Welcome to the discussion on the virtues of crunching on CPU cores as well as GPU devices.


. . A place to continue the back and forth on such virtues from the Server problem thread.

These days, with the changes in data sources, and the increased availability of (cheap) high-multiple-core CPUs it may well make sense for some to look in the direction of not using a GPU.

It is worth remembering that many of the really high core count xeons do not have built in graphics, so one may need to find a motherboard with graphics or a really cheap gpu, low performance if you are not going "fully headless". But don't expect to do any processing on that gpu, just have it driving the monitor when needed...

---

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

IMO it depends how many CPU cores you have available.
I`m running on 4 cores and have another 4 cores free.
It doesn`t hurt to use the free CPU cores and still makes valid science.

Just my 2C.

---

With each crime and every kindness we birth our future.

IMO it depends how many CPU cores you have available.
I`m running on 4 cores and have another 4 cores free.
It doesn`t hurt to use the free CPU cores and still makes valid science.

Just my 2C.

. . I know that pushing a rig to it's utmost slightly increases the power it consumes but only by a percentage. So since you have paid for the hardware and are already using the power why not make the most of it?

. . I like to put every processor cycle to work to achieve maximum efficiency and then I feel the rig is doing "it's bit" for the project.

. . This is why it is good to have multiple approaches to try for optimising your own setups, whether it is an established app like CUDA50 or Petri's Special App or Raistmer's developing SoG app. You can pick one that best suits your own approach and your particular combination of hardware. And positive co-operation along the way with the developers can make any and even all of the approaches that much better.

. . Some people like Grant prefer to tweak the setup to the way they like and just let it run, with minimal fuss, while others enjoy "micro-managing" their rigs to optimise their output and enjoy the involvement. Whichever is your style here is the place to advocate it's virtues.

. . So start your engines ... :)

.

IMO it depends how many CPU cores you have available.
I`m running on 4 cores and have another 4 cores free.
It doesn`t hurt to use the free CPU cores and still makes valid science.

Just my 2C.

. . I have three rigs.

. . My favourite it this i5-6400 with a GTX950 GPU.

. . I have it running three at a time on the 950 supported by one CPU core while the other three cores are crunching. I am running under Lunatics 0.45 Beta (r3430) and this gives me 3 CPU streams and 3 GPU streams.

. . I have happily embraced the MrKevvy/Stubbles school of micro-management and I am streaming nonVLAR work in GPU streams and VLAR/Guppi WUs in the CPU streams and am happy that it produces about 200 WUs per day.

. . It is a sweet rig.

. . My favourite it this i5-6400 with a GTX950 GPU.

. . I have it running three at a time on the 950 supported by one CPU core while the other three cores are crunching. I am running under Lunatics 0.45 Beta (r3430) and this gives me 3 CPU streams and 3 GPU streams.

. . I have happily embraced the MrKevvy/Stubbles school of micro-management and I am streaming nonVLAR work in GPU streams and VLAR/Guppi WUs in the CPU streams and am happy that it produces about 200 WUs per day.

Could you provide:

>  Memory Used
>  GPU Load
>  Power Consumption

Hey folks,

For those who are pro-CPU, I started a new thread in order to remove an unresolved but related tangent from the thread: Panic Mode On (103) Server Problems?

Also, in the spirit of this thread's title, I entitled it:
Philosophy: To DeviceQueueOptimize or NOT (with a focus on: is it "micro managing"?)


"Keep Calm and carry Crunch On"
R ;-)

. . My favourite it this i5-6400 with a GTX950 GPU.

. . I have happily embraced the MrKevvy/Stubbles school of micro-management and I am streaming nonVLAR work in GPU streams and VLAR/Guppi WUs in the CPU streams and am happy that it produces about 200 WUs per day.

Could you provide:

>  Memory Used
>  GPU Load
>  Power Consumption

. . Sure ...

. . Memory: 8GB DDR4 2133 (Single Channel)
. . GPU Load: 95%
. . Power Consumption: ~125W

. . To me the bang for your buck and bang for your Watt numbers are pretty good.

.

Hi,

I have i7-3930K@3800 ht enabled (6+6 cores) and 4*GTX1080

I run GPU with 4 cores and one WU per GPU.
I run CPU with 6 cores.
I leave 2 cores for OS and surfing & other work.

The system is responsive and productive.

If I didn't have that many CPU cores I'd have to run a different configuration reducing first the number of CPU tasks.

The guppi WUs take an hour or so on CPU and under 3 minutes on GPU. No need for queue management here.

Petri

---

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

Hi,

I have i7-3930K@3800 ht enabled (6+6 cores) and 4*GTX1080

I run GPU with 4 cores and one WU per GPU.
I run CPU with 6 cores.
I leave 2 cores for OS and surfing & other work.

The system is responsive and productive.

If I didn't have that many CPU cores I'd have to run a different configuration reducing first the number of CPU tasks.

The guppi WUs take an hour or so on CPU and under 3 minutes on GPU. No need for queue management here.

Petri

. . Hi Petri,

. . Nice to see you here. As you say, with that hardware there is no need to bother queue managing :). But one thing you neglected to mention is that you are running "Petri's special app" [some call it CUDA 8.0) and it is blitzing those WUs ... :)

. . I wonder how it would run on a pair of 970s on a worn out old 2 core Pent-D ?

. . As in the TV show "home improvements" ... "more Power aargh argh argh!" :)

.