TPF diff on SR2 from Windows to Linux

J

juanP

Limp Gawd
Joined
Dec 4, 2010
Messages
137
I was wondering how much TPF diff did everybody get when switching from windows to unix on their SR2's ??

I installed linux but don't see any diff on my dual quads.

I have not used "Tear's Kraken" yet.
 
Benefit - about 12% better times than windows = 19% more PPD by going from windows to BFS/Ubuntu. (as per instructions here)

With Tears wrapper that goes to 28% better PPD than windows.

Make sure C6 state is enabled in the bios under CPU settings.
Checkpoint to 15 mins

Interesting test: c-states has no effect on the x5680 running with turbo off. On the x5660 with turbo on, c-states has a large effect and should be left on.
 
This is from memory, but is going to be very close.

Dual L5640 hex cores @ 3.6, c-state enabled, NUMA disable for all but the Kraken results

6901

Ubuntu/CFS - 13:05/frame
Windows 7 - 12:50/frame
Ubuntu/BFS - 11:35/frame
Ubuntu/BFS/Kraken - 11:10/frame
 
ok i forgot to install BFS. I just did that. now next step in tears wrapper.

on Musky's post it says to download ver 0.2, but on tears original link there is version 4.

should i install the new version or is it V7 client only?
 
Last edited:
juanP said:
ok i forgot to install BFS. I just did that. now next step in tears wrapper.

on Musky's post it says to download ver 0.2, but on tears original link there is version 4.

should i install the new version or is it V7 client only?
Click to expand...

0.2 still works fine. I haven't tried later versions.
 
ok i installed ver 0.2 and running a 6900 WU currently.

lets see how it goes. thanks everybody.
 
I just put 0.4 on my SR-2 today. Ill see how it does with the current WU later tonight or tomorrow. From the readme it seems like 0.4 wont get you much unless you are using V7.

I had a hellofatime getting everything up and running until I saw:
CAUTION: FahCore_a5 is known to be problematic at user-induced shutdowns*.
To be on a safe side make a backup of complete client directory
before hitting Ctrl+C!

If you are going to stop F@H make sure to kill fah6 and not the core.
 
ok my tpf went from 16.4 to 20.13 after adding the bfs and kraken and also enabling the c state.

this is not good at all. what did i miss?
 
no NUMA is disabled.

the only thing diff i did was after enabling c state a diff c6 option showed up in the bios which was set to auto, i changed that to c6.
 
juanP said:
no NUMA is disabled.

the only thing diff i did was after enabling c state a diff c6 option showed up in the bios which was set to auto, i changed that to c6.
Click to expand...

Enable NUMA. The Kraken wrapper makes FAH somewhat "NUMA-aware", so you need it enabled.
 
yea, that might be your issue (as musky says). I don't have anything set specifically with my C-states so I don't know how big a difference it makes. averaging 11:02 right now on a 2692, with similar TPF's on 6900's
 
Activate: AMD said:
yea, that might be your issue (as musky says). I don't have anything set specifically with my C-states so I don't know how big a difference it makes. averaging 11:02 right now on a 2692, with similar TPF's on 6900's
Click to expand...

That is about as good as you are going to get. Since you are using your max multiplier, I don't think c-state matters. NUMA does, though.
 
ok so i enabled NUMA.

so for my SR2@3.6 plus turbo = 3.8 ( on my dual E5620 quads ) i am now getting for a 6900 WU

16.78 --- Windows 7 64bit
16.47 --- Linux ( Ubuntu desktop 11.04)
16.28 --- Linux/bfs/wrapper/numa enabled
 
a bit better... or at least not worse anymore. Is there feedback from anyone else running dual-quads? I don't see any obvious problems at this point
 
You must log in or register to reply here.
Back
Top