• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Just got a Project 2684

W

Wesleynator

Limp Gawd
Joined
Dec 26, 2008
Messages
182
Downloaded to my Linux VM...hasn't even finished a frame yet. HFM.NET does not yet have the project in it's download file. Run 0, Clone 3, Gen 1 so you know this is the new stuff. Someone shout when their Windows client get's one of these. I have both set up and want to try the Windows version when it is available.
 
How many ppd does this machine do on A2 bigadv? That ppd number sounds in line with other A3 units on an I7 at that speed. It sounds like there is no additional bonus for running bigadv A3 units over regular A3 units.
 
Got one of those WUs in one of my VMs this morning. I had to laugh since the time frame was almost an hour. :eek: which gets me 14,300 ppd. Then again I'm only throwing 6 threads at it :p. Maybe when my other VM on my machine finishes its bigadv I'll stop it and see what happens with the 2684 when I tack on a couple more threads.
 
Damn, this doesn't sound too good. Evil told us that the A3 bigadv will yield lower points, but this is a LOT lower. I cringe to see how low my points will be when I get my first 2684 A3 bigadv since I am only at 3.46 ghz on my 930. Luckily for me, my system just finished a bigadv and picked up another A2 bigadv :p
 
If the output is only 14k on a Lynnfield i7 at 3.8, then there's no reason to change from normal A3.

That being said, since -bigadv is usually very sensitive to memory bandwidth, maybe that's what is holding Lynnfield back. I want results from someone with a third memory channel. Either that or someone who has their RAM clocked to DDR3-2133+.
 
Out of curiosity Wesleynator, how big was your bigadv upload? I'm wondering whether the A3 bigadv WU's have gotten a lot bigger from the 100mb A2 bigadv stuff considering that the TPF's are a lot longer. I would have thought that they would have given a bigger bonus for doing bigadv over regular A3 stuff, but Evil did say that the points could be adjusted by Stanford after a while.
 
What is your tpf? You can use the calculator here (http://www.linuxforge.net/bonuscalc2.php) to work out ppd + bonus.

EDIT: I doubt the 16k ppd figure is not including the bonus, because if it isn't, then the tpf is 7 minutes and 42 seconds. Which doesn't sound very bigadv-ish to me.

EDIT2: According to my absolutely awesome and possibly flawed maths, with 16740ppd (including bonus) you should be getting a tpf of about 45 minutes?
 
Last edited:
you should feel lucky your even getting that... I keep getting these crap WU's that are giving me 4kppd on a i7 920 @ 3.7ghz :rolleyes:
 
I'm itching to get home now and see what my systems have pulled.

I think both my -bigadv boxes won't pull till tomorrow, but my windows box is set to grab a windows unit asap.

It sounds like these buggers want more than just 8 threads to give ya the big points.
 
FLECOM said:
you should feel lucky your even getting that... I keep getting these crap WU's that are giving me 4kppd on a i7 920 @ 3.7ghz :rolleyes:
Click to expand...

That ppd really sucks. 4k is how far my ppd dropped on one of my VMs when it downloaded the 2684. :eek:
 
TPF is abour 43-44 min. Normal PPD for this machine on an A2 is 23K+ and TPF's around 35 min. Sorry, not sure about upload sizes and such.
 
Wesleynator said:
TPF is abour 43-44 min. Normal PPD for this machine on an A2 is 23K+ and TPF's around 35 min. Sorry, not sure about upload sizes and such.
Click to expand...

My i7 930 at 3.46ghz with triple channel memory had a TPF on A2 bigadv of around 35:40 to 35:50. So i'm wondering whether memory bandwidth makes that much of a difference since you have a TPF of 35 min with dual channel memory.

As for upload sizes, you can normally see it when your bigadv work unit has completed and is in the process of uploading to the servers. A2 bigadv work units are typically between 95-101 megs from what i've seen.
 
amdgamer said:
My i7 930 at 3.46ghz with triple channel memory had a TPF on A2 bigadv of around 35:40 to 35:50. So i'm wondering whether memory bandwidth makes that much of a difference since you have a TPF of 35 min with dual channel memory.

As for upload sizes, you can normally see it when your bigadv work unit has completed and is in the process of uploading to the servers. A2 bigadv work units are typically between 95-101 megs from what i've seen.
Click to expand...


the memory bandwidth only comes into play when its loading and offloading the WU from the memory.. so you usually wont see it represented in the PPD shown on hfm or any other monitor but by watching the stats sites you will usually see the PPD drop..
 
New concern: what is the deadline and is it a problem? HFM.NET claims my deadline is 2 Days 8 hrs away and my ETA is 2 Days 6 hrs 13 min. This had been running uniterrupted since it started. TPF while running 2 GPU clients has risen slightly to 47-48min. Is this what we can expect or perhaps the deadline isn't reported correctly by HFM.NET.
 
The preferred deadline is 4 days. I'm pretty much in the same boat as you as my WU will complete with 6 hours to spare. It looks like I may have to rethink my VM strategy if this type of WUs become more common.
 
tjmagneto said:
The preferred deadline is 4 days. I'm pretty much in the same boat as you as my WU will complete with 6 hours to spare. It looks like I may have to rethink my VM strategy if this type of WUs become more common.
Click to expand...

This has me extremely nervous as well considering that i'm likely running the smallest overclock of all of the Core i7 users here. However, we are using a VM right now. Is the Windows SMP client now assigning bigadv work units yet? Not having the overhead of running a guest OS in a VM should help things along significantly as I remember you got much higher ppd averages natively in Linux.

See if you can catch how big the upload is when your WU finally finishes as i'm curious about size considering the longer TPF's everyone is experiencing with this new bigadv.
 
amdgamer said:
Is the Windows SMP client now assigning bigadv work units yet?
Click to expand...
I have the -bigadv flag in one of my Windows A3 clients with 8 cores and I haven't received any WUs yet. Maybe others had better luck.
 
Nope. I just got a new WU about half an hour ago, no A3 Windows bigadv goodness for us yet.
 
I just got one of these last night. Same setup as Wesleynator (i7 860 @ 3.8Ghz, 19x200BCLK).

With one GPU client going concurrently and using all 8 threads or -smp 8, it's pulling in 19,300 PPD and a frame time of just over 45:45. The project is 2684, R3, C17, G0. This box normally does 34:00 TPF on 2681/2683 projects.


Deadline is 3 days, 12 hours from now, ETA is 2 days, 20 hours.

Credit for the unit is 8955 points.

I am hoping that the speed advantage of Windows with A3 units holds up when they come to the Win SMP client, because these units seem to be cutting it close.

I might shut down the GPU client and see what happens to TPF in that regard.
 
I just picked up 2 of these on a Linux box with no -advmethods flag. The first one EUE'd immediately. The second has been running for about 9 hours now. HFM is saying 18K ppd. This is on a box that does about 24K ppd on A2 bigadv. I'm showing a little under 49 minutes per frame on it. 3 days, 9 hours total time on a 4 day limit, so it isn't that close on time. I7@3.8, 4GB dual channel at 1300MHz or so. One of my faster I7s will finish in 6 hours, so we'll see if that one grabs one of these as well.
 
With reports of sub-20k PPD on these WUs, it doesn't seem much different than regular A3 WUs.
 
This particular machine gets a little under 15K on regular A3s, so these are a little better.
 
My bone stock dual 5530 machine just grabbed one and EUE'd immediately with an UNSTABLE_MACHINE error:

Code: 
[03:05:42] + Processing work unit
[03:05:42] Core required: FahCore_a3.exe
[03:05:42] Core found.
[03:05:42] Working on queue slot 05 [May 26 03:05:42 UTC]
[03:05:42] + Working ...
[03:05:42] 
[03:05:42] *------------------------------*
[03:05:42] Folding@Home Gromacs SMP Core
[03:05:42] Version 2.21 (May 10, 2010)
[03:05:42] 
[03:05:42] Preparing to commence simulation
[03:05:42] - Looking at optimizations...
[03:05:42] - Created dyn
[03:05:42] - Files status OK
[03:05:44] - Expanded 20078071 -> 30791309 (decompressed 153.3 percent)
[03:05:44] Called DecompressByteArray: compressed_data_size=20078071 data_size=30791309, decompressed_data_size=30791309 diff=0
[03:05:45] - Digital signature verified
[03:05:45] 
[03:05:45] Project: 2684 (Run 6, Clone 0, Gen 0)
[03:05:45] 
[03:05:45] Assembly optimizations on if available.
[03:05:45] Entering M.D.
[03:05:59] Completed 0 out of 250000 steps  (0%)
[03:06:02] mdrun returned 255
[03:06:02] Going to send back what have done -- stepsTotalG=250000
[03:06:02] Work fraction=0.0000 steps=250000.
[03:06:06] logfile size=12787 infoLength=12787 edr=25 trr=1
[03:06:06] logfile size: 12787 info=12787 bed=25 hdr=1
[03:06:06] - Writing 13325 bytes of core data to disk...
[03:06:08]   ... Done.
[03:06:13] 
[03:06:13] Folding@home Core Shutdown: UNSTABLE_MACHINE
[03:06:13] CoreStatus = 7A (122)
[03:06:13] Sending work to server
[03:06:13] Project: 2684 (Run 6, Clone 0, Gen 0)

Again, native Linux and no -advmethods. I don't think these things are quite ready for prime time just yet.
 
APOLLO said:
I have the -bigadv flag in one of my Windows A3 clients with 8 cores and I haven't received any WUs yet. Maybe others had better luck.
Click to expand...

No luck here either. My work pc has had the flag set and hasn't received one. Although if it did, it wouldn't be able to complete it on time either, since it's an i7 860 running at stock speeds.
 
mplee said:
No luck here either. My work pc has had the flag set and hasn't received one. Although if it did, it wouldn't be able to complete it on time either, since it's an i7 860 running at stock speeds.
Click to expand...
Judging from other recent posts, it seems Linux clients are receiving them, not Windows. /shrug
 
APOLLO said:
Judging from other recent posts, it seems Linux clients are receiving them, not Windows. /shrug
Click to expand...

Hmmm........

It seems like the ppd numbers running bigadv are barely more than what a typical Core i7 will see running Windows A3 stuff. Since people are getting these 2684's in their Linux clients now, I think I will just switch over to Windows and run A3's for now once my current 2681 is done tomarrow. Judging from the TPF's you fellow i7 folders are seeing, I would be running way too close to the 4 day deadline for comfort.

*sigh*
 
amdgamer said:
Hmmm........

It seems like the ppd numbers running bigadv are barely more than what a typical Core i7 will see running Windows A3 stuff. Since people are getting these 2684's in their Linux clients now, I think I will just switch over to Windows and run A3's for now once my current 2681 is done tomarrow. Judging from the TPF's you fellow i7 folders are seeing, I would be running way too close to the 4 day deadline for comfort.

*sigh*
Click to expand...

I have one machine that runs bigadv in a vm and based on everyone's experiences so far, it won't be able to complete this one in time.
 
mplee said:
I have one machine that runs bigadv in a vm and based on everyone's experiences so far, it won't be able to complete this one in time.
Click to expand...

*sigh*

I guess i'll let my VM run till it pulls at least one 2684 just for comparison. If you guys running at 3.8ghz in a VM are only managing to pull 48-49 second TPF's, then it doesn't look promising for me since i'm at 3.46ghz. Perhaps Stanford is shifting bigadv folding to far more powerful systems now.

For people who have the threads and run Linux natively, the bonus should work out pretty nice.
 
amdgamer said:
I guess i'll let my VM run till it pulls at least one 2684 just for comparison. If you guys running at 3.8ghz in a VM are only managing to pull 48-49 second TPF's, then it doesn't look promising for me since i'm at 3.46ghz. Perhaps Stanford is shifting bigadv folding to far more powerful systems now.
Click to expand...
It could be that Stanford prefers to keep this limited to dual-socket systems as much as possible, or limit their processing to only high end single-socket processors like Gulftown. Who knows; I still don't understand it but we have seen similar turns of events in the past. Different benchmarking, different intended hardware applications, there could be a host of reasons. My bet is on a different benchmark system, well that's probably going to be the likely official explanation in any case.
 
Something to keep in mind is that A3 core work units are nowhere near as efficient on Linux as Windows for some reason. I switched my native Linux normal SMP client to an XP VM I already have running on that system because of this. I'm currently doing 7400PPD with the client running inside the XP VM on a 6054. More than likely, I wouldn't be getting any better than 6400PPD if it was running in a native Linux client. I have a feeling the points discrepancy will shrink a good bit once the work units are released for the Windows client.

 
Running two 2684 units currently (both running on opensuse 11.2 native)
bigadv2: P2684(R3,C17,G0) Core 2.19, TPF 55:47 (Dual E5504 Xeons 2.0GHz) - HFM PPD 15114.9
bigadv5: P2685(R5,C0,G0) Core 2.21, TPF 48:45 (OC'd I7 930 @3.89GHz) - HFM PPD 17534.7

PPD on bigadv2 dropped about 5000ppd and bigadv5 dropped about 7000ppd vs A2 bigadv
 
APOLLO said:
It could be that Stanford prefers to keep this limited to dual-socket systems as much as possible, or limit their processing to only high end single-socket processors like Gulftown. Who knows; I still don't understand it but we have seen similar turns of events in the past. Different benchmarking, different intended hardware applications, there could be a host of reasons. My bet is on a different benchmark system, well that's probably going to be the likely official explanation in any case.
Click to expand...

All of a sudden, i'm very interested to see what Intel is going to price their upcoming Core i7 970 x6 processor at. I suspect it isn't going to be much cheaper than their current 980x Gulftown as it is clocked only slightly less. If that i7 970 comes at a price point of $500 or less, I am very tempted to upgrade as it will probably overclock like stink as the current Gulftown does. My next folding rig won't happen till the next holiday season, and I am seriously hoping Bulldozer is as impressive as rumored as I really really really want to go back to AMD. This flirting with Intel for folding makes me feel very dirty.
 
I'm not sure the deadlines we're all worried about are as close as HFM.NET might seem to make them. My rig at 3.8 should finish in about 3.5 days even though HFM.NET says it will only have an hour to spare for the 4 day deadline. My hunch is this is because the time in the VM is not accurate.
 
I just got a 2684 on my bigadv machine...so I'll see if it can complete the unit on time. According to the the points calculator, the max TPF to get a bonus within the 4 day deadline is 57:36. Within a couple of hours, I should be able to see if it can hit that or not.
 
mplee said:
I just got a 2684 on my bigadv machine...so I'll see if it can complete the unit on time. According to the the points calculator, the max TPF to get a bonus within the 4 day deadline is 57:36. Within a couple of hours, I should be able to see if it can hit that or not.
Click to expand...

What are the specs for this rig, and are you running Linux natively or via a VM? I assume it isn't your work pc since that is running windows SMP? I'm eager to find out since my 2681 bigadv will be done tonight, which means that it will probably grab one of those 2684's next.
 
amdgamer said:
What are the specs for this rig, and are you running Linux natively or via a VM? I assume it isn't your work pc since that is running windows SMP? I'm eager to find out since my 2681 bigadv will be done tonight, which means that it will probably grab one of those 2684's next.
Click to expand...

It's running on a Dell desktop, with an i7 860. It's in a VM running on Windows 2008r2.

First frame took 1:08:36. Its running with -smp 7 though since the machine still needs to be usable while folding. With your machine, I think you'll be able to finish before the deadline.

My work pc is just doing SMP now, but I did have bigadv running on it within a VM before.
 
Apparently I just picked one up (checking via logmein.com as I am on the road heading home tonight).

My PpD (based on FHM.net) varies depending on the GPU2 WUs I am getting. My PpD (in sig rig) fluctuates between 16K and 27K. Obviously that is a pretty wide spread. I know that 353's and 768's are very CPU intensive. 450's are high PpD GPU2 WUs that are not CPU intensive. Looks like my average is around 18K PpD. This is down 4K from a Big Adv.

edit: my other rig does have a Big Adv WU
 
So let me get this right, these Are Linux bigadv boxes that are pulling beta A3 bigadv units? And these units are puttting out only 17k on an i7? That's almost 10k lower than my current dedicated bigadv box.

Really can't wait fir my SR2 now if thus is gonna be the norm, these might need 16 threads
Posted via [H] Mobile Device
 
You must log in or register to reply here.
Back
Top