Plz help !

What can i do ? Running a GTX 670 , but these guppi workunits won´t notice it.

Got Lunarics 0.44 installed (64-Bit 0.45 crashes everytime), newest nvidia drivers PLUS Vulkan Runtimes , also newest, running on Win 7 64-Bits.

Could someone/anyone give me advice to make these guppies to rush ?

Any ideas welcome.

Greetings

ok i modified the file "mbcuda.cfg" , but there´s no effect...but ... great ? 0.001 to 0.010 percent/per second. What a disgrace.

1 hour gone, and 18% completion of ONE guppi wu.

---

A few things - Lunatics 0.45 installer has not been "officially" released by them yet, they are still at 0.44, so file an error report with them and continue using the applications in the 0.44 installer.
guppis take 50% and 200% longer to run on an Nvidia GPU than "normal" Arecibo tasks.
The newest Nvidia drivers might not be the best thing for your GTX670 when performing calculations.
As far as I'm aware there are no SETI apps that use the Vulkan API, so you are probably wasting your time trying it. One potential problem of using an API for which the application hasn't got the tags is that the API might be trying to grab every command, look and see if it can use it, and so slow everything down.

---

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

good thoughts, i will exclude Vulkan Runtimes (deinstall) and will take a look.

---

The advice that Rob gave is sound. No point in the Vulkan API at this time. I just pulled my old GTX 670's off the shelf and built a new computer based on Windows 10 to see if I could stomach the change in the UI. Installed the Lunatics 0.45 Beta3 platform and running the older Nvidia 358.91 drivers which Windows 10 decided on. There really isn't any reason to run the latest Nvidia drivers on older hardware, there are no improvements to be achieved for distributed computing. I believe there is even less performance because of increased overhead and driver bloat in the later drivers. Run the minimum driver that supports your hardware is my advice. I'm running quite well on the dual 670's though disappointed in the power draw compared to my newer 970's. RAC is building nicely.

---

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

A proud member of the OFA (Old Farts Association)

Said pretty similar things (with more verbosity) in my Cuda 8 prerelease testing. Not sure it will register, but either way you can consider our part done and dusted.

---

"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.

@Keith

You were right :O

The old drivers did the trick, using the recommended drivers from the Asus webpage

nvidia drivers version 353.62 and disable gammacorrection, and switch to performance settings.

Now guppies running with 0.03 to 0.1% percent/per second.

All this new drivers are full overloaded with anything we dont need.

Thanks a lot, keep crunshing.

---

Plz help !
What can i do ? Running a GTX 670 , but these guppi workunits won´t notice it.
Got Lunarics 0.44 installed (64-Bit 0.45 crashes everytime), newest nvidia drivers PLUS Vulkan Runtimes , also newest, running on Win 7 64-Bits.
Could someone/anyone give me advice to make these guppies to rush ?
Any ideas welcome.

Hey T_M,
I might get blasted for this "Short suggestion" to a "Volunteer tester":
you might want to reassign your guppis to the CPU.
see thread: Is it possible to swap a guppi assigned to GPU with a Arecibo assigned to CPU?. Feel free to ask me Qs in that thread.

Here's the full "Long Suggestion" that I hope to get positive feedback (and corrections) on:
Since you are a "Volunteer tester", you might want to consider helping out to improve guppi processing on nVidia's less-than-GTX-750 cards by either:
a. crunching Guppis on GPU in S@H Beta project; or
b. helping Raistmer in this thread: SETI applications for NVIDIA GPU improvement - how you can help

As for me, I'm currently "playing" the Short suggestion (while waiting for My Kevvy's script), and I'm close (I hope) to helping out with the 2nd Long suggestion.

One last point: from what I read here-&-there, the upcoming future will most likely be much more guppi-type WUs (and considering the current non-guppi stuff is from 2010 and has been crunched before with older apps), we might as well be part of the guppi & nVidia solution of the future.

That last point might be mostly wrong and I invite any feedback.
Cheers,
Rob :-}

---

Just got a crapload of "guppy" workunits. Happy crunching everybody!

Just got a crapload of "guppy" workunits. Happy crunching everybody

That should take you no time with a rig like yours:
===>> i7-3770 CPU @ 3.40GHz with NVIDIA GeForce GTX 980 Ti
;-p

---

Calm down.

Guppi is a piece of cake.

See my times.

There is the future -- there is no future but what you make of it.

---

To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones

Calm down. Guppi is a piece of cake. See my times.

All I see is your RAC and your rig with 4 GTX 1080!!!
I would laugh at guppis too with that rig ;-}
...but I'll settle for your new apps. What's the ETA?
Cheers,
Rob :-D

---

Calm down. Guppi is a piece of cake. See my times.

All I see is your RAC and your rig with 4 GTX 1080!!!
I would laugh at guppis too with that rig ;-}
...but I'll settle for your new apps. What's the ETA?
Cheers,
Rob :-D

Hi Rob,

If you'd look at my results you'd see that I have but only a single one 1080. :)

There are some errors to iron out, but with the help of my friends and the long summer vacation and the ultimate understanding of my spouse and my family You all will be eventually enjoying the fruits of my lonely nights at the keyboard etc.

That's what I hope. And I know there are a couple of Heavy Duty professionals ... So I'll get a good night sleep.

I'm awaiting as much as all of the rest of you with NV HW. We know where the slowness comes from. I know what to do with it. But I have a life. *


* Committed to crunching.

---

To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones

You dudes are dedicated and totally insane.
Thanx

---

You dudes are dedicated and totally insane.
Thanx

Thank You!

I know, but that is something that gives and takes. Mostly gives.

---

To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones

It's 230am, and I'm a little bleary-eyed.

On my new machine Big32 I am running 2 x GTX980 with 368.22 drivers. If I read this thread correctly, should I back off to an older driver? If so, which do you guys recommend?

Thanks!

---

I'm running driver: 368.39 without any problem

jear, must be the "old" kepler gpu, that wont´t do it, or drivers system settings.

I can´t reproduce the failure anymore, too many changes have been made.

---

Said pretty similar things (with more verbosity) in my Cuda 8 prerelease testing. Not sure it will register, but either way you can consider our part done and dusted.

Could you please give a status update on the Cuda 8 version ?
I hope it will be an imiprovement on my GTX-750. If you need volunteer testers (either here or over on Beta), please let me know.

Tom

Said pretty similar things (with more verbosity) in my Cuda 8 prerelease testing. Not sure it will register, but either way you can consider our part done and dusted.

Could you please give a status update on the Cuda 8 version ?
I hope it will be an imiprovement on my GTX-750. If you need volunteer testers (either here or over on Beta), please let me know.

Tom

Next mainstream version is likely to be a radical improvement on most Kepler+ class devices, thanks to A massive injection of work by Petri33 (currently in development). While it's unlikely Cuda version will matter as much in the future, in general the Cuda flavour that was released around the time of your device (probably ~Cuda 5-6.5) will remain the best choice, until future applications change to be able to dynamically choose which libraries to use anyway.

That's largely because there has been some performance regression of older generations with newer Cuda versions (unknown reasons). Things are changing pretty quickly though, so much of what I say today could be outdated in a week ;)

---

"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.