Want to say thanks to Frozen CPU for giving me a chance to play a bit and verify your custom Core i5/i7 mount plate, but I got to bench for a weekend. Its not often a toy, yes for me a Phase Unit is a toy. I would only use a phase unit for benching, with how much the unit draws from the wall and the process required to change CPU’s… liquid is my choice. Again, thanks Frozen CPU. Just a disclaimer, this log will be much more relaxed, less editing and very little monitored and logged tests. This log is purely benching and seeing what the OCZ Cryo-Z, Frozen CPU Core i5/i7 mount plate and the hardware I have available can produce. Enjoy…
I didn’t believe it until I opened the box… the OCZ Cryo-Z is real and doesn’t just exist on a web site. I had heard a variety of different things. But the last I knew was they were available from a retailer in California and the units were pickup only. Fast forward… I have a Cryo-Z on its way to me and there is a new mount plate I need to verify with Core i7 and LGA1366. Panic sets in, I have never done phase or sub zero cooling… oh crap. Forums to the rescue, just start reading the stickies and thanks to XtremeSystems I felt somewhat prepared. I was able to finally grab the specs out of the manual:
The specs of the unit and absorbing what I could in the phase section at XS, I was trying not to get my hopes to high for the Cryo-Z. After all, the unit was originally planned for the Core 2 and LGA775 platforms. Yeah… guess they didn’t make it. Enough negativity out of me regarding the vaporware status of the OCZ Cryo-Z, but I seriously thought the were all destroyed or never really existed.
After lugging the box from my doorstep down the stairs to the lab, I was surprised by the shock watch stickers all over, and quickly verified all was good. Unit is out, coming with quite a bit of insulation, backplate, mount screws/springs, manual. Was going to be a few days before the Frozen CPU i5/i7 mount plate and backplate were going to arrive, figured I should figure out how to take the collar and plastic stock mount plate off. I don’t have any available LGA775 machines to test with quickly either.
After a day or two of walking by the Cryo-Z, thinking about how I was gonna get that collar and plate off it was finally time to actually figure it out. I started by twisting, seeing if the collar would screw off… nope. Then I started moving the foam at the head around, sure enough there is a c-clip holding the collar back from slipping off. Quick job of the needle nose, pulling one of the side out, then gently pulling the c-clip loose. There ya go, collar should come right off.
I mentioned this above, but want to make sure everyone caught it… the Cryo-Z does not natively support Core i5 (LGA1156) and Core i7 (LGA1366). The Cryo-Z was originally planned for Core 2 and LGA775. Frozen CPU went and whipped up a mount plate that is a perfect replacement for the plastic stock plate. The Frozen CPU plate is aluminum, with the LGA775 to LGA1366 mount holes, use the stock plate if your rocking AMD. Frozen CPU includes a LGA775 to LGA1366 backplate as well, the included 6-32 screws work just fine with the LGA1156 and LGA1366 backplate from Frozen CPU.
Now that we’ve got the stock plate off, I’ve got to clean that copper head before going any further. Looks like sitting in a warehouse for 2 years aged the copper surface a bit. Vapor’s method of vinegar and salt, a few dips for roughly 90 secs, polished with a brush in between dips. Finish it off with a polish from the Krazy Kloth. With the copper head looking usable again, slip the Frozen CPU mount plate over the copper head and insulation, push the collar back on and ready the clip. You’ll notice a small notch/extrusion in the recess where you have to secure the clip. That extrusion is where you want the opening of the c-clip. I was able to get one spot started in the recess and just used my thumb to work the clip back in securing the collar from slipping over the copper head. If you end up with some pinched insulation, just use your fingers to work the insulation out and the clip under.
With the Cryo-Z all dressed up, time to focus on the board. I didn’t get fancy, just used the foam insulation included and called out in the manual, not bad as far as instructions go, some things are left to draw your own conclusion. I ended up cutting quit the sliced up piece of insulation for the board… but it works. Now with the board insulated, have to craft some insulation for the head. For some crazy reason, I forgot to photo my hack job of 3 pieces of 3/8″ foam. I cut it for the collar diameter, marked holes and used a drill bit to punch out the mount screws. I went for 3 pieces of 3/8″ foam, seemed to do the trick quite well.
Nothing fancy on the bench for this one, just the same rig used for round 1 and 2 of CPU block testing, capable of at least 4.2GHz at 1.315v Vcore:
So I didn’t completely kill the weekend just on reading, copper head cleaning and mount plate photos… I actually got to make a first pass on clocks. Most important of all, I got to break my sub 0ºC cherry and that has to be the best part of the bench log. However, I have a lot of clocking to do… I can’t just rely on the brute force method (raise Volts, raise bclk until it works) of clocking, time to actually learn and tweak. If the investment (time and money) in sub zero is going to be worth it, time to up the OC’ing game. First, lets revisit that whole expectations thing and remember the Cryo-Z was originally designed for Core 2 heat loads, Core i7 920’s can reach well above 250w of heat… OCZ only spec’d a 120w compressor. So don’t expect to see DTS sensor temps of -60ºC, that’s just not going to happen with a Cryo-Z.
I have 3 of the many screen shots to come as I continue running the Cryo-Z and tweaking. But, the first shot shows 100% BOINC load on 8 threads running 4.2GHz with 1.2875V bios set. Next screen is just to show how bumping the Vcore up by roughly 0.1V means to heat load. Now, please remember that RealTemp and the Core DTS sensors on die aren’t great at low and below 0ºC temps are inaccurate –BIOS was -105ºC to 18ºC– so you can’t rely on Tjunction or Tcase for relatively accurate temps. Even still, I’m using RealTemp as my proof of running fully loaded and maintain a temperature colder than 0ºC… don’t take my small victory. 😉
Once the initial twitch of being below zero and the expense I could accumulate in case of disaster happening, I gave in to the text messages from Vapor and decided to start running a few benchmarks. Sad thing is, I only got my 3DMark Vantage score baselined and had to call it a night and weekend.
At the present, time to change the Blue Bench (skinnee labs network name) back over to data logging for radiator tests… but I’m leaving as much in place as possible to get benching again before and during the Thanksgiving holiday. I really want a 15k Vantage score with this setup.