• 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.

smp2

L

lehmann

[H]ard|Gawd
Joined
Feb 8, 2005
Messages
1,362
just as i figured would happen. its only using 25% or my cpu and still loading the old cores same as the rest of console cliets. am i not setting something right
 
lehmann said:
just as i figured would happen. its only using 25% or my cpu and still loading the old cores same as the rest of console cliets. am i not setting something right
Click to expand...
You specified -advmethods when running the client, correct? Without it, you will only get the older WUs. However, as was spelled out in the SMP2 release post, there is still A1 work out there and A1 work will be assigned if there is no A3 work available. Until A1 is phased out completely, one will always risk the chance of getting A1 work.
 
Zero82z said:
This is the most likely issue. Not using -advmethods will mean that you don't receive A3 workunits, but an A1 unit will still load all four cores to 100%.
Click to expand...

Heh, tell that to my Linux clients stuck working A1 units. It's not even averaging a 50% load on each core. It's probably closer to a 33% load on average and the PPD of the machine is reflecting something similar. Only doing about 2300PPD whereas it normally does about 6200 PPD with A2 units with all cores maxed out.

 
Heh, tell that to my Linux clients stuck working A1 units. It's not even averaging a 50% load on each core. It's probably closer to a 33% load on average and the PPD of the machine is reflecting something similar. Only doing about 2300PPD whereas it normally does about 6200 PPD with A2 units with all cores maxed out.
Click to expand...
same, since 2 days I keep receiving a1 on my opterons, 800ppd instead of 3k... yuck :p

but 25% usage seems like you don't have - smp. I made the same mistake, add it to the link, then run the install.bat file in the folder (you'll need to enter your windows user/password), then delete the work folder+queue file you should be pciking up a3/a1 WUs.
 
My VM just finished a bigadv unit and next one sent was an a1 core... ugh... My 2ghz Pentium dual core running Win SMP keeps getting A3, but now my 3.6ghz i7 is getting handed a1s in its VM. There goes my run at high ppd
 
404 - File not found said:
same, since 2 days I keep receiving a1 on my opterons, 800ppd instead of 3k... yuck :p.
Click to expand...
Which Opterons and how many cores total? My Opterons (dual 280s@2.8GHz) are doing worse on the A3 than A1 - believe it or not. With the new core released today I'm getting roughly 800-900PPD. With the smaller A1 WUs I will see a consistent 1100PPD or higher in a VM. I read somewhere else that the A3 core is not yet optimized for AMD architectures, so this is likely the reason we're seeing abysmal performance. A2 is by far the best performing core/WUs AMD has seen so far.
 
SmokeRngs said:
Heh, tell that to my Linux clients stuck working A1 units. It's not even averaging a 50% load on each core. It's probably closer to a 33% load on average and the PPD of the machine is reflecting something similar. Only doing about 2300PPD whereas it normally does about 6200 PPD with A2 units with all cores maxed out.
Click to expand...
My Q9550 is doing 2500-3000PPD on a P2653, and all four cores are pegged.
 
APOLLO said:
Which Opterons and how many cores total? My Opterons (dual 280s@2.8GHz) are doing worse on the A3 than A1 - believe it or not. With the new core released today I'm getting roughly 800-900PPD. With the smaller A1 WUs I will see a consistent 1100PPD or higher in a VM. I read somewhere else that the A3 core is not yet optimized for AMD architectures, so this is likely the reason we're seeing abysmal performance. A2 is by far the best performing core/WUs AMD has seen so far.
Click to expand...
quad-core 1.8ghz Barcelona. Your lower score on A3 is mainly because of the bonus.
 
404 - File not found said:
quad-core 1.8ghz Barcelona. Your lower score on A3 is mainly because of the bonus.
Click to expand...
Or the lack thereof from the looks of it. I'm starting to not like the bonus scheme at all. It makes sense for -bigadv but I am not really seeing it for regular WUs. It is biased towards the newest hardware because of the amplifying effect it has on PPD. Makes me want to shut down my older systems since I'm producing less than before. A four core system should produce more than 800PPD.
 
Or the lack thereof from the looks of it. I'm starting to not like the bonus scheme at all. It makes sense for -bigadv but I am not really seeing it for regular WUs. It is biased towards the newest hardware because of the amplifying effect it has on PPD. Makes me want to shut down my older systems since I'm producing less than before. A four core system should produce more than 800PPD.
Click to expand...
I agree with you. Standford made a post explaining why they prefer fast WU return. Because the WUs are in a sequence. But from a user perspective this is more trouble, keeping the machines working with no interruption. Most folder don't run dedicated machines, if WUs timesout or just tiny amount of points is given for smp2 work, this will turn off a few people.
 
404 - File not found said:
I agree with you. Standford made a post explaining why they prefer fast WU return. Because the WUs are in a sequence. But from a user perspective this is more trouble, keeping the machines working with no interruption. Most folder don't run dedicated machines, if WUs timesout or just tiny amount of points is given for smp2 work, this will turn off a few people.
Click to expand...
Even though I see the reason for the bonus scheme as completely justified from Stanford's POV, its successful implementation doesn't necessarily mandate a lowering of the existing production from older machines. They could just as easily adjusted the bonus calculations to have retained a certain level of performance in order to accommodate older hardware, and granted whatever bonus rewards to superior architectures they deemed reasonable. In the present scheme of things, people with older but still capable hardware will not want to maintain operating their systems 24/7. In any case, it's an age-old argument in F@H whenever major changes are made and there's little we can do to change things.
 
404 - File not found said:
same, since 2 days I keep receiving a1 on my opterons, 800ppd instead of 3k... yuck :p

but 25% usage seems like you don't have - smp. I made the same mistake, add it to the link, then run the install.bat file in the folder (you'll need to enter your windows user/password), then delete the work folder+queue file you should be pciking up a3/a1 WUs.
Click to expand...

I don't have to worry about part of that as I'm not running the Windows clients. I'm running all Linux clients with two in VMs and the third is running on a Linux native machine. I get the best PPD by far out of my native Linux install even though all three machines are Q6600s@3.6. The -smp flag is definitely set on all of them so I know that's not a problem as it's the only way to get SMP work units and that's all these machines have done.

I'm just running into a problem with A1 core work units. They have some of the worst efficiency I've seen and the ones I'm currently working on seem to be P2665s. It's because of the A1 work units I won't switch to the new native Windows client. Lately I've had to deal with enough of the damn A1s on Linux clients. At least with the Linux clients I have a better chance of picking up something better since I can get A2s.
 
I may just keep the NF VM because I don't like the looks of the Bonus system.
 
gwarren007 said:
I may just keep the NF VM because I don't like the looks of the Bonus system.
Click to expand...

I'm going to stick with my VM for a while as well. The latest EVGA VM is working fine for me and I am just going to sit on the sidelines and see how things sort themselves out.

Eventually I'm sure it will be worthwhile to switch over but I'm not feeling too compelled at the moment.
 
I may just keep the NF VM because I don't like the looks of the Bonus system.
Click to expand...
It's especially annoying if you turn your computer off when you sleep. Overnight my ppd for a typical work unit drops about 4-6000 because of the time the computer's offline. But when I'm awake, it almost doubles the ppd of the SMP client in Notfreds.
 
gwarren007 said:
I may just keep the NF VM because I don't like the looks of the Bonus system.
Click to expand...

tjmagneto said:
I'm going to stick with my VM for a while as well. The latest EVGA VM is working fine for me and I am just going to sit on the sidelines and see how things sort themselves out.

Eventually I'm sure it will be worthwhile to switch over but I'm not feeling too compelled at the moment.
Click to expand...

My understanding is that all SMP work units are going to be moving to the bonus point system eventually. For now, you'll still have the option of running the older client which will get A1 and A2 cores only. However, this client will be phased out eventually and you'll be forced into the newer client with the A3 work units and the bonus point system. This may yet be a while as the A3 core is still in what I call open beta since you need the -advmethods flag to get work units for it. If the open beta works out well, you will probably see the the older client deprecated within a couple of months; especially if the crop of A1 core work units is finished soon. Once those are done, there won't be much reason for Stanford to keep the older clients around any longer.

That said, I don't blame you for sticking with older client VMs for now. My PPD has gone to absolute shit since I installed the newer clients. All the A1 work units I've been getting has killed my PPD and the bonuses for the A3s either hasn't kicked in for me for some reason or something is rather screwy right now. On my native Linux machine (Q6600@3.6) I've seen A3s with lower PPD than A2s whereas A2s usually get 5800-6200. Once or twice I've seen HFM.net showing somewhere in the range of 8000PPD on the A3s. However, the points range is all over the place even though that machine is my server and isn't used for much of anything other than streaming files to other machines which is not a CPU intensive process.

At the moment, I don't know what to think of the new A3s. Hopefully I'll see some change before too long.
 
SmokeRngs said:
the bonuses for the A3s either hasn't kicked in for me for some reason or something is rather screwy right now.
Click to expand...
Have you completed 10 SMP workunits using your passkey yet? That's required to qualify for the bonus.
 
I'm wondering about that too. I certainly didn't complete 10 work units on my passkey after the A3 introduction, and yet the first SMP unit I completed got the bonus.
 
2592ppd - E6500, Project: 2662 (Run 1, Clone 183, Gen 89)
8378ppd - X3440, Project: 2662 (Run 0, Clone 380, Gen 66)
10970 as shown in Fahmon

I'll wait untill I have to to upgrade....
 
Mr. Pedantic said:
I'm wondering about that too. I certainly didn't complete 10 work units on my passkey after the A3 introduction, and yet the first SMP unit I completed got the bonus.
Click to expand...
You don't need 10 A3 units, you need 10 SMP units. If you were folding SMP before using your passkey, those units would have counted.
 
The 10 units can be ANY COMBINATION of A2 or A3 units. :cool:

10 A2
10 A3
9 A2 + 1 A3
1 A2 + 9 A3
5 A2 + 5 A3

all valid. :p
 
You don't need 10 A3 units, you need 10 SMP units. If you were folding SMP before using your passkey, those units would have counted.
Click to expand...
I thought I read somewhere something about the counter being reset because of the 80% completion criteria for the release of A3...?
 
Mr. Pedantic said:
I thought I read somewhere something about the counter being reset because of the 80% completion criteria for the release of A3...?
Click to expand...
The bonus refresh would not impact those who already had been qualified for bonus points. The 80% completion criteria also applied to -bigadv and not limited to A3.

It is being done in part because of some errors in the stats system that counted too many units as bonus-qualifying. Also, in an ideal case the >=10 qualifying threshold would be permanent but the success rate for the 80% threshold would be calculated over a moderate timescale (e.g. 2 months). That isn't feasible with the current system, so this operation has the same effect.

The operation will be along the following lines, if you take bonusExpired as the number of bonus-qualifying work units that expired without being completed and bonusFinished as the number of bonus-qualifying work units that were completed.

for each user/passkey:
fractionExpired = bonusExpired / (bonusExpired + bonusFinished)
if bonusExpired > 0, bonusExpired = round(10.0*fractionExpired)
if bonusFinished > 10, bonusFinished = 10

You will note that this preserves the success rate bonusFinished/(bonusFinished+bonusExpired)
 
You must log in or register to reply here.
Back
Top