He likely isn't allocating a cpu core out of his 8 HT on his i7 to the

AP OpenCL 6.04's .

It makes a huge difference for NVIDIA OpenCL tasks.

I am just starting to test a HD7950 for SETI and only stock tasks I get are
AP's no MB's the AP Opencl ATI tasks using 12.4 OpenCL with 13.2 beta 7 Catalyst/Control
take 32 minutes and do not use a full core CPU time is roughly 1/3 the total time. I have been reading here about newer Lunatics AP's and MB's needing a full core for ATI but I will experiment with a stock AP and anonymous MB's
for the HD7950 as I did for the GTX660 I have been running for several months now.

Now I need to find that post on how to do it.

How to do what - free up a core? Go to settings and change 'on multiprocessors use at most x% of processors' to a suitable value (e.g. 80/90 for 6/7 out of 8 cores respectively) If you have local preferences, change the value there.

Thanks William. I changed the local pref to 80 and have 6 cores running now. I take it I can scale that up to get 7 running by going to 85% ?

---

[/quote]

Old James

He likely isn't allocating a cpu core out of his 8 HT on his i7 to the

AP OpenCL 6.04's .

It makes a huge difference for NVIDIA OpenCL tasks.

I am just starting to test a HD7950 for SETI and only stock tasks I get are
AP's no MB's the AP Opencl ATI tasks using 12.4 OpenCL with 13.2 beta 7 Catalyst/Control
take 32 minutes and do not use a full core CPU time is roughly 1/3 the total time. I have been reading here about newer Lunatics AP's and MB's needing a full core for ATI but I will experiment with a stock AP and anonymous MB's
for the HD7950 as I did for the GTX660 I have been running for several months now.

Now I need to find that post on how to do it.

How to do what - free up a core? Go to settings and change 'on multiprocessors use at most x% of processors' to a suitable value (e.g. 80/90 for 6/7 out of 8 cores respectively) If you have local preferences, change the value there.

Thanks William. I changed the local pref to 80 and have 6 cores running now. I take it I can scale that up to get 7 running by going to 85% ?

Not quite - 7/8ths of the available cores is 87.5%, so you need that figure or higher. That's why William suggested 90, as an easy figure - or you could use 88, then your fingers don't have to move so far.

He likely isn't allocating a cpu core out of his 8 HT on his i7 to the

AP OpenCL 6.04's .

It makes a huge difference for NVIDIA OpenCL tasks.

I am just starting to test a HD7950 for SETI and only stock tasks I get are
AP's no MB's the AP Opencl ATI tasks using 12.4 OpenCL with 13.2 beta 7 Catalyst/Control
take 32 minutes and do not use a full core CPU time is roughly 1/3 the total time. I have been reading here about newer Lunatics AP's and MB's needing a full core for ATI but I will experiment with a stock AP and anonymous MB's
for the HD7950 as I did for the GTX660 I have been running for several months now.

Now I need to find that post on how to do it.

How to do what - free up a core? Go to settings and change 'on multiprocessors use at most x% of processors' to a suitable value (e.g. 80/90 for 6/7 out of 8 cores respectively) If you have local preferences, change the value there.

Thanks William. I changed the local pref to 80 and have 6 cores running now. I take it I can scale that up to get 7 running by going to 85% ?

Not quite - 7/8ths of the available cores is 87.5%, so you need that figure or higher. That's why William suggested 90, as an easy figure - or you could use 88, then your fingers don't have to move so far.

Thanks Richard, I will try 88% and see what happens.
Will freeing up a core help my GTS 250 any? Im running lunatics on that machine.
And while Im asking dumb questions, My I7 3770 with the 550TI does not have lunatics installed. When the day comes and they get the go ahead If I install them will I loose crunching AP on a GPU?

---

[/quote]

Old James

Thanks Richard, I will try 88% and see what happens.
Will freeing up a core help my GTS 250 any? Im running lunatics on that machine.
And while Im asking dumb questions, My I7 3770 with the 550TI does not have lunatics installed. When the day comes and they get the go ahead If I install them will I loose crunching AP on a GPU?

Depends on the day ;)

Out of the box, the v0.40 installer doesn't include the OpenCL AP for NV app.
It would need to be (semi-)manually added.

If you wait for the next installer (after V7 release), that one is planned to contain all latest releases, including OpenCL AP for NV.

---

A person who won't read has no advantage over one who can't read. (Mark Twain)

Thanks Richard, I will try 88% and see what happens.
Will freeing up a core help my GTS 250 any? Im running lunatics on that machine.
And while Im asking dumb questions, My I7 3770 with the 550TI does not have lunatics installed. When the day comes and they get the go ahead If I install them will I loose crunching AP on a GPU?

Depends on the day ;)

Out of the box, the v0.40 installer doesn't include the OpenCL AP for NV app.
It would need to be (semi-)manually added.

If you wait for the next installer (after V7 release), that one is planned to contain all latest releases, including OpenCL AP for NV.

Thanks William, I can wait for the one with all the bells and whistles.

---

[/quote]

Old James

James, how much faster are running the Ap's now with one free core?

William. How many time for the V7 release? +/- Weeks?

---

William. How many time for the V7 release? +/- Weeks?

No idea. That is entirely up to Eric - when he believes that the applications have been tested enough and the whole system is sound.
Also provided no further problems with apps are spotted that would lead to a fresh round of testing.
Currently a major problem is to get the scheduler to deliver the best CUDA app to the host. So 'not ready yet'
Anyway anybody running stock will just start processing the new tasks.

---

A person who won't read has no advantage over one who can't read. (Mark Twain)

James, how much faster are running the Ap's now with one free core?

William. How many time for the V7 release? +/- Weeks?

Total time to run is about 2 hours and 55 minutes. So I saved an hour by going to 7 cores.
I dont overclock or run more than one WU on the GPU. Yet anyway. I will ask how to do that when I feel more at ease with tweaking settings. I really need to read my ASUS motherboard manual.

One thing that is strange since running Ap's on the gpu is the temps. Im reading 55 C. doing an Ap and 63 C. when I run a regular work unit.

---

[/quote]

Old James

And the shorty storm continues.

---

And the splitters are still unable to keep up their production. Rgeardless of the number of shorties, they have been able to sustain 50/s+ in the past even when there were shorties going through the system.
For some reason though they haven't been up to the job for over a week.


And i noticed yet another download record on the Cricket Graphs- 662Mb/s after the outage.

---

Grant
Darwin NT

I think the Shorties got together and had a litter of shorter shorties

I think the Shorties got together and had a litter of shorter shorties

+1

---

With the shorties I am surprised there was a steady ~200Mb download rate during the outage.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

Uh oh. May be problems with the upload server. I've got WUs backing up for the first time since the move.

---

Join the PACK!

Uh oh. May be problems with the upload server. I've got WUs backing up for the first time since the move.

I see the same thing over here. Not good...

---

....
And i noticed yet another download record on the Cricket Graphs- 662Mb/s after the outage.

why 220 mb/s download bits from the servers during the outage, that isn't seti traffic, it is something else on top of seti.

---

Uh oh. May be problems with the upload server. I've got WUs backing up for the first time since the move.

Yep.
It shows green on the Server Status page, but inbound traffic has dropped off & i've got uploads accumulating here as well.

---

Grant
Darwin NT

is the server full?? all backed up!!!

---

or Washington DC?!?

---

A few uploads are starting to go through, at 2-5kB/s.
As though they just teleported everything back to the Lab...


EDIT- well they were uploading for a while. Uploads appear to have come to a halt again.

---

Grant
Darwin NT