Folding on CPU decreases choke whine from GPU?
The title says it all:
I got a GTX 550Ti for my "My System Specs" rig after discussing in this thread. The card has a very quiet fan - comparable to case fans - but it emits a high-pitched whining sound while folding, probably from chokes.
I found something odd while measuring the new F@H PPD / Watt of my main rig which has said GTX 550 instealled. If I fold SMP on my CPU + GPU3 on my GPU, the whining sound is much quieter than when I'm folding GPU3 only. If I fold GPU3 only, the whining sound is very noticeable, and annoying because it changes frequency often. However, If I fold with both CPU and GPU, the whining sound decreases to the point where I barely notice it. The SMP doesn't seem to be starving the GPU client, as the PPD reported from HFM.NET is about the same as folding GPU only, and just to make sure I raised the priority of GPU FahCore_15 to "below normal" while the SMP is at its default "idle" priority.
Can you offer any explanation to this black magic? If I fold GPU only, I'd have to underclock the card to somewhere around 825-850 core to quiet down those chokes to the same point as if I were folding with CPU + the GPU at its normal OC of 970 core. (memory clock doesn't affect the whining, and shader clock is locked to Core x 2)
Essentially my rig now gets QUIETER if I fold more on it! ...how does loading the CPU make my GPU quieter?
different loads on the psu? :)
play with the shader/clock speeds on the 550 to lessen the whine :)
i had the gtx 550 Ti and did not notice any whine at all and I was folding only on GPU, i had the MSI cyclone the one that has also been reviewed here. Must be something wrong with the card, the power distribution system is not optimized for high loads for long periods of time.
Are you sure the whining isn't coming from your PSU/motherboard? It could be a C-state/idle incompatibility issue with your PSU, very common between high quality motherboards and PSU's. Or your PSU might not like being fully loaded on one rail while being idle on the other. (not sure if the M12 600W is a true dual rail or a single rail Seasonic says dual, but it's not true half the time)
Thanks for the replies. I may try playing with this on another rig to test out those PSU and / or motherboard theories. I'm a bit wary of plugging this brand new but not exactly shiny card into a CoolerMaster ElitePower 400W though (the only other PSU I can play with at the moment).
Oddly enough the card doesn't whine at all with OCCT GPU test going, despite the card reaching higher temperatures than with folding. For games I haven't tried as much because my closed-back headphones almost completely drown out the whine (even with no sound playing through them). I'll see if maxing out the card's OC has any effect on the card.
Anyway I won't return the card over this because it's still an improvement over my Radeon 4870 which sent its electrical noise through the headphones :haha: At least this card doesn't do that... Now if the constant F@H while OC'd to 1GHz core happens to kill the card over the next week or so that's a different story :bananafunky:
Heh, so that CoolerMaster may not be a kills-components unit, but it surely is an inefficient one - compared to my Seasonic I'm seeing 10-20W more out of the wall to feed it.
Anyway as noted above, I tried the CoolerMaster PSU. It was hard to listen with the Seasonic because it was in my case, but I have my CoolerMaster hooked up outside the case. When folding on the CoolerMaster, I hear the whine from the GPU and the PSU. It's in fact the same whine at least subjectively - I don't exactly have the resources of my school's research lab here. Also, with the PSU whining to the same tone as the GPU, I can hear it while the computer is idle. If I fold GPU + CPU on the CoolerMaster, the whining sound stays at about the same volume but changes tone ever so slightly, compared to the sound getting softer on the SeaSonic. Folding with CPU only doesn't result in any whine - this is the same for both CoolerMaster and Seasonic.
EDIT: I tried again with the Seasonic by listening near the exhaust of the PSU. Here's my result:
Hmm Gigabyte tech support is willing to consider the coil whine a defect and issue RMA if I test it in another computer (which I just did, by plugging it into what used to be my main rig). That *should* mean Canada Computers will give me a refund for it if I send in the transcript of my communication with them.
Any suggestions for a replacement which has coils that won't protest my use of the card in F@H?
Yeah, I guess that's the issue when there's no full line of "guaranteed no whine" parts, and any lemon in the line can cause it. Yes MSI has that little "military grade" gimmick but nothing says army-grade VRM circuits aren't allowed to hum. I'd do the conformal coating trick on my GPU and PSU if the chokes weren't the kind with the metallic cover.
I may have a lot more fun seeing if I can build an inline filter to place between the PSU and GPU (by hacking a PCI-E adaptor) to silence the thing. Both will cost time and effort / money, except the time spent trying multiple RMAs just builds frustration. At least making a Mythbuster-like contraption is fun. (I apologize if I'm getting people to run away or pull out electric circuit theory / applications books right now.)
Band-stop filter - Wikipedia, the free encyclopedia
Low-pass filter - Wikipedia, the free encyclopedia
I assume what I'm trying to stop is some ripple occurring at 15-22 kHz (given the audible pitch of the noise)? My key that a buffer will help is that the ripple is sent between the PSU and GPU.
|All times are GMT -7. The time now is 01:20 PM.|