17th Folding@Home Holiday Race: The race is over and Mark's Marauders win.

[biodoc]

Welcome to the 17th Annual intra-team Holiday Race between members of Team Anandtech. Participation is voluntary and the goal is to have a bit of fun while contributing to the scientific goals of the Folding@Home project. Just before the start of the race, I'll split the enrolled participants into 2 teams and choose captains for each team.

Here are the rules:
1: Have fun.
2: Let others have fun.
3: Crunch as much as you can, as often as you can and as efficiently as you can.
4: Recruiting members to the teams is allowed.
5: (Sid's rule) Whistle while you work; DON'T whistle while you sip.
6: (Sid's rule) No pouting.

Comment: Sid (Insidious) is a member of the TeAm. He has, unfortunately, not been active for quite a few years ... but we keep his two rules anyway.

The race starts January 1, 12:00 UTC and ends January 21, 12:00 UTC.

• Please post in this thread if you intend to take part (see below for more info).
• If you need help installing or running the F@H app go to this link or ask for help here. The TeAm AnandTech number is 198.
• If you don't have a passkey, get one here. If you have a passkey you'll get bonus points.
• You can join any time during the race and I'll decide which team you are assigned.

For those who want to participate:
Please post your forum name and your Folding@Home name in this thread and estimated PPD.
If you have a one or more GPUs, this database will give you an idea of expected PPD. The same site also has a CPU PPD database.
Please also indicate if you will participate for the full month or just part of the month.
I will put together the initial composition of the teams just before the start of the race.

@Pokey is the captain of the Moonshot Gang team
@Markfw is the captain of Mark's Marauders team

Enrolled participants (forum name - Folding@home name - Est. PPD :

biodoc - biodoc - 19 million
Endgame124 - endgame124 - 7.5 million
Assimilator1 - [TA]Assimilator1 - 1.9 million ppd
Markfw - Mark_F_Williams - 77 million
Pokey - Pokey_TA - 65 million
drnickriviera - DrNickRiviera - 2 million
salvorhardin - salvorhardin - 2.3 million
TennesseeTony - 10esseetony - 6 million
Neurodog - Neurodog - 2.5 million
Cellarnoise - Cellarnoise2 - 10 million
mmonnin03 - mmonnin - 30 million
IEC - IEC - 6 million
The Hive - Bikthar - 4 million
jdiddy - johnnevermind - 12 million
crashtech - crashtech - 31 million
emoga - emoga - 3.5 million
kiska - kiska - ?

 

[Endgame124]

Endgame124 - endgame124 - 7.5 million - full month

 

[Assimilator1]

Well I bring a nice little surprise this year, I'm finally upgrading my RX 580 to an RTX 3060 Ti I bought cheap from a mate . I just need to install it!

Assimilator1 - [TA]Assimilator1 - 1.9 million ppd - full month

Approx double what I've been doing the past few years
So according to the database link above a 3060 Ti is about 200k ppd slower than a 3060! What's that about then?? When the Ti GPU is a better spec than the vanilla! Is the latter's higher RAM clock? [edit] Nope! The Ti has a 256 bit bus vs the 128 bit bus of the vanilla (+125 MHz according to techpowerup).

 

[Markfw]

Mark_F_Williams est ppd somewhere over 60 million ppd. full month

 

[drnickriviera]

Shouldn't it be everyone vs Mark? I'm not sure i'd be much help, but i'll throw my name in as a wildcard because my ppd will vary from 0 to 4mppd.

 

[Pokey]

OK, Pokey folding as Pokey_TA
60-70M PPD. (hopefully closer to 70). Full Month

 

[Pokey]

I'm not sure i'd be much help, but i'll throw my name in as a wildcard because my ppd will vary from 0 to 4mppd

the more the merrier and every credit helps.

 

[Markfw]

Now that I am almost fully back (missing 2 boxes, should be fixed Monday) I can say more like 65-70m.

 

[Assimilator1]

So anyone know why a 3060 would be quicker than a 3060 Ti in F@H? (according to that database, see my previous post). Even though the Ti has a more powerful GPU and nearly double the memory bandwidth....

 

[StefanR5R]

Look at the details of each GPU in this database, the explanation is almost obvious:
a) Linux and Windows perform differently.
b) Check out the "best PPD" and "worst PPD" entries for one GPU under one OS. There is a _huge_ span.
c) Also, there is considerable variation in the "previous 12 weeks" graph.

Edit:
Conclusion: The "recent average folding performance" datum is highly variable. They should also show the standard deviation to make this a bit more obvious. Also, the datum is highly prone to measurement error. The samples which they have available for different cards are not the same.

To compare two cards, it would be better to do that under the same OS and, most importantly, for just one and the same project ID.

 

[Markfw]

So anyone know why a 3060 would be quicker than a 3060 Ti in F@H? (according to that database, see my previous post). Even though the Ti has a more powerful GPU and nearly double the memory bandwidth....

You don't say what OS you are running. Linux is much better for F@H. My 3070 (closest to 3060TI I have) says 4.5 million ppd

 

[Pokey]

So anyone know why a 3060 would be quicker than a 3060 Ti in F@H? (according to that database, see my previous post). Even though the Ti has a more powerful GPU and nearly double the memory bandwidth....

It isn't faster. I have run them side by side and the diff is almost imperceptible at times, but If you use @SefanR5R's methodology you can see it. The ti is faster, but in real life the diff is hard to see. The numbers in that data base are submitted by all manner of rigs with all manner of variables. Nothing scientific about it as far as I'm concerned.

 

[mmonnin03]

Tips on getting FAHControl to connect to other PCs on the local network? I've tried giving access to my main PC, a range of IPs and even 0.0.0.0/0 but it just says connecting. Windows connecting to Linux. Also Linux to Linux. Anyone have an example config.xml? It's been years since I've ran FAH at all.

FAHControl is also a PIA on Ubuntu 20. The v7.7 seems to work https://foldingforum.org/viewtopic.php?f=108&t=36830

 

[Markfw]

As you can see below, for your remote computers, under configure, click remote access tab, then use what you see under the allow line (assuming you have a 192.168.1.xx local net)

Then on your main computer, click the add button, on the lower left, then enter the remote IP address, and give it a name.

And ignore the total estimated points per day, its never even close to the credit you will get. At the moment I MIGHT be getting 70 million ppd.

 

[biodoc]

Tips on getting FAHControl to connect to other PCs on the local network? I've tried giving access to my main PC, a range of IPs and even 0.0.0.0/0 but it just says connecting.

Once you enter in the IP range and password, you'll need to reboot for those settings to take effect.

 

[StefanR5R]

If there are port filters (vulgo 'firewalls') active, the PC which runs FahControl needs to allow outbound traffic to port 36330, and the PC which runs FahClient needs to allow inbound traffic to port 36330.

In FahClient's config.xml, the following options configure the 'remote command server':

Spoiler

Remote Command Server:
  command-address <string=0.0.0.0>
    The address to which the command server should be bound.

  command-allow-no-pass <string=127.0.0.1>
    IP address ranges that are allowed access to the command server with out a
    password if the 'password' option is set. These addresses will also have to
    be allowed IP based access.

  command-deny-no-pass <string=0/0>
    IP address ranges that are not allowed access to the command server with out
    a password if the 'password' option is set. Overriden by
    command-allow-no-pass

  command-enable <boolean=true>
    Set to false to disable the command server.

  command-port <integer=36330>
    The port to which the command server should be bound.

  eval <string>
    Evaluate the argument as a script.

  password <string>
    Set a command server password. Warning, setting a password disables the
    default IP address blocking.

  script <string>
    Run commands from a script file.

(Taken from FAHClient --help.)

Of these, I only use command-allow-no-pass myself and grant passwordless access from the entire LAN, which in my case is in the 192.168.0.* range:

  <!-- Remote Command Server -->
  <command-allow-no-pass v='127.0.0.1 192.168.0.0/24'/>

And then there is also a HTTP server in FAHClient. Section in the manual:

Spoiler

HTTP Server:
  allow <string=127.0.0.1>
    Client addresses which are allowed to connect to this server. This option
    overrides IPs which are denied in the deny option. The pattern 0/0 matches
    all addresses.

  connection-timeout <integer=60>
    The maximum amount of time, in seconds, a connection can be idle before
    being dropped.

  deny <string=0/0>
    Client address which are not allowed to connect to this server.

  http-addresses <string=0:7396>
    A space separated list of server address and port pairs to listen on in the
    form <ip | hostname>[:<port>]

  https-addresses <string=>
    A space separated list of secure server address and port pairs to listen on
    in the form <ip | hostname>[:<port>]

  max-connect-time <integer=900>
    The maximum amount of time, in seconds, a client can be connected to the
    server.

  max-connections <integer=800>
    Sets the maximum number of simultaneous connections.

  max-request-length <integer=52428800>
    Sets the maximum length of a client request packet.

  min-connect-time <integer=300>
    The minimum amount of time, in seconds, a client must be connected to the
    server before it can be dropped in favor or a new connection.

  threads <integer=88>
    Sets the number of server threads.

HTTP Server SSL:
  certificate-file <string>
    The servers certificate file in PEM format.

  crl-file <string>
    Supply a Certificate Revocation List. Overrides any internal CRL

  private-key-file <string>
    The servers private key file in PEM format.

I have this in my configs:

  <!-- HTTP Server -->
  <allow v='127.0.0.1 192.168.0.0/24'/>

Actually I'm not sure if both or just one of these is needed.

 

[StefanR5R]

Anyone have an example config.xml?

Spoiler: config.xml from a dual GPU computer, some presumably unnecessary options removed

<config>
  <!-- HTTP Server -->
  <allow v='127.0.0.1 192.168.0.0/24'/>

  <!-- Remote Command Server -->
  <command-allow-no-pass v='127.0.0.1 192.168.0.0/24'/>

  <!-- Slot Control -->
  <pause-on-start v='true'/>
  <power v='full'/>

  <!-- User Information -->
  <passkey v='0123456789abcdef0123456789abcdef'/>
  <team v='198'/>
  <user v='xii5ku'/>

  <!-- Folding Slots -->
  <slot id='0' type='GPU'>
    <paused v='true'/>
  </slot>
  <slot id='1' type='GPU'>
    <paused v='true'/>
  </slot>
</config>

Spoiler: config.xml from a headless dual-socket computer, 128 threads, some presumably unnecessary options removed

<config>
  <!-- HTTP Server -->
  <allow v='127.0.0.1 192.168.0.0/24'/>

  <!-- Remote Command Server -->
  <command-allow-no-pass v='127.0.0.1 192.168.0.0/24'/>

  <!-- Slot Control -->
  <pause-on-start v='true'/>
  <power v='full'/>

  <!-- User Information -->
  <passkey v='0123456789abcdef0123456789abcdef'/>
  <team v='198'/>
  <user v='xii5ku'/>

  <!-- Folding Slots -->
  <slot id='0' type='CPU'>
    <cpus v='64'/>
    <paused v='true'/>
  </slot>
  <slot id='1' type='CPU'>
    <cpus v='64'/>
    <paused v='true'/>
  </slot>
</config>

The <pause-on-start v='true'/> and <paused v='true'/> switches are in there because I prefer to un-pause each folding slot explicitly after FahClient started. If you prefer that the client proceeds to download work and fold immediately when the client started, just omit these lines.

As for the passkey, see #1. Edit: Keep your passkey secret.

CPU slots should have a thread count which is as high as possible. E.g. one 32-threaded slot gives considerably higher PPD than two 16-threaded slots combined. My example from a 128-threaded computer nevertheless has got two CPU slots, because recent FahCores do not scale beyond 64 threads.

On computers with one or more GPUs, leave a spare core (not just a spare thread) for each GPU feeder. Or do it like me and don't define a CPU slot on GPU-equipped computers at all.

 

[mmonnin03]

Once you enter in the IP range and password, you'll need to reboot for those settings to take effect.

This def did it on 2 PCs.

I've got a working config.xml to start now with 127.0.0.1,10.0.0.0-10.0.0.120 as I've got a 2nd router behind the Verizon FiOS router so my router is 10.0.0.1
Thanks everyone. Not sure when I'll start. Early this week, on time if I can get enough remote access before hand, or around the 8th as I'll be away for like 2.5 weeks.

Anyone use next-unit-percentage still to save a bit of QRB time if ya have a fast connection?

 

[biodoc]

Anyone use next-unit-percentage still to save a bit of QRB time if ya have a fast connection?

Yes, I use 'next-unit-percentage' set to 100.

 

[Markfw]

Yes, I use 'next-unit-percentage' set to 100.

mine is set to 99.

 

[Assimilator1]

Look at the details of each GPU in this database, the explanation is almost obvious:
a) Linux and Windows perform differently.
b) Check out the "best PPD" and "worst PPD" entries for one GPU under one OS. There is a _huge_ span.
c) Also, there is considerable variation in the "previous 12 weeks" graph.

Edit:
Conclusion: The "recent average folding performance" datum is highly variable. They should also show the standard deviation to make this a bit more obvious. Also, the datum is highly prone to measurement error. The samples which they have available for different cards are not the same.

To compare two cards, it would be better to do that under the same OS and, most importantly, for just one and the same project ID.

I didn't have time to take a deep dive into the database, I just looked at the average windows score of the 3060 vs the 3060 Ti, I just looked at it as is. So it's not obvious under those circumstances, would've been worth mentioning that from the outset .
a) I wasn't comparing windows to linux, but windows to windows.
b) & c) and edit. That explains it then, ta.

You don't say what OS you are running. Linux is much better for F@H. My 3070 (closest to 3060TI I have) says 4.5 million ppd

Windows, I'm not installing and learning a whole new OS just for F@H (even though it appears quite substantial, nearly double?!).

 

[Markfw]

I didn't have time to take a deep dive into the database, I just looked at the average windows score of the 3060 vs the 3060 Ti, I just looked at it as is. So it's not obvious under those circumstances, would've been worth mentioning that from the outset .
a) I wasn't comparing windows to linux, but windows to windows.
b) & c) and edit. That explains it then, ta.

Windows, I'm not installing and learning a whole new OS just for F@H (even though it appears quite substantial, nearly double?!).

If you install linux mint, you don't really need to learn a whole new OS, its pretty easy. This is coming from a linux noob, even though 18 of my 20 boxes are linux. If you build a new PC try it. linux cinnamon mint 19 for any hardware that is not the latest. I say that because mint 20.3 has a few odd things, so 19 is best but will not work on the latest hardware.

 

[salvorhardin]

salvorhardin on f@h. I think 2.3m ppd, just added a 6700xt and it'll be for the full month.

 

[StefanR5R]

the explanation is almost obvious:

@Assimilator1, I guess my choice of words of "almost obvious" is on a level of expressions like "almost pregnant". :-) I could have said "rather well obfuscated". The database numbers look super precise, while they aren't.

 

[TennesseeTony]

TennesseeTony - 10esseetony - 6 million - full month