hotwired

linuxrouter thanks .  rocking along at 28 min frames

That is excellent TPF. What kind of system configuration do you have?

linuxrouter

touche

Hey guys!

I got VMWare folding set up with BigAdv, but FahMon tells me I'm only getting 1277PPD out of it.  So far, I've finished 2 WUs with it over a span of 12 hours, and the current WU looks like it is worth 470 points... and its 7 hours in.

I've got it set up on 7 cores, affinity set to 0-6, and GPU folding on 7.


This isn't normal, is it?

 


Those are the new Core a3 work units. If you have a passkey setup and have completed 10-work units by the preferred deadline, then you should qualify for bonus points. The base point value is low on these work units, but the bonus points can be fairly high depending on processor and smp configuration. HFM will report work unit and PPD values including both the base and bonus points.


Do you know what TPF you are getting and what project # you are running?

 


No clue on either.  I'll install HFM later tonight.

touche

linuxrouter

touche

Hey guys!

I got VMWare folding set up with BigAdv, but FahMon tells me I'm only getting 1277PPD out of it.  So far, I've finished 2 WUs with it over a span of 12 hours, and the current WU looks like it is worth 470 points... and its 7 hours in.

I've got it set up on 7 cores, affinity set to 0-6, and GPU folding on 7.

This isn't normal, is it?

 


Those are the new Core a3 work units. If you have a passkey setup and have completed 10-work units by the preferred deadline, then you should qualify for bonus points. The base point value is low on these work units, but the bonus points can be fairly high depending on processor and smp configuration. HFM will report work unit and PPD values including both the base and bonus points.


Do you know what TPF you are getting and what project # you are running?

 


No clue on either.  I'll install HFM later tonight.

 

 
i7 smp 8 @4.2 mem at 1200 winsmp A3 core

its been interesting testing both with one machine .
just amazing not a software snag one amazing.

i so want to pair 920s and oc .  evga (hint hint)

moretorque06

The 1.1 has been crunching away great for me, however, I noticed there is a 20min pause between when it finishes the work unit and when it starts uploading it.

The pause in question is in bold. The machine is an i7-860 @ 3.5ghz

That is a fairly common issue with the bigadv work units that has been reported multiple times at foldingforum.org. I ended up switching to a new file system because of this issue. With the previous file system, the delay was as much as 1 hour. The only way I know of to minimize the delay is to use a Ramdisk. I run a Ramdisk on my bigadv systems. This is what my log shows:


[13:48:23] - Writing 95957393 bytes of core data to disk...
[13:48:26]   ... Done.
[13:48:26] - Shutting down core
[13:48:26]
[13:48:26] Folding@home Core Shutdown: FINISHED_UNIT
[13:51:44] CoreStatus = 64 (100)
[13:51:44] Sending work to server
[13:51:44] Project: 2681 (Run 4, Clone 11, Gen 70)

If I can find a file system that will further improve the performance with the folding client, I will look into switching to it.

theGryphon

A feedback: the smp2 seems to like declaring as many cores as you'd be using for folding. If you declare 4 cores in the VM and disable the cores using the affinity switches and use "-smp 4" the TPF is awful; more than 28 minutes on 6020. If you use "-smp 2", TPF is much better; more than 12 minutes on 6020. The best I found was to declare 2 cores in the VM and use "-smp 2"; then TPF is less than 11 minutes on 6020...

Oh, I'm using image 1.1...

Thanks for the feedback. I have been considering adding in several VMX files to cover the various processor configurations.

What PPD are you getting? What clocks?
I'm getting around 2200 PPD on a3 WU's on two cores of my Q6600 OC'ed to 3.7GHz. That's exactly what I was getting with notfred's image.
See my feedback at #520 above. LR's 1.1 image likes that you declare 2 cores at VM, if you didn't do so already...

A general semi-offtopic question: what is the strategy for upcoming Gulftowns? As far as I understand the VM will not work with more than 8 cores. So is native Linux the only way?

leftygof

I guess that's just the difference are the newer a3 units which is why notfred's was a little higher...nevermind

If you have the latest version of HFM.NET, it should show the PPD including bonus points. The PPD may still be less than the a2 work units though. From what I have read, the overclocked dual-cores at 3.8GHz + see higher PPD than the a2 work units. However, on my 2.0 GHz dual-core, PPD is less than the a2 WU.


Otherwise if you uncheck advmethods and check MPICH_NO_LOCAL options in the web config, you should be able to pickup a2 work after finishing the current unit.

Do you think enabling MPICH_NO_LOCAL has an adverse effect on a3 WUs?

Edit: Hey I've got a lot of 9's going on, lol...

Hey LR on HFM my credit is showing NaN ???? wth??

is this correct? it is showing project P6024 GRO-A3 core but the credit is NaN

I edited your 1.1 image to run on @6600 4 cores and memory used.

It download a A3 wu, and it is showing folding on VM and HFM just the credit is NaN

Thanks LR

WB

linuxrouter

wb488641

Hey LR on HFM my credit is showing NaN ???? wth??

is this correct? it is showing project P6024 GRO-A3 core but the credit is NaN

I edited your 1.1 image to run on @6600 4 cores and memory used.

It download a A3 wu, and it is showing folding on VM and HFM just the credit is NaN

Thanks LR

WB

Hey WB,

Try unchecking the virtual machine option in the HFM client config. This option caused some issues with my VM monitoring via HFM.

sweet that did the trick!

thanks

leftygof

wb I have to do that every once in a while,,,,I don't really know why to be honest maybe a network thing

thanks, that is good to know. i don't like the A3 core.

farinhir

So I am running an E8400@3.73Ghz. I have been using notfreds, but I wonder if I would see a benefit from changing over to bigadv. Any suggestions?

To run -bigadv, you need an 8-core capable setup. An overclocked i7 or dual quad works well for -bigadv. With the dual core, you have the option to run either the regular core A2 work units worth 1920 points or run the new beta SMP2 work units.

You can set the -smp flag to lower values, but that applies to the smaller work units such as the 1920-pt core a2 and SMP2 work units. However, the -bigadv work units are much larger. These work units have a deadline of 4-days and it takes 2-3 days to complete with an overclocked i7 on average. It is possible to run bigadv with -smp 7 and still meet the deadline with an i7.

If you want, you can enable advmethods and set the number of cores to 2. With this, you should be able to get SMP2 work provided that some work is available. The SMP2 work qualifies for bonus points like the -bigadv work units.