Good day,
I am wondering why my Gtx Titan is doing so bad. My two Gtx 580 is beating it. What did I do wrong?

---

Referring to this system
--- Microsoft Windows Server 2012 R2
Standard x64 Edition, (06.03.9600.00)
7 Mar 2016, 4:57:24 UTC
ID: 7937552
Details | Tasks
Cross-project stats:

BOINCstats.com Free-DC Emily home 12,936.91 221,790 7.6.22 GenuineIntel
Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz [Family 6 Model 26 Stepping 5]
(8 processors) NVIDIA GeForce GTX TITAN (4095MB) driver: 361.91 OpenCL: 1.2 Microsoft Windows 7
Enterprise x64 Edition, Service Pack 1, (06.01.7601.00)
7 Mar 2016, 4:44:34 UTC

---

elec, how many work units are you running at a time on that card?

Also, why did you choose cuda 32 for that card?

Might have been better to go with cuda 42 or even 50

Are you also running any CPU task at the same time?

Edit..

What is your CPU to GPU ratio?

Yes those GPU results take a long time with way to much CPU time being used on them.

Now 1 thing I did spot is that it looks like you're running Anonymous platform via Lunatics, but you have selected the Cuda 32 app instead of the Cuda 50 app.

Also is the card running cool enough and at it's proper speeds (not throttling down)?

Cheers.

Yes those GPU results take a long time with way to much CPU time being used on them.

Now 1 thing I did spot is that it looks like you're running Anonymous platform via Lunatics, but you have selected the Cuda 32 app instead of the Cuda 50 app.

Also is the card running cool enough and at it's proper speeds (not throttling down)?

Cheers.

Thank you for the advice, I will chose cuda 50 and see the difference. The gpu shows 50% load with two units going.

---

You're under utilizing that card. It should be doing minimally 3 tasks per card at a time, probably 4 to get the utilization up into the high 90's.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

You're under utilizing that card. It should be doing minimally 3 tasks per card at a time, probably 4 to get the utilization up into the high 90's.

Thank you for the advice I made that change, but card is running 4 instances and looking at 85% load.

---

Keep an eye on your temps.

What are your times?

...

What are your times?

His times are comparable to the cores on my i5's, way to slow, the card must be idling (or cooking). :-O

[edit] And his CPU time used for doing those tasks is the same as it takes my GTX 660 to spit out a w/u doing 1 task at a time (something not right there).

Cheers.

Yeah his times are way off, maybe the titan is throttling down??

Yep, that was exactly 1 of my 1st thoughts, or trying to do too many w/u's at once. ;-)

Cheers.

by chance you are not using double precision mode are you? that could slow things down a bit.

---

Double precision isn't used for the tasks here. ;-)

Have you got GPU-z and either MSI Afterburner or EVGA Precision X on that setup and what are they reporting?

Cheers.

I notice you are using the stock project settings for the card. Those are primarily aimed at entry to mid level cards. Yours is high end. You should be running with enhanced processing parameters. You should modify the MBCUDA.CFG file in the project directory and use these parameters.

processpriority = abovenormal
pfblockspersm = 18
pfperiodsperlaunch = 200

Read the commentary in the file to understand your options. You might want to bump pfblockspersm up to 22-24 even. pfperiodsperlaunch could go to 300 or 400 also. I wouldn't bother shifting processpriority to high though unless you use the computer only for crunching.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

I notice you are using the stock project settings for the card. Those are primarily aimed at entry to mid level cards. Yours is high end. You should be running with enhanced processing parameters. You should modify the MBCUDA.CFG file in the project directory and use these parameters.

processpriority = abovenormal
pfblockspersm = 18
pfperiodsperlaunch = 200

Read the commentary in the file to understand your options. You might want to bump pfblockspersm up to 22-24 even. pfperiodsperlaunch could go to 300 or 400 also. I wouldn't bother shifting processpriority to high though unless you use the computer only for crunching.

But doing that will not solve his current problem, will it? ;-)

Let's get the main problem fixed 1st before do the fine tuning, ok? :-)

Cheers.

Yes, we need to get it running well with just 1 task before we start having him pile on others.

Something is causing this GPU to slow down significantly.

The first thing I see is only 4Gb of memory, it could very well be getting choked feeding the card.

The first thing I see is only 4Gb of memory, it could very well be getting choked feeding the card.

That part is due to 32 bit address space limitations, not a problem with the system.

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

But even when he was running only 1 GPU at the very beginning, it was still running extremely slow.

I think he should just run 1 at a time until we can figure out why it's running so slow.

Having him run more than 1 at a time is only adding to the confusion of what is going wrong.

Exactly how many cpu work units are you trying to run at the same time?
If you are trying to run to many, you might be starving your gpu. You might have to free up cpu core if that is the case.

---

Bruce