Bulldozer PPD

J

jebo_4jc

[H]ard|DCer of the Month - April 2011
Joined
Apr 8, 2005
Messages
14,567
Some people have been curious about this.

I have an 8120 running at 255 x 16 = 4080MHz with 4GB DDR3 running at something like 1400MHz. Turbo is on but I doubt it's engaging under FAH load. This is on Ubuntu.

Code: 
 Project ID: 6051
 Core: GRO-A3
 Credit: 481
 Frames: 100

 Name: Bulldozer 8120
 Path: [URL="file://\\BULLDOZER\fah\"]\\BULLDOZER\fah\[/URL]
 Number of Frames Observed: 300
 Min. Time / Frame : 00:02:38 - 21,833.1 PPD
 Avg. Time / Frame : 00:02:40 - 21,425.0 PPD
Code: 
 Project ID: 6055
 Core: GRO-A3
 Credit: 481
 Frames: 100

 Name: Bulldozer 8120
 Path: [URL="file://\\BULLDOZER\fah\"]\\BULLDOZER\fah\[/URL]
 Number of Frames Observed: 100
 Min. Time / Frame : 00:03:43 - 13,021.0 PPD
 Avg. Time / Frame : 00:03:45 - 12,847.8 PPD

Code: 
 Project ID: 6099
 Core: GRO-A3
 Credit: 1588
 Frames: 100

 Name: Bulldozer 8120
 Path: [URL="file://\\BULLDOZER\fah\"]\\BULLDOZER\fah\[/URL]
 Number of Frames Observed: 300
 Min. Time / Frame : 00:07:13 - 27,736.7 PPD
 Avg. Time / Frame : 00:07:50 - 24,526.7 PPD

Code: 
 Project ID: 7500
 Core: GRO-A3
 Credit: 529
 Frames: 100

 Name: Bulldozer 8120
 Path: [URL="file://\\BULLDOZER\fah\"]\\BULLDOZER\fah\[/URL]
 Number of Frames Observed: 300
 Min. Time / Frame : 00:02:18 - 29,602.8 PPD
 Avg. Time / Frame : 00:02:28 - 26,653.8 PPD
Code: 
 Project ID: 7504
 Core: GRO-A3
 Credit: 644
 Frames: 100

 Name: Bulldozer 8120
 Path: [URL="file://\\BULLDOZER\fah\"]\\BULLDOZER\fah\[/URL]
 Number of Frames Observed: 300
 Min. Time / Frame : 00:03:07 - 25,219.2 PPD
 Avg. Time / Frame : 00:03:12 - 24,240.5 PPD
 Cur. Time / Frame : 00:03:13 - 24,038.6 PPD
 R3F. Time / Frame : 00:03:12 - 24,183.9 PPD
 All  Time / Frame : 00:03:13 - 24,038.6 PPD
 Eff. Time / Frame : 00:03:13 - 24,038.6 PPD

Code: 
 Project ID: 7511
 Core: GRO-A3
 Credit: 785
 Frames: 100

 Name: Bulldozer 8120
 Path: [URL="file://\\BULLDOZER\fah\"]\\BULLDOZER\fah\[/URL]
 Number of Frames Observed: 300
 Min. Time / Frame : 00:03:47 - 25,383.7 PPD
 Avg. Time / Frame : 00:03:54 - 24,253.3 PPD

I haven't gotten a single bigadv yet since I started it up about a week ago.

According to a kill-a-watt it is drawing approx 240W from the wall.
 
jebo, you mentioned Ubuntu. I assume you followed musky's guide but I am wondering if you are running with or without the kraken? I realize the kraken is more for 2P and 4P systems but I saw some improvements, with certain kernels, by using it on a 1P system.

I am just wondering, for reference, as musky is sending me his 8120 to play with on my Crosshair V motherboard.
 
no kraken here, but I might give it a shot before I reinstall ubuntu
 
not bad ...probably comparable to adding 2 cores to my thuban at same speed....
 
Don't have the code snippets seen with the OP, so this link is posted vs. dumping a wall of text in here.

The machine is a FX-8120 / Zalman CNPS9500 copper, Asus Sabertooth 990FX and 16gb Corsair Vengeance DDR3 1600. The AMD-supplied cooler butts right against the slot 0 RAM with barely 1/8" separation, hence the Zalman I had lying around.

Everything is at stock / default, running CentOS 5.7 64-bit current distro kernel.

I used the Stanford "Linux" installation instructions.
 
Well, just pulled up my logs and and its certainly much quicker than my x6, well over double in some cases. Given that A3 performs nearly the same under windows and linux, *nix must have had some recent tweaks to get more out of the chip.

It will be interesting to see how brother Tobit and brother Jebo get on with the addiotnal tweaking
 
I don't think the Kraken makes any difference. At least, not on a 7504 WU
 
Great, thanks jebo for some solid numbers.
Though still a disappointment for me, all the listed WU give me better PPD on my 1090@4029MHz which, to be fair, has faster RAM (mostly bigadv take a benefit from this) and probably higher power consumption.

Jebo, your times on the 6055 seems to be way of, they should be very close to the 6051. I've had both these WU on one of of my rigs and on that (AMD X4) the avg. TPF diff is 3 seconds.

Some numbers of mine, also running Ubuntu with no special tricks really:

Code: 
 Project ID: 6099
 Core: GRO-A3
 Credit: 1588
 Frames: 100

 Name: TC tor-A1090T
 Path: \\rog\fah\
 Number of Frames Observed: 200

 Min. Time / Frame : 00:07:07 - 28 323 PPD
 Avg. Time / Frame : 00:07:10 - 28 027 PP

Code: 
 Project ID: 7504
 Core: GRO-A3
 Credit: 644
 Frames: 100

 Name: TC tor-A1090T
 Path: \\rog\fah\
 Number of Frames Observed: 300

 Min. Time / Frame : 00:02:59 - 26 929 PPD
 Avg. Time / Frame : 00:03:01 - 26 484 PPD
 
We need someone to test newer kernels with scheduler fixes for bd...
curious as to any changes if at all in ppd...
 
Patriot said:
We need someone to test newer kernels with scheduler fixes for bd...
curious as to any changes if at all in ppd...
Click to expand...
This is my plan once musky sends me his FX-8120 to play with.
 
Patriot said:
We need someone to test newer kernels with scheduler fixes for bd...
curious as to any changes if at all in ppd...
Click to expand...

The log I gave earlier is for kernel 2.6.32-71.29.1.el6.x86_64, I rather doubt it has the BD caching patch. Will check the notes and see how to apply it.
 
If somebody can tell me if an optimized ububtu kernel exists and how to install it, I would be happy to test.
 
Meh. Zero performance improvement in any of Phoronix testing with the patch.
 
True, but the fact that the patch is 0 for 10 doesn't bode well.
 
So, the FX-8120 I was going to test for musky arrives today. Should I bother trying to get bigadv to fold on it in light of the recent news or just try to OC it and tweak it the best I can for SMP units?
 
I can tell you this: mine has been running non stop for weeks without pulling a single bigadv WU. The Bigadv shortage has hit this rig hard. So, you might not get a bigadv even if you try.
 
jebo_4jc said:
I can tell you this: mine has been running non stop for weeks without pulling a single bigadv WU. The Bigadv shortage has hit this rig hard. So, you might not get a bigadv even if you try.
Click to expand...
Dumb question, did you remember to set the work size to big in the config? I forgot to do this when I first setup the dodeca, as soon as I changed the work size to accept big, I had no problem getting a bigadv.
 
Yeah I checked that. There might be something goofy about this ubuntu install though.
 
I also have some saved off Linux bigadv units I can test for 10 to 20 frames. Maybe I can test for shits and giggles, *shrugs*
 
Just installed the FX-8120 that musky sent me into my Crosshair V and OC'd it to 4 GHz (222x18). Currently getting around 15.8K PPD in Windows on a P6986. I will install Linux when this unit finishes.

2cxx6xz.jpg
 
2v and 6ghz ......cmon...get that dry ice out....
 
I folded a 7504 unit overnight at 4.4 GHz and it ran fine in Windows so I decided to push the overclock a bit more this morning and continuing with Windows testing
til I decide on which OC I want to keep. I think I can still get a few more MHz out of it. Once I am happy with an OC, I will install Linux to see how it truly folds.

Currently folding a P6099 @ 240.79x19 (4.575 GHz) - DDR3-1600 @ 8-8-8-24.

Code: 
 Project ID: 6099
 Core: GRO-A3
 Credit: 1588
 Frames: 100

 Name: io - Bulldozer Test -- Win
 Path: \\io\fah\smp\
 Number of Frames Observed: 15

 Min. Time / Frame : 00:08:53 - 20,309.3 PPD
 Avg. Time / Frame : 00:08:55 - 20,195.6 PPD

34e95km.jpg
 
Tobit said:
I folded a 7504 unit overnight at 4.4 GHz and it ran fine in Windows so I decided to push the overclock a bit more this morning and continuing with Windows testing
til I decide on which OC I want to keep. I think I can still get a few more MHz out of it. Once I am happy with an OC, I will install Linux to see how it truly folds.

Currently folding a P6099 @ 240.79x19 (4.575 GHz) - DDR3-1600 @ 8-8-8-24.

Code: 
 Project ID: 6099
 Core: GRO-A3
 Credit: 1588
 Frames: 100
 
 Name: io - Bulldozer Test -- Win
 Path: \\io\fah\smp\
 Number of Frames Observed: 15
 
 Min. Time / Frame : 00:08:53 - 20,309.3 PPD
 Avg. Time / Frame : 00:08:55 - 20,195.6 PPD
Click to expand...
for comparison, reposting my results above from a 4ghz (I think) OC running DDR3 at roughly 1400MHz in Ubuntu

Code: 
 Project ID: 6099
 Core: GRO-A3
 Credit: 1588
 Frames: 100

 Name: Bulldozer 8120
 Path: [URL="file://\\BULLDOZER\fah\"][COLOR=#ffcc66]\\BULLDOZER\fah\[/COLOR][/URL]
 Number of Frames Observed: 300
 Min. Time / Frame : 00:07:13 - 27,736.7 PPD
 Avg. Time / Frame : 00:07:50 - 24,526.7 PPD

Tobit, what BIOS settings are you using?

I think I should do what you have done...OC in Windows then install Linux.
 
I'm now playing with the FX-8120 @ 4.6 GHz in Linux.

Currently folding a 6901 bigadv bench unit and getting a TPF of 26:46 for a PPD of 35.5K.
 
whats the average for the 2600k? i haven't payed much attention to the points since they changed them.
 
I applied Borislav Petkov's kernel patches to fix the AMD F15h cache aliasing issue to my 3.1 kernel and I have noticed a small improvement on the same bench unit. However, the 2600K still blows the Bulldozer out of the water.

TPF is now 26:31 for 36K PPD on a 6901.
 
I've been playing around, and I just picked up a 6904 on my bulldozer.

At 4ghz, I'm getting a TPF of 1:29:26, and it's not going to make the preferred deadline. I'm a bit surprised by this. Could it be because I'm running out of memory?

Top says I only have 65MB of memory free......
 
jebo_4jc said:
Top says I only have 65MB of memory free......
Click to expand...
That could be misleading as Linux will use unused memory for filesystem caching.

Better command would be: free -m

Code: 
# free -m
             total       used       free     shared    buffers     cached
Mem:          3963       3555        408          0        819       1142
-/+ buffers/cache:       1593       2369
Swap:         5719          0       5719

Quick and dirty way of displaying the amount of total, free and used physical memory (RAM) in the system (-m displays in MB), as well as showing information on shared memory, buffers, cached memory and swap space used by the Linux kernel.
 
jebo, sounds like a bad WU.

I want to do some more Linux testing before I return this CPU to muksy. Skyrim's arrival and some other things keeping me in Windows is preventing this. I will get back to it soon.
 
rmdashrrootsplat said:
That could be misleading as Linux will use unused memory for filesystem caching.

Better command would be: free -m



Quick and dirty way of displaying the amount of total, free and used physical memory (RAM) in the system (-m displays in MB), as well as showing information on shared memory, buffers, cached memory and swap space used by the Linux kernel.
Click to expand...

mine:

Code: 
             total       used       free     shared    buffers     cached
Mem:          3909       3875         34          0          1        133
-/+ buffers/cache:       3741        168
Swap:         3904        161       3743

so, I really am running out of memory, right?
 
You must log in or register to reply here.
Back
Top