Try "Extend desktop to this monitor" in desktop settings.
Look if Physics + SLI is turned off in Nvidia control panel.
Mine did werird thing with it.
Like using SLI in games while i dont even have a SLI bridge.
(and yeah it did work in SLI without the bridge :P) hmz..
No mutch from 20k to 24k in 3dmark2006....

---

...Was someone saying it is better to disable one CPU?

Yes, there are a few threads on this topic.

The data transfer to the GPUs will be faster if you switch to 88% of processors (7 of 8), but you'll have to try yourself if the gain in GPU throughput outweighs the loss of one CPU core for BOINC processing.

Gruß,
Gundolf

---

Computer sind nicht alles im Leben. (Kleiner Scherz)

SETI@home classic workunits 3,758
SETI@home classic CPU time 66,520 hours

...Was someone saying it is better to disable one CPU?

Yes, there are a few threads on this topic.

The data transfer to the GPUs will be faster if you switch to 88% of processors (7 of 8), but you'll have to try yourself if the gain in GPU throughput outweighs the loss of one CPU core for BOINC processing.

Gruß,
Gundolf

Is there anyway to get Boinc to run CPU crunchers at Low (lvl 4) and GPU crunchers at Below Normal (lvl 6) ?
That way the GPUs can still be fed, while the CPUs get the rest.

---

...Was someone saying it is better to disable one CPU?

Yes, there are a few threads on this topic.

The data transfer to the GPUs will be faster if you switch to 88% of processors (7 of 8), but you'll have to try yourself if the gain in GPU throughput outweighs the loss of one CPU core for BOINC processing.

Gruß,
Gundolf

No.. no longer needed.

Have a small look in my profile and DL Fred's nice prog TThrottle.
This prog can increase the priority of the CUDA tasks to 'normal', so only CPU tasks are disturbed if working on the PC. Max. GPU performance.


Also, I saw some Core i7 without HT enabled. (Only 4 threads)
So the CUDA WU start is faster.

IIRC, HT on/off give you ~ 15 % performance difference (with CPU WUs).

But, msattler or other Core i7 user can say more about this..

---

Is there anyway to get Boinc to run CPU crunchers at Low (lvl 4) and GPU crunchers at Below Normal (lvl 6) ?
That way the GPUs can still be fed, while the CPUs get the rest.

AFAIK, the CUDA tasks are at 'lower than normal'.
You see this in the first ~ 3 sec. if a new CUDA task start in TaskManager.
After this this priority go down to 'low' (but stay in the background at 'lower than normal', if opt_CUDA_6.08_V12_app).

But Windows isn't smart enough for to disturb in the priority hierarchy.
If a prog (for example boinc.exe) have activity - which have normal priority - all under 'normal' will be disturbed.

So you need to increase the priority of the CUDA tasks to 'normal', for to disturb only the CPU tasks.

TThrottle!

---

I downloaded Throttle and will check it out later. Thanks and you really have a great profile page.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

I downloaded Throttle and will check it out later. Thanks and you really have a great profile page.

I have found that leaving a spare CPU slot available seems to help.......
More time left over for the bloated Boinc manager to do what it has to do.......

And, since the Cuda is vastly faster than what the i7 can do.....anything to bolster it's progress is good.

I played with process assignment manually, and did not find it to really alter things much.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

Kind of cool watching my Cuda cards in the same machine head to head. Seems the GTX260 is easily twice as fast as my 9600GSO. The 260 passes the 9600 like the 9600 passes the CPU's. Can't wait for Santa....and maybe a 285.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

I would buy a manufacturer OCed GTX260-216.
For example the GIGABYTE SOC have higher GPU/shader/RAM speed than a stock GTX285. But maybe half prize?

Or wait some weeks/months to the new GTX3xx series (Fermi chip).
I guess, GTX360 - double performance/same wattage - than GTX260-216.

---

Here is my old post because of the GTX260-216/275/285 speeds..

[http://setiathome.berkeley.edu/forum_thread.php?id=54583&nowrap=true#948705]_message 948705




Maybe someone with a GTX285 (stock/OCed) can jump in for to post the calculation times..


IIRC.. for a normal AR 0.44x WU, my OCed GTX260-216:

GIGABYTE SOC -> ~ 540 sec.
(Intel Core2 Extreme QX6700 @ 3.14 GHz - DDR2 944/5-5-5-15) (CPU/mobo/RAM OCed) (4x MB + 1x CUDA)

EVGA SSC -> ~ 565 sec.
(AMD Phenom II X4 940 BE @ 3.0 GHz - DDR2 1066/5-5-5-15) (CPU/mobo/RAM stock) (nothing on CPU, 4x CUDA)

The system RAM is also involved in the whole/also pure GPU calculation time.

I changed the system RAM from 800/5-5-5-18 to 1066/5-5-5-15 on the AMD mobo and the calculation got a ~ 2 % speed up.
At a RAC of ~ 13,000/OCed GTX260-216 this are + ~ 260 RAC.
So my whole GPU cruncher got + ~ 1,000 RAC.

---

Does anyone know what wording to put in to make it return after 1 minute? Now that it is working I don't want to screw it up.

<cc_config>
<options>
<report_results_immediately>1</report_results_immediately>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

---

Official Abuser of Boinc Buttons...
And no good credit hound!

This ~ 60 sec. report_results_immediately was only with Crunch3r's 'old' BOINC mod.

If you insert this in the cc_config.xml with stock BOINC client:

<report_results_immediately>1</report_results_immediately>

The report will be ~ 1 - 6 sec. after UL.

---

So for anyone wondering. I think the problem with my 2 cards was the cc_config and not plugs for my spare card. I have since unplugged the extra monitor disconnected and restarted computer many times and it still uses both cards but computer knows there is only one monitor.

---

Official Abuser of Boinc Buttons...
And no good credit hound!