[HowTo] Building an overclocked Core2Duo box for OSx86.

[bofors]

TPM which I have noticed is talked about in several places in these forums but only have 4SATA ports as aposed to 8 on the non TPM boards as mentioned several times these boards also support EFI so this is my main choice but compatability with Apples EFI could still be an issue...

 

I do not think the TPM version of the Bad Axe was actually released and the regular version supports EFI (all modern Intel boards do).

 

If anyone here who has this board do you have TPM or non TPM and what difference does this make to 10.4.x and more importantly 10.5.x

 

I really doubt that TPM will ever become a requirement for OSx86 here, but some people have been working with flashing TPM chips themselves which could allow the use of unhacked OS X. But this is probably a long way off at best.

 

Also am I better of purchasing the conroe chip or the allendale which is considerably cheaper or will 10.5.x require the conroe chip in its cpu instructions

 

Motherboard -D975XBX

CPU -E6600 / E6400

Memory -2x 1GB DDR 667

 

Conroe and allendale use the same instructions set, there is no 10.5 issue here. Bad Axe board are said to overclock the E6600 better than the E6400, so if maximum overclocking and CPU performance is a priority go with the E6600.

 

Being based in NZ prices may be a little higher than what others are paying in US / UK it may be just cheaper for me to bite the bullet and buy Apple

 

There is nothing wrong with buying Apple when they provide what you want at a fair price. The Mac Pro is still on my menu of options too, it is a sweet machine at an awesome price (I had expected it to cost about $1000 more).

 

But now is a terrible time to buy RAM and prices are likely to remain absurd for months (after the holiday season).

 

Regarding NZ, you should talk to Simon (sbeehre), he is a Kiwi too. He had Kiko (an Aussie) buy his Bad Axe and ship it to NZ.

[aliquis]

You people who order such expensive machines just (?) to install OS X on them, you are so confident future versions will work? I'm thinking about ordering an iMac just for Mac OS but I already have a decent Athlon64 3000+ with 1.5GB ram so I would skip it if I didn't had to.

 

Buying a new "random PC" would also be an option if Mac OS X ran perfectly but it would suck to just notice later that it isn't supported longer, because then I would be looking for a mac again :/

[mifki]

Well im happy with my conroe mac, and i also got a new iMac, so im happy.

 

I agree with wat bofors said about the price and evrything. In the end i think its just a matter of preference, if you want a cheap PC get a hacki if you want o spen dheaps get a MAC

[bofors]

People thinking about getting a Bad Axe should know that the Bad Axe 2 will be released this month and perhaps in a few days:

 

http://www.vr-zone.com/?i=4111

[joe75]

I'm a firm believer that a hack is not a mac. If people have a hard on for Apple they should get a real mac. No matter how compatible x86 equipment is, there will always be something to hack or screw with. It may be fun or challenging to play with but if all you want is a mac, just get one

[joe75]

Cooling needs :http://www.frozencpu.com/

 

My cooling : http://www.frozencpu.com/products/1418/cpu...ml?tl=g40c14s48

 

: http://www.frozencpu.com/products/4064/vid....html?tl=g40c16

 

I put on a 50 on the Jing Ting and have only one 92, Enermax high output on the Noctua

[Alessandro17]

I'm a firm believer that a hack is not a mac. If people have a hard on for Apple they should get a real mac. No matter how compatible x86 equipment is, there will always be something to hack or screw with. It may be fun or challenging to play with but if all you want is a mac, just get one

 

Maybe some people simply can't get the kind of Mac they would like, like a Merom Macbook or a mid-range Desktop.

The Mac Pro might be good value for money, but you need to upgrade it (especially the RAM-very expensive) and besides 2 Xeons CPUs might not be the best choice for a desktop/workstation.

[martin1976]

I'm a firm believer that a hack is not a mac. If people have a hard on for Apple they should get a real mac. No matter how compatible x86 equipment is, there will always be something to hack or screw with. It may be fun or challenging to play with but if all you want is a mac, just get one

Well, the way I particularly see it is, I could buy any mac I want. (Really, I could). But I'm such a geek that I couldn't refuse the idea of actually putting together my own. Yeah, I do like macs, but what I really like is OS X and not mac hardware. Don't take me wrong, macs are aesthetically perfect, but honestly I get a "hard on" from building my own {censored} rather than buying whatever apple decides I should use. My

[aberracus]

im sure i have paid less than 1000 for my ihack (wich is just my new computer, i sell the one i had before so invested like 200 dollars really, i keep renewing my hardware almost bimonthly , never with the top, just the second one. (or third )

 

Anyway look at my signature the xbench comparision with a real Mac Pro, my machine is faster and cheaper i love apple and macs but here in my country the tag price got a 50% extra, so theres no way i coudl buy a real mac like my ihack.

[joe75]

Those are all valid points except maybe the "aesthetically perfect" part. I will only build my own, but I would never build just for mac Like I said, its fun to play with. Apple hardware is limited and overpriced, but I see the minis as a decent value if you have to own Apple. I think the future will hold more hacks and less compatibility once Leopard arrives. I also believe Apple should look at this community and realize they should start to focus on being more x86 compatible to get into the bigger market share they want.

[bofors]

Anyway look at my signature the xbench comparision with a real Mac Pro, my machine is faster and cheaper...

 

One of the major reasons this is true is again FB-DIMMs. I have already established that FB-DIMMs are much more expensive than DDR2, twice the price last time I checked. But now I have noticed that the performance of FB-DIMMs even with quad channel architecture is really bad (at least in benchmarks).

 

Here are some references:

 

http://arstechnica.com/news.ars/post/20040416-3668.html

http://www.maconintel.com/news.php?article=209

 

In short, the added latency of FB-DIMMs kills performance. AnandTech measured this here:

 

http://www.anandtech.com/mac/showdoc.aspx?i=2816&p=11

 

 

So, even CAS 5 DDR2-667 has over 50% more bandwidth than quad channel FB-DIMMs (at about 25% lower latency).

 

The ability to overclock DDR2 RAM makes the performance difference even more significant in the hackintosh systems that I am considering.

[aberracus]

Excelent article Bofors.. so then why apple went with FBDimms? and are they requisite for woodcrest?

[bofors]

Yes, the Woodcrest chipset requires FB-DIMMs. Intel is responsible for this, Apple had no choice.

 

Given this situation, a machine built with the upcoming Q6600 (Kentsfield, quad-core), which is supposed to be released in January, might really kick the Mac Pro in price performance (which is really amazing, because at $2500 the Mac Pro is still a sweet deal).

[joe75]

Latency is only going to get worse With DDR3 CS6 or 7 will be the norm. AMD is going to try and combat this by adding a mid bridge between the northbridge and cpu, it will process ram information and boost it through the system. Intel has talked about adding a on chip controller similar to what AMD has now, to speed up latency. Benchmarks and actual latency speeds aside, bandwidth is what matters. You win some and you lose some

[bofors]

So, even CAS 5 DDR2-667 has over 50% more bandwidth than quad channel FB-DIMMs (at about 25% lower latency).

 

The bandwidth of quad channel FB-DIMMs (CAS 4 533 MHz) is compared to dual channel CAS 5 DDR2-667 on Linux here:

 

http://www.phoronix.com/scan.php?page=arti...m=495&num=5

 

The results are not nearly as bad as AnandTech's, quad channel FB-DIMMs are about equal in bandwidth to the dual channel DDR2. Part of the improvement is likely do to AnandTech's use of CAS 5 FB-DIMMs (but I am not sure if they did).

 

Either way, quad channel FB-DIMMs is no way near providing twice the bandwidth of dual channel DDR2. Until this situation is resolved, it certainly appears that one is much better off going with a Kenstfield for a quad-core machine.

[bofors]

This test compares dual channel CAS 5 667 MHz FB-DIMMs with dual channel CAS 5 DDR2-800 on Windows:

 

 

http://www.gamepc.com/labs/view_content.as...6800&page=7

 

So, even if FB-DIMM bandwidth doubled here with quad channel one would still only match DDR2 performance. However at best, by Phoronix's data, we would only expect to see only about 35% increase when moving from dual channel to quad channel with FB-DIMMs:

 

http://www.phoronix.com/scan.php?page=arti...m=495&num=9

[amiloboy]

Bofors, just thought I'd give some quick info for you.

 

I have a Gigabyte GA-965P-DS3 board, with an Allendale E6300, and Corsair XMS2 DDR2-675 RAM. I have managed to get the FSB to a max of 510 Mhz on this motherboard. The only thing causing instablility was the RAM. I highly recommend this board, it is overclocking amazingly!

 

Enjoy your building

 

 

 

Does your 965P motherboard work with OSX 10.4.6 ?

 

Thanks

[bofors]

Considering how expensive memory is in general (absurd at the moment and probably not coming back down to "pre-Conroe" prices until January), it is important to know exactly what you need (and as what this report clearly shows, what you don't need).

 

X-bitLabs.com has a series of articles on Conroe Memory titled "Choosing the Right Memory for Core 2 Duo Platform": http://www.xbitlabs.com/articles/memory/

 

I just went through the first one and found some useful information: http://www.xbitlabs.com/articles/memory/di...mory-guide.html

 

I am going to just cut and past the highlights, but note that in most of the real world benchmarks the most expensive memory was less than 5% faster than the slowest:

 

 

http://www.xbitlabs.com/articles/memory/di...ry-guide_4.html

 

As you can see, memory types with different theoretical bandwidths do not differ much in practice. For example, there is a 100% difference in theoretical bandwidth between DDR2-533 and DDR2-1067 whereas the difference between the practical results obtained with those memory types is 17% at maximum.

 

This poor performance of fast DDR2 SDRAM is due to the architecture of Core 2 Duo systems in which memory is connected to the CPU via the chipset and two sequential buses. In this design it is not the bandwidth of dual-channel high-frequency memory that becomes the bottleneck, but the Quad Pumped Bus that connects the CPU with the chipset's North Bridge. Its maximum theoretical bandwidth is 8.5GB/s in Core 2 Duo systems, which only equals the bandwidth of dual-channel DDR2-533 SDRAM. That's why we don't see a really big performance growth if we use memory faster than DDR2-533.

 

It seems it doesn't make any sense to use memory faster than DDR2-533 on the Core 2 Duo platform. This is not quite so. Memory access latency decreases along with frequency, which can be seen in practical tests.

 

 

 

http://www.xbitlabs.com/articles/memory/di...ry-guide_5.html

 

 

http://www.xbitlabs.com/articles/memory/di...ry-guide_7.html

 

Here, DDR2-1067 SDRAM enjoys a nearly 50% advantage over DDR2-533 SDRAM on the Core 2 Duo platform. Yet you should keep it in mind that this is only a special case. In a majority of applications the positive effect from faster memory is negated by the limited bandwidth of the front-side bus that connects the chipset's North Bridge with the CPU.

...

Using higher-frequency memory modules makes the system costlier, yet doesn't lead to any significant performance increases. This is true, in part: memory faster than DDR2-533 can only provide a maximum of 5% performance growth in a majority of widespread applications. The problem is in the front-side bus which is only clocked at 266MHz as yet.

 

But it doesn't mean fast memory is completely useless for owners of Core 2 Duo systems.

...

We only wanted to see what effect on the overall system performance the memory subsystem parameters would have if the FSB was overclocked. We set the FSB frequency at a rather typical value of 400MHz (50% above the default). At that frequency the FSB bandwidth grows from 8.5GB/s to 12.8GB/s. In theory, this should make the use of dual-channel DDR2-800 worthwhile.

 

 

http://www.xbitlabs.com/articles/memory/di...ry-guide_8.html

 

The measurements of the CPU-memory bandwidth agree with the results we've got in the previous section. When the memory frequency is higher than 800MHz, the bandwidth indeed stops to grow further which coincides with the maximum bandwidth of the FSB at 400MHz.

 

 

The memory latency tests confirm the point: the latency almost stops to decrease after an 800MHz memory frequency. So, it is clear that DDR2-1000 SDRAM cannot bring about a significant performance gain in comparison with DDR2-800 SDRAM on Core 2 Duo platforms even when the FSB is overclocked to 400MHz.

 

 

http://www.xbitlabs.com/articles/memory/di...ry-guide_9.html

 

 

http://www.xbitlabs.com/articles/memory/di...y-guide_11.html

 

The memory bandwidth on Core 2 Duo systems is limited not by the characteristics of DDR2 SDRAM modules but by the bandwidth of the bus that connects the CPU with the chipset's North Bridge. This is why your changing the memory frequency or timings is going to have a small effect on performance of a non-overclocked Core 2 Duo system (but the frequency influences performance somewhat more than the timings do).

 

More interesting are the results of the overclocked platform. In this case, there is more sense in using fast memory and the optimal memory frequency divisor is 1:1 (FSB:DRAM) as has been shown in our tests

 

http://www.xbitlabs.com/articles/memory/di...y-guide_12.html

[bofors]

Now, regarding the above data, can anyone explain why the absolute value of the SuperPi and WinRAR scorces DECREASES when overclocking the FSB from 266 (blue) to 400 (orange)???

[bofors]

bofors, about the mem. i'm running mine @ 800 4-4-4-12 perfectly stable and it comes 667 stock.

 

Yes, by XBit Labs this again seems to be way to go. Use "value" CAS 4 DDR2-667 and then overclock to 800 or so 1 : 1 with the FSB by raising the vDIMM if necessary.

 

Again, the Consair ValueSelect RAM I already have appears to be exactly what I want (and I bought 4 GB of it in January for much less than today's market price). I now have Evercool copper heatspeaders for that RAM and will be getting Artic Silver adhesive to attach them (instead of using the supplied 3M thermal tape).

[greenman100]

bofors, your analysis is mostly correct, except that downward dividers do not yet exist, that is, you cannot run memory slower than FSB.

 

This leaves us with two options, if high processor clocks are needed

 

Cheap Conroe, expensive RAM (e6300 + $230 corsair) = $400, ~3.4ghz@475mhz FSB

Expensive conroe, cheap RAM (e6600 + $180 valueRAM) = $450, ~ 3.4ghz @ 350mhz FSB

 

it is evident ot me at least that going from 350 --> 475 is a nice jump, speedwise

 

 

then again, of course, this only matters if youre pushing for high clocks...

[bofors]

Now, regarding the above data, can anyone explain why the absolute value of the SuperPi and WinRAR scorces DECREASES when overclocking the FSB from 266 (blue) to 400 (orange)???

 

I am know guessing that XBit lowered the multiplier in attempt to keep the CPU running at the same speed (i.e. they were trying to isolate the affect of raising the FSB from 266 to 400 alone, not with the usual increase of CPU clockspeed).

 

EDIT: On closer examination of the XBit article, I am doubting that is the explanation because they make no mention of it. Rather they seem to indicate that the 400 MHz overclocked-data is running at the same multiplier of 7 (2.8 GHz): http://www.xbitlabs.com/articles/memory/di...ry-guide_8.html

 

A couple other people noticed this discrepancy in the articles discussion thread, but no explanation is proposed: http://www.xbitlabs.com/discussion/3321.html

[bofors]

Cheap Conroe, expensive RAM (e6300 + $230 corsair) = $400, ~3.4ghz@475mhz FSB

Expensive conroe, cheap RAM (e6600 + $180 valueRAM) = $450, ~ 3.4ghz @ 350mhz FSB

 

Yes, I have noticed this sort of "balance" before. A key factor in the economics is the amount of RAM to be used. Originally, I was planning on using 4 GB, but now I am thinking about trying to work with 1 GB instead. This would mean using expensive 2 X 512 MB RAM but saving $100 on the "cheap" Conroe system.

 

I am also starting to think more about that the real Allendale, the E4300. It will be like the Penitum D 805, a bargain for overclocking. It should be about $140 and given it's specs: 1.8 GHz and 800 MHz FSB, we determine that it has a mulitplier of 9, just like the E6600. So cheap memory should work with it too: http://www.dailytech.com/article.aspx?newsid=3372

 

On the other end, Kentsfield effectively doubles the core to RAM ratio. Even at say $550 for Q6600, it would certainly make sense if one needed to work with 8 GB given the gross expense of 4 x 2GB (~$1100).

 

Unfortunately, nether the Q6600 nor the E4300 are expected to ship before January at the earliest.

 

EDIT: It is now being reported that the Q6600 will go for $851: http://www.reghardware.co.uk/2006/10/18/in...d-core_pricing/ That is way too much for what is essentially two E6600 chips. One might as well just buy the unlocked QX6700 at $999 instead.

[bofors]

Xbit Labs has another article in there Conroe memory series that is worth noting, this one compares DDR2-800 (which probably isn't necessary unless one is extreme-overclocking an E6300): http://www.xbitlabs.com/articles/memory/di...y/ddr2-800.html

 

Xbit compares that overclocking capacity of the following 2 x 1GB kits:

 

Corsair TWIN2X2084-6400C4

Crucial Ballistix BL2KIT12864AA804

GeIL GX22GB6400UDC

G.Skill F2-6400PHU2-2GBHZ

Kingston HyperX KHX6400D2LLK2/2G

Mushkin 996523 XP2-6400 DDR2

 

In short, only the Consair and Mushkin overclock well, the other fail to get above 1000 MHz.

 

 

http://www.xbitlabs.com/articles/memory/di...ddr2-800_2.html

 

 

http://www.xbitlabs.com/articles/memory/di...ddr2-800_7.html

 

But at some $380 the Mushkin is much more expensive than the Consair for $280

[bofors]

AnandTech has posted a new DDR2-667 RAM overclocking article:

 

http://anandtech.com/memory/showdoc.aspx?i=2862&p=2

 

Here are the results:

 

 

The results are somewhat surprising with the A-DATA DDR2-667 memory reaching DDR2-970 at 5-5-5-18 timings. This represents a 45% increase in memory speed that only required an increase in voltage to 2.2V.

 

http://www.newegg.com/Product/Product.asp?...N82E16820211061

 

The Transcend JetRam DDR2-667 and G.Skill DDR2-667 were both able to reach DDR2-910. This represents a 36% overclock and is what we expected out of our memory choices on average. The Transcend JetRam memory was listed at $76 per 1GB module at ZipZoomFly when we purchased ours, reminding us of the "good old days". We did a last minute check today and it is sold out now.

 

http://www.zipzoomfly.com/jsp/ProductDetai...oductCode=83410

 

There are never any guarantees in overclocking and our Corsair Value Ram represents this premise. We could not get the memory to run at DDR2-800 at the 4:5 ratio and could only reach DDR2-750 in our overclocking results. This required a change to 5-6-5-18 timings that resulted in performance just slightly better than our stock DDR2-800 settings.

 

This seems to contradict what AnandTech previously reported here: http://www.anandtech.com/mb/showdoc.aspx?i=2797&p=17 However, do to comments on NewEgg, I suspect that Corsair may be now using lower grade chips in the ValueSelect DDR-667 (I have been running 4-4-4-12 at 667MHz at 1.8V on the 4GB I bought in January with no problems): http://www.newegg.com/Product/CustratingRe...N82E16820145098

 

EDIT: AnandTech's data on the Corsair Value Select seems to be wrong, I have overclocked mine (after attaching heatspreaders) to 934 MHz and I can easily run 800 MHz at 4-4-4-12.

 

Overall, our DDR2-667 memory selections, sans the Corsair Value Ram, performed admirably and scored within a single percent of our much more expensive DDR2-800 memory at stock settings.