I was reading the release notes and this was in there:

http://us.download.nvidia.com/Windows/285.27/285.27-Win7-WinVista-Desktop-Release-Notes.pdf

• GeForce 500 series: GPU power levels appear to fluctuate unnecessarily while
  navigating the Firefox browser. [857542]
  
  
• In the Release 280 drivers, some users reported a noticeable fluctuation in clock speeds while engaging in various tasks on the PC. With the Release 285 drivers, adjustments have been made to reduce the sensitivity to levels similar to the R275 driver.
  

All of which I found pretty interesting. For one, I use Chrome, so this could by why the Firefox didn't affect me accept at random times when I would open Firefox for Private Browsing whilst transitioning some of my login credentials for bank accounts, etc over to Chrome.

For two, it sounds like the R275 behavior is close if not exactly what some downclocking issues people were having...

For these reasons, I have began testing the new 285 BETA drivers and will keep an eye on WUs. For my initial viewpoint, I do see the already calculated remaining times on my GPU scheduled WUs to be dropping a few seconds at a time. (always a good sign)

For these reasons, I have began testing the new 285 BETA drivers and will keep an eye on WUs. For my initial viewpoint, I do see the already calculated remaining times on my GPU scheduled WUs to be dropping a few seconds at a time. (always a good sign)

Do make sure that you're not confusing driver behaviour with the SETI server behaviour described in the adjacent Shorties estimate up from three minutes to six hours after today's outage! thread.

I thought this was related to using BOINC 6.12.34? I am still on 6.12.33.

It is BOINC client independent. The problem affects all new tasks downloaded within the last few days.

---

I read more and understand. My WUs do not "have hours for completion times and only taking minutes to complete", so I do not think the new timings are related to my BETA driver test. I do appreciate the insights though :)

This is only my 3 post on testing newer NVIDIA drivers and don't mind being the "Guinea Pig" forward in the future BETA and WHQL releases.

So far my WUs have all returned successful with no errors, however waiting for validation with wingmen.

I will continue to post my results

It is BOINC client independent. The problem affects all new tasks downloaded within the last few days.

Not all new tasks, Only all new GPU tasks downloaded in the last few days,

Claggy

It is BOINC client independent. The problem affects all new tasks downloaded within the last few days.

Not all new tasks, Only all new GPU tasks downloaded in the last few days,

Claggy

I beg to differ. Granted I run Lunatics apps, so that might be the cause, but my CPU tasks are overestimated by a factor of about 3.5. They are finishing with a raw DCF of ~0.3. The GPU tasks are MUCH worse, usually finishing with a DCF of about 0.02, so they are off by a factor of about 50.

---

It is BOINC client independent. The problem affects all new tasks downloaded within the last few days.

Not all new tasks, Only all new GPU tasks downloaded in the last few days,

Claggy

All tasks, downloaded since Tuesday maintenance, by users with anonymous platform.

It's just FAR more noticiable on GPU with factors in the region of 25x as on CPU which is more like 4x - after DCF adjustments the CPU tasks should be fine again, 'just' the GPU ones will continue to suffer.

It is BOINC client independent. The problem affects all new tasks downloaded within the last few days.

Not all new tasks, Only all new GPU tasks downloaded in the last few days,

Claggy

I beg to differ. Granted I run Lunatics apps, so that might be the cause, but my CPU tasks are overestimated by a factor of about 3.5. They are finishing with a raw DCF of ~0.3. The GPU tasks are MUCH worse, usually finishing with a DCF of about 0.02, so they are off by a factor of about 50.

I agree, i'm wrong, i've looked at the changeset again, no mention of it's just for GPU's,
I hadn't experienced it on the CPU at Seti Main as i have no CPU tasks here on my Main host, and my other hosts are stocked with CPU work from before the change.

Claggy

so, maybe this is affecting me :)

I reread the thread again, and I am definitely off by more than a factor of 10, in fact, more like off by a factor of 100.

Average processing rate 302.11901926551

Measured floating point speed 2538.59 million ops/sec
(moving the decimal point) to be 2.53859 GFLOPS

So with that being said, should I be adding

<flops>2.53859e09</flops>

to my app_info.xml?

so, maybe this is affecting me :)

I reread the thread again, and I am definitely off by more than a factor of 10, in fact, more like off by a factor of 100.

Average processing rate 302.11901926551

Measured floating point speed 2538.59 million ops/sec
(moving the decimal point) to be 2.53859 GFLOPS

So with that being said, should I be adding

<flops>2.53859e09</flops>

to my app_info.xml?

Goodness gracious, no - check this post from Joe on how to calculate flops and see if it is in line with the 302e09 APR shows - it usually is but APR may have drifted off.
THEN put in

<flops>302.11901e09</flops>

or what you get from the calculation, if off by more than a few %

edit: your host(s) are hidden, so we can't assist with calculations.

the wrong assupmtion that that 2.5e9 speed from the CPU has anything to do with the speed of the GPU is what brought us into the mess in the first place :D

Ah hah, that makes sense.

Also deciding not to mess with the <flops> as my DCF for Seti seems to be very close to 1.

<duration_correction_factor>0.970445</duration_correction_factor>

Update:

Neither my GTX570 or GTX560Ti has returned invalid results due to the driver update. I have also not witnessed any downclocking.

I am running:

• BOINC 6.12.33
  
• Lunatics .37 x64
  
• GTX570 is with the the i7 9300 with HT (running 8 WU on the CPU)
  
• GTX560Ti is running GPU only
  
• (2 GPU WUs / GPU on both rigs)
  

I had a driver crash last night, but after a pause and restart of BOINC, the clocks were at default level. This was while it was running a Collatz task.

---

I had a driver crash last night, but after a pause and restart of BOINC, the clocks were at default level. This was while it was running a Collatz task.

Did you reboot after you upgraded your driver? Even though most times it doesn't seem like it is required. I do it anyway...

edit: your host(s) are hidden, so we can't assist with calculations.

I allowed my computers to be seen.

Can you calculate my flops for the 570 and 560?

I used 302.11901e09 for the 570 and 232.65208e09 for the 560.

These do seem to have put the completion estimates to where they were before.

Thanks...

edit: your host(s) are hidden, so we can't assist with calculations.

I allowed my computers to be seen.

Can you calculate my flops for the 570 and 560?

I used 302.11901e09 for the 570 and 232.65208e09 for the 560.

These do seem to have put the completion estimates to where they were before.

Thanks...

Do what you must, but most sage advice lately has been against using manual flops entries......

As to the new drivers, I am kinda watching the Lunatics boards to see if there is any input on their Seti crunchability before installing them.
280.26 is just working so well here that I hate to change without verification.

---

"Time is simply the mechanism that keeps everything from happening all at once."

Do what you must, but most sage advice lately has been against using manual flops entries......

Under normal circumstances, I would tend to agree with you, however it seems that this was the only way to get my estimates back down to normal times so that BOINC even request WUs. Otherwise, I was going 3 or 4 hours before BOINC would even make a WU get request (during those 3 to 4 hour periods, BOINC was stating that it was not requesting new tasks even when my cache was empty). Now BOINC is making WU requests and my cache is slowly building up (slllllllllllllllllowwwwwwwwwwwwly being the operative word) :)

Although most of these posts are WAY off topic from the BETA 285.27 topic this thread was originally meant for.

As for "crunchability", so far so good and seem to be on par with the 280.26 (which actually breaks my normal rule of "if it works, don't mess with it").

I will role back to the WHQL series at the first sign of a problem.

Do what you must, but most sage advice lately has been against using manual flops entries......

Under normal circumstances, I would tend to agree with you, however it seems that this was the only way to get my estimates back down to normal times so that BOINC even request WUs. Otherwise, I was going 3 or 4 hours before BOINC would even make a WU get request (during those 3 to 4 hour periods, BOINC was stating that it was not requesting new tasks even when my cache was empty). Now BOINC is making WU requests and my cache is slowly building up (slllllllllllllllllowwwwwwwwwwwwly being the operative word) :)

Although most of these posts are WAY off topic from the BETA 285.27 topic this thread was originally meant for.

As for "crunchability", so far so good and seem to be on par with the 280.26 (which actually breaks my normal rule of "if it works, don't mess with it").

I will role back to the WHQL series at the first sign of a problem.

Hmmmm....
I don't understand why no work requests if the cache was empty. DCF should not matter much at that point I would think. Unless you are doing another project and some debt issues came into play.

Anyways....awaiting more reports on the new drivers.
Jason's been playing with them and says so far they appear to be a little faster on some ARs and a little slower on others....but no big swing either way.

---

"Time is simply the mechanism that keeps everything from happening all at once."

Running 1 far too old driver on the GTS250 (175.xx?) and later one on the GTX470,
but too old for Einstein, according to a message in BOINC.

Looked at 4 MB WUs, running on the 5870s, longer, but appeared to be vlars
Time to upgrade those, I read the posts made earlier, always helpfull.
(My 2 ATI 5870 drivers are up to date, when I got home, the sound of a jet, was comming out of the room, Milkyway, 70 seconds, 1 at a time, but 99% load and 520Watt, power use). (These are 'shorties and take 4 hours on CPU).

Back on topic, though..........
First, upgrade the 470.

---