Folding Forum

The description of your case fans indicates they are not on PWM controllers, so you have several choices...

The Noctua NF-A14 PWM has a max speed of 1500 RPM (1200 with the LNA cable) vs 1000 for the Fractal fans. That gives it a max airflow of 82.5 CFM (67.9 with LNA) vs 66 for the Fractal. Noise and airflow are virtually identical with the LNA, and you'd have more OC headroom without the LNA. You may decide to plug them into a 4-pin MoBo slot instead of the 3-pin case controller, and have them controlled by the BIOS program or Gigabyte software (if available on that MoBo). A similar setup on my MSI MoBo works very nicely.

The NF-A14 FLX is a 3-pin model that tops out at 1200 RPM, with the same flow and noise as the PWM version with LNA. You can step it down with either of 2 included LNAs for less flow/noise.

The NF-A14 ULN is a lower-performing 3-pin model that you probably don't want in that rig.

A plan that I would try is a pair of the PWM fans in the front, and the FLX versions in the bottom and rear. Start with the LNA on the front pair (works with a Y cable for both fans) and the rear exhaust. Let the bottom fan blow full force onto the GPUs.

Since your current fans are equivalent to the FLX fans, you might try the front PWM pair alone at first.

Thanks for all that, jrweiss! I will look into all the fans you've mentioned for sure later on down the road and decide which may be best for my build. I really do appreciate all the help you've been giving me! I got the Noctua fans installed onto my radiator now. My CPU temps are about the same and my GPU temps seem to have gone up by 1-2C for whatever reason. I have to say though, when I hold my hand over the CPU radiator exhaust at the top of the case, the Noctua don't seem to be pushing nearly as much air as the stock Corsair fans did. However, the Noctua fans sure are WAY quieter than the stock Corsair fans!

Are the fans controlled, or at full speed? If controlled, temporarily override to max if possible, and see what happens with temps and noise.

The fans are being controlled through Corsair Link and the software is reporting that both fans are spinning at around 1425 RPM. I cannot seem to get the fans to spin any faster than that, but it's close enough to max speed anyway (max speed being 1500 RPM). One thing is for sure, these fans sure are a heck of a lot quieter than the stock H100i fans! It still doesn't feel like they are pushing as much air out of the top of the case as the stock H100i fans though. However, I have noticed an overall drop in CPU temps of about 2-3C in CPUID HWMonitor. My GPU has also dropped down to around 63C now after finishing the WU it was working on and starting a new one. Perhaps the last couple WU my GPU was processing were extra heavy or something. I also played around with the fan curves of my GPU in Precision X so that the fans spin at 10% higher than the temperature of the GPU. So if the temp is 70C the fans will spin at 80%. That seemed to help a bit too. Right now my PC is quite noticeably quieter and a few degrees cooler as of this moment.

I have also been monitoring my points throughout the day as my CPU and GPU finished their WUs. It seems like the total PPD today was about 68,000 for my CPU and GPU combined. I am looking at the total points being reported on my contributions page after the completed WUs for the day have shown up there. Not sure if this is normal given my hardware and it does not seem like my GPU has gotten any "bonus" points today. The EP that the control panel showed I may get ended up being pretty much dead on what I got on my contributions page.

So now I have these stock H100i fans that I have nothing to do with. I don't want to put them anywhere else in my case really since they're so darn loud.

The appearance of less airflow may be due to the lesser turbulence caused by the NF-F12 fans. The stator vanes will smooth out the airflow before it gets to the radiator, thus increasing efficiency. Smooth flow will feel different than turbulent flow...

Remember that component temps will also depend on room air temp. The best easy measure of overall cooling performance is the temperature differential between case intake air and the component temp. I have also seen significant changes in component temps due to changes in F@H WUs.

For the old fans, attach them to the LNA cable from the Noctua package, and see if the noise becomes tolerable. If so, you might be able to use them in a future build. Otherwise give them to someone who needs them...

The appearance of less airflow may be due to the lesser turbulence caused by the NF-F12 fans. The stator vanes will smooth out the airflow before it gets to the radiator, thus increasing efficiency. Smooth flow will feel different than turbulent flow...

Ah... Makes sense. I do feel quite a bit more heat coming out of the top of the case even though the airflow doesn't feel quite as strong. My CPU temps are about the same as they were with stock and these new fans are WAY quieter than the stock, so either way I'm happy. Also, I cannot see the Noctua fans through the window of the case when I'm standing up or even when I'm sitting down in my chair, so they don't even ruin the aesthetics of the interior of my case at all! It's a win-win!

Remember that component temps will also depend on room air temp. The best easy measure of overall cooling performance is the temperature differential between case intake air and the component temp. I have also seen significant changes in component temps due to changes in F@H WUs.

My room temperatures are typically the same on any given day and throughout the day. I don't open the windows much because on nice days, I'm typically not inside until later in the evening. That, and my PC sits only three feet away from my apartment's thermostat, so my home controls the air based on how hot my PC is, haha!

For the old fans, attach them to the LNA cable from the Noctua package, and see if the noise becomes tolerable. If so, you might be able to use them in a future build. Otherwise give them to someone who needs them...

Well, I have tried playing with these old fans a bit now. They are much too loud without the LNA cable and don't push nearly enough air when I do use the LNA cable... I guess these stock H100i fans are just made of suck which is kind of disappointing seeing as how "premium" the H100i is supposed to be. You would think they would have included something like their SP120 fan series instead.

I have been monitoring my points throughout the day today! It seems like I got 101,000 points today! So right now I am getting 68-101K PPD it seems. Hopefully I will get to see more in the 100K range. I got my GTX 750 Ti in the mail today and I'm about to install that now! I'm excited!

Well, I have my GTX 750 Ti installed now and it's cranking through it's first WU as I type this. It's actually working through it's first WU much faster than I had expected. Not sure if it's just an easy WU or if the 750 Ti really is that efficient. Anyway, just to make sure I have my software setup correctly, here are the new slot settings I'm using:

0 CPU
Number of CPU threads = 6 (I left 2 threads for the GPUs to use, but I wasn't sure if that is necessary or not)

1 GPU
GPU Index = -1 (I keep setting this to 0 but the software keeps changing it back to -1 for some reason)

2 GPU
GPU Index = 1

I like the Folding@Home Web Control panel as it is easy to leave open on the screen so I'm able to walk by and checkup on my folding every now and then. Right now it seems like I have the slots setup correctly as on the FAH Web Control panel it's showing three slots running with the labels - "CPU:6", "GPU:0:GK104 [GEFORCE GTX 770]", and "GPU:1:GM107 [GEFORCE GTX 750 TI]". However, I wasn't sure if I needed to leave two CPU threads free (one for each GPU to use) or if I could leave only one CPU thread free for both GPUs to take advantage of. As it is right now, moving the CPU from 7 to 6 has impacted the CPU folding performance. Would you all recommend I leave my settings as they are or make some tweaks?

CPU:6 is good.

-1 (autodetect) defaults to gpu-index=0, so the client isn't saving the default value to config.xml. However, it is possible that the client got the two GPU slots mixed up. Doesn't matter on a fully dedicated setup since all Fermi+ NVidia FahCores fold the same way, but since the 750Ti is the 24/7 folding GPU and the 770 is running the display, you might want to (auto)pause folding on the 770 when you're using your computer. Pause either GPU slot and check GPU utilization to find out which GPU is actually doing the work, then set gpu-index=1 accordingly.

AFAIK, 750Ti is the current Points/Day/Watt champion by a comfortable margin. As far as pure PPD is considered, the 770 should definitely yield more points, but IMHO 750Ti is a great low wattage folding GPU.

PS: FahControl allows you to pause individual folding slots.

Napoleon wrote:PS: FahControl allows you to pause individual folding slots.

To do that, you have to right-click the slot, itself. That may not be obvious.

Each GPU can certainly use a whole CPU core, but most of the time it's in a wait loop until the GPU needs a data transfer. So it's a matter of experiment to see which way gives more PPD overall -- CPU:6 or CPU:7. There may be more projects available for CPU:6, since 7 can give problems for some projects. If the CPU is hyperthreaded, it only has four actual FPUs to do the work, so there may be little difference in PPD anyway. Whether using 7 for CPU folding slows the GPUs more than using 6 slows the CPU has not been widely reported here, so it's an interesting experiment. If you want to do the comparison, remember to run each configuration for several days to even out the different projects.

Of course, running CPU:7 and trying to do other stuff as well will cause more delay to GPUs than on a dedicated rig doing nothing but folding, so there's another variable to factor in (or out!).

Each GPU can certainly use a whole CPU core, but most of the time it's in a wait loop until the GPU needs a data transfer. So it's a matter of experiment to see which way gives more PPD overall -- CPU:6 or CPU:7. There may be more projects available for CPU:6, since 7 can give problems for some projects. If the CPU is hyperthreaded, it only has four actual FPUs to do the work, so there may be little difference in PPD anyway. Whether using 7 for CPU folding slows the GPUs more than using 6 slows the CPU has not been widely reported here, so it's an interesting experiment. If you want to do the comparison, remember to run each configuration for several days to even out the different projects.

A whole CPU core, or a whole thread? I'm running an i7 4770K which has 4 physical cores and 8 threads. So if I'm running CPU:6 then I'm essentially only leaving one full core for both GPUs then, aren't I? Since each core counts as two threads? In that case, should I only run CPU:4 to let my GPUs have more power since those are the components that earn the most points / do the most folding? I know one thing for sure, there is a slightly noticeable difference between CPU:7 and CPU:6 after adding the GTX 750 Ti. The ETA is about a hour to a hour and a half longer for the CPU on average per WU now.I also noticed that my GTX 770 is working on a new WU that seems extremely abnormally heavy as it's ETA is 1.08 days... I have never seen an ETA that high on any WU on any of my slots before... The highest I've seen before now was like 11 hours. Is there that much of a difference between different WU?

-1 (autodetect) defaults to gpu-index=0, so the client isn't saving the default value to config.xml.

Ah, okay, that explains why the software kept changing it back then. It seems like the slots are defaulting to the correct GPUs though, so no worries then!

AFAIK, 750Ti is the current Points/Day/Watt champion by a comfortable margin. As far as pure PPD is considered, the 770 should definitely yield more points, but IMHO 750Ti is a great low wattage folding GPU.

Well, you and everyone else has been telling me the GTX 750 Ti was a great choice for a budget folding card between $100 and $200 which would also not overpower my PSU. I'm really glad I bought it! It's been blowing through those WU since I hooked it up late last night!

Yes, there is quite a range of WUs for both CPUs and GPUs. Don't worry too much about minute-to-minute or hour-to-hour performance. In fact, my 2 [very low power] rigs together vary from 30K to 95K PPD, but are consistent week-to-week or month-to-month.

With the 4770K, you should be able to give 1 thread to each GPU, and use CPU:6 for the CPU slot. The GPUs require CPU access every X seconds, depending on the WU. The SMP WUs just take too long (relatively) to relinquish the CPU time if they control all the cores & threads.

Thanks for that clarification, jrweiss. It just surprised me when I saw an ETA of 1.08 days on my GTX 770 slot. Like I said, I had not seen an ETA on any of my WU over 11 hours until then and that was on my CPU slot which is the least powerful slot. So I didn't expect to see such a high ETA on my most powerful slot, haha! Anyway, that ETA has to be way off since the WU is already 25% complete and it's only been two hours. There is no way it is going to take another 21 hours to get through the other 75% at this rate! Also, something kind of peculiar happened just now. at 7:30PM, one hour after my last WU completed, I was at 469,000 points but then when I looked a few minutes ago, it showed 472,000 points. So it just jumped about 3,000 points out of the blue when no other WU have been completed and I had refreshed a few times since 7:30 and it showed no change... Is that one of those bonuses? Do they take several hours to show up sometimes?

Stanford's stat server gets updated on every hour, 3rd party stats like EOC and Kakao are updated every 3 hours, so there is a little bit of lag with the crediting. In the past, the Stanford stats server has been down occasionally, so sometimes the stats update can be further delayed. Not to worry though, it'll catch up with the backlog later on, and you may see a spike in points when the backlog gets credited all at once.

The client's PPD/PPWU prediction is just that - a prediction based on the performance during folding of the WU. Sometimes it takes a few percents of progress until the estimate stabilizes. Particularly so if you haven't folded a WU from the specific project before, and the client doesn't have any performance history to rely on.

Stanford's stat server gets updated on every hour, 3rd party stats like EOC and Kakao are updated every 3 hours, so there is a little bit of lag with the crediting.

Ah, I see. That explains where this other 4,000 points came from which I just got randomly again! So, it would seem that yesterday, I got 114K points for that day! That's exciting!

A question I have now - is there a way, other than choosing which disease you'd like to focus on, to actually choose which WUs you would rather work on? This is more of a curious question. Also, I've been looking into getting a NZXT Kraken G10 with a NZXT Kraken X41 to cool my GTX 770. This would have to be further on down the road though since that would cost me another $140. From what I'm seeing in people's YouTube reviews and other reviews from around Google, this solution would let my GTX 770 run much cooler than the stock ACX cooler.

I've also been developing a slight fear that running my components at full load 24/7 will eventually lead to something breaking faster than it would have otherwise. Will running everything full load 24/7 significantly decrease the life of any of my hardware?