Dominator Pro 437 Up & Running

[DNA]

System Configuration:
Lian Li PC75B Case
(2) Rear Fan, Panaflo 80MM, 40CFM – Intake
Rear Fan, Panaflo 120MM, 115CFM – Intake
Floor Fan, Panaflo 120MM, 115CFM – Intake
(2) Front Fan, Panaflo 80MM, 40CFM – Intake
AMD Athlon 64 3200 Venice, LBBLE 0516 GPDW, 295 x 10Mhz
DFI nF4 Ultra-D Motherboard
(2) Crucial Ballistix PC3200 256MB, 2-2-2-6, 1T, 2.7V
(2) Hitachi SATA II 80G (RAID 0)
Hitachi SATA II 80G (Backup)
Abit RX300SE Video Card
Lite-On 1673 DVDRW
LS 120 Floppy Disk Drive
Antec Neopower 480W Power Supply
Dominator Pro 437W TEC CPU Block
Mean Well S-600-24 Power Supply – 5 1/4-in. Bay-Mount
Lian Li Expansion Drive Bay Grill – Mean Well Exhaust Vent
Swiftech Pump Relay, Version II – Mean Well AC Power Control
(2) Airplex Evo 240 Fan Grill
(2) Black Ice Extreme II Radiator – Internally Top-Mounted
(4) Sanyo Denki 120MM,102 CFM Fan – Exhaust
Mean Well SD-50A-12 DC-DC Converter
(2) AquaXtreme 50Z-DC12 Pump @ 14.0V
Dual 3 ½-in. Bay Reservoir
Masterkleer General Purpose Clear PVC Tubing 7/16-in. ID, 3/16-in Wall
Crystalfontz Model 634 4-Line LCD
Digital LED Volt & Amp Meter w/Shunt on TEC Power Supply Output
Coolant Temperature Metal Probe/LCD – (+/- 1C Accuracy)
* Cold Plate Temperature Metal Probe/LCD – (+/- 1C Accuracy)
* Probe embedded in cold plate. Probe tip located over CPU center area.

Conditions:
- All case panels installed.
- All (10) fans running at 12V.
- TEC PS Output = 19.9V @ 24.6A
- CPU Core Voltage (V) = 1.54 – as reported by CPU-Z via BIOS
- CPU Core Power (W) = 126 - estimated
- Pump Heat Dump (W) = 23 - estimated
- Room Temp (C) = 20.3
- Load Time = 60 Min.
- CPU loaded using Prime 95 with large FFT setting, maximum heat.
- All readings taken at end of noted time period.

Results:
- Coolant Temp (C) = 30.4
- Cold Plate Temp (C) = -14.6
- CPU Temp (C) = 5 - as reported by Motherboard Monitor via BIOS

 

[DFI Daishi]

what are youur coolant temps like, with that monster pelt and only a BIX2?

with two panaflo 120mm U1A 115 CFM fans cooling my BIX2, and my CPU, GPU, 226 watt pelt and my 80 watt pelt, i have to run my fans at fulll to keep my coolant temps much below 40C.

i was really thinking that to run one of those dominator pro blocks, you'd need to use a 3 fan rad to keep the fan noise sane with reasonable coolant temps, and based on the fans you are using, i think that i was more-or-less correct.

 

[gclg2000]

Let's see some pics.

 

[Jesus_Faction]

pics now kthxbye

 

[DNA]

what are youur coolant temps like, with that monster pelt and only a BIX2?

Not setup to montor water temp yet. Running (2) cooling loops thru DP. Each loop has a BIX II and 50Z pump. Hotside temps are in the high 20s to low 30s range.

 

[zeebs]

picsplskthxgetemupkthxplsyeahkbye

 

[DFI Daishi]

Not setup to montor water temp yet. Running (2) cooling loops thru DP. Each loop has a BIX II and 50Z pump. Hotside temps are in the high 20s to low 30s range.

ah. i should have read your setup specs more carefully. nice hotside temps, though. with hotside temps that low, i'm guessing that you can keep that processor in the range of -20C OCed and under load.

 

[DNA]

Let's see some pics.

Next on my things to get list - digital camera

 

[zeebs]

Awesome I can't wait to see it.

 

[DNA]

ah. i should have read your setup specs more carefully. nice hotside temps, though. with hotside temps that low, i'm guessing that you can keep that processor in the range of -20C OCed and under load.

I hope your right.

 

[DFI Daishi]

I hope your right.

well, i mean weird things can happen, however i am basing that on the performance curves for the style of pelt that you are using.

 

[DNA]

]

 

[DFI Daishi]

looking good.

any further performance updates?

have you gotten some way of accurately measuring CPU temps as of yet? i believe that you previously mentioned you compunurse probes failing to report below 0C.

 

[DNA]

looking good.

 

[DFI Daishi]

please do note that installation of a thermal probe via your current slution or via AMD's reccomendations will result in reduced cooling performance.

i have a bare core, and i attach my thermal probe to the side of the core itself.

another common alternative is to attach the thermal probe to the back of the chip package.

your coolant temps are just beautiful. you're paying for it in noise level, but MAN are those nice temps.

 

[DNA]

i have a bare core, and i attach my thermal probe to the side of the core itself.

another common alternative is to attach the thermal probe to the back of the chip package.

Are either of these techniques applicable to the AMD 64 core?

Refer to Pg. 37 of the AMD Thermal Design Guide, http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/26633_5649.pdf
Rather than cut a channel in the coldplate for the thermocouple wire, I would drill a hole from the coldplate perimeter to its centerpoint. A second hole would be drilled at the coldplate's centerpoint so as to intersect the first hole. Both hole sizes would kept to an absolute minimum so as to just accommodate the thermocouple wire routing in one axis and the CPU top case contact on the other axis. Do you believe this approach would degrade the TEC's cooling performance?

 

[DFI Daishi]

Are either of these techniques applicable to the AMD 64 core?

the first is only applicable if you remove the IHS. given how heavy your TEC is, that is most likely unwise.

the second is most likely feasible. i don't have an a64 on hand to fiddle with, however if suspect that if you seperate the leads coming off of a thermal probe of the sort that a compunurse uses, you could thread each individual wire through the maze of pins, and still be able to seat the chip into the socket. if that is the case, then a little thermal epoxy to hold the probe to the back of the package is all that you need.

 

[DNA]

the first is only applicable if you remove the IHS. given how heavy your TEC is, that is most likely unwise.

the second is most likely feasible. i don't have an a64 on hand to fiddle with, however if suspect that if you seperate the leads coming off of a thermal probe of the sort that a compunurse uses, you could thread each individual wire through the maze of pins, and still be able to seat the chip into the socket. if that is the case, then a little thermal epoxy to hold the probe to the back of the package is all that you need.

What your suggesting may be doable on a Socket A CPU, but don't think it is on a Socket 939. Have you ever seen the pin size/shape and spacing on the A64?

 

[DFI Daishi]

What your suggesting may be doable on a Socket A CPU, but don't think it is on a Socket 939. Have you ever seen the pin size/shape and spacing on the A64?

not first hand, no.

i can feed a twinned pair of wires in between the pins of my socket 462 processor, and i did not think it unreasonable that you would be able to feed two single wires through the pins of your socket 939 processor.

however, as i said at the start, i have not handled a s939 processor first hand, so i don't really know for sure.

as for your above comment about it being just a small hole to the center of the core, referring to the AMD guidelines: i have read that document as well (i like to know just how badly i am violating manufacturer's reccomendations , or planning to violate them in this case), however you should keep in mind that the vast majority of heat from the processor goes directly to the area immediately above the core, regardless of the effects of the heat spreader, and therefore anything that is between the pelt and the core in that region is definately impeding your cooling capacity somewhat. when you're working with less than a square centimeter of area to start wtih, even a 6 square mm bite out of it out of it represents a measurable percentage.

if you were even stopping just shy of the core, then the effects would be greatly reduced.

 

[DNA]

not first hand, no.

i can feed a twinned pair of wires in between the pins of my socket 462 processor, and i did not think it unreasonable that you would be able to feed two single wires through the pins of your socket 939 processor.

however, as i said at the start, i have not handled a s939 processor first hand, so i don't really know for sure.

as for your above comment about it being just a small hole to the center of the core, referring to the AMD guidelines: i have read that document as well (i like to know just how badly i am violating manufacturer's reccomendations , or planning to violate them in this case), however you should keep in mind that the vast majority of heat from the processor goes directly to the area immediately above the core, regardless of the effects of the heat spreader, and therefore anything that is between the pelt and the core in that region is definately impeding your cooling capacity somewhat. when you're working with less than a square centimeter of area to start wtih, even a 6 square mm bite out of it out of it represents a measurable percentage.

if you were even stopping just shy of the core, then the effects would be greatly reduced.

Without actually measuring it, my guess is that the A64 top case heatsink area is ~ 30mm square, much larger than that of the XP. The Type K thermocouple wire could easily be routed through a 1.5mm diameter hole. With a hole size no larger than 1.5mm and the contact end of the thermocouple measuring < 1mm in diameter, do you still believe it will impact cooling. The thermocouple contact end hole would be filled in with a highly thermal conductive epoxy, Omegabond 101, so as to just expose the sensing end of the wire.

 

[DFI Daishi]

i was not talking about the size of the lid/IHS/whatever you care to call it, but rather the size of the core itself, which is quite small. http://www.pcinpact.com/images/bd/news/10406.jpg

i'm saying that according to arctic silver, it's okay to just have the central area of the heatspreader covered in thermal compound, becausethe majority of heat follows the path of least thermal resistance straight from the core, to the base of the cold plate, withough actually spreading out much via the heat spreader.

if you increase the thermal resistance over the core by making a gap in the cold plate for the thermocouple, then heat will then take an alternate path using some of the rest of the area of the heat spreader, however the added thermal resistance of taking that longer path will definately increase the opperating temperature of the core to some degree, as compared to the situation where heat is able to follow a less resistive path.

you should still have quite good cooling, just not the best possible cooling. i feel that you are sacrificing a bit of performance to have highly accurate temperature monitoring. i don't have numbers to back all of this up, just conclusions drawn from AS' comments and observations. i guess that there is no way to know for sure other than to try both and see, however in my own case, i prefer to have better performance and approximate temperature monitoring than better monitoring and reduced performance.

 

[mikelz85]

what's the max OC you are getting with the huge pelt? Is that 12C a real load temp? Seems really high to me for a pelt that huge.

 

[DNA]

System Configuration:
Lian Li PC75B Case, Aluminum Wheel Set
(2) Rear Fan, Panaflo 80MM, 40CFM – Intake
Rear Fan, Panaflo 120MM, 115CFM – Intake
Floor Fan, Panaflo 120MM, 115CFM – Intake
(2) Front Fan, Panaflo 80MM, 40CFM – Intake
AMD Athlon 64 3200 Venice, LBBLE 0516 GPDW, 295 x 10Mhz
DFI nF4 Ultra-D Motherboard
(2) Crucial Ballistix PC3200 256MB, 2-2-2-6, 1T, 2.7V
(2) Hitachi SATA II 80G (RAID 0)
Hitachi SATA II 80G (Backup)
Abit RX300SE Video Card
Lite-On 1673 DVDRW
LS 120 Floppy Disk Drive
Antec Neopower 480W Power Supply
Dominator Pro 437W TEC CPU Block
Mean Well S-600-24 Power Supply – 5 1/4-in. Bay-Mount
Lian Li Expansion Drive Bay Grill – Mean Well Exhaust Vent
Swiftech Pump Relay, Version II – Mean Well AC Power Control
(2) Airplex Evo 240 Fan Grill
(2) Black Ice Extreme II Radiator – Internally Top-Mounted
(4) Sanyo Denki 120MM,102 CFM Fan – Exhaust
Mean Well SD-50A-12 DC-DC Converter
(2) AquaXtreme 50Z-DC12 Pump @ 14.0V
Dual 3 ½-in. Bay Reservoir
Masterkleer General Purpose Clear PVC Tubing 7/16-in. ID, 3/16-in Wall
Crystalfontz Model 634 4-Line LCD
Digital LED Volt & Amp Meter w/Shunt on TEC Power Supply Output
Coolant Temperature Metal Probe/LCD – (+/- 1C Accuracy)
* Cold Plate Temperature Metal Probe/LCD – (+/- 1C Accuracy)
* Probe embedded in cold plate. Probe tip located over CPU center area.

Conditions:
- All case panels installed.
- All (10) fans running at 12V.
- TEC PS Output = 19.9V @ 24.6A
- CPU Core Voltage (V) = 1.54 – as reported by CPU-Z via BIOS
- CPU Core Power (W) = 126 - estimated
- Pump Heat Dump (W) = 23 - estimated
- Room Temp (C) = 20.3
- Load Time = 60 Min.
- CPU loaded using Prime 95 with large FFT setting, maximum heat.
- All readings taken at end of noted time period.

Results:
- Coolant Temp (C) = 30.4
- Cold Plate Temp (C) = -14.6
- CPU Temp (C) = 5 - as reported by Motherboard Monitor via BIOS

 

[fore1337]

wow! 6.5C under load overclocked. very nice.

Do you think that meanwell has enough in it to also power a GPU TEC?

 

[DNA]

wow! 6.5C under load overclocked. very nice.

Do you think that meanwell has enough in it to also power a GPU TEC?

The Mean Well SE 600-24 Power Supply that I am using is rated at 600W. If you take my current CPU TEC power demand of (19.9V. x 24.4A.), that's ~ 486W. It leaves ~ 114W for possible future increases in power. You need to estimate your GPU TEC power consumption. Also, your GPU TEC must be rated to operate on the same voltage as your optimized CPU TEC operating voltage. That fact alone is pretty risky.

Without knowing your specific GPU TEC power demand, my guess is that this power supply probably will not meet your needs.

Are you quite sure that watercooling your GPU on a separate pump/radiator loop might not be a better solution? I have limited TEC experience but I do know if you can't dissipate the TEC hot-side heat load, your not going to be pleased with the CPU load temperature results regardless of your CPU TEC cooling capability. Hope that helps.

 

[fore1337]

ah. i should have read your setup specs more carefully. nice hotside temps, though. with hotside temps that low, i'm guessing that you can keep that processor in the range of -20C OCed and under load.

That's a big difference -20C (on paper) vs 6C (actual)
Any ideas on why this is?

Are you quite sure that watercooling your GPU on a separate pump/radiator loop might not be a better solution? I have limited TEC experience but I do know if you can't dissipate the TEC hot-side heat load, your not going to be pleased with the CPU load temperature results regardless of your CPU TEC cooling capability. Hope that helps.

That is definately food for thought. I've always been pushing my GPUs until they pixellate in games... but my core temps never breach 60C. I thought if I lower that temp down a bit I could squeeze more MHz out of em'

I'm just trying to get my head wrapped around the TEC solution.

Have you seen that wintschlabs is now working on 24V, 437W GPU blocks? Too bad they locked themselves into an NV-68 only solution.

If they come out with an ALL GPU block similar to the DangerDen Maze-4 I would probably pick one (or two) up. My real direction would be to put 2 of these guys together in an SLi or crossfire configuration... and try to get it to work with a single Meanwell 600w/24A PSU. Using a TEC & H20 loop independant of the CPU loop.

edit:
thanks so much for sharing your experience.

 

[DNA]

That's a big difference -20C (on paper) vs 6C (actual)
Any ideas on why this is?

Yes, I believe I need to make some changes in my coolant flow loops. With a coolant temperature of ~ 30C, the best that I can hope to achieve with my current TEC power setting and total heat load under max load conditions is ~ -4C, 30-[(1-150/334)*61]. I'm ~ 10C higher than that at present.

As for this -20C on paper, there are too many variables in that estimate. Without the massive overclock I'm running, I'd be down to <-10C.

 

[DNA]

Seems really high to me for a pelt that huge.

Based on what? Show me your math.

 

[DFI Daishi]

That's a big difference -20C (on paper) vs 6C (actual)
Any ideas on why this is?

well.......his cold plate temps are a lot closer to my -20 figure than his core temps are. based on my own experiences, i honestly expected his core temps to be closer to his cold plate temps than they are turning out to be.

his actual cold plate temp of -13.5C is not so far off of -20C, depending on the overclock that he ended up with. i honestly did not think that he would be able to scale up as close to 3.0 GHz as he has, given what people with phase change are hitting. all in all, this demonstrates how well the heat-hardening of the current line of processors has reduced the benifit of using extreme cooling on them.

having that cold plate temperature probe located where it is will have some effect on his overall cooling ability. if it wasn't for the weight of that cooling solution, i would also say that removing the IHS would be a good idea to try and get core temps closer to the cold plate temp. given the danger of a core-crush, though......

If they come out with an ALL GPU block similar to the DangerDen Maze-4 I would probably pick one (or two) up. My real direction would be to put 2 of these guys together in an SLi or crossfire configuration... and try to get it to work with a single Meanwell 600w/24A PSU. Using a TEC & H20 loop independant of the CPU loop.

well.....i would look hard at using a 24 volt 120 watt 40x40mm pelt in conjunction with a maze 4-1 or a silverprop cyclonefusion SLT block to do the job. thing is, i don't know that the maze 4 low profile comes in a pelt config.

 

[Zoran]

I honestly think the power supply won't be enough to power anything else... Not even the 437W moster... as it's running only @ 342W!!! that's probably the reason why the paper temps are much lower than the real ones.

Z