Ga-x58a-ud3r

[justr]

Great information justr! I really appreciate it. I will read the information on the forum you directed me too. I have already done some basic testing using Prime95 and HWMonitor. At the 211 bassclock and 19x multiplier running Prime95 "Torture Test" with the second test option selected I get some confusing results. HWMonitor shows the CPU core voltage as 1.36V (no test running and idle). Temps are all 6 cores around 45 degrees C. As soon as I start the test the CPU core voltage drops down to 1.344. It does not go above that until I stop the test and return to idle. The temps during the test eventually rise to about 70-72 Degrees C. I see that the Xeon x5650 specs you pointed me to (thanks again) show 1.35V as the top end of the "VID voltage range". Is that a hard limit? In other words if I'm showing the results I mentioned above are there any alarm bells going off?

I take to heart your cautionary tone with regard to overclocking when you don't know what you're doing. In principal I agree. But, you have to start somewhere and I guess that's where I am, with your help, trying to get better informed. The anandtech forum you linked me to seems to have a great deal of info, but in my initial perusing, I can find not step by step approach or methodical guide to this process. Maybe I have just not done enough reading yet. I really appreciate your input justr.

*Note- It makes no difference if I have BClock at 200 or 211 (at 19x multi) the HWMonitor shows idle CPU core Voltage as 1.36Volts. When Prime95 begins the voltage drops for both but it drops more for the lower 200 BClock to 1.325Volts.

The uptight version: Hitting 1.36 volts is out of spec, and Intel would tell you that you risk damaging the chip or system.

The reality: Plenty of folks are running these chips with that much voltage, and there are not a lot of reports of acute damage (though degradation might take months to years to show up).

Personally, I wouldn't be very comfortable going that high. I try to keep my own chip at 1.3v or lower... for ease of mind, and cool temps. I think I have a killer chip (good luck), so I bet I'd get 4.4 Ghz on all cores at 1.35, but I think it's silly risking that.

A couple questions:

- When you installed the CPU, did you reset the BIOS to "load optimized defaults" or whatever the language is? You should, because otherwise it's possible your mobo is still running the chip at i7-9xx settings (which are not identical to X5600!), and this could be dangerous? You will need to load optimized defaults, and then go back and make the Hackintosh specific bios changes.

- Is your CPU voltage in bios set to an actual value, or to something like "auto"? If it's the latter, I definitely recommend setting this value manually. Same for things like dynamic vcore, QPI/VTT core voltage, ram/DIMM voltage.

CAUTION: The following guide is for i7-9xx chips, so DO NOT use its voltage limits or relationships (uncore 2x +1 for instance) with your overclock. They may not be applicable to Westmere 95w X5600 chips. Instead, use it for the logistics of finding a stable overclock at low temps/voltages. This is essentially how you overclock: http://www.techreaction.net/2010/09/07/3-step-overclocking-guide-bloomfield-and-gulftown/

Bottom line: You're probably better off doing this manually. By enabling a higher multi x22, you'll be able to hit CPU frequencies at lower voltages and temperatures, as well as put less stress on other components of your system. And you won't be running voltages out of spec by accident... as you kinda are doing now!

 

[justr]

Quick USB 3 update:

I installed the USB 3 kext via Multibeast, and, interestingly, my new Toshiba Canvio 1TB external drive was recognized and did run at the expected speed (~110 MB/s read and write) compared to when it was plugged in to a USB 2.0 port (~30 MB/s). A 17GB folder transfered in about 3 minutes.

The interesting part is that when I first plugged the drive in, the system would not recognize it (neither mounted, nor showed up as connected in Disk Utility) no matter how long I waited. When I unplugged it and plugged it back in... all of a sudden it mounted instantly and operated at the correct speed. I tried it in both USB 3 ports at that point, and it worked flawlessly.

Next I tried my USB 3, 3TB Seagate Expansion Desktop (which I'd been trying to get working with USB 3; it works fine over USB 2). It mounted. Then it refused to operate faster than a snail's crawl... like USB 1.1 speeds, with pauses, crashes, etc. Then I tried that drive with the USB cable that came with the Toshiba-- same outcome. So it looks like the USB 3.0 kext works with at least some external devices, but not others; I'm guessing the individual SATA-->USB 3.0 bridges are implicated in the inconsistency.

 

[RandomNumber]

The uptight version: Hitting 1.36 volts is out of spec, and Intel would tell you that you risk damaging the chip or system.

The reality: Plenty of folks are running these chips with that much voltage, and there are not a lot of reports of acute damage (though degradation might take months to years to show up).

Personally, I wouldn't be very comfortable going that high. I try to keep my own chip at 1.3v or lower... for ease of mind, and cool temps. I think I have a killer chip (good luck), so I bet I'd get 4.4 Ghz on all cores at 1.35, but I think it's silly risking that.

A couple questions:

- When you installed the CPU, did you reset the BIOS to "load optimized defaults" or whatever the language is? You should, because otherwise it's possible your mobo is still running the chip at i7-9xx settings (which are not identical to X5600!), and this could be dangerous? You will need to load optimized defaults, and then go back and make the Hackintosh specific bios changes.

- Is your CPU voltage in bios set to an actual value, or to something like "auto"? If it's the latter, I definitely recommend setting this value manually. Same for things like dynamic vcore, QPI/VTT core voltage, ram/DIMM voltage.

CAUTION: The following guide is for i7-9xx chips, so DO NOT use its voltage limits or relationships (uncore 2x +1 for instance) with your overclock. They may not be applicable to Westmere 95w X5600 chips. Instead, use it for the logistics of finding a stable overclock at low temps/voltages. This is essentially how you overclock: http://www.techreaction.net/2010/09/07/3-step-overclocking-guide-bloomfield-and-gulftown/

Bottom line: You're probably better off doing this manually. By enabling a higher multi x22, you'll be able to hit CPU frequencies at lower voltages and temperatures, as well as put less stress on other components of your system. And you won't be running voltages out of spec by accident... as you kinda are doing now!

@justr, I sent you a PM.

 

[justr]

PM replied.

Also, note I've edited the settings. The uncore should not be 2x +1 the memory multiplier, that's the old Nehalem OC value. For Xeon 5600 the uncore multi should be 1.5x the memory multi by default, and it is reportedly safe to run it up to 2x the memory multi. Most people suggest there are little to no actual performance benefits from 1.8x up to 2x (or beyond) the memory multi for the uncore multi. And if you go beyond 2x while running really high memory multis, bclk, or voltages... you'll definitely encounter stability issues at best, and perhaps risk your chip.

So:
sys mem multi 6x means you should have an uncore multi of between 9x and 12x
sys mem multi of 8x means you should have an uncore multi of between 12x and 16x

That's the short version. I'm not responsible for anyone blinding following these settings, BTW! Always consider that I-- or anyone other individual-- could be mistaken, could misread his own settings, or could have committed a typo. That's why it's wise to get to these values via your own knowledge and the consensus of reading both official documents and the experience of other users. Chips and mobos have been killed in establishing these things!

 

[Going Bald]

Thanks Guys!

GB: Can you confirm that the universal USB3 kext in Multibeast is actually running drives at full USB3 speeds? I've used that kext in the past in Mavericks... and it never actually works. USB 3 shows up in System Profiler, but a couple seconds after I hook up a USB 3 drive and start a file transfer... the transfer drops down to more like USB 1 speeds, or crashes altogether! The same drive works on the USB 2 ports at USB 2 speeds./QUOTE]
Actually, it does not. You need the NEC / Etron / Renesas d720200 driver from earlier MultiBeast on the GA-X58A boards not sure which one, probably ML. On mine I get read/write rates in the 60's with the USB3.0 Passport drive - same as I get with it on the USB3.0 from the CPU generic driver on the Z87I-Deluxe. I think the low rate is the HDD, not the port or driver.

EDIT: just double checked and the Generic USB3.0 is what is running - removing it and rebooting causes the USB3.0 hub to disappear from the System info.

So, yes, the Generic USB3.0 driver does work. The Passport drive gets read / write rates below 30 when connected to USB2.0 port.

 

[RandomNumber]

PM replied.

Also, note I've edited the settings. The uncore should not be 2x +1 the memory multiplier, that's the old Nehalem OC value. For Xeon 5600 the uncore multi should be 1.5x the memory multi by default, and it is reportedly safe to run it up to 2x the memory multi. Most people suggest there are little to no actual performance benefits from 1.8x up to 2x (or beyond) the memory multi for the uncore multi. And if you go beyond 2x while running really high memory multis, bclk, or voltages... you'll definitely encounter stability issues at best, and perhaps risk your chip.

So:
sys mem multi 6x means you should have an uncore multi of between 9x and 12x
sys mem multi of 8x means you should have an uncore multi of between 12x and 16x

That's the short version. I'm not responsible for anyone blinding following these settings, BTW! Always consider that I-- or anyone other individual-- could be mistaken, could misread his own settings, or could have committed a typo. That's why it's wise to get to these values via your own knowledge and the consensus of reading both official documents and the experience of other users. Chips and mobos have been killed in establishing these things!

Thank you justr! With your help I've managed a much healthier overclock, and learned a little too!

* Edit - I have an inquiry related to this but thought it worthy of it's own Topic. Please take a look here:
http://www.tonymacx86.com/overclock...cpu-speed-being-reported-see-explanation.html

* Note - I solved the issue in this related post by deleting the nvram.plist file in Extra.

 

[justr]

Yup, I figured out the USB 3 crashing out problem was isolated to something in my Seagate Expansion Desktop 3TB drive, probably the SATA-USB3 bridge, firmware, or something. Shame, because that's my photo library backup.

With a Toshiba Canvio 1TB (2.5"), I get read/write of around 110MB/s using the Black Magic app.

The funny thing is that when I first plug the Toshiba in, it won't mount. It's as if the USB 3 system gets initialized only when a drive goes in, but won't actually run. Second and all subsequent plugs mounts the drive and runs it at full USB 3 speeds.

 

[fruitbat04]

I see there are still tons of those x5650's available on eBay for £65 so I am tempted. I'm sitting on a i7 920 at 3.8ghz with no huge desire to upgrade, but being able to keep all the RAM and board do make it tempting.

Does anyone who's done it have any feedback to the real world differences, as well as stuff like handbrake encodes of which I do a fair amount. It generally takes about 2 hours for an HD rip at the moment.

 

[Hackintoshron]

This might not be the right place for this question, but does anyone here know how to overlock a real Mac Pro 2010's Xeon processors? There has got to be some way to do this.

 

[justr]

I see there are still tons of those x5650's available on eBay for £65 so I am tempted. I'm sitting on a i7 920 at 3.8ghz with no huge desire to upgrade, but being able to keep all the RAM and board do make it tempting.

Does anyone who's done it have any feedback to the real world differences, as well as stuff like handbrake encodes of which I do a fair amount. It generally takes about 2 hours for an HD rip at the moment.

Yeah-- video encoding and large batch processes that use all the threads you can provide (batch RAW --> JPG w/ hundreds of 20+mp files) get an almost perfect efficiency increase from 4 -> 6 cores (8 -> 12 threads), and a near perfect scaling for frequency. Let's say take the theoretical increase and then hit it with 80% efficiency.

There are LOTS of handbrake comparisons for various processors at various speeds as measured in fps for identical files.

So (please understand this is back of the envelope stuff here, but it mostly works*)...

3.8 ghz x 4 cores = 15.2

4.2 ghz x 6 cores = 25.2

25.2 - 15.2 = 10.

(10/15.2) x 100% = ~65% theoretical increase if you get perfect efficiency. Let's knock this down to, say, 80% of perfect efficiency with respect to scaling frequency and cores (a safe bet for good video encoding software). That gives you around 52% increase in computing power.

That fits with my own sense of the increase in multi-threaded apps like FCPX/Compressor/Handbrake/Aperture... about 50% faster (remember, that doesn't mean "twice as fast," which would be a 100% increase in computing power; encoding times will not be halved). It also matches most of the benchmarks that can make use of threads/cores. My Geekbench went from 12,500 to 19,000, my cinebench 11.5 went from 6.5 to 10.

For most tasks you won't see this increase, since most tasks use at best a few threads. For encoding video in a well-written application, you'll get near-full benefits of the extra cores. I'd guess that your encode time for the 2hr/120min video to drops to something like 80 minutes, or 1hr 20min give or take 10-15 minutes.

*All of this assumes the bottleneck in your work is the CPU and not the GPU or I/O like SSD/HDD/RAID-- or poorly coded software.

To the poster above: No, this is definitely not the place for getting clues to overclocking a real Xeon Mac. There was at one point a tool for very small OC on very specific models. Wouldn't be my cup of tea messing with the thermal profile and potential stability of a system that doesn't allow much tinkering... even if the tool is still around and you are lucky enough to have a mac that it will run on.