EK-FC5970 waterblock installation features
High temperatures of ATI HD 5970 power elements when cooled by fullcover waterblock and possible reasons for it
During usual WC loop building I got acquainted with ATI HD 5970 Sapphire video card, which was planned to by cooled be EK-FC5970 waterblock. I won′t speak much about the testbed — the important thing is it includes EK waterblock for a ATI HD 5970 card. Stock cooling system was replaced with fullcover type waterblock. A separate loop for video card cooling was built:
Laing D5 pump ” Video card waterblock EK-FC5970 Acetal+Nickel ” TFC Xchanger Double Radiator 240
The radiator was mounted outside the case and cooled with two Enermax UCMA12 fans at 1500 RPM.
After a successful start I then began testing procedures. Furmark stress test showed the following:
Water temperature went up to 36°С while the ambient temperature was 23°C. GPU temperatures are quite satisfactory — 50-55°С. However, temperatures of video card’s power elements 84-86°С do not justify watercooling of the card.
To my mind the temperatures of video card′s power elements VDDC phase looked a bit suspicious.
The following elements are responsible for powering the video card:
- 3 × Volterra VT1157SF: GP_0 powering (vddc phase 1, 2, 3 gpu_0)
- 3 × Volterra VT1157SF: GP_1 powering,( vddc phase 1, 2, 3 gpu_1)
- 2 × Volterra VT1157SF: memory(GDDR5 IMC) controller powering, common phase for both GPs. (vddci phase 1 and 2)
- 1 × Volterra VT232WF: memory powering (VDD+VDDQ)
- 1 × Volterra VT232WF: memory powering (VDD+VDDQ)
- 2 × Infineon n-channel MOSFETs (042N03LS + 119N03S): PCI Express PLX Technology PEX8647 bridge chip powering
- 1 × AOSMD (AO)Z1024DI low-power integrated buck regulator
Taking–off the waterblock:
Thoroughly examined the pressing of thermal pad imprints to find out the strange case. Here is a pressing imprint of a CPL-2-50 IC (memory two-phase power regulator for vddci phase 1 and 2):
Lifting the thermal pad up and instead of chip imprint I see a small step in block surface:
I suppose that it is made on porpoise because the chip should be higher then the other ones near it. Chip height comparison:
They are the same height. Why on earth did Eddy make that step for that chip, I don’t know. According to some info, this hapless chip was a bit higher than the neighboring chips in some early card batches, but other cards, including Asus card that was tested, use chips of the same height. In my opinion it wasn′t that hard for Eddy to include a 0.5 mm thermal pad instead of 1 mm pad for a CPL-50-2 chip — it would solve all the problems concerning chip height without any harm done to the cooling and safety.
73°C on the vddci is bearable but I did add the second pad on CPL-50-2. Now how about those vddc phase temperatures: the imprints are fine — the pressing is good. I mount the waterblock with adding a second t-pad for memory power regulator and what do I see … the temperatures haven′t change a bit.
Eddy promised to figure that problem out, but no official statements about new revision of fullcover waterblock were made.
Further performance comparison of EK waterblock with same products from other manufacturers revealed that EK-FC5970 is one of the best waterblocks for HD 5970 cards. The answer to the question of high temperatures of power elements lies in the circuitry of the graphics card. When compared to HD 5870 the amount of phases of HD 5970 each GPU was reduced from 4 to 3 which is one of the reasons for high temperatures.
Despite some bugs of Ek waterblock, EK-FC5970 can be highly recommended for purchase.
No comments