What is a "frame"

[Qinsp]

Did a search, and didn't find a definition.

So...

What is frame?
Are frames the same between clients, or specific to client?
How do frames relate to WU's?

[P5-133XL]

A frame is the incremental amount done between reporting the progress of the core, in the log, while completing a WU. It is normally 1% of the WU. There have been a very few old projects that broke that rule causing problems for the monitoring programs that assumed the 1% rule but it has been a very long time since that has happened.

A frame is consistently the same across all clients.

[PantherX]

TPF = Time Per Frame.

A Frame is a single percentage of a WU.

Every WU for F@h Clients is divided into 100 Frames which gives 100% of the WU.

The "length" of the frame (aka TPF) varies between Projects and their WUs. Thus when we compare WUs of similar Projects, we ask the TPF to see how fast/slow the system is. The lower the TPF, the faster the WU gets completed and the more PPD you get, vice-versa.

[Qinsp]

Thanks!

Yes, I've seen a lot of TPF references too. Two birds with one stone.

I have a 4 core Windows machine that has a 16:05 TPF on SMP jobs. Sound right?

[PantherX]

Please tell us the Project number, CPU Model, CPU Speed so we can give you similar values Of course a dedicated machine will have lower TPF than a non-dedicated one.

[RAH]

If its a Project 6701/6702 that should be about right.
My Q8200@2.88 is 14:10 or so. While folding GPU also.
If its an i5/7, and a normal SMP (ie 6054) , then slow.
If its an i7 doing big-adv, then WOW!
Check the third party tools, and use FahMon, or HFM.net to figure PPD, and such.

[Qinsp]

A) There is a 6072 job on an AMD Phenom X3 3.4ghz that is 7:10. (500,000 steps)
B) There is a 6702 job on an AMD Phenom X4 3.2ghz that is 15:10. (2,000,000 steps)
C) There is a 6025 job on an AMD Phenom X4 3.2ghz that is 6:00. (500,000 steps)
D) There is a 6701 job on an Intel i7 1.8ghz (notebook) that is 16:05 (2,000,000 steps)

added for reference:

E) 6701 on 1090T @ 3.6ghz = 9:30.

A is an overclocked and core unlocked X2.
B & C are twins, made the same day, identical in all ways. They are machine controllers.
D is a HP notebook that failed to run the CAD software and is being punished.

A, B, C are work computers, so F&H will be interrupted in the morning. Most likely paused all day.

Is there a way to get the 500,000 step jobs? I want to figure out a way to run A/B/C when I go home, then shut them off in the AM.

[PantherX]

TPFs of A, B, C looks fine. TBH, I am surprised at D since it is a laptop and its TPF seems "too fast" but I haven't any idea with laptop SMP folding. BTW does the "work computers" mean that they are located in the office? If yes, then do you have a written permission to run them as stated in the EULA?

Pausing isn't a problem as long as you finish the WU and submit them before the preferred deadline.

Unfortunately, there isn't any sure way to get the 0.5M steps WU. However, currently all SMP2 WUs except 6701, 6702, 6041(?), 6042(?) are 0.5M steps. This may change in the future.

[P5-133XL]

Since WU's have different point values, the frame times are only part of the equation as to efficiency. Steps are really irrelevant for they have little external meaning. The best metric we currently have is not TPF but rather PPD (Points per day) which incorporates points as well as frame times. May I suggest that you get a 3rd party monitoring program like HFM.net that automatically calculates PPD?

Also, since you are apparently new to this, you may wish to get a Passkey and configure it into your clients. While not required, it can significantly improve the points you are receiving for the SMP WU's.

You say some of your machines are work machines. You really should get official written permission to run folding on them. There are potential risks to folding without permission such as the risk of your job if someone take offense. Written permission is self protection even if the manager changes. verbal isn't allwys sufficient.

[Qinsp]

Thanks!

The i7 notebook speed of 1.8ghz is misleading. IIRC, i7 notebooks run faster when needed and when plugged in. 2.5ghz? or something like that.

No problem with the EULA. I'm the boss. Or at least when wifey ain't around! While my employees don't like me messing with "their" machines, what I do after hours doesn't affect them at all.

What's the correct way to pause the FAH SMP client? Is there a "stop when complete" command also?

It's the command prompt, and not a service. I called the executable fah6.exe per instructions.

Yes, I gave them all a passkey.

[bruce]

Stop when complete is a flag called -oneunit.

[Qinsp]

Oh, and it would be great if there was a way to target runtime. There are probably a number of folk who want to run only at night, and it would be better if they could finish the WU than let it sit all day.

[Qinsp]

These machines sit in a temperature controlled laboratory. Running multiple 200w+ computers on "kill" during lab hours isn't really bright. We aren't supposed to go over 72°F, or under 68°F, because it affects the measurements. So even the F@H didn't alter the reading on the machines, I still would shut them down. At night, I'm paying to heat the building for the morning anyways. So in a way, it saves power.

[Zagen30]

TPFs of A, B, C looks fine. TBH, I am surprised at D since it is a laptop and its TPF seems "too fast" but I haven't any idea with laptop SMP folding.

I believe it. I have a 1.6 GHz laptop i7 and it gets frame times on 670x of around 18:30. I don't think a 2:00 boost from a 200 MHz boost is too unreasonable.

The i7 notebook speed of 1.8ghz is misleading. IIRC, i7 notebooks run faster when needed and when plugged in. 2.5ghz? or something like that.

Yeah, but I'm pretty sure that's only when one or two CPU cores are in use and the other cores can be shut off temporarily to reduce power draw/heat. I believe that if all the cores are loaded, such as when the SMP client is running, TurboBoost will not engage since there's not enough thermal headroom. I'm not actually sure the laptop could survive all 4 cores running at 2.5 GHz without a much better cooling system (apologies if your laptop's cooling system is good)- my i7 hovers around 90C when folding with no vent blockage, and I'm certain it's running at 1.6 GHz. A 40% speed boost would certainly kill it.

[Qinsp]

Sounds right. I do know that it sounds like a freakin' jet taking off right now.
It's a HP dv8, and I do not recommend it. It's very poorly built, over-priced, and unreliable. It's already needed a new motherboard at 6 months, some of the features don't work, it's crash prone, etc.

The one good thing about it, is that it's fast. Not just for a notebook, it's fast for a desktop IMO. If it wouldn't crash and eat my CAD files, I'd be using it on my desk.