@Justr, I've sent you a couple of PM's but I think the board is having issues...
Because of your tip I snagged a
X5650 from Ebay for $60 w/ free shipping.
Can you give me details of your bios settings on this board (specially voltages for the Xeon) and any other changes you made for this build? I also got a brand new Cooler Master Hyper 212+ to go with it.
Any special config for OSX? How's your power management?
Thanks!
I think as long as we stick to the GA-x58a-id3r this is still valid. We already covered pretty much everything there is to do for installation on the previous pages.
OK, if everyone is cool with it, I'm happy to include that info in this thread! As a side note, I never got the PMs... so something must be acting up on the board.
I went from i7-950 to L5639 (6 core, low power) to X5650. I was already on the latest BIOS firmware version for the board. My steps were:
1. Install CPU.
2. Load BIOS.
3. Load Optimized Defaults (like when you first built the system; important because voltages are not the same as i7-9xx).
4. Go back and set the Hackintosh specific BIOS settings we need to use to boot (don't remember these off the top of my head)
5. Boot into OS X and make sure everything was stable enough, run a few benchmarks to have a baseline.
6. Boot into Windows, run stability tests, benchmarks for baseline scores.
OC (only showing where I made changes; Hyper 212+ w/ MX-4 thermal paste):
- CPU Multi, AKA "Clock Ratio": 20/22 (either setting gives me 22 in both OS X and Windows on all cores)
- Advanced CPU Core Features: All enabled (so HT, Turbo Boost, C States, Energy Savings-- all of them)
- QPI Clock Ratio: x36
- Uncore Clock Ratio:
x13 (2x RAM ratio + 1) This should actually be x9 to x12, or 1.5x to 2x the memory multiplier. Technically Xeon 5600 uses 1.5x the system mem multi for uncore by default; 2x +1 was the old nehalem/i7-9xx value!
- BCLK: 190 (I've fiddled at same voltages from 185 to 193)
- System Memory Multiplyer: 6*
Voltages:
- Load Line Calibration: Level 1
- CPU VCore: Normal (1.21250v)
- Dynamic VCore: +0.10000v)**
- QPI/VTT: 1.2950v
- CPU PLL: 1.800v
- IOH Core: 1.120v
- DRAM Voltage: 1.660V***
Everything else left as it was, generally set to auto
*My RAM (6x 2GB Corsair XMS 3 1600) has never OC'd very well, so I've generally used low multis. Honestly, most benchmarks show that there isn't a big advantage to a high DDR3 RAM OC when it comes to real world performance, and I want my RAM to be rock stable (I also initially tested my RAM overnight using Memtest as around 1333... and I'm too lazy to do that again with higher clocks)
**Though it looks like my VCore should go up to 1.3125v, both BIOS, Windows, and OS X show a highest voltage of 1.28v under full load (6 cores @ 4.200 Ghz or X22 multi), and close to 1.04v when idle (6 Cores at x12 multi)
***As above, BIOS, Windows, OS X all report RAM voltage as 1.64. I only set 1.66 because that's the lowest setting above 1.65, which is what my RAM is rated for.
One Peculiarity: OC behavior is not identical for me in Windows vs. OS X. In Windows 1 or 2 threaded tasks can hit an X23 multi for two cores at 4.4 Ghz. In OS X I never get that X23 multi, only x22, even if I'm running a single-thread test such as the ones offered in Cinebench r11.5 and r15. My single thread scores in these are significantly higher in windows (~1.4) compared to OS X (~1), which is odd. My multi core benchmarks, however, are actually a little higher in OS X than in Windows (in both OS's all six cores stay at x22 or 4.20Ghz).
WARNING to folks reading this: Do keep in mind that OC gives you the power to put too much voltage through your chips, or the wrong ratio of voltages (some values should be within certain ranges of one another, like DRAM and QPI/VTT being within 0.5v of one another), or supply too much heat to the chip or MOBO components. Also, keep in mind that the best methodology is not to just copy someone else's BIOS settings for OC, because each chip (even identical models) and each Mobo (even identical models) have different tolerances. You'll find people with X5650 who are stable at 4.4-4.6Ghz on all cores, and people who struggle with 4.20 Ghz on all cores with identical settings. Also, the final frequency of the chip doesn't necessarily = some kind of stability or even benchmarking ability. When I had my L5639, which had a very low CPU multi (x16 up to x18 w/ turbo), I was able to run my mobo BCLK at 212 totally stable! And as a result, my benchmarks were higher with the L5639 @ 3.8 Ghz than my X5650 @ 3.8 Ghz when the X5650 was using a 22x multi and a BCLK of 172. There's no way in hell I am going to try to get my X5650 stable at 212 BCLK and 22x multi at reasonable voltages or temperatures!