![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Just got a Project 2684
- Thread starter Wesleynator
- Start date
- Joined
- Oct 23, 2007
- Messages
- 6,105
Dedicated boxen (pc-k7b in sig) has now also pulled a 2684.
Code:
[01:31:24] + Processing work unit
[01:31:24] Core required: FahCore_a3.exe
[01:31:24] Core found.
[01:31:25] Working on queue slot 03 [June 2 01:31:25 UTC]
[01:31:25] + Working ...
[01:31:25]
[01:31:25] *------------------------------*
[01:31:25] Folding@Home Gromacs SMP Core
[01:31:25] Version 2.19 (March 6, 2010)
[01:31:25]
[01:31:25] Preparing to commence simulation
[01:31:25] - Ensuring status. Please wait.
[01:31:34] - Looking at optimizations...
[01:31:34] - Working with standard loops on this execution.
[01:31:34] - Created dyn
[01:31:34] - Files status OK
[01:31:36] - Expanded 20075477 -> 30791309 (decompressed 153.3 percent)
[01:31:36] Called DecompressByteArray: compressed_data_size=20075477 data_size=30791309, decompressed_data_size=30791309 diff=0
[01:31:36] - Digital signature verified
[01:31:36]
[01:31:36] Project: 2684 (Run 7, Clone 17, Gen 0)
[01:31:36]
[01:31:36] Entering M.D.
[01:31:50] Completed 0 out of 250000 steps (0%)Make that 19k.....
I'm going to probably -oneunit this boxen and install windows on it. Normal A3 here I come!
- Joined
- Oct 23, 2007
- Messages
- 6,105
my stats on these 2684's, VMware player 3.0 using linuxrouter's VM image. 5gb ram allocated to VM.
i7-1 @ 3.9, ddr3 @ 1488mhz - TPF 48min 17sec - 18,356.5k PPD - 61,550 credit (w/2gpu clients)
i7-2 @ 3.6, ddr3 @ 1368mhz - TPF 50min 53sec - 16,967.7k PPD - 59,957 credit (no gpu clients)
[knock on wood] i have not had any crash and flush with these, so far they have processed ok.
Not sure what to do after...guess I'll play wait and see for the next day or two. Once I get closer to finish date/time, I may close the client, restart with -oneunit, and then go straight to the Win client. Going to see if any word gets out in the next 24-36 hours about # of A2's left to go out in the wild before determining whether to shut the VM's down or not.
i7-1 @ 3.9, ddr3 @ 1488mhz - TPF 48min 17sec - 18,356.5k PPD - 61,550 credit (w/2gpu clients)
i7-2 @ 3.6, ddr3 @ 1368mhz - TPF 50min 53sec - 16,967.7k PPD - 59,957 credit (no gpu clients)
[knock on wood] i have not had any crash and flush with these, so far they have processed ok.
Not sure what to do after...guess I'll play wait and see for the next day or two. Once I get closer to finish date/time, I may close the client, restart with -oneunit, and then go straight to the Win client. Going to see if any word gets out in the next 24-36 hours about # of A2's left to go out in the wild before determining whether to shut the VM's down or not.
- Joined
- Oct 23, 2007
- Messages
- 6,105
Posted a thread over at the FF, guess we'll see what they say. 
I know what you mean. I just wish Stanford assigned a greater credit to these new WUs seeing how unstable and risky they are. It is counterproductive to the project overall if people see a disincentive to run them. At least with the same or more creidt of A2 -bigadvs, more people would work through these issues until Stanford resolved them.I've moved both my rigs to SMP A3. I've lost too many 2684's and am now gun shy. I'd rather take a hit on PpD and be stable than lose a 3 day long WU after 2.5 days
- Joined
- Oct 23, 2007
- Messages
- 6,105
Surprisingly mine are still chugging along. Once I get a flush though, the VM's are gone, I'm not even going to play around losing 3 days worth of crunching on these PMS WU's.
Last edited:
Yup..........
Reward risk as the real world does. In life, there is no such thing as NO for an answer but rather the question of "how much?"
- Joined
- Oct 23, 2007
- Messages
- 6,105
Well I had to close the 1st one, I control+C'd the VM then powered it off. Did my business (was cloning HD partitions to eliminate my smallest drive)....load up the VM, start it,...FLUSH.
I was about to shut it down for good but it grabbed an A2 so it has earned a 'temporary reprieve'.
I was about to shut it down for good but it grabbed an A2 so it has earned a 'temporary reprieve'.
I still haven't lost any of these (and this makes me think unstable OC are the cause of the WU flipping out).
I like them because they only upload every 4 days or so....
The issue witht eh lower ppd, and it seems better ppd with a mid size A3 WU on windows I may switch.
I'm on the fence.
I like them because they only upload every 4 days or so....
The issue witht eh lower ppd, and it seems better ppd with a mid size A3 WU on windows I may switch.
I'm on the fence.
Well I had to close the 1st one, I control+C'd the VM then powered it off. Did my business (was cloning HD partitions to eliminate my smallest drive)....load up the VM, start it,...FLUSH.
I was about to shut it down for good but it grabbed an A2 so it has earned a 'temporary reprieve'.
That is exactly why I havn't added the -oneunit flag yet. My current 2684 is at 99% right now and i'm afraid that i'll lose it if I ctrl-C to close it to start it back up with that flag. Since you just grabbed an A2, i'm hoping that maybe i'll luck out and grab one as well.
I'm pretty certain that unstable OC's may be the reason for some people having issues, but I really think there is a problem based on what we are seeing here, at eVGA's forums, and at the official FF's. Overclocks that have been rock stable folding nothing but A2's should still be stable for A3's considering that many of us had no problems running A3 advmethods via the same core.
If I recall, I had no problems with my overclock of 3.67ghz when I was doing A3's, and had to ultimately downclock to 3.46ghz to get it stable for A2 bigadv. I also find it wierd how so many people have had strings of failures with the 2684 WU all during the first 0-1%. I had a string of 6 consecutive failure EUE and flushes within the 1% myself before it grabbed another 2684 that was completely stable. When an overclock is marginally unstable, it seems like something only fails after it has been running for a long time, hence why Prime95 might be stable for a few hours, but can often fail when you stress test overnight.
We had the same issue with OC when A2 -bigadv. Clocks that people swore where stable because they folded other SMP WU with no problems. People lowered their OCs after huffing and puffing and everything was fine.
They are squeezing everything they can out of these procs. A stable OC for one kind of WU is not always going to be the same for other.
I have the same conversation that say their system is prime 95 stable but crashes when folding.
Also, if a OC is not stable for that WU, it wigging out in the first few frames is not a shock.
They are squeezing everything they can out of these procs. A stable OC for one kind of WU is not always going to be the same for other.
I have the same conversation that say their system is prime 95 stable but crashes when folding.
Also, if a OC is not stable for that WU, it wigging out in the first few frames is not a shock.
You're right, it's the WUs that are unstable. Whenever a new run of WUs are released that have EUEs in a string, it's not the machines they run on that are responsible, especially with systems that have proven stable for months at the same settings. I also experienced a short string of 3-4 WUs in a row that EUEd before the 0% mark - it was nearly instant. No way that's a clock issue.
The interesting thing is that A3 bigadv in Linux doesn't seem to be loading up all 8 threads like A2 bigadv was inside the VM. I'm sure running an emulator like program is part of the this, but the utilization percentage seems to jump around a lot more with A3 bigadv with dips down to 98% per physical core being more common. If so, it might not be pushing the processor as hard. By comparison when running Windows A3 SMP stuff, all cores would get pegged to 100% and stay there without fluctuating according to Core Temp and using the Windows hardware monitor. We'll have to see though as Bruce over at the FF's replied to my thread and stated that the PG is looking into it, but needs more data from more people. When I asked him about when A3 bigadv will come to Windows, he said the PG's official position is soon but that they need to figure out these issues first.
Yeah, all 6 failures occured within the first 1% and it seems odd that it finally grabbed a stable 2684 that just finished without a problem and was submitted 30 minutes ago.
On that note, my computer picked up another 2684 A3 bigadv WU and appears to be chugging away without a problem so far. If it doesn't crash within the first 1%, I feel pretty good about it. Maybe the PG has been making some progress over the last few days trying to resolve these issues
- Joined
- Apr 7, 2006
- Messages
- 10,477
2684 is much more RAM intensive than the 2681-83 A2 WUs were for me.
On my linux it was using 6.5gb of RAM and 900mb of Swap at one point. So for those of you that are only running 6gb that may be problematic.
On my linux it was using 6.5gb of RAM and 900mb of Swap at one point. So for those of you that are only running 6gb that may be problematic.
I thought A3 bigadv required much less memory? If I remember correctly, the ram requirements of A2 bigadv are right around 2gigs or so. I think Evil mentioned in a thread that is gone that A3 bigadv has much lower memory requirements.
I have 12gigs od ram and I think I have at least 7500megs dedicated to the VM.
I finished one on an I7 with 4GB of RAM. I haven't watched memory usage, but these must not be that bad.
- Joined
- Apr 7, 2006
- Messages
- 10,477
Maybe i got a bad unit, cause i was shocked.
One of my -bigadv systems running native Linux cannot complete these WUs. It hasn't uploaded results for over half a week. I have removed the OC and it's now running at stock frequency. It probably won't make the bonus deadline but I don't care. I just want to see if it can even complete one P2684 at all. If it cannot, then Stanford definitely needs to go back to the drawing board.
One of my -bigadv systems running native Linux cannot complete these WUs. It hasn't uploaded results for over half a week. I have removed the OC and it's now running at stock frequency. It probably won't make the bonus deadline but I don't care. I just want to see if it can even complete one P2684 at all. If it cannot, then Stanford definitely needs to go back to the drawing board.
I'm sorry to hear that. I remember the heartbreak I felt when I was browsing the eVGA forums(looking for 2684 info due to my recent string of problems) a few days ago and people were talking about not being able to finish these new 2684's on time on their stock clocked i7's.
We know that Stanford has been working trying to resolve these stability issues, but i'm not sure if your rig will prove anything if you took off the OC for an already stable WU unless you run it for a while to give it a chance to pick up a potentially unstable WU.
- Joined
- Apr 7, 2006
- Messages
- 10,477
Yes, this is true. Most of them have run with no issues on my systems.
I have one running right now and its using less than one gb.
I was only stating that one example because its the only one that failed on me.
The maximum is 57:36 according to the calculator. I wasn't even close to making it on my pc.
OK, without the OC, my system with all the issues is doing a bit over 55:00 per frame. That's cutting it very close. If I see it progressing fine after 24hrs, I might raise the frequency a bit to take off a minute or two from the TPF.
/crosses fingers
Just grabbed one on one of my faster I7's (27K ppd on A2 bigadv):
Looks like about a 7.5k ppd drop on this machine. It runs Ubunutu 9.04 native, 190x20, 6 GB memory @1520. It appears that the A3 bigadv's are less affected by slower memory than A2's. This machine is getting abot 700 ppd more than another I7 at about the same speed with memory at around 1100. With A2 bigadv, the difference was over 3K ppd.
This machine did EUE immediately on 3 in a row before it started processing the fourth. The same machine has processed around 40 A2 bigadv units over the last few months without a single EUE.
Code:
Project ID: 2684
Core: GRO-A3
Credit: 8955
Frames: 100
Name: topraider
Path: \\192.168.1.170\fah\
Number of Frames Observed: 5
Min. Time / Frame : 00:46:19 - 19,538.0 PPD
Avg. Time / Frame : 00:46:24 - 19,485.4 PPD
Cur. Time / Frame : 00:46:19 - 19,518.6 PPD
R3F. Time / Frame : 00:46:24 - 19,483.5 PPD
All Time / Frame : 00:46:24 - 19,483.5 PPD
Eff. Time / Frame : 00:46:42 - 19,358.3 PPDLooks like about a 7.5k ppd drop on this machine. It runs Ubunutu 9.04 native, 190x20, 6 GB memory @1520. It appears that the A3 bigadv's are less affected by slower memory than A2's. This machine is getting abot 700 ppd more than another I7 at about the same speed with memory at around 1100. With A2 bigadv, the difference was over 3K ppd.
This machine did EUE immediately on 3 in a row before it started processing the fourth. The same machine has processed around 40 A2 bigadv units over the last few months without a single EUE.
Last edited:
This chip might, but my motherboard in this rig doesn't have have a multiplier setting if I remember correctly. My memory is probably also a limiting facter on FSB. Unless you back it down to 6x, you are running well over 1600 at 211. I am surprosed at the difference in your two, but I would guess it has more to do with memory speed from the higher FSB than the higher FSB itself. I think that the highest FSB I was ever able to get (12x multiplier, memory at 4x) was around 195, so my boards are also holding me back.
- Joined
- Oct 23, 2007
- Messages
- 6,105
Both of my i7's finished their WU's this morning, the A2 one grabbed another A2, the A3 grabbed another A3.
I really wonder how many A2's are left at this point?
I really wonder how many A2's are left at this point?
Wesleynator
Limp Gawd
- Joined
- Dec 26, 2008
- Messages
- 182
After the first 2684, I've had 3 A2 units since. I'm hopefully the A2's will last until the Windows A3 bigadv units get here.
Yes, it was a folding crash that froze the system. This is a system that hasn't had a crash in months. The temperature is very cool today and all previous 2684s on this system completed without a hitch. I am stumped...
Sorry to hear your woes. For now, I'm sticking to A3. I lost too many WUs to instability that I just could not figure out how to resolve. The PpD is lower, but losing one 2684 A3 more than makes up for any loss in PpD.
tjmagneto
[H]ard DCOTM x2
- Joined
- Aug 6, 2008
- Messages
- 3,299
I'm just wrapping up a pair of bigadvs right now. Sure hope I don't get any 2684s.