well seriously then, glad we could help.

Kind regards Vyper

---

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group

The number of ULs shouldn't be a prob.
From my experiences the size of WUs on HDD could make BOINC unuseable.

But, it's look like the 6.10.40+ versions have solved this prob.

Looks like not all will agree with this observation:

> If you have many uploads outstanding, your BOINC Manager will
> eventually complain it can't connect to the client. With all the
> problems people have with that. They inadvertently clicked the
> Transfers tab and can't get BM to react anymore.

I personally never had so much results pending because to get so many results I would need to get too many tasks first and BOINC would freeze on that much before task will bring result to upload.
But this could be problem (if this has place) with downloads inhibition removed.
In short, the presence of this bug should be tested.

I have current ~ 2,400 results backlogged in the UL overview of my BOINC Manager and I can look to all overviews without a prob. Of course, sometimes I get a message like 'please wait, need to communicate with client', but after max. ~ 10 sec I have a refreshed overview. BOINC V6.10.18 .

If I see 'only' the counter counting backwards, also if I switch on one results for an UL try, all well with the BOINC Manager (but of course, current because of the disabled UL server no UL possible).

---

Good morning/afternoon/evening all.
I would just like to say to all that have given me kudos - you are all very welcome!

Also I see that the Upload server is back online (thanks to whomever fixed that)
I wonder how long it will be til I get my 4800 tasks uploaded from the weekend
:-D

---

Hello all, well thanks to you all i am cautiously optimistic that my 480 is now working ok.
My only concern is that the only work i am receiving now is 12 hour cuda wu's and 16 hour regular wu's. Someone has suggested i use rescheduler, if this is true i will download and try it but have not used it before so i'm a bit nervous lol.
Am going to watch this machine very closely in the next few days, if there is nothing to worry about after 3-5 days i will invest in more.
Vyper i do appreciate all your help and advice i really do but if my home team caught me i would be tried for treason, they've been a great source of info too and with a team stomp on the horizon am sorry but there's no way i could leave them now.
Thank you one and all and SciManSte for your pointers, have a good day.

---

Hello all, well thanks to you all i am cautiously optimistic that my 480 is now working ok.
My only concern is that the only work i am receiving now is 12 hour cuda wu's and 16 hour regular wu's. Someone has suggested i use rescheduler, if this is true i will download and try it but have not used it before so i'm a bit nervous lol.
Am going to watch this machine very closely in the next few days, if there is nothing to worry about after 3-5 days i will invest in more.
Vyper i do appreciate all your help and advice i really do but if my home team caught me i would be tried for treason, they've been a great source of info too and with a team stomp on the horizon am sorry but there's no way i could leave them now.
Thank you one and all and SciManSte for your pointers, have a good day.

The Rescheduler tool will help.
http://lunatics.kwsn.net/12-gpu-crunching/cpu-gpu-rebranding-perl-script.msg19454.html#msg19454

It takes units called VLARs (Very Low Angle Range), which take a long time on a GPU, but crunch normally with a CPU, from the GPU, and gives them to the CPU's. This means that the GPU work you do will go much faster. I am glad you got everything working.

Steve

---

Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website

Thank you Steve, i will be reading up on the subject of rescheduler this evening.
I have a good idea of how excited you were when you got 2 working in an awesome set up too, its been worth the late nights and sore eyes.
Happy Crunching Sir !

I run mine every hour automatically. The default is every four hours, which should be fine. It would be nice to have it integrated, but for now it is a stand alone. I have to run and minimize it after every reboot, or if I accidently shut it off. That was the quickest link I could find to it.

That link came from Sutaru, thanks!

Steve

---

Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website

ToxicTbag,
Your computers are hidden so I couldn't check for myself but, what are you running that takes 12 hours on GPUs and 16 hours on CPUs? Even my little celeron dual was faster than that. My new E5400 is running an AP in around 19.

---

PROUD MEMBER OF Team Starfire World BOINC

I run mine every hour automatically. The default is every four hours, which should be fine. It would be nice to have it integrated, but for now it is a stand alone. I have to run and minimize it after every reboot, or if I accidently shut it off. That was the quickest link I could find to it.

That link came from Sutaru, thanks!

Steve

Ok, I know how to make It run after a reboot, I could do that with one hand tied behind My back, as that's too easy to do, 30 years of computer usage is handy sometimes. :D

I hadn't tried putting in the start up folder yet. When it unzips, it has drivers with it that I figured it needed when loading. Tomorrow, I'll try to auto load it. If it works, then I will use it that way. I hope I never stop learning!

Steve

---

Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website

I run mine every hour automatically. The default is every four hours, which should be fine. It would be nice to have it integrated, but for now it is a stand alone. I have to run and minimize it after every reboot, or if I accidently shut it off. That was the quickest link I could find to it.

That link came from Sutaru, thanks!

Steve

Ok, I know how to make It run after a reboot, I could do that with one hand tied behind My back, as that's too easy to do, 30 years of computer usage is handy sometimes. :D

I hadn't tried putting in the start up folder yet. When it unzips, it has drivers with it that I figured it needed when loading. Tomorrow, I'll try to auto load it. If it works, then I will use it that way. I hope I never stop learning!

Steve

;-)


I tried the autostart with ReSchedule tool, but after every reboot of my PC I got a strange error in it and it didn't worked (WinXP Home 32bit).
I guess it need a delayed execution after reboot in my OS. But I don't know how I could do it. :-#

So all the time I make a reboot the first activity is always to click on the icon on the WinDesktop. :-D

---

I run mine every hour automatically. The default is every four hours, which should be fine. It would be nice to have it integrated, but for now it is a stand alone. I have to run and minimize it after every reboot, or if I accidently shut it off. That was the quickest link I could find to it.

That link came from Sutaru, thanks!

Steve

Ok, I know how to make It run after a reboot, I could do that with one hand tied behind My back, as that's too easy to do, 30 years of computer usage is handy sometimes. :D

I hadn't tried putting in the start up folder yet. When it unzips, it has drivers with it that I figured it needed when loading. Tomorrow, I'll try to auto load it. If it works, then I will use it that way. I hope I never stop learning!

Steve

;-)


I tried the autostart with ReSchedule tool, but after every reboot of my PC I got a strange error in it and it didn't worked (WinXP Home 32bit).
I guess it need a delayed execution after reboot in my OS. But I don't know how I could do it. :-#

So all the time I make a reboot the first activity is always to click on the icon on the WinDesktop. :-D

Sutaru,
One possibility is to use a windows script file and place that in your startup folder.

You would need to create a file e.g. reschedule.vbs which contains something like :-


Dim WshShell, oExec
Set WshShell = CreateObject("WScript.Shell")

WScript.Sleep 20000

Set oExec = WshShell.exec("c:\reschedule.exe")


The sleep command is in milliseconds so sleep 20000 delays for 20 seconds etc and then reschedule is executed.

Hope this helps.


John.

---

GPU Users Group

Or a batch file named "Reschedule.bat"

Del Reschedule.log
ReSchedule.exe /Autorun 95
Del D:\BOINC\client_state2010*.xml
exit

NOte that this batch file even cleans up the left behind client_state2010*.xml files.

Then call the batch file from a Scheduled Task which runs the batch file at any time you wish. I use 00:00AM and 12:00PM for twice a day.

[edit]You should edit the paths to match your system. My rescedular stuff is in a folder named Reschedular in the Boinc working directory.

---

Boinc....Boinc....Boinc....Boinc....

You are welcome!


You can't imagine the booster? ;-)
I guess.. only guessing, you would see at least halving of crunching times with a real CUDA_Fermi_app. Double RAC!

The nVIDIA site say, the GTX480 have 480 CUDA cores.
But this are the shader cores. 480 shader cores, are 60 CUDA cores.

So if the CUDA V12b app detect only 15 CUDA cores and only use 15 CUDA cores, then theoretically you have 45 unused CUDA cores.
Then it would mean you would see a quarter crunching times. Fourfold RAC..

But I don't want arouse too much anticipation.. :-)


If current the Fermi chip is only used to 50 or 25 %, you could imagine that the temps will increase later with 100 % usage, or? ;-)


Maybe someone with more knowledge about the CUDA V12b app and the Fermi chip could jump in here? Which performance (RAC) would be to expect with a real CUDA_Fermi_app and 100 % usage of the Fermi chip?

Im intrigued by this as I was unaware of the difference between shaders and CUDA cores - thx for the heads up.
I sent my problematic 295 back and was ghoing to get another, but you have me thinking
All my calculations are based on the following site
http://www.nvidia.co.uk/object/graphics_cards_buy_now_uk.html

Once I have my money back, would I be better going for another 295 or a 470 - the card would be used purely for crunching??? Im afraid that the 480's are out of my budget, but the 470's have come down in price enough to make them a possibility

---

Im intrigued by this as I was unaware of the difference between shaders and CUDA cores - thx for the heads up.
I sent my problematic 295 back and was ghoing to get another, but you have me thinking
All my calculations are based on the following site
http://www.nvidia.co.uk/object/graphics_cards_buy_now_uk.html

Once I have my money back, would I be better going for another 295 or a 470 - the card would be used purely for crunching??? Im afraid that the 480's are out of my budget, but the 470's have come down in price enough to make them a possibility

I'm afraid there is no difference between shaders and CUDA cores.

There are, however, significant differences between the Fermi (GTX 470/480) architecture and earlier CUDA cards. One big difference is that the Fermi cards have 32 shaders per Streaming Multiprocessor: the earlier cards only had 8 shaders per SM (this caught out BOINC in the early days, which reported a low estimated speed for Fermis because it still assumed only 8 shaders per SM).

The old GTX 295 had two separate logical GPUs, each with 30 multiprocessors or 240 shaders.

The new GTX 470 has 14 of the new multiprocessors, for a total shader count of 448 in a single logical GPU.

[Data from Comparison of Nvidia graphics processing units at Wikipedia]

It would probably be true to say that the current SETI application (see Running SETI@home on an nVidia Fermi GPU) doesn't yet make most efficient use of a GTX 470/480, but hopefully a new version will follow soon after the CUDA 3.1 development tools and 256-series drivers are released to the public.

I'm afraid there is no difference between shaders and CUDA cores.

There are, however, significant differences between the Fermi (GTX 470/480) architecture and earlier CUDA cards. One big difference is that the Fermi cards have 32 shaders per Streaming Multiprocessor: the earlier cards only had 8 shaders per SM (this caught out BOINC in the early days, which reported a low estimated speed for Fermis because it still assumed only 8 shaders per SM).

The old GTX 295 had two separate logical GPUs, each with 30 multiprocessors or 240 shaders.

The new GTX 470 has 14 of the new multiprocessors, for a total shader count of 448 in a single logical GPU.

[Data from Comparison of Nvidia graphics processing units at Wikipedia]

It would probably be true to say that the current SETI application (see Running SETI@home on an nVidia Fermi GPU) doesn't yet make most efficient use of a GTX 470/480, but hopefully a new version will follow soon after the CUDA 3.1 development tools and 256-series drivers are released to the public.

My 295 cost me £250, and Im seeing 470's for a shade under £300. I know that the 470/480 is not fully utilised yet, but if Im going to splash out on a card purely for crunching , Im now thinking that the 470 looks a good bet? Especially once it is fully integrated into seti?

Wikipedia says for the 470 448:56:40 vs 480:60:48 for the 480 at £125 more, so its not too crippled in comparison...

What do you think?

---

The latest client does have full support for the Fermi cards and from my usage it does work well. Nvidia has also release their 256.15 drivers (beta) that improve double precision floating point operations - the CUDA 3.1 DLL's (found in the toolkit) are required and not integrated into this build. Maybe Lunatics will take this task on in the future.

Todd

---

The latest client does have full support for the Fermi cards and from my usage it does work well. Nvidia has also release their 256.15 drivers (beta) that improve double precision floating point operations - the CUDA 3.1 DLL's (found in the toolkit) are required and not integrated into this build. Maybe Lunatics will take this task on in the future.

Todd

Ofcourse we will ;)

//Vyper

---

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group

ToxicTbag,
Your computers are hidden so I couldn't check for myself but, what are you running that takes 12 hours on GPUs and 16 hours on CPUs? Even my little celeron dual was faster than that. My new E5400 is running an AP in around 19.

Hi perryjay
Apologies for not answering sooner have been buisy.
I am a relative novice to crunching and am having some real problems, i do not ask questions unless i have to and i have tried to make full use of the message boards and search engines.
I have 3 machines and decided to get a 480, SciManStev gave me some great pointers and have subscribed to the thread.
I am running w7 64 bit on an i7 930 6gb ram, have installed boinc 6.10.56 and lunatics .36 no astropulse sse4.1 intel only.
Downloaded and used Todds files and edited app_info accordingly, first time it worked great tasks were completing and validating ok then it slowed to a crawl followed by errors.
I have been at my screen for allmost 4 days now and i cannot seem to get it working.
There are a few of us on the OcUk team who really want to get these working myself and MadMac included.
Kudos to all the volunteer testers but i am plain stuck.
Vyper if your watching you said a few days ago i could show my appreciation by joining the gpu users group.
I give you my word i will join and your welcome to the output of a 480 if i can get it working, i would have to create a new user as i cant leave my team.
My computers are visible perryjay....i was just trying to hide the embarrassment of a non working 480 lol!

Thanks ToxicTBag,
Sorry for being so nosy, I hope you get this running. You are in the hands of some of the best on these boards as far as the new 480s go. If anyone can get you back up to speed they can. Never worry about asking questions here. Most of us are more than willing to answer or point you to the place to get an answer.

---

PROUD MEMBER OF Team Starfire World BOINC

ToxicTbag,
Your computers are hidden so I couldn't check for myself but, what are you running that takes 12 hours on GPUs and 16 hours on CPUs? Even my little celeron dual was faster than that. My new E5400 is running an AP in around 19.

Hi perryjay
Apologies for not answering sooner have been buisy.
I am a relative novice to crunching and am having some real problems, i do not ask questions unless i have to and i have tried to make full use of the message boards and search engines.
I have 3 machines and decided to get a 480, SciManStev gave me some great pointers and have subscribed to the thread.
I am running w7 64 bit on an i7 930 6gb ram, have installed boinc 6.10.56 and lunatics .36 no astropulse sse4.1 intel only.
Downloaded and used Todds files and edited app_info accordingly, first time it worked great tasks were completing and validating ok then it slowed to a crawl followed by errors.
I have been at my screen for allmost 4 days now and i cannot seem to get it working.
There are a few of us on the OcUk team who really want to get these working myself and MadMac included.
Kudos to all the volunteer testers but i am plain stuck.
Vyper if your watching you said a few days ago i could show my appreciation by joining the gpu users group.
I give you my word i will join and your welcome to the output of a 480 if i can get it working, i would have to create a new user as i cant leave my team.
My computers are visible perryjay....i was just trying to hide the embarrassment of a non working 480 lol!

I was wondering what you 480 temps are. I remember Todd saying that after 80°C, they back throttle and bog down. Mine are running at 46°C, and have driven my RAC up 10000 credits in 1 week in spite of the outages, and me being out of GPU work for over 24 hours. And thank you for the kind words.

Steve

---

Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website