Plissken, I reinstalled my NB cooler with AS. I can hold the upper portion in my fingers but it's hot. The lower portion near the board is so hot I can only hold it for about 15 seconds and then I have to let go. I bought a Noctua NB cooler to address it, but my Tuniq Tower must be modded with a grinding tool to be able to turn it 90 degrees without hitting any caps. Until then I don't OC.

In addition there are several VRMs that are scorching hot on my P5B-E. They now have copper HSs on them as well, and they are still pretty hot but much cooler than before.

I wonder how many major heat sources are on a typical AM2 mobo.

Hey vg30et, how long are you going to keep up your torrid pace? Not that that's a bad thing, mind you.

Noticed the same thing with the VRMs on my gigabyte mobo. Saw them reach the lower 70's according to the infrared thermometer during a memtest with a 40% overclock and fans on low. It's probably a little cooler with speedfan controlling the fans in Windows.

I've had to remove FAH from one server going production and we have several more due up very soon. My ppd will start to decline but I hope to be able to get a x2 3600+ Brisbane folding shortly. If I can find a G0 stepping Q6600 and get some WAF points, that will will be my 2nd dedicated folding machine at home.

http://www.sidewindercomputers.com/thhrthsifori.html

Hey vg30et, here's a great buy on one of the heavy duty AM2 coolers. Check out his oddball little HSs for VRMs and that, even a sizable NB stick on cooler is only $1.25. Real cheap NMB 80mm fans. Some interestng copper NB jobs. At cost priority USPS shipping, I love this place.

waf points?

Wife Acceptance/Approval Factor

aristide1 - thanks for the link. They do have some nice stuff at really great prices.

Wife Acceptance Factor -- ahh, yes, I'll remember that one.

I'm running a 3.4 GHz Pentium D and a X1950Pro. I measured some surprising (to me at least) F@H results:



For the last two months I've been running the 1 console client and 1 GPU client. Yesterday, I dl'ed FahMon to find out my average ppd. I also decided to try the SMP client for fun, and WOW! I was surprised! I thought I'd post this to show the differences between 2 Console clients vs. 1 SMP client. In my case the SMP is almost three times faster than 2x Client versions. So I'm ditching GPU folding and exclusively using the SMP client. My goal is to get back on the top producers list.

Well looks like I have to increase my PPD otherwise Plissken is going to pass me in the near future. =)

You guys running the SMP Client under VMWare, how much RAM is each client using? I have a QX6700 here and I'm running two virtual machines with around 700 MB Ram and I'm wondering if it's enough. What are your experiences?

I'm so tempted to buy a new Q6600 but with new stepping coming in August I don't know if it's a good decision to buy it in july after the price cut.

why are you running two clients?

Because the free version of VMWare only supports two CPUs. So I have to run two clients to use all four cores. I don't want to run the Windows SMP client because it doesn't support resuming and the computer isn't running 24/7. So that's not an option.

Aha,so SMP client only appropriate for 24/7 PCs? so basically an SMP Folding machine has to have a huge CPU folding all the time....huge leccy bill here I come

Not when you're using the Linux client through a virtual machine. The Linux client supports resuming, so you can turn off your PC whenever you want and you can resume at the point where you stopped. Besides that, you get a higher PPD than using the win client. At least that's what I read at folding forums.
Here is a good tutorial for setting up VMWare and folding:
http://rage3d.com/board/showthread.php?t=33880616
But beware of the team number =)

thanks

it doesn't resume? strange...I've used the smp client in linux, in linux through Vmware, and in windows. The vmware isn't much more efficient than using the windows smp client. native linux is surely the best, but I can't undervolt in linux. I'm not sure what you mean though. I start and stop the client all the time in windows, and it picks right up where it left off. It'd probably fold faster using one Win smp client than two linux SMP clients in vmware. You're basically running 4 folding cores on each pair of cpu cores. Maybe it's about the same as just one Windows client though, I can't be sure. Still, the windows client resumes fine for me.

edit: you shouldn't have the pc off for long anyway, the deadline for the SMP units is 4 days.

I tried running 2x xubuntu FAH SMP clients under vmware with my q6600 but it wasn't as fast as a single Windows SMP client. A single Linux FAH SMP session was producing around 1300ppd (100% cpu usage on 2 procs) but when I fired up the second session, both dropped to ~750ppd (100% cpu usage on 4 procs).

I assigned 512mb RAM to each vm session, made sure that there was no memory sharing between the sessions and adjusted proc affinity to the appropriate vm sessions but was still getting poor performance under vmware.

The cheap frys bundled ECS mainboard could be the culprit though...

Hello,

The SMP clients are designed to run on 4 cores, but function on 2. I would just run one client on a quad core.

well, like they said, in the free version of vmware, they can only set it to use 2 cores. To me, two sessions of vmware would just be too much to have going on, not to mention it wont dynamically adjust to cpu load from other programs. Using vmware basically dedicates the processing power to the client, it will bog down the system, at least it did for me.

I won't pass you anytime soon. I built an E4300 for my mom, right now I'm testing it with FAH It's going away shortly, will lose about 600 ppd ...or maybe I leave it installed?

Is that 4300 at stock and folding 24/7? How much more might a 4 mb cache increase the PPD (eg, with a 6320)?

I just started folding with an E6320 running at 2.4GHz. I don't know how many PPD I'm getting, but I finished a (1700 pt. ) WU in about 33hrs.

It's at stock, and probably closer to 800+ ppd if I leave it 24/7, which is basically is right now. My 6300 OC is getting around 1200 ppd, and the Athlon XP (4200? no OC) at work around 600 ppd.
All running the Windows SMP client with -advmethods and -forceasm switches.
I don't know how much extra cache will help, probably not much.

Plissken - how high is that OC on the 6300? I've got an Opty 165 at 2.7 (2 x 1 mb cache) and I'm probably averaging a little over 800 PPD with it folding 24/7.

Disphenoidal - Thanks for the info. That will definitely help with the comparisons. Are you running Linux?

No, Windows.

2.84GHz (7x406), system in tagline.

well aparently the SMP client isnt for every box yet. my xeon box has 2 hyperthreaded cpu's and the smp client thinks thats 4 cpus.. and loads it up accordingly. thus it takes 6 days to do a 5 day WU.. thus missing the deadlines.
and it seems this is a known bug on the FAH forums. so I go back to running 2 clients until they fix that problem.

but i have 1 smp client cranking nicely for most of the day.

Yeah, sadly the Prestonia core 2.4ghz Xeons can't make the SMP deadline but I went ahead and ran 4 fah console clients on it instead of 2. I believe that gives a ~15% ppd boost over just 2 clients.

The Potomac/Irwindale core 3.0ghz Xeons on the other hand don't seem to have any problems completing the SMP units. The additional clock & cache seems to make a huge difference.

Doh! Should have had more caffeine!

The cache question matters to me because I'm looking between a single-channel DDR board with a 6320 versus a dual-DDR board with a 4400. Is the $50 (pre-July price cuts) difference in the processors worth it from a points standpoint, and would it outweigh the performance decrease from single-channel DDR, since the processor could hold more in the cache and it wouldn't have to read from the main memory as much? I think the 4 mb cache chips have access to more of the higher point projects, but that could change at any time. Still, 1700 points in 33 hours is pretty darn good in my book.

it's not a problem. The F@H SMP client launches 4 folding cores, it does this on every machine. On this dual core X2, there are 4 active cores running at 23-25% cpu usage each.

How big is the price difference between the single and dual-channel boards? I didn't think there were any current boards that didn't support dual-channel memory.

Well, since I'm trying to save some $$ and recycle as much of my old hardware that I can, I've come across two boards that would work with my 2 x 512 DDR 400 and would allow for future upgrades down the road:

This ASRock board with integrated graphics (NewEgg link here),
and this ASRock board that's also DDR2 capable (NewEgg link to a similar board here).

The first is micro-ATX with integrated video (great, since I wouldn't have to buy a new vid card) but only supports single-channel DDR2. The second is aout 16-17$ more expensive and would require a vid card. I'm looking for a vid card anyway, but I wanted to hold off since the currently available cards are pretty much power hogs that would require a new power supply. Both boards will both support Kentsfield CPUs, so I can get one when the price comes down more.

I guess the hardest part of the decision is trying to figure out if it would make more sense to get a 4300 and OC the crap out of it on the bigger board or get a 6320 with more cache and OC it to whatever it will take, figuring the more cache will negate the dual-channel loss somewhat. There's about a 50$ difference between the processors. I dunno. I'd be curious to read what you folks think.