[HowTo] Building an overclocked Core2Duo box for OSx86.

[looper]

I bought everything for my first hackintosh (a copy of your first box) from NCIX- I live in Vancouver. Doesn't suprise me that they don't take US credit cards, they seem very cautious types.. I paid for some drives in cash and they checked over the bills like I was trying to pass counterfeits for sure! A reputable place however, and usually well stocked (if a bit expensive)

 

[bofors]

I have got some new hardware in, but I am still testing stuff out.

 

Here are the preliminary results:

 

- 4 x 512 MB RAM modules will not work on the 1333 strap (just like 4 x 1 GB), apparently this is a known problem at XtremeSystems. So, I am sending all of my Corsair 5400UL back. Instead, I ordered a set of PC2-8000 (DDR2-1000) Crucial Ballistix from NewEgg because it supposed to be good (but not the best) Micron D9 RAM (Micron owns Crucial) and it is on sale: http://www.newegg.com/Product/Product.asp?...N82E16820146563

 

- I have a Xeon 3060 running at 3.6 GHz with no problem on the 1066 strap. Going above 3.6 GHz seems tricky and I think that that BadAxe1 is holding this chip back (I had expected to do better because of my excellent results with the E6300). So, I am planning to send the two BadAxe1's that I ordered back to Mwave and I have ordered two BadAxe2's from ClubIT instead. The BadAxe1 really can not take full advantage of high performance RAM either because the vDIMM is limited 2.2V, the BadAxe2's vDIMM goes all the way up 2.8V (higher than you should probably consider using).

 

- I am thinking about ordering some ThermalRight HR-05 heatsinks for the Northbridge. The HR-05 SLI is supposed to have better performance (and is actually a little cheaper at NewEgg), but I do not think it will fit next to a Big Typhoon very well. The initial results of the Noctua NH-U6 have not been spectacular: http://www.xtremesystems.org/forums/showth...3957&page=2

[bofors]

I just read a report that an Ultra V-Series "500W" PSU like I am using has died and taken a new BadAxe2 motherboard with it: http://www.xtremesystems.org/forums/showpo...p;postcount=175

 

I suspect the PSU may have been overloaded with a high power graphics card though. The Ultra V-Series "500W" PSU are overrated and actually put out only 460W. They should not be used to power anything like an Nvidia 8800GTX.

 

EDIT: A low power graphics card was in use when that Ultra died, so that was not the cause of failure.

[joe75]

The Ultra V-Series "500W" PSU are overrated and actually put out only 460W. They should not be used to power anything like an Nvidia 8800GTX.

 

EDIT: A low power graphics card was in use when that Ultra died, so that was not the cause of failure.

 

Ultra are junk

 

That sucks about your ram not working right, and I know you're an Intel fan but, Intel chipsets will always hold your overclock back

 

 

The HR-05 is a little better because of the extra holes in the fins which allow for better airflow and faster cooling but with the right fan the noctua can perform just as well.

 

I see you like xstremesystems.org, have you looked around 2cpu or bleedinedge?

http://www.bleedinedge.com/forum/

http://forums.2cpu.com/

[greenman100]

I have got some new hardware in, but I am still testing stuff out.

 

Here are the preliminary results:

 

- 4 x 512 MB RAM modules will not work on the 1333 strap (just like 4 x 1 GB), apparently this is a known problem at XtremeSystems. So, I am sending all of my Corsair 5400UL back. Instead, I ordered a set of PC2-8000 (DDR2-1000) Crucial Ballistix from NewEgg because it supposed to be good (but not the best) Micron D9 RAM (Micron owns Crucial) and it is on sale: http://www.newegg.com/Product/Product.asp?...N82E16820146563

 

- I have a Xeon 3060 running at 3.6 GHz with no problem on the 1066 strap. Going above 3.6 GHz seems tricky and I think that that BadAxe1 is holding this chip back (I had expected to do better because of my excellent results with the E6300). So, I am planning to send the two BadAxe1's that I ordered back to Mwave and I have ordered two BadAxe2's from ClubIT instead. The BadAxe1 really can not take full advantage of high performance RAM either because the vDIMM is limited 2.2V, the BadAxe2's vDIMM goes all the way up 2.8V (higher than you should probably consider using).

 

- I am thinking about ordering some ThermalRight HR-05 heatsinks for the Northbridge. The HR-05 SLI is supposed to have better performance (and is actually a little cheaper at NewEgg), but I do not think it will fit next to a Big Typhoon very well. The initial results of the Noctua NH-U6 have not been spectacular: http://www.xtremesystems.org/forums/showth...3957&page=2

 

 

Quite a change of heart from the old days eh??

 

throwing down for $300 memory and $300 processors

 

better get you a new PSU before that Ultra blows up

[REVENGE]

I would trust that Ultra PSU...to replace my old Dell's old 250W PSU i would LOL!

[bofors]

That sucks about your ram not working right...

 

It is no big deal, I wanted to send it back to NewEgg anyways.

 

... and I know you're an Intel fan but, Intel chipsets will always hold your overclock back...

 

I really not an Intel fanboy, I just try to find the right hardware for my needs, which just so happens to be Intel for reasons of EFI and perhaps stability.

 

The HR-05 is a little better because of the extra holes in the fins which allow for better airflow and faster cooling but with the right fan the noctua can perform just as well.

 

I suspect they are similiar in performance too, but the Noctua NH-U6 is still not available in the US as far as I know, and the HR-05 looks a little cheaper (I am planning to buy three).

 

I see you like xstremesystems.org, have you looked around 2cpu or bleedinedge?

 

I do read 2CPU, but since I have ruled out dual Woodcrest machines on the basis of no overclocking motherboards and FB-DIMMs latency, there it is of little interest to me now. I will be going with Kentsfield for more power, perhaps in a few months, when the prices come done and the new models are on the market. In particular I am interest in the entry level Xeon Kentsfield, the X3210, but I think the price of $690 is still way to high (I think about $500 would be fair): http://www.macnn.com/articles/06/09/25/qua...re.intel.xeons/

 

Otherwise, XtremeSystems seems to be the leading overclocking forum and it serves my needs pretty well.

[bofors]

Quite a change of heart from the old days eh??

 

throwing down for $300 memory and $300 processors

 

Sort of, but there are some reasons for this too.

 

First of all, the E6300 build was my first overclock, so I wanted to limit my risk. The results turned out pretty good, these Conroes are very fast and I had no problem getting a decent overclock on a E6300 with a BadAxe1. Now I am more comfortable spending money on higher quality processors (but so far the results (3.6 GHz) have been less than expected (~3.9 GHz). Since I write my software, I can try to turn to take advantage of the larger 4 MB cache on these more expensive Conroes too. If I can figure out a way to use it (and I think I can), I should get a lot more than the 10% performance max. improvement that AnandTech got with "off-the-shelf" software.

 

I actually got a discount on the Ballistix RAM that I ordered from NewEgg after bitching about the 5400UL being mis-spec. and such. So, I am not paying $300 for it and it looks like the best deal on market now anyways. I was considering buying some more Corsair ValueSelect (Elpida IC's) because I am very pleased with my results on the two sets I already have, but I thought the price was little high at $200 (I paid about $185 before). But I want to see first hand want difference high quality RAM will actually make on my software too. So, I may or may not be buying some more of the expensive stuff.

 

Second, there have been two major breakthroughs in OSx86 in the last month: 10.4.8 (with a choice of kernels even) and Titan / Natit. In particular, getting Natit to work with dual displays in extended desktop mode is a big deal to me. It means I can continue to save money by running two smaller panels, instead of having to lay out some $800 or more on a 24" or larger panel. Furthemore, because of OSx86, I have decided it is time to sell my PPC Mac gear which is probably worth as much as I have spent here on three BadAxe1 / Xeon 3060 boxes.

 

Finally, my "professional" situation has changed in the last month or so too. I am using a lot more computer power now. Really, I am close to the point where spending $1100 or so for a QX6700 actually makes good economic sense. Furthermore, my financial prospects look much better in the near term, so I feel like I am going to be having a much larger budget for hardware in perhaps a few months.

 

... better get you a new PSU before that Ultra blows up

 

Well, it turns out that the killer Ultra was new out of the box. From what I have read, this is when most PSU fail (other than dying from old age). So, I am really not worried about the Ultra I have been running for the last weeks but rather the two new ones sitting here in boxes (I have actually got four of these Ultra V-Series 500 W PSU's now). Basically, I am going to try to test them before using them on valuable hardware. But statisticall speaking, I still feel very safe. I found no one else claiming this PSU blew up and destroy something (and there must thousands of these in service). Nonetheless, I have emailed Ultra on this issue already and I am going to be reaming them over the safety of these PSU's. If there are more cases of this they know of, Ultra had better tell me.

 

However, I am building another box today with another refurbished BadAxe1 rev. 305 and Xeon 3060 (L629A780). But this second "open box" BadAxe1 was missing some parts, in particular the 4-pin ATX to 8-pin EPS power adapter (which is needed to make an ATX PSU work on with Bad Axe's 8 phase power system, it will not boot unless all 8 pins have juice). I am going to order one (or two) adapters (and perhaps a very high quality Enhance 500W PSU) from eWiz, but in the meantime I picked up a BFG Tech 650 ATX/EPS (4 + 4 pins) PSU at CompUSA. I also want to see if my cheap Ultra PSU is holding back my overclock too.

[joe75]

I told you before that I just have bad opinions for Ultra, if you have no problems then their good for you. Don't test your PSUs with a tester, instead just use a multi-meter, plug it in and go through checking all the rails, this will allow you to put the power directly in the individual parts and if it's gonna blow thats when it would happen. I looked at the same BFG before but turned it down for my Hiper that was only $99

[Alessandro17]

A bit OT here, sorry.

 

What do you people think of this PSU:

 

http://www.tecnoware.com/uk/dettaglio.asp?idf=41&ids=13

 

I bought the 650 W version. It was the best I could find in a hurry.

[bofors]

I told you before that I just have bad opinions for Ultra, if you have no problems then their good for you.

 

I got email from Ultra on this already (damm fast), they are claiming this is an "isolated" case. Although I am grilling them for more information, I have certainly have not seen any other reports of an Ultra V-Series 500W destroying hardware.

 

Don't test your PSUs with a tester, instead just use a multi-meter, plug it in and go through checking all the rails, this will allow you to put the power directly in the individual parts and if it's gonna blow thats when it would happen.

 

Plug it in to what, a motherboard?

 

I looked at the same BFG before but turned it down for my Hiper that was only $99

 

I am not planning on keeping this BFG Tech 650W for $160, I will be taking it back to CompUSA in a week or so. While it looks a little more stable in BIOS, the 12V rail is little low.

 

I got almost identical results from building a second 3060 (L629A780 ) box today with another BadAxe1 rev. 305. It seems to be 100% stable at 3.6 GHz (on either 1066 or 1333 strap), but requires a vCore of about 1.55V. The only differences seem to be this system can be stabilized by raising the vFSB and BIOS reports that runs some 10 degrees cooler than the first one.

 

Both of these systems will overclock a little above 3.6 GHz on the 1333 stap, but are not very stable and will not reach 3.7 GHz. I have tried overclocking above 400 FSB on the 1066 strap with ClockGen (Winodws) with the same results.

 

A bit OT here, sorry.

 

What do you people think of this PSU:

 

http://www.tecnoware.com/uk/dettaglio.asp?idf=41&ids=13

 

I bought the 650 W version. It was the best I could find in a harry.

 

I have not heard of Tecnoware before, but since it appears that they only produce PSU's I would guess they are probably good. But I think it is rather odd that they do list the standard "rails and amps." chart for their PSU online. Otherwise, I would recommend googling for reviews on them and this PSU in particular.

[Alessandro17]

Thanks Bofors

[joe75]

Plug it in to what, a motherboard?

Plug it into the wall and go through putting the probes into the ends of the jacks, you should see a little higher voltage readings then when you tested it plugged into your board. It is usually not recommended to do this but if something is gonna go wrong this will force it

[bofors]

Plug it into the wall and go through putting the probes into the ends of the jacks, you should see a little higher voltage readings then when you tested it plugged into your board. It is usually not recommended to do this but if something is gonna go wrong this will force it

 

I had tried to test one of these Ultra's by just plugging into the wall with only a fan before, but that did not work. Anyways, I have found a multimeter now and will give it a shot.

 

EDIT: Ok, did get a reading of some 5V off a pair of pins in the 20 pin motherboard connector, but I do not think the PSU turn all the one when disconnect like this. The fan does not spin. I think I need to short some mother pins to trick it into turning on.

 

EDIT2: Ok, I found this diagram, it look like I need to connect pin 14 to ground to turn the PSU on: http://pinouts.ru/Power/atxpower_pinout.shtml

[joe75]

http://www.ochardware.com/articles/psuvolt/psuvolt2.html

 

Sorry for not being more descriptive, I gotta work on that

[bofors]

That's a good reference. Anyways, I tested my other two Ultra V-Series 500W PSU out this way and they seem fine. I checked the +3.3 V+, +5 V and +12V rails with the multimeter and also powered a 120 x 38mm 100+CFM Sunon fan too. Of course, this is not the same as putting a real load on the PSU.

[joe75]

Of course, this not the same as putting a real load on the PSU.

 

The reason I said to do it that way is because you're putting power to separate transformers to see whats the weak link. You're not suppose to do that because PSUs are designed to support equal power distribution, thats why you have to jump it to get it to work without being plugged into the board.

Any PSU is only as good as the parts that make them, if nothing blew up when energizing individual plugs, their as stable as the company can guarantee.

 

Just a warning to anyone that ever jumps a PSU for testing or any other reason, you do NOT want to get shocked by DC power, you will not forget it

[bofors]

You're not suppose to do that because PSUs are designed to support equal power distribution, thats why you have to jump it to get it to work without being plugged into the board.

 

Joe, I agree that testing a PSU without load does put some unusual stress on it (because they are designed to run with a normal, balanced load). But this unloaded stress will not necessarily trigger every failure which would have been cause under load. JonnyGuru blew-up two PowMax Assassins, but it took about 300 W of cross-load to do so: http://www.jonnyguru.com/PSU/Assassin/Powmax_Assassin.html

[bofors]

Temperature Monitor 4.0 works on my Conroe / BadAxe systems for CPU temps. now!:

http://www.versiontracker.com/dyn/moreinfo/macosx/19994

[Alessandro17]

Interesting article:

 

http://www.geekpatrol.ca/2006/12/geekbench...-december-2006/

 

If there is any truth in that article, a hackintosh with a slightly overclocked Intel Core 2 Duo E6600 should easily outperform the most powerful Mac Pro.

[bofors]

I have got some initial results from my third Xeon 3060, but now with a BadAxe2. The BadAxe2 is clearly a better board than the BadAxe1. It has much better BIOS options, one can take the 1066 strap to 500 FSB, raise vDIMM to 2.8 and there are finer choices for vMCH and vFSB. The BadAxe2 also appears to more stable, I can boot OS X at 3.8 GHz (FSB 425 x 9). However, OS X is not very stable above 3.6 GHz while Windows clearly is. This can be attributed to the microkernel-like archetecure of OS X versus the monolithic structure of Windows. While OS X may be easier to port, it is a little less stable.

 

My results clearly indicate that a multiplier of 9 is unnecessary to reach the limit of air-cooled Conroes (especially for OS X). Unless you can take advatage of a 4 MB cache, the best deal is the Xeon 3050 (the Xeon version of the E6400) with a multiplier of 8. Likewise, the refurbished BadAxe1's I have gotten for half price effectively offer the same performance (3.6 GHz) as the BadAxe2's for running OS X here (running quad-core Kentsfield may be another story though).

[bofors]

Interesting article:

 

http://www.geekpatrol.ca/2006/12/geekbench...-december-2006/

 

If there is any truth in that article, a hackintosh with a slightly overclocked Intel Core 2 Duo E6600 should easily outperform the most powerful Mac Pro.

 

The 3.0 GHz MacPro has a GeekBench score of 372.3.

 

Here are my results with a Xeon 3060 (3.66 GHz, FSB 407 MHz), BadAxe2, PC2-8000 CAS 5 Crucial Ballistix (DDR2-814, 3-3-3-4, 2.2 V):

 

Geekbench 2006 (build 242).  Email geekbench@geekpatrol.ca with feedback.

System Information
 Geekbench Version:		 Geekbench 2006 (build 242)
 Geekbench Platform:		Mac OS X x86 (32-bit)
 Geekbench Compiler:		GCC 4.0.1 (Apple Computer, Inc. build 5363)
 OS:						Mac OS X 10.4.8 (Build 8L2127)
 Model:					 Hackintosh
 Motherboard:			   ACPI
 Processor:				 Intel(R) Xeon(R) CPU			3060  @ 2.40GHz
 Processor ID:			  GenuineIntel Family 6 Model 15 Stepping 6
 Logical Processor Count:   2
 Physical Processor Count:  1
 Processor Frequency:	   3660 MHz
 Bus Frequency:			 1064 MHz
 Memory:					2048 MB

Integer Performance
 Emulate 6502
single-threaded scalar   224.2 (rate: 1.0, result: 423.9 MHz)
multi-threaded scalar	446.9 (rate: 2.0, result: 844.3 MHz)
 Blowfish
single-threaded scalar   320.9 (rate: 1.0, result: 132.4 MB/sec)
multi-threaded scalar	640.5 (rate: 2.0, result: 264.2 MB/sec)
 bzip2 Compress
single-threaded scalar   313.6 (rate: 1.0, result: 48.9 MB/sec)
multi-threaded scalar	625.0 (rate: 2.0, result: 96.9 MB/sec)
 bzip2 Decompress
single-threaded scalar   392.1 (rate: 1.0, result: 145.8 MB/sec)
multi-threaded scalar	803.1 (rate: 2.0, result: 289.3 MB/sec)

Floating Point Performance
 Mandelbrot
single-threaded scalar   247.8 (rate: 1.0, result: 1.8 Gflops)
multi-threaded scalar	495.2 (rate: 2.0, result: 3.5 Gflops)
 Dot Product
single-threaded scalar   447.1 (rate: 1.0, result: 2.3 Gflops)
multi-threaded scalar	871.4 (rate: 2.0, result: 4.5 Gflops)
single-threaded vector   211.8 (rate: 1.3, result: 3.0 Gflops)
multi-threaded vector	403.0 (rate: 2.5, result: 5.8 Gflops)
 JPEG Compress
single-threaded scalar   222.2 (rate: 1.0, result: 20.6 Mpixels/sec)
multi-threaded scalar	445.0 (rate: 2.0, result: 41.2 Mpixels/sec)
 JPEG Decompress
single-threaded scalar   213.5 (rate: 1.0, result: 35.5 Mpixels/sec)
multi-threaded scalar	421.2 (rate: 2.0, result: 69.8 Mpixels/sec)

Memory Performance
 Read Sequential
single-threaded scalar   593.9 (rate: 1.0, result: 7.4 GB/sec)
multi-threaded scalar	369.2 (rate: 0.3, result: 2.2 GB/sec)
 Write Sequential
single-threaded scalar   994.4 (rate: 1.0, result: 7.6 GB/sec)
multi-threaded scalar	539.0 (rate: 0.3, result: 2.0 GB/sec)
 Stdlib Allocate
single-threaded scalar   379.7 (rate: 1.0, result: 13.4 Mallocs/sec)
multi-threaded scalar	109.6 (rate: 0.3, result: 3.9 Mallocs/sec)
 Stdlib Write
single-threaded scalar   254.5 (rate: 1.0, result: 6.5 GB/sec)
multi-threaded scalar	283.7 (rate: 1.0, result: 6.6 GB/sec)
 Stdlib Copy
single-threaded scalar   517.1 (rate: 1.0, result: 5.6 GB/sec)
multi-threaded scalar	602.6 (rate: 1.1, result: 6.2 GB/sec)

Stream Performance
 Stream Copy
single-threaded scalar   328.0 (rate: 1.0, result: 4.1 GB/sec)
multi-threaded scalar	327.7 (rate: 1.0, result: 4.1 GB/sec)
single-threaded vector   303.4 (rate: 1.0, result: 4.1 GB/sec)
multi-threaded vector	296.0 (rate: 1.0, result: 4.0 GB/sec)
 Stream Scale
single-threaded scalar   338.5 (rate: 1.0, result: 3.9 GB/sec)
multi-threaded scalar	343.5 (rate: 1.0, result: 4.0 GB/sec)
single-threaded vector   317.2 (rate: 1.1, result: 4.3 GB/sec)
multi-threaded vector	292.9 (rate: 1.0, result: 4.0 GB/sec)
 Stream Add
single-threaded scalar   278.8 (rate: 1.0, result: 3.6 GB/sec)
multi-threaded scalar	337.6 (rate: 1.3, result: 4.6 GB/sec)
single-threaded vector   319.2 (rate: 1.2, result: 4.4 GB/sec)
multi-threaded vector	325.1 (rate: 1.3, result: 4.6 GB/sec)
 Stream Triad
single-threaded scalar   279.8 (rate: 1.0, result: 3.6 GB/sec)
multi-threaded scalar	339.9 (rate: 1.3, result: 4.6 GB/sec)
single-threaded vector   274.7 (rate: 1.3, result: 4.7 GB/sec)
multi-threaded vector	256.6 (rate: 1.3, result: 4.6 GB/sec)

Overall Score:   394.2

 

I could probably run the GeekBench at about 3.8GHz too, but OS X is not 100% stable at that speed.

[joe75]

I could probably run the GeekBench at about 3.8GHz too, but OS X is not 100% stable at that speed.

It's not stable on your system or not stable on anything

[bofors]

It's not stable on your system or not stable on anything

 

My results suggest that OS X (or a least the lastest Mifki-Vitaly-Daemon kernel) is not stable much above 3.6 GHz. I say this because I can boot OS X at 3.8 GHz (and probably a little higher) on a BadAxe2, but I have to slow it all the way down to 3.66 GHz for 100% stability. I had thought OS X would have been 100% stable at 3.75 GHz by my BadAxe1 experience.

 

EDIT: I have another BadAxe2 board I am going to be working with tomorrow, so we will see if I get the same results.

 

I am using PowerFractal (set at the highest "Maximum Count" of ~16.7 million) to test for 100% stability. This loads both cores for about half an hour: http://www.versiontracker.com/dyn/moreinfo/macosx/10540

[Alessandro17]

Impressive results, but you are using a very expensive RAM as well.

 

And BTW, do you mind a question?

Any particular reason why you have chosen a Xeon 3060 over a core 2 duo E6600 (for instance)?