Updates From Guru3D Folders

That's some quick turn around time! Sounds like you might have more and more of a super folder machine!

 

only thing stopping me going linux is that ive never done it before

 

Is a Linux-based client supposed to be faster?

 

yer a fair bit when it come to the bigadv client the other client smp has a bosted

 

Interesting...I might give it a try on a Mint installation.

 

'k, first thing, sorry for the double post.
I just thought no one'd be peeking inside this thread if I just edited my last post a.k.a. there's nothing new in it. =b

Anyway, been trying the 6.02 Uni client on a Linux Mint.
It's a C2D E6550@3.02GHz and 4x1GB DDR2 Corsair CMX@861MHz.
Compared to my laptop's i5-520M@2.66GHz (2-core w/ HT), it completes 1% steps at an average of 15 minutes each from my laptop's 45 minutes.

Is this normal (performance per clock-wise, 3.02GHz vs 2.66GHz), or the Linux version really is that faster?

I know my C2D has a faster clock, but I thought the i5 should have an architecture advantage, along with SSE 4.1 & 4.2 + HT; E6550 only has up to SSSE 3, IIRC.
Well, unless the client doesn't really use SSE 4.x optimizations.
Both machine uses the exact same FAH settings, both run two instances at the same time, with the exception of versions; 6.23 for laptop (Windows 7 64-bit), 6.02 for desktop (Linux Mint LXDE 32-bit).
I haven't tried getting Mint on my laptop to direct compare performances since I don't want to fiddle with partitions when the laptop is where most of my (real) work is being done.

 

are they the same projects your are working on or are they different

 

The C2D E6550 works on 7028 & 7019, while i5-520M works on 10720 & 10722.
All of these projects are still running, but the C2D ones are close to finish now.
IIRC I've folded 70xx projects before with the laptop, and it was nowhere near as fast; at best it was 35+ minutes per 1% step.

 

sound like it down clocking have you run cpu z to make sure it not are you folding on the gpu on the laptop

 

Actually it stays stable at 2.66GHz, both cores.
I do check for speeds and temps with Speccy and/or CPU-Z from time to time.
This CPU has a TB max of 2.93GHz, and normal clocks of 2.40GHz.
Maybe the client doesn't use SSE 4.1 & 4.2 optimizations much?

And no, I don't fold with the GPU; the laptop's taken quite a beating as is.
I don't like having the GPU running close to 90C 24/7 when not gaming. =b

The C2D just finished 10722, with 10720 soon enough.
Quite a feat, considering it didn't take more than a day.
After 10720 is finished, I plan to install a 64-bit Mint to test out 6.34 SMP.
Too bad no GPU computing on Linux. =/

 

i never done those projects on all my systems running smp 10722, 10720 remember there can be a big different between points on different project's, those 107XX sound like big work units so they just get lower ppd i do't think the instructions sets make much differents.
maybe a silly question but is smp using all cores and may help if you post some of your log

 

Um, maybe I should've been more specific.

These are not SMP WUs.

6.02 on Linux doesn't support SMP, as well as 6.23 on Windows doesn't support SMP.
And even if I'm wrong and they actually do support SMP, I didn't use -smp.
I simply run two instances with different Machine IDs on each system.
In both systems the clients are set to receive normal WUs.
Maybe it's also worth noting that I use console versions (no GUI whatsoever) on both systems, and that I don't set them up as services on both systems.

Here are the logs for both instances on my laptop.
http://paste2.org/p/1775886
http://paste2.org/p/1775890

If there's anything on my stats in EOC being given points today, that's probably the last 2 projects (10720 & 10722).


[edit]
Heck, I didn't even realize I've worked on 10722 on this laptop before.
And it averaged at almost 2 hours per 1% steps?!
[/edit]

 

well you i5-520M will only use 1 thread of 4 i would guess that the c2d just has a more powerful thread/core than the i5-520M giving you a better TPF but i would not expect it to be 15min to 45 min difference

 

I didn't either, but that's just how it went.
Not that it's bad, though.
At least now I know I can relegate folding jobs to that desktop from my laptop.
Now the C2D's starting a big WU on 6.34 with -smp -oneunit; just finished installing Mint 64-bit.
If it takes too long, then I'll just 'revert' back to normal folding with two instances & normal WUs.

Anyway, slightly off-topic, that C2D ran with an 80GB Seagate IDE drive.
I'm also interested in getting GPU folding runs on that desktop, but I'll try it later; gotta catch a few winks first.
Having to set up wrappers and all that screams "Oh sh!t..." all over my brain.


[edit]
The C2D is actually running quite well with that big WU SMP.
Takes around 12~14 minutes per 1%; it's already on 30-ish %.
But I'm pretty sure it'll be hit by a harder WU the next time around, if my previous experience with an i5-2500K was anything to go by.

Here's the log.
http://paste2.org/p/1776518

Gah, I'm so itching to install a 64-bit Linux on this laptop.
I'm a bit curious whether the OS does affect the folding speed.
[/edit]

 

I've decided to install the CPU client and permit it to run for the next 24hrs. Unfortunately, that's the best I can do for now.

 

Just got a whiff regarding GPU folding support in Linux.

http://foldingforum.org/viewtopic.php?f=38&t=20039

If the pace holds, we might see Linux GPU support in around...a year.

 

gdgd more options is all ways good

 

Well, loaded F@H GPU.....and didn't like the results. GPU almost immediately hit 60C from an idle temp of 32C, with the fan set to 80%. Rather unsettling to see a near 30C spike within seconds of F@H starting.... I've still got F@H CPU going for now. It's slow....very slow.... Was showing nearly 2 days to complete 7704....finally down to under 1 day. I'll see what I can do to keep my GPU temp reasonable and stable.....looks like I'll be able to leave F@H CPU going (only a single core) for now as it doesn't consume near the power of my 460....but the GPU will have to be a weekend thing at best.

 

how the air flow in your case my two 460 run at 46c 48c at 100% fan at 850core

 

Airflow is fine. This is the first time the card has reached 60C on F@H. Prior to stopping F@H last year....it was stable at 52-54C.