Your i7 is considerably faster at folding than a Core 2 Quad CPU. You can't compare your frame times to people who are running C2Qs.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
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.
A3 core/ SMP2 PPD & config thread
- Thread starter Kendrak
- Start date
Mr. Pedantic
[H]ard|Gawd
- Joined
- Sep 19, 2009
- Messages
- 1,707
I doubt it's twice as fast. Your Q9550 is actually clocked higher than my i7, and I take almost half as long to do a frame. Hyperthreading isn't supposed to make a difference; so what does? I don't really think the nehalem architecture is that much better than the Core 2 architecture.
You'd be wrong about that. Nehalem-based CPUs do considerably better than Core 2 CPUs at folding, especially with the A3 core. 10k PPD is pretty common with overclocked i5 CPUs, and 14-15k with overclocked i7 CPUs.
Based on the info out there, what SHOULD I be getting on my QX9300?
Given that the other laptop mentioned above I getting higher PpD I have to assume something is wrong.
Other Specs: 4GB RAM, Win7 Ultimate x64. I left it alone for an hour and frame times did NOT improve
Given that the other laptop mentioned above I getting higher PpD I have to assume something is wrong.
Other Specs: 4GB RAM, Win7 Ultimate x64. I left it alone for an hour and frame times did NOT improve
Memory subsytem. If the A3 core is more memory intensive, that would explain the difference. I don't know the technical differences between the A1/A2 cores versus the A3 to be absolutely certain.
You should be getting at least 4-5k PPD.
- Joined
- May 21, 2008
- Messages
- 5,840
OK, just to be sure, is the new SMP GPU friendly yet? Im running 4 GPUs on this rig and thats the bulk of my points until I get bigadv going. I'd be adding a stock Q66, but if its not worth it ill let it wait
- Joined
- Jan 31, 2003
- Messages
- 14,173
so my Q9550 at 1.3K PPD is severly under than?!?
Yes. If that 1.3K PpD is being reported by HFM.net and you have it configured to calculate bonus points then something is wrong.
You'll probably need to use a priority-changing program to decrease the SMP client's priority, but aside from that you're golden.
@ Vaulter. Make sure you get the version 2.15 a3 core. 2.14 is not very sociable with GPUs and I couldn't get it to respond to any priority changing either.
Nice frame times on that unlocked 720 there OvrrDrive. I'm getting as fast as 8 minutes with my 3 core at 3.35 Ghz. When it is running the 2 HD4890s as well the times jump to about 11-12 minutes.
Nice frame times on that unlocked 720 there OvrrDrive. I'm getting as fast as 8 minutes with my 3 core at 3.35 Ghz. When it is running the 2 HD4890s as well the times jump to about 11-12 minutes.
I'm having the same problem Tobit with that R710 I recnetly got using CentOS 5.4. HFM.net is saying the PPD is about 3600 and that just does not sound right.
I have made sure that the passkey is in there properly and that the config.cfg is not mangled (which happened once by using open office to edit it in linux)
Maybe a bug in HFM.net?
I have made sure that the passkey is in there properly and that the config.cfg is not mangled (which happened once by using open office to edit it in linux)
Maybe a bug in HFM.net?
If it's any consolation, you're not the only one. I have 3 SMP2 clients and all of them are under 1500 PPD, sometimes under 1000 PPD. The weird thing about it is they're all very different hardware configurations. The only thing they have in common is XP as an OS and they all run at least 2 GPU clients. I'm going to be stopping at least two of these clients and either reverting to the original SMP client, running uniprocessor clients instead (not much difference in PPD) or shutting down the systems altogether. It just doesn't make sense to run them any more because the new bonus scheme indicates to me that they're not adequately productive.
- Joined
- Jul 5, 2004
- Messages
- 1,855
I'll throw in my results also. (with fahspy)
3Ghz Core2's at work and they are pulling ~2.5kppd
W3520 @ 3.8Ghz ~25-30k ppd (with me still messing with it, and learning -bigadv)
AMD 965 @ 3.6Ghz 1k ppd (WTF!! core GRO-A3 pulling 5 min frames)
that 965 was getting 6k ppd with notfreds, i'm just hoping fahspy is just wacked
3Ghz Core2's at work and they are pulling ~2.5kppd
W3520 @ 3.8Ghz ~25-30k ppd (with me still messing with it, and learning -bigadv)
AMD 965 @ 3.6Ghz 1k ppd (WTF!! core GRO-A3 pulling 5 min frames)
that 965 was getting 6k ppd with notfreds, i'm just hoping fahspy is just wacked
It might be indeed and sometimes I also see high strangeness with FahSpy compared to HFM.net. But, most of the times both monitors aren't too far off, so I think what I'm seeing is probably the weird bonus calculations that amplify like crazy under a certain threshold in TPF. With GPU clients consuming CPU ccles like there's no tomorrow, your SMP clients get starved and exceed that magic TPF threshold, losing out on the majority of the bonus. It is biased towards very new architectures and low resource consuming machines, ie., the less GPU clients, the better.
You can't. A3 is for SMP2 clients only.
EvilAlchemist
2[H]4U
- Joined
- Jan 11, 2008
- Messages
- 2,732
Something is wrong with your point per day calcuations.
My X3350 @ 2.66 Ghz ( 45nm C2Q) TFP - 00:05:55 - PPD - 6,476
I am at almost 6 min per frame and getting 6.4K
Recheck your Bonus Calcuation Settings.
EvilAlchemist
2[H]4U
- Joined
- Jan 11, 2008
- Messages
- 2,732
What type of CPU are you running in those systems?
If it is a dual core - then yes, SMP1 (A1 cores) or uni-processor client is your best bet.
If it is a quad core - then running SMP2 (A2 -smp 2) flag is a good option.
The reason A1 works well is because it is horrible at using the processor fully, leaving room for the GPUs to get CPU cycles.
I am using WinAFC and have actually tried using High Priority to no avail.
I am running 2.15 A3 cores on two laptops. Both are getting poor performance. It seems to be running just fine on EVERY desktop I have installed this one. So just to confirm, the performance issues, at least for me, is limited to my mobile implementations of SMP2.
- Joined
- Jul 5, 2004
- Messages
- 1,855
The_Heretic
Certified [H]
- Joined
- Jun 22, 2001
- Messages
- 15,110
The 2.15 A3 core is. I've had nothing but GPU loading problems on two boxes using the VMs despite trying various affinity programs to compensate. The new A3 core doesn't have the issue(s).
I am defnitely not getting this and have never even come close
I may need to go back to Linux VMs on this box as the PpD loss isn't worth it.
Sounds right to me.
That is honestly very bizarre. I don't know why you would see that kind of performance.I am defnitely not getting this and have never even come close
I may need to go back to Linux VMs on this box as the PpD loss isn't worth it.
Mr. Pedantic
[H]ard|Gawd
- Joined
- Sep 19, 2009
- Messages
- 1,707
Wow, really? Even with the environment variables my GPU takes about 25% of the ppd of the SMP client.
He's also using a GTX260, not an ATI card.
- Joined
- May 21, 2008
- Messages
- 5,840
does that ATI fix still work? would that help?
The environment variables are the fix. I presume he's already using them since he mentioned it in his post.
I don't see an option to enable or disable CPUs in the BIOS unless I overlooked it somewhere.
Which benchmark do you recommend? I could try anyone's suggestions.
all my Linux SMP2 clients deliver abysmal performance too (less than 1k ppd)
Mr. Pedantic
[H]ard|Gawd
- Joined
- Sep 19, 2009
- Messages
- 1,707
3DMark, PCMark, SuperPi, wPrime. Most benchmarking sites and archives will have lists according to benchmark, architecture, and clock speed, so it should be easy to figure out whether something's wrong. Most of these benchmarks are pretty quick - 3DMark benches are only 5-10 minutes long at most, and everything else is much shorter - the SuperPi 1M benchmark should only take 5-10 seconds.