I just found this machine on the Leaderboard: https://setiathome.berkeley.edu/show_host_detail.php?hostid=8568434

I counted 10 gtx 1080Ti's when I examined a gpu task that was pending. And it lists 13 GPUs. It looks like it is running "stock" Seti under Linux. If it hadn't listed Linux I would have assumed it was running stock seti under windows :)

I think we have met our first repurposed BitCoin Miner. :)

Tom

He/She/It is at 27th place at this moment.

And the cpu is a very high clocked cpu on its baseline. Plus its a "K" which means its unlocked. 4c/8t 4.2GHz.

Tom

---

A proud member of the OFA (Old Farts Association).

As I remember it, you already have a high GPU MB and a CPU on order. You can use those and at least one Gtx 750Ti or higher card to get your "feet" wet one way or another.

I would also put at least one Gtx 750TI or higher on your e5-2670 MB and run it with the Lubuntu(Linux) and Seti(Cuda90) script from the Unix/Linux Questions and Answers message thread. You will be pretty impressed by the performance.

Tom

If you get the raft of gtx 1050Ti's? you ordered you should certainly put two in both your Mining Rig and in your e5-2670 boxes. I had two gtx 750's running in an older Xeon I have and the production even under Linux/CUDA90 was impressive. They were spitting out tasks near every 10 minutes. That was from a 12-14 minute range under Windows.

Since the two higher memory Gtx 1050Ti's has a bunch more cores than the gtx 750Ti you should get even better results. And maybe even go twice as fast as whatever you were getting under Linux/CUDA90 when you move up to Linux/CUDA91.


Tom

---

A proud member of the OFA (Old Farts Association).

Tom,

Can you point me to a place to lean how to get this version of Linux and CUDA installed for Seti?
I have little to no experience with Linux.

It sounds like it has better performance than Windows, correct me if I'm misunderstanding.

Dave

Linux is not the faint of heart. Took me 10 days of non stop learning and a lot of handholding to get it installed and running.

---

Hummm...

Found this on the net...

https://www.techspot.com/review/1683-linux-vs-windows-threadripper-vs-core-i9/

Interesting, first time I have ever seen head to head comparison between Linux and Windows on the same platform.
In some cases Linux smokes Windows but in others Windows wins.

I guess at the end of the day as the author states, its all about what you're doing with your OS that counts with a dash of hardware.

Is there a SPEEDTEST for Seti@home that has both a Linux and Windows version?

Also, I wish I had a Seti APP that would show me all my machines and their state at a glance.

Tom,

Can you point me to a place to lean how to get this version of Linux and CUDA installed for Seti?
I have little to no experience with Linux.

It sounds like it has better performance than Windows, correct me if I'm misunderstanding.

Dave

With the CUDA90 or CUDA91 it has better performance than windows. Otherwise it is apparently a horse race. I would start with CUDA90 because it is a baseline. If you can't get it to work, probably need to stop experimenting.

Take a look at this: https://setiathome.berkeley.edu/forum_thread.php?id=83274&postid=1952639#1952639

It would probably be a good idea to set it up on seperate HD especially if you can't get the dual-boot to setup and share with Windows. To share with windows you pretty much need to shrink the Windows partition and to have a basic non-GPT type windows install. It is simpler to use a different HD.

Another reason to set it up on a seperate HD is you can't crash your Windows setup if you do. Yes, I prefer plugging and unplugging the HD's to change OS's. Some people can do it more eligantly.


HTH,

Tom

---

A proud member of the OFA (Old Farts Association).

Everything I am doing is with crap PC's. A stick of orphan RAM here, an extra HD there. Diving into Linux with a dedicated HD was my plan. Cant wait to get those GPU's!

Tom,

I put that file you suggested (app_config.xml) in C:\Program Files\BOINC and it didn't set my CPU/GPU ratio. Did I do something wrong?

That file lives in the project data directory, which under windows defaults to c:programdata/Boinc/setiathome
This is normally a hidden folder.
Under linux the location varies, most common are off /var/lib/boinc-client and /user-home/boinc/
Exact spellings may vary along with capitisation so one has to look....

Earlier you asked about Libux vs Windows performance. I've found the two about the same when running stock applications. Linux will run on just about anything, and can be faster to instal, with my last junk bench build taking about 30 minutes from pile of bits (including a bare hard disk) to crunching, and another ten to running optimised apps. Compared to the hour or more to get windows installed on the same pile (apart from another bare hard disk). Then there is the highly optimised GPU application that Tom mentions, it uses features of CUDA that are only usable under Linux so will not work uner windows, but you do need to have modern nvidia gpus

---

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

Thank you Rob!

Can someone point me to the plethora of technical data on Seti@home and Boinc? I cant seem to find the "Everything you always wanted to know about looking for Alien Radio signals with massive distributed computing but were to stupid to ask". section on either website.

Thank you Rob!

Can someone point me to the plethora of technical data on Seti@home and Boinc? I cant seem to find the "Everything you always wanted to know about looking for Alien Radio signals with massive distributed computing but were to stupid to ask". section on either website.

http://seticlassic.ssl.berkeley.edu/about_seti/about_seti_at_home_1.html

http://seticlassic.ssl.berkeley.edu/screensaver/index.html

The above are urls for exactly what "we" are doing when we crunch data.

So what do you mean by "technical data"? Server status?

I suggest you go to this message area and take a look: https://setiathome.berkeley.edu/forum_forum.php?id=10

This where the more technical discussions about crunching Seti at Home data happen.

This might be what you are looking for? https://setiathome.berkeley.edu/sah_about.php

Tom

---

A proud member of the OFA (Old Farts Association).

And https://en.wikipedia.org/wiki/Search_for_extraterrestrial_intelligence

---

A proud member of the OFA (Old Farts Association).

Thanks Tom!
Got the mining rig CPU today.
Slowly getting my system on line again.

I'll give updates as I make them!

By the way, this is me...

https://www.linkedin.com/in/dave-nyberg-ab72628/

Dave

Well,

I'm not impressed yet.
I found two of my GPU's were defective so I only have 4 total including the one on the mobo.
Too early to say if they are crunching better than the rippin cores are.
Look for the array. Its called MINER.
Need more GPU's. Need the better ones from China.

ZZZZzzzzz

550W to run the whole frame w/ 4 GPU's.

More later.

Well,

I'm not impressed yet.
I found two of my GPU's were defective so I only have 4 total including the one on the mobo.
Too early to say if they are crunching better than the rippin cores are.
Look for the array. Its called MINER.
Need more GPU's. Need the better ones from China.

ZZZZzzzzz

550W to run the whole frame w/ 4 GPU's.

More later.

Sounds like a "send it back" or ask for a refund for the two that are defective :(

Its the usual issue with used equipment. You act in good faith and you hope the other person does too. If you got this through ebay, after contacting the seller I would ask about the "DOA" issue to eBay.

Tom

---

A proud member of the OFA (Old Farts Association).

Hummm... cant seem to figure out how to attach a PIC here...

Had to turn the beast off for the night.
I feel like I was in a server room all day with that huge power supply screaming. Its capability is no where near being used (Yet)
Now I know why server rooms are so well sound proofed.

So far as I can tell, even with all 4 GPU's dancing, it was impossible that they were outperforming the CPU cores/threads.
They were gonna take about 3 times as long to do the same thing they could do plugged into a full PCIeX16 socket in my other systems. :-( (About 20 min)

My real fear here is that the full 16 lanes of PCI down to one is poo-poo.
Funny thing is, they run hot and heavy current for not doing so much...
Hummm...

Dunno.

Might try and see if i can get Linux breathing tomorrow on another HD.
I will try and attach the PIC to you in a PM Tom.

Later crunchers,

Dave

Hummm... cant seem to figure out how to attach a PIC here...

You can't. You have to host the picture someplace like Google drive or Drop box and then post the url here.

Tom

---

A proud member of the OFA (Old Farts Association).

So far as I can tell, even with all 4 GPU's dancing, it was impossible that they were outperforming the CPU cores/threads.
They were gonna take about 3 times as long to do the same thing they could do plugged into a full PCIeX16 socket in my other systems. :-( (About 20 min)

My real fear here is that the full 16 lanes of PCI down to one is poo-poo.
Funny thing is, they run hot and heavy current for not doing so much...
Hummm...

I can't find anything with 4 gpus on your machine listing.

What gpu's are you playing with?

Tom

---

A proud member of the OFA (Old Farts Association).

Were you looking at the "estimated time" or the actual execution time.
When you first kick off a new computer the estimates are way over the top, so it is far batter to look at the actual run time.
Also what application was being used - it is known the the (stock) SoG application requires a lot of very fast communication between the CPU and GPU for it to run properly so four GPUs may be swamping the PCIE bus channel

---

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

Interesting Bob. Thanks for any input.

Any clue about the PCIe bus GEN setting in BIOS? I noticed its stuck at 1 instead of auto or 2 or 3.
This mobo has a really weird BIOS anyhow but the specs for these NVIDIA Quadro 4000's says, "Interface Type. PCI Express 2.0 x16."
I wonder if using the X16 to X1 riser cards limits the throughput down to 1 or less?
I need a way to scientifically/technically test this? Got any ideas on some software that can tell me what the heck is going on here with this config?

I will run the system again today of course now that the sun is paying for the electric bill and keep an eye out for the final completion times as you say.
I've never done anything like this so I am a little lost just yet.

Also, This app_config.xml file Tom keeps telling me about... Does this have any impact on performance or is it just tinkering with the textual numbers on the user interface display??

Dave