-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
The Thing
- Thread starter Norcalsteve
- Start date
- Joined
- Jan 9, 2006
Sounds smart to me, Posidon!
@Norcalsteve:
Matrox is a well known graphics company - they go way back. That's why Ubuntu set up our video drivers with no problem. When I was entering data for a benchmark test, they requested the graphics for Tanker, and had a list to choose from. Matrox was on that list.
OP
- Joined
- Sep 19, 2009
- Location
- Crestview, FL
- Thread Starter
- #163
Dang, thanks for reminding me, Adak... I gotta nab a copy of Server 2008 and do some benches for OG and the bench team... IS there an end to that contest for the benching score submission? I want to wait to make sure I got my set up working well through this week... and heck, that OC guide may come out for my board soon too 
- Joined
- Jan 9, 2006
I'd let Tear and Co. get more experience with their O/C, Steve. With two borked boards, which are not yet proven to be restorable, I'm not sure that it's quite as safe as you might hope.
IMOG's post requesting bench team help, has the link to the Windows Server trial offer. You have to sign up, and yada, yada.
IMOG's post requesting bench team help, has the link to the Windows Server trial offer. You have to sign up, and yada, yada.
OP
- Joined
- Sep 19, 2009
- Location
- Crestview, FL
- Thread Starter
- #165
Good to know... And yes, I am stoked that there is a potential OC capability, but I am sure I will hold off.
OP
- Joined
- Sep 19, 2009
- Location
- Crestview, FL
- Thread Starter
- #166
performance question...
So after my first 2 p6903's I was getting anywhere from 14:08-14:10, then got me a p6901, and now I am back to p6903... My TPF for the last two have been lower, i.e.~14:45. Is that just because the diffrent run/clone/gens run diffrent speeds within the same project? or should it be faster? I noticed the 30 sec increase in PPD when I adjusted my BIOS settings to "optimal"... but yet, theKraken works fine now (sofar) from one unit to the next, which is a good thing. in general... does this look good for ppd/kracken, no OC?
So after my first 2 p6903's I was getting anywhere from 14:08-14:10, then got me a p6901, and now I am back to p6903... My TPF for the last two have been lower, i.e.~14:45. Is that just because the diffrent run/clone/gens run diffrent speeds within the same project? or should it be faster? I noticed the 30 sec increase in PPD when I adjusted my BIOS settings to "optimal"... but yet, theKraken works fine now (sofar) from one unit to the next, which is a good thing. in general... does this look good for ppd/kracken, no OC?
- Joined
- Jan 2, 2008
Look in home/fah/work for the wudata_0x.log file. Open it and look for a string of text, similar to "DD Step 166... loadimb. : force 0.3%." The higher the imbalance, the longer the TPF. At least with my 4P, load balancing does not fully engage on about one out of every four bigadv work units. When that happens, TPF can be higher by as much as 45 seconds.
OP
- Joined
- Sep 19, 2009
- Location
- Crestview, FL
- Thread Starter
- #168
ahhh, i see... that explains the 30-45 increase
Code:
DD step 20504619 load imb.: force 12.4%
At step 20504620 the performance loss due to force load imbalance is 5.3 %
NOTE: Turning on dynamic load balancing- Joined
- Jan 2, 2008
There is a normal TPF variation with P6903 and P6904 work units, which is typically a maximum of 15 seconds. At least from my experience, when the TPF exceeds 15 seconds over the usual, it means a high cores load imbalance. Checking top or System Monitor always shows that Thekraken is running, regardless of the frame times. I don't know whether to attribute the imbalance to the OS or Thekraken. Sometimes stopping and restarting Folding will fix the imbalance, but usually not.
OP
- Joined
- Sep 19, 2009
- Location
- Crestview, FL
- Thread Starter
- #170
Oh, another thing I could try....
When I re-installed my linux to make it EXT3... I followed the instructions as to NOT update it (writing issues after the "....Done" even in ext3)... I could try to only update the Kernal, as the noticeable increase in time happend after I reinstalled to ext3... I thought I seen a guide somewhere as to using a spesific kernal...
curently running 2.6.35-22-generic
When I re-installed my linux to make it EXT3... I followed the instructions as to NOT update it (writing issues after the "....Done" even in ext3)... I could try to only update the Kernal, as the noticeable increase in time happend after I reinstalled to ext3... I thought I seen a guide somewhere as to using a spesific kernal...
curently running 2.6.35-22-generic
- Joined
- Feb 25, 2004
- Location
- N of splat W of Torin
Just to compare note's Steve, its not just you. I have slowed down on every WU.
Here is what I have.
1) 11.10 ext4 but on an SSD. So it should not matter(But it looks like it may)
2) No Kraken currently.
3) Very low temps. So that's not an issue.
So my thoughts were even with load balancing being a factor times should not drop so dramatically. So that points me to delays transferring the Wu. Which points me to ext3(Which you have Steve)
Things I will change...I'll time a oneunit for the weekend,then:
1) My bench is done so I can transfer over.
2) I have a Molex to 8 pin, so I'll add the third 8 pin connector.
3) Re Install Ubuntu 11.10 with ext3 on the SSD, I plan on adding the updates.(Its the way I roll
)4) Install fah according to the guide and see if it puts it in a different place then I have it.
5) Run the client for a few frames then try Kraken again.
Then report back here.
Here is what should happen(and I really hope I get a 6903 so its an exact comparison) My frame time's should be exactly the same, as the first time I started. Kraken should reduce the frame times in the same manner as before.
Last edited:
- Joined
- Jan 2, 2008
Are you referring to TPF becoming longer within a discrete work unit between download and finish or are you comparing more than one work unit out of several?
Also, have you checked the load imbalance entries in the wudata log? If all load balancing is properly engaged and the computer is doing nothing but Folding, the load imbalance should be 3% or less.
- Joined
- Feb 25, 2004
- Location
- N of splat W of Torin
Unfortunately I am away a bit, and really only have time on the weekends to physically be at the rig. I could probably get that info out of HFM. I'll have to check on that...Also its more productive to just let it run below peak performance rather then risk a lost WU.
If its not at 3% or less what does that indicate? Remember I am not running Kraken. I can understand losing tpf to load balancing but I don't get the degradation with each passing WU.(3 complete so far 1 working) I would have expected ups and downs.
I am comparing 3 6903's the tpf has slowed on each.
Last edited:
- Joined
- Jan 2, 2008
A low percentage on load imbalance means that the workload is more evenly distributed between the cores. The better the distribution, that faster the work is accomplished.
If you are only getting 45 second variations, it's not that big of a deal. It's just a matter of how fanatical you are (yes, I'm pointing the finger at me). HFM is the best way to monitor production, as measured by time per frame. Work units within the same project tend to be very consistent.
If you are only getting 45 second variations, it's not that big of a deal. It's just a matter of how fanatical you are (yes, I'm pointing the finger at me
). HFM is the best way to monitor production, as measured by time per frame. Work units within the same project tend to be very consistent.
- Joined
- Feb 25, 2004
- Location
- N of splat W of Torin
I'm chuckling, yeah, I know what load unbalancing is. I thought you were going to tell me what more then 3% indicates. I am pretty sure I am in the 6% range.
I made that post just to let Steve know I was seeing the same thing. Also what I was going to try to do to remedy it.
I said I expected peaks and valleys on load imbalance...I got it this morning, my tpf went down so that gives me some assurance that it's not hardware. Just the nature of all these cores trying to even out.
That 3% seems like it would be more in line with the Kraken running, no?
Thanks for the help L
I made that post just to let Steve know I was seeing the same thing. Also what I was going to try to do to remedy it.
I said I expected peaks and valleys on load imbalance...I got it this morning, my tpf went down so that gives me some assurance that it's not hardware. Just the nature of all these cores trying to even out.
That 3% seems like it would be more in line with the Kraken running, no?
Thanks for the help L
Last edited:
- Joined
- Jan 2, 2008
I misunderstood - sorry, hope I did not come across as condescending
Here's what I meant in other terms: when the system is running right, it's usually at 1% or less load imbalance. If work units run slower than the same series of units completed previously, it's my experience there will be a load imbalance way above 1%, close to 10 or even higher.
Here's what I meant in other terms: when the system is running right, it's usually at 1% or less load imbalance. If work units run slower than the same series of units completed previously, it's my experience there will be a load imbalance way above 1%, close to 10 or even higher.
OP
- Joined
- Sep 19, 2009
- Location
- Crestview, FL
- Thread Starter
- #177
Well, updating the Kernel to the latest actucally added 5 sec to my TPF (but that could be W/U variance)... I dunno, I just wont touch it, and be happy with what I got... Unless there is something in the stock BIOS I can do to the memory, since that seems to be the only thing you can play with, as you cannot overclock it as is. GAH! What am I talking about, tinker too much... must... stabalize... PPD... I am so OCD!
OP
- Joined
- Sep 19, 2009
- Location
- Crestview, FL
- Thread Starter
- #178
Has anyone messed with IPMI and how the heck to log into it when connected to the IPMI LAN port? I want to brows some settings with out having to restart and go into BIOS since I am folding... My current set up is
4p >> LAN cabled to my Laptop >> ICS'ed to my WIFI card...
this is so I can get internet (router to far from my 4p for cabling, and WiFi USB adapter will not work on my 4p)... Since I dont need Internet for the 4p, aside the up/download of W/U's. Can I just unplug it from the LAN port, and into the IPMI lan port? (I have tried that though but I may be missing a step)... when I try to log into the comp with whatever IP address ipconfig tells me on windows, nothing pops up.
4p >> LAN cabled to my Laptop >> ICS'ed to my WIFI card...
this is so I can get internet (router to far from my 4p for cabling, and WiFi USB adapter will not work on my 4p)... Since I dont need Internet for the 4p, aside the up/download of W/U's. Can I just unplug it from the LAN port, and into the IPMI lan port? (I have tried that though but I may be missing a step)... when I try to log into the comp with whatever IP address ipconfig tells me on windows, nothing pops up.
- Joined
- Jan 2, 2008
Steve, unless you are averse to stringing long cables throughout the house (attic, crawlspace, basement), you can get long Cat5 cables very cheap at Amazon with their house brand. They are good quality, too. I never messed with IPMI on my 4P, as I just preferred the simplicity of LAN, running through through switches to my router (computers on three different floors). I monitor all the computers through my primary multi-tasker, which is Win7.
