I, like many here, bought a CM Gemini II when they were basically giving the buggers away after rebate.

My first tests, on a S939 4600+, were disappointing. It was either loud or underwhelming.

I tried it again today on an E6750, but with just one fan: A Silverstone FN121 which is nearly silent at full-blast and really silent when moderated. I mounted it on the "CPU side" of the cooler (the one with the base), as opposed to the position which hangs over the RAM.

The temps (on a one-hour OCCT run) peaked at 57/55. I gave up with the stock cooler when it hit 61 on one core after like five minutes.


The problem I could see is twofold:

1. One fan leaves about one-third of the heatsink (about 2 heatpipes) unfanned. That could lead to uneven cooling. Indeed, there's a 4-6 degree spread between cores at almost all times; it used to be as little as one or two degrees. I can generally not heat Core 1 over Core 0; probably could if I found a workload that redlined Core 1 only. The idle temperatures are not wow-inducing-- 40/35 (although ambient is 28)

2. The GeminiII design is supposed to blow across other parts of the mainboard. I'm worried that if there isn't that flow, it will actually trap the heat that 6Gb of DDR2-800 is belching out (which is nontrivial )

Should I be worried? I've just set the stock cooler aside for the time being. I don't really notice the difference in sound or performance, but there's something comforting in knowing that my CPU fan can actually go to zero on a cold day. (My PC at work uses a 65W 4200+, and the fan rarely spins up, period!)

Update: Now with pic attached. You can see what's covered.

I also discovered that the wussy FN121 frequently simply doesn't spin up at the low voltage-regulated characteristics the board is throwing it. A little tap gets it spinning. While I wouldn't mind a fanless setup, this ain't how I intend to get it!

I threw a spare Power Logic PL12B12H (83CFM, ~2500RPM) on, and while it spins up every time, it's a much heavier-duty fan. Its "slowed to minimum" setting is 1200RPM-- the same as the high setting of the FN121-- and at full-blast (when you force it on), it's impossible to ignore.

I was considering getting a 4-pin (PWM-controllable fan) to alleviate this problem.

The Scythe Kama PWM ( Scythe Kama PWM 120mm Fan ) looks very comparable, specs-wise, to the FN121 (although negligibly lower in capacity).

Arctic Cooling also makes a similarly-specced fan (AF12025PWM), which seems to benefit from a notch higher rating, a very fancy-sounding bearing technology, and a much longer expected-time-to-failure, but I'm sort of not liking that "extra head in case you want to extend the PWM signal to another fan" cable, as it's just another cable to manage.

Any other thoughts? Bearing in mind I'd rather keep this adventure under $15, and better yet, under 10 (most of it, of course, will be postage :/ )

Any other suggestions? Should I just hope that the machine stays cool enough that the Power Logic fan never spins up (from what I can tell, the smart-fan feature is supposed to hit 100% duty cycle at 65C, but they don't say on which sensor)

It seems like all the options sort of suck:

The FN121 can't be regulated. (it also looks like crap because of my messy conversion to 3-pin plug)

The Power Logic spins up from a quiet 1350RPM to a noticable 1900 under load, and isn't much cooler than the FN121 at 900-1200 RPMwas in exchange for that noise (like 2-3C)

A new fan would cost me ~15.- including shipping.

I still don't know if I'm making a horrible hot pocket for my precious RAM.

Going back to stock, with its PWM-able but quiet (despite a 0.60A rating!) fan, and "officially adequate" heatsink, looks more and more compelling, especially if it means less wrestling around the GeminiII itself inside the case.

Attached Thumbnails

 

I have played with 3 different Gemini II's, and my experience has been consistent with all three. They all seem to not be perfectly flat.

The first time I installed one, my temps were high so I pulled it back off (what a pain!) and only about half of the CPU was even touching it.

I lapped it, and checked it with a square to see if it made a difference, and it actually helped quite a bit. I am running it right now on an x2 4200+@ 2.7, and load temps are 52-53.

The only thing I could think of is this is why they were so cheap for a while.

Oh and by the way, my fan is on the CPU side as well.

Quote: Originally Posted by Hak Foo The FN121 can't be regulated. (it also looks like crap because of my messy conversion to 3-pin plug)

(almost) Any fan can be regulated -- lower volts should result in slower blade spin. Have you tried simply giving it less power? Unless something has been built into the fan that waits until it has full voltage to turn on (unlikely), you should be able to cycle down the voltage little by little to achieve various speeds. You should also be able to cycle up the voltage, but this is like OC's your cpu, only with fewer checks and ballances...

Quote: I still don't know if I'm making a horrible hot pocket for my precious RAM.

Probably not. The fan itself is still moving air that will spread all around your board for a short distance, and that should help move air around your RAM as well. Not as effective as direct flow, but there should be not air pocket development.

Quote: Going back to stock, with its PWM-able but quiet (despite a 0.60A rating!) fan, and "officially adequate" heatsink, looks more and more compelling, especially if it means less wrestling around the GeminiII itself inside the case.

Snowman's hit the mark on that one, I think. Lap the sink -- you should see a remarkable difference.

-godling

Quote: Originally Posted by godling (almost) Any fan can be regulated -- lower volts should result in slower blade spin. Have you tried simply giving it less power? Unless something has been built into the fan that waits until it has full voltage to turn on (unlikely)

I meant, "it can't be regulated in a functional sense". The way my board wants to regulate the fan involves giving it too little voltage to spin up reliably from a cold start. Maybe if I waited long enough, it would heat to the level where it gave it 100% voltage and THEN it would spin, but that sort of "wait until it hits 65C" mentality seems a poor approach.

Quote: Probably not. The fan itself is still moving air that will spread all around your board for a short distance, and that should help move air around your RAM as well. Not as effective as direct flow, but there should be not air pocket development.

Okay, good.

Quote: Snowman's hit the mark on that one, I think. Lap the sink -- you should see a remarkable difference.

All I have is like 150 and 220 grit and this weird plasticy stuff designed for model use. I suppose I can spend some time lapping it as best as possible, but I don't expect much.

Edit. Update:

I don't even have the plasticky stuff anymore, so I tried a Half-Assed Lap Job (tm) with 220 grit paper wrapped around an old one-by-two.. In an attempt to clean it up after that, I tried polishing it with toothpaste, but Tartar Control Colgate lacks a significant abrasive factor.

With the Power Logic fan, case closed, it peaks at 55/52 in an OCCT run,versus 57/55 in a previous test case (which, however, might have been with the Silverstone fan when it actually turned-- I should have taken better notes). The "imprint" of the core in the thermal grease is much closer to the one you see on the stock sink now-- more of it was purged than before.

I sent out for an Arctic Cooling AF12025PWM so I can regulate it properly.