I see 6.11.1 has been released to beta testing. I've just checked my email but can't see any release notes. Any news on what this new .11 series contains?

A lot. ;-)

By the way, that they're available on the dl page doesn't mean that they're alpha's per se. They could be internal test builds.

---

Still no GPU tab :(

Still no GPU tab :(

There's no CPU tab either, so why the need for a GPU tab?

---

Within Boinc's Advanced Preferences there is a processor usage Tab, under which is the solitary GPU control, "Use GPU While Computer is in Use".

I would like to be able to keep a low cache for GPU projects, and a higher cache for CPU projects. At the minute this cannot be done. For GPU projects it is important that tasks are reported immediately, whereas with CPU tasks this is undesirable in some cases.

Some people want to run GPU projects on one card but not another, using their slower card for display and their faster card for crunching. Such advanced configurations are difficult to arrange via the cc_config file, and requires advanced user skills.

At some stage the introduction of separate CPU and GPU Tabs with several selections might facilitate such configurations.

One of the better introductions was the Snooze GPU (from the taskbar). Would be nice to have a snooze CPU function as well. Sometimes we just want to use the CPU for other programs, but would prefer to allow the GPU to crunch. Presently the only way to do this is suspend individual projects and then restart them. However some projects use both CPU and GPU (well, did at one stage).

Would be nice to have a snooze CPU function as well. Sometimes we just want to use the CPU for other programs, but would prefer to allow the GPU to crunch. Presently the only way to do this is suspend individual projects and then restart them. However some projects use both CPU and GPU (well, did at one stage).

All GPU apps use the CPU. There isn't an OS yet that can run applications on only the GPU, so the CPU is always used. The GPU can't run applications as we know them, it can only run kernels.

The CPU is used to translate the contents of any task to these kernels, then it transports them to the GPU's memory. The GPU will run these kernels, put the outcome back in its memory, after which the CPU will come, pick that up, translate it back to understandable stuff and store it on disk. Rinse, repeat etc.

So you can't just snooze the CPU, as the program doing all this translating etc. for the GPU is then no longer running either.

---

Just add a feature to suspend all non-GPU projects. You can already suspend individual projects, so all you have to do is amalgamate this into one handy feature.

Would be nice to have a snooze CPU function as well. Sometimes we just want to use the CPU for other programs, but would prefer to allow the GPU to crunch. Presently the only way to do this is suspend individual projects and then restart them. However some projects use both CPU and GPU (well, did at one stage).

All GPU apps use the CPU. There isn't an OS yet that can run applications on only the GPU, so the CPU is always used. The GPU can't run applications as we know them, it can only run kernels.

The CPU is used to translate the contents of any task to these kernels, then it transports them to the GPU's memory. The GPU will run these kernels, put the outcome back in its memory, after which the CPU will come, pick that up, translate it back to understandable stuff and store it on disk. Rinse, repeat etc.

So you can't just snooze the CPU, as the program doing all this translating etc. for the GPU is then no longer running either.

Yes, but we'd like to be able to suspend CPU tasks only, (as opposed to CPU usage) GPU tasks will still keep crunching as long as you only run programs on the CPU with a low priorty,

Like last night i did some video encoding for my PSP, i had to suspend all my CPU projects like CPDN and Einstein, and all my Seti CPU tasks while leaving GPU tasks unsuspended,
since GPU tasks tend to have a thread that runs at a higher priorty to feed the GPU, as long as i ran my encoder at low priorty, the GPU ran at full or close to full speed,

Claggy

Yes, but we'd like to be able to suspend CPU tasks only

Come on , that's stretching it a bit, isn't it? if you want to do other things, just exit BOINC. With every button and option added it isn't going to make BOINC any easier for everyone.

---

I would like to be able to use foldit (a most excellent program) and still use my GPUs; I have 4 GPUs in one system! Foldit uses the CPUs but not the GPU.
When I am working with large office files it would be really useful to be able to free up the CPU cores for a few minutes (when a client needs something urgently) without interrupting the GPU tasks.

I'm still waiting for BOINC to be able to make a decent cup of tea.

Wait outside.

I would like to be able to use foldit (a most excellent program) and still use my GPUs; I have 4 GPUs in one system! Foldit uses the CPUs but not the GPU.
When I am working with large office files it would be really useful to be able to free up the CPU cores for a few minutes (when a client needs something urgently) without interrupting the GPU tasks.

Forgive me if I'm wrong, but doesn't BOINC run the apps in low priority so if a program with a higher priority wants the CPU the BOINC app stops while the higher priority app utilises the CPU? Doesn't this do what you want?

BTW, why is the notices tab first and why has the log tab gone and you can only access the messages under the advanved menu?

I see lots of comments about "Attach to Project" vs "Add Project".

If it's not broke...

WRT "Consider aborting it" - Please, NO!
The Lattice Project, for one, would not be too pleased.
They are running with small time limits and tasks that last much longer than expected; it would be to their downfall.

Perhaps something more hands on (useful) such as, Refer To Project for task duration and Run Time Advice, or a better suggestion if someone can come up with one.

Thanks,

The 6.10.58 client as witnessed whilst off-line already self aborts unstarted overdue tasks. Presume this discussion to be solely relating to those that have already started and are overdue. All depends on the Lattice policies and practices. Do they reissue new copies on the spot when overdue? I'm still having my ear open to learn if a client that crunches overdue task i.e. have been started, to connect to the respective project to tell of this fact so it can hold off on reissuing... that would constitute an efficiency improvement, though no idea how much. When so, the "consider to abort..." may need an all together complete revisit.

All depends on the Lattice policies and practices. Do they reissue new copies on the spot when overdue?

Yes, Lattice issues another copy immediately when one exceeds the deadline.

BobCat13, do you mean that when a task has not been started it gets resent, or some form of acknowledgement/instruction from the server to say, continue crunching same task?
When I run Lattice tasks they all go well over the deadline. They don’t reset; at least the task is not dumped and reissued (same task or new one), and I don’t start crunching it from the start again.
Lattice advise crunchers to continue running tasks past their deadline, as most go over the estimated runtime, are just as useful and people get normal credit. I think in the case of Lattice it is just down to their app and their scheduler that the % complete is incorrect. Looking at running tasks, the expected runtime might now be more accurate, for some systems, but their deadlines are still too short; I have 2 tasks at 28h on a Q6600 (stock, 4GB XP system). The tasks are at 1% complete and the estimated time remaining is about 326h. These will both finish after their deadline of the 14thAug.
Lattice may deliberately set short runtime cutoffs to force Boinc to continuously run the Lattice tasks, rather than stopping to do other tasks because switching tasks would crash the Lattice tasks/reset them.
Whatever the situation, it would be better if the projects handle their own "abort task" messages, preferably without the word abort; crunching is not a rocket mission or a birth control campaign, so perhaps Boinc should not employ such brutally worded generic messages when individual tasks reach such deadlines.
Also, if the system loses its time, Boinc aborts everything! This is a clear situation and the user should be given the choice to end all tasks, or change the time.

What I am saying is that if a Lattice task has passed it's deadline while you are running it, the project sends another copy to a different user. They do not reset or cancel your task as they are willing to accept late returns, but they cannot wait forever before sending another copy as the task that has passed deadline may never be returned by the user (they may have clicked on reset project or detach project).