I took some time off, maybe a year or two, but have started doing some crunching in the last month or so. It seems like my RAC isn't what is used to be. For my GPU's I have two GTX 460 768MB and one GTX970 4GB. Each card is running 3 tasks with 0.2 CPU's. I also have 4 CPU tasks running as well. Do the jobs just take longer than they used to or do I have some settings that are off? Thanks in advance!

Welcome back to the project.

First issue you have noticed is the credit given for work. That is a can of worm. We can it CreditScrew (Credit New). A calculation devised by someone that determines how much credit you get for a work unit. Unfortunately, it's a flawed system and the amount given is going DOWN and not up.....

Second,

The server is having you try different cudas to find the fastest one for your system. While you were gone, there has been a new source of data that includes work that starts with BLC. These work units, unfortunately are terrible with the normal cuda applications.

Raistmer developed the SoG app which process these new work units better but require a full core for each work unit. This will also be a app that the server will eventually try on your system.

Your choices are to allow the server to continue to cycle through the different apps and hope that it settles on SoG or download the latest version of the lunatics installer and make the choice for yourself as to what is install on your computer.

http://mikesworld.eu/download.html

Someone is going to come along and tell you that you should use the Alpha App from Petri using the Cuda 9.0 but you need to convert your system to Linux and it's no problem (quick search of Number Crunching will show how many threads and time dealing with trying to install it, should give you an idea on that statement.)

Choice is your. Good luck.

---

It is unfortunate you are still trying to run the GTX 460's. That limits you from running the OpenCL SoG app which the fastest on all task types now. With the 460's in the system, you will be limited to running the CUDA 50 app for both the 460's and the 970. And yes, the tasks are much harder to process now. They doubled in size in the last couple of months.

I would recommend downloading the Lunatics installer to get the most optimized apps for your cards. You can get it at Mike's World

---

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

A proud member of the OFA (Old Farts Association)

The SoG app minimum OpenCL version is 1.2 and the 460's only supported OpenCL 1.1 so that SoG app is out of the equation.

---

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

A proud member of the OFA (Old Farts Association)

I wonder if it would be better to just take out the 460's. If sounds like they might not be worth the power required. I tried the Lunatics installer a few weeks ago and didn't notice a difference but that could be because of the card mismatch. I have two other 460's in storage, maybe I can sell the 4 and get another 960. Thanks for the ideas, I'll let you know how things work out.

---

Have you considered taking the 460s out and just running the OpenCL SoG applications on the 970? I'm running a 970, one task at a time with an i5- 3570K (I x GPU + 3 X CPU tasks), if that's any indication to you. Edit....you beat me to it!

---

Don't take life too seriously, as you'll never come out of it alive!

I wonder if it would be better to just take out the 460's. If sounds like they might not be worth the power required. I tried the Lunatics installer a few weeks ago and didn't notice a difference but that could be because of the card mismatch. I have two other 460's in storage, maybe I can sell the 4 and get another 960. Thanks for the ideas, I'll let you know how things work out.

I think your assessment of the power required to support the 460's for the amount of work they would produce is correct. I would remove them and just run the 970 on the SoG app. It is a good idea to think about the latest Pascal cards too. They have a very good power per output rating.

Look at Shaggie's chart to see which cards give the most bang for the buck for production versus power consumption.

GPU FLOPS: Theory vs Reality

---

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

A proud member of the OFA (Old Farts Association)

One things others may not have noticed. With masively increased computational demands of the current crop of applications three simultaeneous tasks per gpu is too much. The 460 will manage one and the 970 just about 2. Reduce to one per gpu, release one cpu core for each gpu task you are running , again, this due to the way the cuurent crop of applications work.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

I think I'm in business. Did my first WU after making all the changes and it came in pretty quick. Will need to run more today to get an average but it's looking good. Thanks for all the help everyone!