Life After the Honeymoon
This is the first of my entries into what it has been like living and working with the computer I built back in January 2014. For the most part everything has been working normally and without any major trouble.
The initial round of discoveries through ownership I made related to how the pairing of Intel Core i7-3770K – Ivy Bridge CPU with the z68 chipset. While both are 1155 socket products they do have their hardware preferences. To that end the 3770K can only reach a clock multiplier of 39 on a z68 chipset, meaning that it cannot clock over 3.9 GHz. This in itself defeats the reason for owning an unlocked and overclockable CPU. At any rate I wasn’t too bothered about this since the CPU was still a solid performer.
I began to noticed that kernel_task in OS X was using a unusually high amount of CPU cycles at idle, 10% of one of eigth threads. Since Firewire busses are dependent on the CPU for processing this usage is reflected through kernel_task. I double checked of the same card with my 2010 Macbook Pro and saw that it was running at 5% on a slower four thread dual core system. In both tests I checked by having my Focusrite audio interface on or off. After further research I pinned it to the integrated Firewire controller on the motherboard and its behaviour in OS X’s Console logs. It was displaying an invalid GUID which would continue flagging during system uptime. This does have the knock on effect of affecting stability and increases the potential for kernel panicks. I experimented with adjusted the HPET interrupt values of the system worked on trying modifications for the x58 chipset. I finally admitted defeat and opted for the safe option of buying a more compatible part. I chose the only suitable candidate by virtue of its Texas Instruments XIO2213AZAY chipset Syba SD-PEX30009 1394A PCI-Express x1 Card. This combination has been tried and tested for the specific uses of pro audio and video.
Quiet, Cool, and Dust Free – March to May 2014
During this time I also started developing a dislike towards certain frequencies the water pump and cooler fan would emit. I chose to buy some light acoustic insulation from Akasa and apply to the inside of the case. Since the computer was already fairly quiet I did not have to install thick insulation that would only serve to keep more hot air inside the case. I also began to replace more of the case fans with Antec Truequiet 120mm Case Fans and a 140mm for the top exhaust port, which isn’t quite as big as the wombat sized port on the Death Star.
I decided to take a more preventative stance towards dust management and clean up by purchasing three SilverStone FF121B Filters to install on the remaining intake fans. Having never used additional air filters before I was happy to report that the effect fan pressure and cooling was negligible.
By the time May came around I was considering further acoustic and cooling upgrades. Mainly in changing out the fan on the hydro cooler with a low noise static pressure fan. This manifested itself in the Corsair Air Series 120mm PWM High Performance Edition High Static Pressure Fan in addition to pair of their SP120 Quiet Edition High Pressure Fans. The result was a push-pull configuration on the hydro cooler and the other Quiet Edition fan being placed next to the CPU by nature of its static pressure design it would be better suited to the narrow location between the case side wall and the CPU cutout.
A Bridge Over the River Ivy – May 2014
Fast forward a month later and I began to notice that the warmer weather was bringing disproportionately warmer operating temperatures. This triggered another round of research and inspections that yielded a varied and interesting set of results. I first began with an inspection of the water cooler to check for leaks, broken or cracked hoses, and any abnormal noises from the pump that would indicate a failure. I could not locate any obvious points of failures and moved on to the other components. One interesting factoid that jumped out was Intel’s switch from soldering the Integrated Heat Spreader (IHS) to the die to instead using fairly low grade thermal paste in combination with adhesive to connect the IHS to the die. This results in inferior thermal performance when compared to Sandy Bridge CPUs. Combined with the higher density of the 22nm architecture employed in the Ivy Bridge design and you face a noticeable rise in operating temperatures. As a result the Ivy Bridge CPUs could not be overclocked to the same speeds as Sandy Bridge. By slightly warmer temperatures I mean idle temperatures that are in the 40°C range and load temperatures that are in the 80-90°C range. An ambient room temperature increase of 3-5°C does not translate into hellish inferno unleashing itself on your unsuspecting computer. Something is definitely wrong here.
It was during this round of research that I came across a modification that a friend of mine had mentioned. The dangerous art of delidding a CPU by removing the IHS and exposing the bare CPU die underneath. In practice this can either involve razor blades or hammers and vice grips. I opted for the razer blade and gently cutting around and under the IHS and PCB to break the adhesives tyrannical grip. The aftermath is the usual set of niceties involving the clean up and removal of old thermal and paste and adhesives that leave you with a clean slate to apply the “good stuff”.
Hooray! I didn’t destroy $320 worth of CPU. The computer has been reassembled and ready for testing. While the idle temperatures are marginally lower by around 2-3°C the overall result is not delivering the kind of result I was hoping for. In the spirit of being thorough I reapplied the thermal paste on both the CPU and IHS several times while also tinting the heatsink on the water pump. The results were similar and the operating temperatures unacceptably high.
The next step was to approach the possibility of cooler failure by switching it out with the stock Intel cooler that came packaged with the 3770K. Operating temperatures immediately dropped to something that resembled normal by bringing the temperatures at idle to 35°C. I followed up with Corsair and started and RMA ticket.
I should note that the GPU temperatures were staying consistent with the warmer weather only raising the idle and load temperatures by 2-3°C.
I decided to ditch the mixed Ivy Bridge on z68 chipset by going ahead and buying Intel Core i7-2600K – Sandy Bridge CPU and directly emulating the success of my previous Hackintosh build. The first CPU that arrived was dead on arrival and was immediately returned. The replacement happily settled into its new home and mild 4.1 GHz overclock.
The next step was to find a replacement CPU cooler. I was eager in trying out the tried and tested approach of buying a big block of metal and strapping some fans to it. My weapon of choice is a Zalman CNPS10X Optima Heatsink, mainly because of the price and its ability to mount two fans in a push-pull configuration.
The installation was straight forward with the biggest take home being that the heatsink is a prime candidate for a future lapping workshop since the base wasn’t exactly on the smooth and shiny copper side.
The results from this modest heatsink coupled with the Corsair Air Series fans in a push-pull configuration are great! Idle temperatures are now in the 28°C neighbourhood with load temperatures varying on applications. Sustained Prime98 testing yielded 65°C with a room temperature of 23°C. Testing with Cinebenchresulted in temperatures of 55°C. The best thing in addition to the improved cooling is that it is noticeably quieter than the hydro cooler. The only downside is that it does consume a hefty amount of case real estate making it a little trickier to move around. The new Firewire card has reduced the CPU overhead required by 5% in addition to the extra ports and Firewire 800 capabaility. Except for Bluetooth I have a contemporary Mac Pro matched for connectivity, features and a fair degree of performance.
Stay tuned for the next installment in my quest for maximum computer workstation nerdage!