Scratch Build – In Progress Project RabidCicada (Mineral Oil Computer) - 20th October 2010

So another Idea popped in my head. Maybe I'm overtightening the freezer13 and it's causing seating problems with the processor?

Is it easy to do that and cause proc problems?

 

90 degrees plus?! Centigrade?! Bulldozer's Thermal Cutout point is 75! No wonder things aren't working if it can't even reach the point where it'll start conducting heat properly!

 

So...Where are people getting this information. I like playing around and custom building but I still lack a lot of this knowledge.

Where do you get the thermal cutout point?

Where is a datasheet or something for the FX-8150?

 

Experience. Highest I've had this chip without cutout is 74 degrees. I'm not too happy about that fact either.

It's similar to the Phenom II architecture in that respect. Maximum safe temperature stated by AMD is 62 degrees, for both the Phenom II and the FX series.

 

Thanks much for the info Tundra. I guess I got misled because I see other people talking about having IX on 8150. Wonder how they did it.

New plan:


Gonna try using my trusty new Heatgun and IR thermometer to heatset the IX

Hope it works. If not then I'll get to using ceramique.

 

Drastic, but shouldn't, in theory, damage the chip, so long as you don't exceed 100+ degrees for too long.

 

Thats what I'm thinking. Tired of this stuff not working.

Gave a test run already on two heatsinks pressed against each other. Learned that an IR thermometer suffers from bad glancing readings on a reflective surface. Point as orthogonally as you can at the metal surface. On non-reflective surfaces you should be alright.

I thought I was having trouble getting the test Heatsink over 50c then pointed directly 90degrees at the heatsink and was at 132c

Just gotta hope not to bake the motherboard and have my external fan controller ready to spin up the HeatsinkFan when I know the IX has melted.

 

I'll probably give the real attempt a shot later today or tomorrow.

 

Ouch. Just be careful not to subject the chip to too much heat for too long. It should bear it better while it isn't running, but I'd still be cautious.

 

Well...I just got an unexpected response from the IX guys. Ill be holding off until early next week. Chris from IX. Said that hell talk to a guy whose done bulldozer + IX and get back to me with some info. Made my weekend.

 

Hey,
Did you try keeping a fan plugged into the CPU fan header on the mobo? Might not boot without a signal.

 

Yeah...I tried faking out the mobo with the stock HSF fan plugged in (While I had the freezer13 setup). Didn't work.

 

Almost final HW in week long Test Run

Alrighty. Been a long time.



Here's where things are:
Total setup
I've got almost final HW in the case and it's setup running Folding@Home on GPU and CPU for a heat test and week long stability test run. My goal is to find the temperature equilibrium point with 1 8800gtx gfx card and the FX-8150 cpu running full bore.

I bought a prolimatech megahalem heatsink but couldn't use it because it was too large to fit in the case and allow the motherboard tray to slide in and out like it was designed to. I acquired a special edition freezer13 from arctic cooling and I'm running with arctic ceramique 2. From talks with Legoman666 I will likely have to re-apply every couple of months but that doesn't bother me one bit.

Indigo Xtreme -- I didn't use it -- Good product though
I had originally planned on using Indigo Xtreme after talking with Legoman666 and doing plenty of research elsewhere. I ended up having too much trouble. I burned through 4 installs (2 purchases worth) trying to get it working. It looks like the Bulldozer/FX architecture just wont get high enough to melt the IX. In fact, it wont even post BIOS without a properly applied heatsink. I don't know if it's a safety shut-off or it simply conks out because of heat but it just wont get hot enough. Another issue I would have run into is that the Freezer 13 is not a full coverage heatsink. The metal base that contacts the processor does not do so on ALL 4 SIDES and you need the pressure applied on all sides to properly distribute the molten IX. So after a lot of experimentation, and some nice emails from the IX guys (Thanks Chris), it just looks like it's not achievable. I originally thought it might be the sabertooth 990fx board being protective of the processor, and I haven't completely ruled that out...but I think it's the processor itself (with supporting comments from others and talking with IX guys).

On a side note I did a test with the heatgun and chose not to burn my cpu to death. Maybe I'll try another time. I put two heatsinks together witht he IX between and used the heatgun and my Infrared thermometer. I actually got the heatsink so hot that I unsoldered the heatpipes from the heatsinkbase...And that's some high temp solder.

Heatsink Itself
I had a heck of a time finding a good heatsink. I wanted the prolimatech megahalem and I wanted downflow (blowing towards the bottom of the case) in the case from the cpu fan but I couldn't find a larger heatsink that: fit, used a 92mmfan (orignally wanted 120mm), and was configurable to blow down the case. I finally settled for the special edition freezer13 (nickel plated) which is a respectable heatsink and used it to blow oil towards the power supply (crossflow).

Forward plan
Gonna run it about a week in it's current state with Folding@home 24/7. If the gfx card lives then I will probably switch over to the mineral oil computer as my main rig and put in my two gtx580's! I also need to polish the layout and cabling etc. Looking forward to it.

 

Lookin' awesome. Good work. Now come fix mine!

 

looking good rabid, needs some case lights though!

 

Equilibrium temperature after 5 hrs had processor at 50c and gfx card at 55c. The fan controller heatsinks actually got VERY hot. If you touched the top of the lid directly above the fan controller it was hotter than I think is healthy. I think I'll probably add one air fan to push air through the lid, add some breathing holes near the fan controller face plate.

The cpu and gfx temps are acceptable and would run fine but not exactly what I wanted.

I still have a little temperature stratification in the oil. The top is noticably warmer but not near as bad as last test run without fans. I should engineer oil flow better on next pass. I probably should draw top (hot) oil into the pump for better circulation and work on a better directed flow.

What I find surprising is that the temp of the cpu was higher than I expected but I can rationalize it by knowing that all other components are dumping heat into the oil and I'm not exchanging it at a high enough rate. I could add more silent fans to the radiator, capable of 9 but only populated by 3 so far. Fix top oil intake for the pump, and up the flow rate by upgrading pump.

 

Moar fan and a slightly higher flow-rate would be easy enough, I know you can probably hit 5.2+ on that Bulldozer if you can keep it cool enough. Under liquid and more exotic cooling; I've seen it done.

 

Got familly over now so I don't have much time...but...I was right about the fan controller. Melted the fan-controller-side connection from power supply to the fan controller. I guess when I shut it off heat peaked real fast and the connector melted...lol.

Luckily everything else is still good. Moar fan indeed. And I'm lucky enough to have just bumped into a man who deals with hydraulics and pump stuff...so I may just end up with a new pump approach here within a week or two.

 

are the fans that are submerged in oil connected to the fan controller? or is it just for your radiator fans?

if the submerged fans are connected to it than its no wonder it melted. id run those directly from 12v since you wont be slowing them down anyways.

if thats the way you already have it wired then disregard my crazy talk

 

The fans in the mineral oil are being run through the fan controller right now. I thought that would be ok because it's rated for 30w each channel and I made sure the fans drew that or under at 12v. The afb1212ghe's can draw more than 30w when run above 12v.

I guess I should have just run them directly from the power supply...

I thought maybe I would have to tune the but running speed, but full power seems like the way it's gonna stay. Thanks for the recommendation Wicked. I'll switch over to that.

I think the fan controller itself is still operable...I just think the connector came "unconnected" internally where it melted. If not I'll order another sunbeam rheobus.

Much appreciate on the design thoughts.