• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."
  • Community Question: What makes a good motherboard?

Welcome to the Distributed Computing Forum ... say Hello

Page 7 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Fardringle

Diamond Member
Oct 23, 2000
8,802
325
126
Hi. I'm (not) new here. My name is Fardringle and I am an addict.

It has been 10 minutes since my last stats fix and I feel pretty good so far. I think I might even make it a whole hour before my next "fix".

Whew! I'm glad I got that off my chest! Thank you for letting me share...
 

AshlayW

Junior Member
Sep 6, 2018
18
35
61
www.sashleycat.com
Hello you delightful individuals. This is not sarcasm, this is genuinely my Pleased Greeting because I am Happy.

There is a World Community Grid / Folding at Home team here?

I have... a rather considerable deployment of processors that are in need of a focus to aim towards, and a leader board to work on.

If you have Biscuits or Cookies, I might be inclined to throw some of that Compute Power your way.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
20,881
9,077
136
Hello you delightful individuals. This is not sarcasm, this is genuinely my Pleased Greeting because I am Happy.

There is a World Community Grid / Folding at Home team here?

I have... a rather considerable deployment of processors that are in need of a focus to aim towards, and a leader board to work on.

If you have Biscuits or Cookies, I might be inclined to throw some of that Compute Power your way.
We have both. I am currently number one in WCG for our tam. I am number 2 on folding. Number 30 in the world for folding. Please join !
 
  • Like
Reactions: AshlayW

VirtualLarry

Lifer
Aug 25, 2001
49,495
5,583
126
We have both. I am currently number one in WCG for our tam. I am number 2 on folding. Number 30 in the world for folding. Please join !
Maybe... mention our teams to join?

I believe that we are team #198 in F@H, and I don't recall if we are TeAm Anandtech, or just Team Anandtech, in WCG.
 
  • Like
Reactions: AshlayW

TennesseeTony

Elite Member
Aug 2, 2003
3,940
2,913
136
www.google.com
Hello you delightful individuals. This is not sarcasm, this is genuinely my Pleased Greeting because I am Happy.

There is a World Community Grid / Folding at Home team here?

I have... a rather considerable deployment of processors that are in need of a focus to aim towards, and a leader board to work on.

If you have Biscuits or Cookies, I might be inclined to throw some of that Compute Power your way.
Greetings and much welcome! We are quite pleased you are considering joining our TeAm!

Indeed, we have a Folding Team, team 198, as V-Larry mentioned, and we have teams created in nearly all other projects, including of course our beloved World Community Grid. When choosing a team for a BOINC project, merely search for Team Anandtech. I look forward to seeing your name in our weekly DC stats posted every Sunday. :)
 
  • Like
Reactions: AshlayW

Fardringle

Diamond Member
Oct 23, 2000
8,802
325
126
Hello you delightful individuals. This is not sarcasm, this is genuinely my Pleased Greeting because I am Happy.

There is a World Community Grid / Folding at Home team here?

I have... a rather considerable deployment of processors that are in need of a focus to aim towards, and a leader board to work on.

If you have Biscuits or Cookies, I might be inclined to throw some of that Compute Power your way.
Hi AshlayW! Welcome to the TeAm. (Dunno why we capitalize the name like this: TeAm AnandTech. But it's kind of a tradition. :))
 
  • Like
Reactions: AshlayW

AshlayW

Junior Member
Sep 6, 2018
18
35
61
www.sashleycat.com
Thanks guys and gals and everyone in between. By the way, I am just finishing up a challenge on my own Team (SashlecatCrunch) then I will join and put my, uh... farm to work... My farm ain't the biggest but it's all mine! Crunchy tasty Science.

CPUs at my disposal:

24/7:
TR 1920X (12c/24t)
4x R7 2700 (32c/64t)
R5 2600X (6c/12t)
R5 3400G (4c/8t)
Athlon 3000G (2c/4t)

part time (overnight):
TR 2950X (16c/32t)

GPUs:

24/7:
GTX 1660 Super
GTX 1650 Super
RX 590

Part time (overnight):
RTX 2080 Ti

:D

Edit, I am just going to grab a slice of this pie then I'll be ready. because I want a slice of that pie. Pie is good.

Hey so does Anandtech have a pie?
 
Last edited:

Endgame124

Senior member
Feb 11, 2008
431
266
136
Thanks guys and gals and everyone in between. By the way, I am just finishing up a challenge on my own Team (SashlecatCrunch) then I will join and put my, uh... farm to work... My farm ain't the biggest but it's all mine! Crunchy tasty Science.

CPUs at my disposal:

24/7:
TR 1920X (12c/24t)
4x R7 2700 (32c/64t)
R5 2600X (6c/12t)
R5 3400G (4c/8t)
Athlon 3000G (2c/4t)

part time (overnight):
TR 2950X (16c/32t)

GPUs:

24/7:
GTX 1660 Super
GTX 1650 Super
RX 590

Part time (overnight):
RTX 2080 Ti

:D

Edit, I am just going to grab a slice of this pie then I'll be ready. because I want a slice of that pie. Pie is good.

Hey so does Anandtech have a pie?
Well, we don't have pie, but we do have weekly DC stats:

And given we have the stats, we could certainly bake a pie with them. .... I might have to look into that.

Edit: Come to think of it, we do have some pi here:
 
Last edited:
  • Haha
Reactions: TennesseeTony

Immy74

Junior Member
Oct 9, 2020
1
0
6
I've been doing distributing.net since rc5-64. I've always used cpu moo client and cuda moo client (when it became available). Now question is, why am I getting a lower key count using cpu+gpu vs just gpu by self? 16k cpu+gpu, 23k gpu only.
 

Icecold

Senior member
Nov 15, 2004
542
378
136
I've been doing distributing.net since rc5-64. I've always used cpu moo client and cuda moo client (when it became available). Now question is, why am I getting a lower key count using cpu+gpu vs just gpu by self? 16k cpu+gpu, 23k gpu only.
Usually when CPU+GPU is lower than GPU only it means that there is not enough CPU resources available to fully utilize the GPU. If you run GPU-z you can see what the load is on the GPU when it's running - if it's close to 100% when not running CPU as well, but is lower than that running both it would tell you that there aren't enough CPU resources.
 

StefanR5R

Diamond Member
Dec 10, 2016
3,716
4,006
106
@Immy74, hello and welcome! :-)

What @Icecold said. It's been a while since I last ran the dnet client (via Moo! Wrapper). But at least the CUDA application, like virtually all CUDA applications, most certainly has got a CPU thread running which continuously polls GPU state, and occasionally copies data. An interrupt handler of the GPU driver may be involved too. Hence, keep a core of the CPU spare for this. If you have an Intel CPU with Hyperthreading, it may even be beneficial to leave 50% + 1 threads of the CPU unused. Or leave the CPU unused entirely (except for the CUDA application) in order to get maximum boost clock for it.

The rationale is that the GPU is so much more productive at dnetc, that it hurts less to reduce CPU allocation for the CPU-only client, or leave that one off entirely, than it hurts to give insufficient CPU capacity to the GPU feeder thread.

There is also an OpenCL version. Unless memory plays tricks on me, that one works better than the several CUDA versions at moowrap.net. My boinc data directories have files in them which tell me that I enforced the use of the OpenCL client instead of the CUDA client as recently as July 2020. Besides better performance in general, it is possible that the dnetc Nvidia-OpenCL version also uses less CPU than the CUDA version.
 
Last edited:

VietOZ

Member
Aug 3, 2019
81
114
66
There is also an OpenCL version. Unless memory plays tricks on me, that one works better than the several CUDA versions at moowrap.net. My boinc data directories have files in them which tell me that I enforced the use of the OpenCL client instead of the CUDA client as recently as July 2020. Besides better performance in general, it is possible that the dnetc Nvidia-OpenCL version also uses less CPU than the CUDA version.
OpenCL does perform better on Moo, only slightly though, and that's comparing with CUDA 7. CUDA version is actually using less CPU than the OpenCL version. OpenCL you'd need 1 thread per task on either linux or windows.
 

ASK THE COMMUNITY