........ Christ. What does Paul think!?

[quote]........ Christ. What does Paul think!?[/Surprisingly]

Don't use development versions ... If I am the Paul you are thinking about ... :)

If I am not the Paul you are thinking of ...

Don't use development versions ...

(Sorry, only sense of humor I have)

Without getting too deep into the weeds. I think that BOINC has long had a flawed development process ... when I was in Federal Civil Service we took part in a program called SEI (not sure what they call themselves now these days, ah, still call themselves the same so there is the link if you like...) ... and they did a bunch of research (Carnegie Mellon Univ.) and came up with a model of software development. And, in that model, the lowest level, level 1, was characterized by chaotic processes, inability to hit schedules, and in general bad software ...

Surprisingly, MS, FOr example, is a classic level one organization ... distain for processes, documentation, schedules, and rigorous engineering ... BOINC is much the same ... hackers running about making change with little thought of what comes next.

When I suggested writing down the processes I was mocked ... the point is not that the process has to be large, or complex, but it has to be written down ... if it is not written down, it does not exist ... got that from Tom Clancy ... so you know it has to be right ...

At any rate, the same goes for a design document ... another concept they find funny ... but, if you don't have a design ... how do you know when you have it right? We don't build a building without a design, or a car, or even a computer ... but software? Infinitely more complex, and we can just wing it because we are so good at it ...

Probably more than you wanted to know ... :)

But, that is what I think ...

That's not your earnestness? Are you sure M$ is running a code and fix process, without documentation? There are lot of concepts developed and used in IT industry, and you think M$ selected the crap of it? Please explain ...

LoL don't tell me you did not know that.

M$ allways did that, making a product and sell it before was ready.
Hence the many crashes and blue screens or the updates who crashed the whole computer :D.

Only with the new vista ( win 7 ) they seem to finally saw the light.

Now they actually start listening to the grand public that things are not working like it should. And fixing them xD.

I agree its not a small project we are talking about millions of lines with code so small errors can and will be made.

But in general M$ is learning from its mistakes in the paste ;) (finally)

But back to the newest beta versions yes i test them all and hoping a stable running one is present with them.

As for now i think its best not to install them over running projects but in a virtual machine like i use and test if it installs well or runs projects ( one by one ) steady.

Now i know that i need to test also with optimized apps for projects also.

The problem with boinc is that you never know its for u the good version till you really tested it in every way.

But basically some are better then others and function with less problems then others, thats why i am not scared to test these beta's as well.

Maybe one version surprises us again and will be the next steady one :D

Fixed in 6.6.9....

These new releases include the following changes:

* Account manager list in the attach to account manager wizard.

* The manager remembers the last email address/user id used in
the attach to project and attach to account manager wizards.

* The attach to project wizard now changes a project's name to a
light gray if it isn't sure your platform is supported for the project.

* Fixes various issues with the core client automatic back-off
mechanisms.

* Various fixes for long term and short term debts.

* Detection of GPU driver versions (Windows Only)

* New preference for whether BOINC is allowed to use a GPU if
the machine is in use.



Please take a look at the new features. One thing of note is that if you click on a project that is grayed out a dialog should appear telling you that, that project may not support your platform. However, it won't stop you from attaching.

____________
Crunching for the benefit of humanity and in memory of my dad and other family members.

JM VII says that the CPU scheduler is still not calculating LTD correctly in 6.6.9 ... be wary ... :)

That's not your earnestness? Are you sure M$ is running a code and fix process, without documentation? There are lot of concepts developed and used in IT industry, and you think M$ selected the crap of it?

Probably more than you wanted to know ... :)

Well, he asked ;)

Concerning the topic: it looks like the 6.6.x code is slowly maturing.. there's hope yet they'll get it right some day.

And concerning the not-topic: I think MS' software is better than most people admit.. so it's still easy to get mad at them for various things.

Like M$. New version, new bugs.

At the time, no cuda WU is running. Three WU are "Ready to start". One is "waiting to run"

What is that? What can I do?
____________
Constant dripping wears away the stone. :)

Like M$. New version, new bugs.

At the time, no cuda WU is running. Three WU are "Ready to start". One is "waiting to run"

What is that? What can I do?

Like M$. New version, new bugs.

At the time, no cuda WU is running. Three WU are "Ready to start". One is "waiting to run"

What is that? What can I do?

Because all versions that work (to this point) don't distinguish between GPU and CPU work the CPU scheduler is not scheduling the tasks. One of the ways to "force the issue" is to select all the projects and suspend them, then un-suspend GPU Grid... then the other projects ... the GPU Grid tasks should continue to run.

I also suggest making sure that the resource share is set for GPU Grid high enough that you always have LTD building ... in the case of 8 cores, 15% share ... quads, 26-30% duos, 51% ... when 6.6.x gets working we should not need to do this ...

Everything Paul has said is good advice...but I recently had this problem and none of the above worked. Ultimately, only a complete machine shutdown and restart got the "ready to start" workunits actually to begin crunching.

At the time, no cuda WU is running. Three WU are "Ready to start". One is "waiting to run"

Because all versions that work (to this point) don't distinguish between GPU and CPU work the CPU scheduler is not scheduling the tasks.

I also suggest making sure that the resource share is set for GPU Grid high enough that you always have LTD building ... in the case of 8 cores, 15% share ... quads, 26-30% duos, 51% ... when 6.6.x gets working we should not need to do this ...

I guess I shouldn't say this too loudly, but 6.6.9 is working perfectly for me thus far (24 hours). Retrieves new work units in a timely fashion and suspends CUDA when system is in use, leaving CPU WUs running. Maybe it'll implode tomorrow.

If you are running one of the 6.6.x versions you will see new messages and different syntax from the client as it fetches work. They are trying to sort out the fetch and scheduling policies. To this point there is one flaw or another in the system. I can't keep up with them all, and don't think that it is important that I need to do that ...

So, that is why you see CUDA mentioned ... but, that syntax I think will change again in later versions as they need to accommodate the ATI cards which are not CUDA, and OpenCL which is right around the corner ... so, that would mean that the syntax would have to extend more ...

Just for your musing, SaH has tasks and applications that will run on the CPU and the Nvidia GPU ... Milkyway has a volunteer that is working on an application for the ATI GPUs (I am running tasks on my brand new 4870 and it is pretty sweet), and the new intern may be working on an OpenCL version so that they don't have to have as many different versions ...

The code is converging ... and one of the next two might be usable. The major impediment right now might be that they are not listening enough to JM VII ... but what do I know ... :)

I am actually looking forward to the first 6.6 version that works better than 6.5.0 ... which may be 6.6.10 or 11 only time will tell ...

There is more tweaking going on, mostly around LTD handling. Maybe 6.6.11 or later.