My gripe du jour is that the current '6 WUs/cpu [=core?]/day'...

That's not per day, it's at any given time. Once the computer has returned a few, it'll get more. That quota reduces the number of WU's 'currently out there' that the database needs to keep track of, and it helped remedy a major 'no work sent' issue for all of us.

The *daily* quota is in the thousands.

---

>helped remedy a major 'no work sent' issue for all of us.
Well I hope someone's getting WUs, my messages tab is glowing red with 'no work sent/reached per-cpu limit'.

---

Cheers,

PeterV

.

This is what frustrates me running the GPU WU's, right now my i7 is running 6 Regular WU's & 1 Milkyway WU with 4 Milkyway WU's in a running/waiting state. So The GPU WU's are taking up 2 Cores to run 1 WU !!!

It's like theres no happy medium right now on mt Box's, 1 run or 7 run ...

Thats why I put the ATI card in one of my slowest machines ... if I lose a core on it I don't lose as much ...

CP has given us a kludge that kinda sorta works almost perfectly. But, we are fighting BOINC with a strap on booster and the inconsistencies are what hang us up. Heck, in a way, we have the same issues with CUDA it is just a little bit smoother ... but, the only way to keep your CUDA cards going is to kludge your resource shares and to use configuration files or a beta version of the Manager to get it to work kinda sorta there too ...

Sadly I am not sure that the current development plan is generalizing the Resource Scheduler enough that when they do get around to adding ATI cards with (CAL) their API we won't live through the 6.6.x nightmare all over again ... and we will have a third opportunity to go through this again when OpenCL is added to the mix ...

At any rate, I sympathize ... :)

On the running / paused, when I set the value to n2, the tasks are launched, but they are actually suspended until the running two tasks are done. Then the next task in line starts as the other finish. RIght now I have nearly 20 tasks in running state, but only 2 are actually running ...

Lest anyone think I am complaining, I am not ... only commenting ... I do appreciate the hard work and the effort by CP and others to bring us this advance and am really pleased at the ability of the system to produce work. There is room for improvement, but, this *IS* a major advance ... when the CUDA app comes out I will likely try it to see if I can improve processing times using those resources too ... May be it will be time to consider adding another GTX 295 card to my farm (will need a new MB too) ...

LoL its so funny with these intel dudes
you don't have 8 cores, you have 4 cores who pretend and acting like they are 2 cores.
yes its maybe annoying that MW see exactly what it is, but the application is running like it should there is only 1 core not 2.
hyperthreading is just a trick which fools a operating system to think there are 2 cores while its just 1.

Okay, thanks for that Startling Revelation Dick Tracy ...

Okay, thanks for that Startling Revelation Dick Tracy ...

Okay, thanks for that Startling Response Dick ...

---

me@rescam.org

LoL its so funny with these intel dudes
you don't have 8 cores, you have 4 cores who pretend and acting like they are 2 cores.
yes its maybe annoying that MW see exactly what it is, but the application is running like it should there is only 1 core not 2.
hyperthreading is just a trick which fools a operating system to think there are 2 cores while its just 1.

Here's a good trick - my first ever Dell hyperthreading is right now churning out WU's far faster than the biggest farm I ever had.

The trick is in the HD4870 hidden inside it ;)

---

Okay, thanks for that Startling Revelation Dick Tracy ...

Okay, thanks for that Startling Response Dick ...

Your Welcome Dick, Less ... :)

Okay, thanks for that Startling Revelation Dick Tracy ...

Okay, thanks for that Startling Response Dick ...

Your Welcome Dick, Less ... :)

Less the fetish.

---

me@rescam.org

Does the GPU App work with Vista32 or Vista64 ????

Does the GPU App work with Vista32 or Vista64 ????

Yes. You'll just need the ATI Catalyst drivers installed. 8.12 seems to be the version of choice around here.

---

So far I have two 4830's, 4850, and a 4870 running with Winxp32 8.12 19b.
I have two more 4850's in the mail. Waiting to get a 4890 for my machine...

4830's seem to be the best bang for the buck ($89.00). Both of mine overclocked to 650 (max) and run at 51c...

http://www.newegg.com/Product/Product.aspx?Item=N82E16814161260

http://www.newegg.com/Product/Product.aspx?Item=N82E16814161259

http://www.newegg.com/Product/Product.aspx?Item=N82E16814102810


They all run quiet under full load.

So far I have two 4830's, 4850, and a 4870 running with Winxp32 8.12 19b.
I have two more 4850's in the mail. Waiting to get a 4890 for my machine...

I have a 4850, a souped up 4850, and a 4870. I'm tempted to go for another 4870 now, but I think I'll take your advice and wait for the 4890 to come out.

I want to got to 850Mhz and beyond! :P

---

I want to got to 850Mhz and beyond! :P

To Infinity and Beyond!!

---

Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

Hello

A friend got a 4870x2 and he doesn't do that much point... less than another one with a 4850.
I wonder why is it like that...

He has a Q6600 on XP32b with the driver 9.1

Do we have to change something in the app_info ? Any ideas ?

There is the detail of units : http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=54555&offset=20

Also its computer is quite slow when the GPU is crunching... (It is a bit slow to deplace windows)

Thank for your help

Hello

A friend got a 4870x2 and he doesn't do that much point... less than another one with a 4850.
I wonder why is it like that...

He has a Q6600 on XP32b with the driver 9.1

Do we have to change something in the app_info ? Any ideas ?

There is the detail of units : http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=54555&offset=20

As you see in the task details the power management downclocks the GPUs to 500MHz. Furthermore the usage is only about 50% on crossfire systems (use both GPUs, but not to the full extent).
Maybe it is a driver issue. You could try Cat 9.2 (look to the readme for instructions to get that to work, some dll files need to be copied and renamed) which is supposed to have improved multi GPU support for crossfire systems (including X2 cards).

Also its computer is quite slow when the GPU is crunching... (It is a bit slow to deplace windows)

That gets definitely better with version 0.19e. I promise ;)

I can't get an opt CPU app to run on my comp. All the wu's just corrupt.
I shut down the running client and put the files in the folder. Restart the client and it runs the stock app. Shutdown the client delete the stock app all wu's corrupt on restart.

PhenomII
Vista64 bit

I tried the SSE2 and SSE3 zips of the .19 files. It doesn't seem to read the app_info file.

Can someone set me straight please.

I can't get an opt CPU app to run on my comp. All the wu's just corrupt.
I shut down the running client and put the files in the folder. Restart the client and it runs the stock app. Shutdown the client delete the stock app all wu's corrupt on restart.

PhenomII
Vista64 bit

I tried the SSE2 and SSE3 zips of the .19 files. It doesn't seem to read the app_info file.

Can someone set me straight please.

Shut boinc down. Delete the stock app. Add the opti app w/app file. Restart boinc. Make sure that boinc exits running. If you start it up too soon it will cause them all to error, not recognizing the opti app.

Sometimes I've had the wu's error, since they want to run off the wrong app. When the next bunch come in it should be fine. SSe 2 should definately work, sse3 may also for you. Vista uses sse4.1, Xp sse2 or sse3.

---

Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

What is the optimal settings to change in appinfo.xml for a quad 6600 running a 4870x2? Or should I leave it as is?

I'm particularly interested in these:

<flops>1.0e11</flops>
<avg_ncpus>0.5</avg_ncpus>
<max_ncpus>4</max_ncpus>
<cmdline></cmdline>

I've noticed an increased wall time if I increase maximum concurrent WU over 2.

At one point I had 10 WU conccurently crunching and the wall time went even higher!

Have you read the readme file? I'd drop the avg_ncpus to 0.1, max_ncpus to 2 and cmdline n2.

That should help.