Thinking about making a comeback

Man of Honour
Man of Honour
Joined
27 Apr 2004
Posts
107,330
Location
In bed with your sister
As title. Seriously considering making a comeback. I miss the banter and the fun in here.

Having been away for a while, I've no idea what the current optimum clients are so I thought I'd ask for a bit of advice. I'll be running on my gaming rig (which does very little in the way of gaming and, as such, lies idle for long periods).

Spec is:
i7 3770K
3 x GTX670
32GB RAM

You didn't expect me to have a low spec rig now, did you? :p

I've had a quick look around and seen something about high value beta GPU clients which sound like just the ticket.

Any help would be... well.... helpful.

PS: I'm at work and won't be home for a couple of weeks so don't expect anything to happen too soon.
 
F@H has been better on Nvidia GPUs for a long time now - one of the reasons I've always stuck with them.

SLI won't be an issue, I'll just run them separately for Folding and on the odd occasion I want to game, I'll switch off the clients and engage SLI and then reverse the procedure afterwards.
Heat may well be an issue. I won't know until I try it. If it is, I'll take one of the cards out and bung it in the media/file server and make it a dual purpose rig.
 
From what I can see from the sticky, it would appear the simplest way to get running is the 'FAH Tracker'. Does this autodetect your hardware and download the optimum clients for you or are there other clients for higher end hardware?

I'll need to dig up my passkey from somewhere.
Edit: Found it. Can't wait to get home now and get started.
 
Last edited:
I'm thinking I might split the 670s up. I was going to re-use my old gaming rig as a media/file server. I was going to run the i7 950 on 2 cores and underclock and undervolt it to run more economically. I have 6 3TB drives and a 128GB SSD to bung in it. I was also going to run it headless to save on power.

What I'm thinking now is to take the middle 670 out of the new rig and bung it in the old one and keep all cores running with a bit of an overclock and run it as a second folding rig and download/unpack/storage/server machine at the same time. This should keep the heat down a bit on the new rig and boost my PPD a bit. Sometime in the future I might add another 670 to the old rig.

**** me, it doesn't take long for the bug to bite again. I haven't even started yet and I'm planning new hardware :o
 
I don't even want to *think* about the power consumption! I cringe when I think about my 4 GPUs

Put it this way - my electric bill would probably cover a small mortgage ;)

That's one of the things that stopped me before but I'm in a better place financially now so shouldn't be an issue as long as I don't go mad like I did before.
 
What PSU you using to run 3-way SLI?

OCZ ZX Series 1250W '80 Plus Gold' Modular Power Supply [OCZ-ZX1250W-UN]. Plenty :D

670s has very reasonable energy consumption but temps do hit 80c under full load. Maybe point another 120mm casefan towards the top GPU which will be hottest and use a side panel fan for exhausting the hot air out completely, when you wash your hands you can use this additional free heat to completely dry your hands and compete with Dyson :p ;)

Thanks for the tip. I was thinking about airflow. I'll try your suggestion. It is my intention to watercool the 670s eventually (CPU is already watercooled) so airflow won't be an issue after that.
 
The sleeper has awakened :D

Got it up and running now. Tried originally with GPU Tracker but it wouldn't run the GPU clients properly so I binned that and went with FAHControl which seems to have worked fine. I just started it up and let it do its own thing.

Current estimated PPD is about 88,500. From what was posted earlier I was hoping for a bit more. Don't know if it requires tweaking or if it's just the WUs. The SMP client is Folding a p7034 and is estimating 12,800 PPD (3770k at 4.4GHz). Two of the 670s are Folding p8072 and estimating about 25,000 PPD while the third GPU is Folding p8073 and estimating 26,000 PPD. The GPUs are at stock.

Temps seem fine - max CPU temp is about 73c (watercooled with fans turned right down) and max GPU temp about 83c. Bit of a racket but not as bad as I was expecting tbh. Will be better when I get around to watercooling the 670s.
 
Not sure if you've already done it, but.....

Restrict your CPU to six cores (in the client and using something like Bills Process Manager) and dedicate the remaing two to the GPU's (using Bills Process Manager - use two HT Cores for the GPU's).

Did a trial today and found it makes next to no difference. I get almost identical PPD whether I set the SMP client to use 6 cores and set the affinity as described or just leave the SMP client using 8 cores and let Windows sort out the affinity on its own. If anything, my SMP is fractionally better while the GPUs are exactly the same.

I've found using GPUTracker for the CPU and FAHControl for the GPU's works better - this way the CPU should get between 18k to 20k and the GPU's should get 25k to 26k (not the best wu's available for the GPU's atm but the better wu's will increase the GPU's ppd upto around 35k).

Thanks for that. It's made a big improvement on my SMP score - now giving me ~19,600 PPD.

Also, maybe better if you run your rig without the side panel off and manually set the GPU Fan's (EVGA Precisionx or MSI Afterburner) to run at 90% at around 70c (trial and error with this to find a happy medium).

Done that - thanks again. Set fan curve in EVGA precision and it's keeping the hottest of the 670s to 70c with a fan speed around 73%. A bit noisier but the cards are cooler.

My 670's (oc'd) are currently averaging around 28k on the 807x's and the 3770k (oc'd to 4.7GHz - Noctua D14 with temps at 60c) is averaging 24.6k on the 6981's.

I've applied a modest OC to the 670s (GPU clock offset +100 and mem clock offset +299). Getting another 2,500 or so PPD out of each card, taking my total estimated to ~103,000 PPD. What have you got yours running at? Wondering how much more I can get out of these.
 
I did a bit of reading and didn't know about the throttling at 70c thing so I've adjusted my fan curves accordingly and the fan on the hottest card is now running at 80% to keep the temp at 68c. I've upped the OC on the cards to +123 and +400 (1103 boost and 1702 (3404) memory). Stable so far. PPD now over 105,000.

Getting a bit noisy in here now, the sooner I get the cash together for watercooling blocks the better :p
 
Got myself a power monitor today so I've hooked the rig up to that to see how much power it's drawing. Early indications are that it's pulling about 540W at full tilt. This is a lot less than I thought with an OC'd 3770K and 3 x OC'd GTX670s and shows that my 1250W PSU is total overkill :o

Going to leave it on the monitor for 24 hours to see how it averages out over the course of a day.
 
I've just got up (banged my toe last night and woke up early and couldn't get back to sleep) and it's now pulling 644W. It seems to have grabbed some better WUs though - estimated PPD = 147,000.

tpS5YU4.jpg

Now that's more like it :cool:
 
Very Nice :D

Almost as good as my 680's but at a lot less cost.

My 680's on the same wu's are averaging just over 42k across all four.

The rig is running around 192.5k PPD :D

Are you using the Client-type 'advanced' flag or 'beta' flag?

BTW: best keep an eye on the GPU's - those WU's become very unstable with high OC's

Also, what make/model are your 670's and what core speed are they running at? (I will check mine in about an hour and let you know what they are in.c speeds)

I'm not using any additional flags - I just did a basic set up. Do you think adding flags would benefit?

The cards are Palit (standard, not Jetstream). I confess to not understanding EVGA PrecisionX to be honest. Two of the cards are running at 1201MHz and the other one is running at 1254MHz. I don't understand how they're running at different speeds, they all have the same settings applied (sync is checked in PrecisionX).
 
Back
Top Bottom