[sea]

There are reports of the 2,6 conroe version (non-smp) to overclock to 3,5 air cooled. What are possible approaches to overclock a board without bios/efi support other than pll/vcore regulator modifications. Where is the cpu stock frequency read. - sea

[Spliffdaddy]

The 'stock frequency' is the result of the Mulitiplier X Front Side Bus speed. Intel locks the multiplier so it cannot be changed. It's been that way since the Pentium2 era. The only way to raise the core frequency is to increase the FSB speed. Most chipsets support higher-than-stock FSB speeds. You just need to have some way of changing those settings....either through the BIOS setup menu - or from within the OS.

There's typically no difference in the CPU core between the slowest and fastest retail offerings of a given Intel processor line. Provided the core stepping is identical, you should be able to achieve clockspeeds similar to the fastest 'retail' processor - even if you have the slowest version.

I have a PentiumD 805 (dual-core 2.66GHz - 20X multiplier & 133FSB...effective 533MHz data rate, 4X) that I bought from newegg.com for $97. It clocked easily to 3.9GHz using the craptastic factory heatsink.

Gone are the days when Intel tested each core from their yield in order to determine its maximum clockspeed. The yields don't vary enough to matter nowadays - pretty much any processor will exceed the clockspeed of even the fastest 'retail' version available - again, assuming the core stepping is identical.

How do you overclock a Mac? No idea. Boot into Windows and use one of the overclocking apps Though you may not be able to tweak the core voltage, RAM, and data bus speeds / multipliers.

[Lateralus]

Originally Posted by Spliffdaddy

How do you overclock a Mac? No idea. Boot into Windows and use one of the overclocking apps

Bingo.

I suspect overclocking Intel Macs will be difficult until EFI makes it way into the PC market and programmers start developing Windows applications to tweak hardware through EFI.

[mduell]

With the multiplier locked on all of the Core chips except for the Core 2 Extreme, the only avenue for overclocking the Xeon is through the FSB.
With an Intel chipset and firmware from Apple, I see zero potential for overclocking the FSB.

The software overclocking apps for Windows are for motherboards where the firmware/BIOS supports overclocking the FSB and the manufacturer encourages it (trying to draw in the OCing crowd).

[iaw4]

I think many of us would be happy to do even basic overclocking. according to general reviews, virtually every 2.66GHz Conroe processor seems to do 3.4GHz with standard cooling and a lot of effort, and 3.0GHz without any effort, mostly just a FSB frequency increase. I presume that EFI firmware does not allow changes in the FSB---is this correct?

one general question---is there a technical reason why the FSB must be increased at BIOS (ahem EFI) time, and why no after-OSX running product can do this?

I do not own an Apple Mac Pro, but was thinking about purchasing one. Yet I find it pretty difficult to justify. I will probably buy yet another linux machine (already own some) that costs half, and runs only one dual-core processor at 3.2GHz. Most of my tasks (e.g. latex) are not parallelizable, so my guess is that this will be a faster day-to-day machine. I would have paid twice the price to get all the nice apple goodies, the software and hardware. That is, I want a system that can at least perform similarly for my single-threded applications at twice the price. For Mac Pros, without any basic OC, I can't see such performance, until I go to three times the price, and this is just too much for me.

/iaw

[seanc]

Originally Posted by Spliffdaddy

The 'stock frequency' is the result of the Mulitiplier X Front Side Bus speed. Intel locks the multiplier so it cannot be changed. It's been that way since the Pentium2 era. The only way to raise the core frequency is to increase the FSB speed. Most chipsets support higher-than-stock FSB speeds. You just need to have some way of changing those settings....either through the BIOS setup menu - or from within the OS.

There's typically no difference in the CPU core between the slowest and fastest retail offerings of a given Intel processor line. Provided the core stepping is identical, you should be able to achieve clockspeeds similar to the fastest 'retail' processor - even if you have the slowest version.

I have a PentiumD 805 (dual-core 2.66GHz - 20X multiplier & 133FSB...effective 533MHz data rate, 4X) that I bought from newegg.com for $97. It clocked easily to 3.9GHz using the craptastic factory heatsink.

Gone are the days when Intel tested each core from their yield in order to determine its maximum clockspeed. The yields don't vary enough to matter nowadays - pretty much any processor will exceed the clockspeed of even the fastest 'retail' version available - again, assuming the core stepping is identical.

How do you overclock a Mac? No idea. Boot into Windows and use one of the overclocking apps Though you may not be able to tweak the core voltage, RAM, and data bus speeds / multipliers.

I have that CPU too

I've left mine at 2.66ghz though since I only have the Intel heatsink and didn't think it would be able to cope with anything higher.

Here's an amazing article: A 4.1 GHz Dual Core at �79.95 - Can it be True? | Tom's Hardware UK and Ireland

[Lateralus]

I see no reason for leaving it at 2.66. Intel's Pentium D heatsinks were excellent. I wouldn't push for 4.1GHz with it, but 3.3~ would be fine.

[seanc]

Hmm, maybe I'll give it a try later. Their review did say it crashed at 3.33 at full load with the Intel heatsink though.

[Lateralus]

They may not have increased the voltage. Regardless, it never hurts to tinker though.

After a bit of playing around, I had my Pentium D 930 running at 3.93GHz with the stock Intel heatsink and it was solid.

The 8xx series doesn't overclock as well as the 9xx series since the 8xx are still built at 90nm so they produce more heat than the 65nm 9xx chips. But they still overclock very well with a little effort.

[seanc]

Off I go to try and gain 700mhz then, I'll try not to derail this thread any further.

Edit: Seems to work fine, first time the bios clocked it down to 2.68 and informed me that it had failed to boot, then I set the settings to PCIE - ASYNC and it's working fine.... for now. I should really start my own thread, but unfortunately this isn't PCNN

[iaw4]

I hope we are talking about the same thing---not overclocking pentiums per se, especially old netburst architecture ones (rather than new core architecture ones) that I believe were never ever used by Apple. yes, one can do a lot of crazy stuff with OCing.

traditionally, server chips overclock more than desktop chips. however, xeon 5100s (woodcrests) tend not to be overclocked, as they have found very little use on the desktop, except in apple's mac pro case. all conroe chips seem to overclock easily by 20% with the standard intel or any other cooling.

So, I am simply interested in learning whether it is possible to crank up the FSB speed of the Mac Pro with Intel Xeon Woodcrest, either through EFI or through any other reasonable software only method. Does anybody know?

/iaw

[seanc]

Well until anyone works out how to manipulate EFI by getting into it's settings or using/creating a program then it doesn't seem possible.

[Strezi]

Hi there,

I just started to research the topic as a new iMac owner. It's just to tempting to let my processor to idle at 2.16 GHz

Succesfully downloaded the intel EFI dev package and reading the specs at the moment. My initial impression is that EFI might not have control over the FSB speed. More likely is down to the apple firmware which can not be overriden from th EFI level. Anyway, I'm really keen on making some progress...
Perhaps a firmware update can solve this, but right now it's too early to say.

[christoh]

We've just released an overclocking tool for the Intel Mac Pro under Mac OS X 10.5.

Regards,

Christoph

[Lateralus]

Wow. Very nice.

[bearcatrp]

I'll be giving this a shot tonight with my 2.0Ghz macpro with 8gb ram. Hope this works as replacement cpu's are still expensive.

[Tenacious Dyl]

I gave it a whirl last night with my 4x2.0 Mac Pro, and safely got each core to 2.256 GHz with no stability or ram erroring issues. xBench returned the same results, as the timing on the comp is changed, but Cinebench returned slightly better results on the single and multiple cpu tests. The gain a score of 7,100 for all processors compared to just over 7,400 for all processors, not really worth the worry of more power draw or heat.

As it stands, I very rarely have all cores on full load, and the few times a task would do that (encoding something for example) my hard-drives usually just lag behind slightly and I still have some of the cores left idle. Thus, over clocking really doesn't do much if all I can get is an extra .256 GHz per core.

[bearcatrp]

Works just like it says. Went from 2.0 to 2.340 without errors. Above that the errors started. Encoding from mpeg2 to dv went from 16:50 to 15:00 on a 1.99gb file. Not bad for free. Been wanting to buy 4 core chips to replace these dual core but there still expensive. This will do for a little bit.

[bearcatrp]

Can you folks who do overclock there mac pro give a detail description of your mac pro? Curious if different ram configurations help or not. Mine is a mac pro 2.0Ghz with 8gb ram (8X1GB) modules.Tnks

PS. If you can, do a comparison doing something that will tax the cpu's and see what difference it makes.

[Tenacious Dyl]

I have 5 GBs of 667 (A pair of 2 GBs positioned to take advantage of the dual-channel, and the stock pair of 512 MB sticks)

As posted above, I ran only the processor intensive task of Cinebench (R10) and had some gain but nothing too impressive at a stable overclock from 2.0 GHz per core to 2.256 GHz per core. Hope this helps!

[bearcatrp]

With you only getting up to 2.256, am thinking the more ram, the higher you can overclock but this is just speculation. I have a 2.0ghz like you but can get a bit higher with 8gb ram. If someone with more than 8gb ram can post there results, might help confirm this.

[Drakino]

Make sure to pay attention to this part of the article about overclocking. It seems from their experience, the system time in OS X will begin to drift due to the overclock. So if you have time sensitive apps, do not overclock.

Also, putting the machine to sleep then waking it will undo the overclock, and require a manual unload of the kernel module used by the app to do the overlocking.

Overall looks like a good utility to have and it comes with a very good writeup from the same ZD folks that wrote it. Amazing how even the hacking tools for Macs are polished :-)

[reader50]

There have been some questions and a Report to the staff, regarding if christoh's post was spam.

christoh, aka Christoph H. Hochstaetter, is the author of the linked article on ZDnet. Lateralus spotted this right away, and waived the 6-month rule in this case. Not all of us noticed the clues, and christoh was mistakenly banned for about a day, then reinstated.

The post is not spam.

If I owned a Mac Pro, I'd offer something on-topic to the thread in this paragraph. Once they release an overclock tool for my Quad G5, I'll be sure to offer more useful commentary.

[mkerr64]

If one uses this program to over clock what frequency should one do?
and also does this void applecare?

[mduell]

You should do whatever you can get away with and remain stable.
Yes, it would void your warranty.

[Lateralus]

Originally Posted by mduell 

Yes, it would void your warranty.

How is a software overclock going to void the warranty?

No, it wont mkerr. I suppose it could if you call up Apple to tell them that you're doing it. But I don't see why anybody would do that.

[UnixMac]

Ran my 3.0 Mac Pro Quad Core to 3,203MHz and my "Geekbench" score went from 388 to 389.. not much of an improvement but then maybe it's not taxing enough. What is the benchmark du jur these days for taxing the Mac?

I'll try to download

[mr. burns]

I'll give this a shot and see how well it improves my Photoshop test time on my August 2006 Mac Pro.

[UnixMac]

also computer began to act odd after installing it.. not sure if it's related. The CD-Drive would not open/eject and when the computer was shut down it would almost act like it's in sleep with the power light blinking and the two monitor power lights steady on.

Again, not sure if it's the software or some other issue I'm having..

[angelmb]

Are your Mac Pro that 'slow' yet to even ponder about overcloking them?

low sugar Apple Juice composition (1% or less)
33% sarcasm
33% rhetorical question
33% aqua

[mr. burns]

Who can pass up a free speed boost? :]

[UnixMac]

Originally Posted by mr. burns 

Who can pass up a free speed boost? :]

I think I will until they make a version that works on the initial production Mac Pro's..

[mr. burns]

"ZDNet uses three Mac Pros as test machines. One comes from the first Intel/Mac Pro generation (Mac Pro 1.1) with 65-nanometer processors and 1333-MHz front side bus."

[revMedia]

So far it seems like the speed gains are minimal. I would have thought the Xeon procs mixed with the massive heatsinks on the chips would have allowed for some better gains.

[Simon]

Originally Posted by revMedia 

So far it seems like the speed gains are minimal. I would have thought the Xeon procs mixed with the massive heatsinks on the chips would have allowed for some better gains.

The limit to overclocking these Xeons isn't given by CPU heat dissipation. It's actually the memory bus. More specifically, it's the AMB on the MP's FB-DIMMs.

[angelmb]

Originally Posted by Simon 

The limit to overclocking these Xeons isn't given by CPU heat dissipation. It's actually the memory bus. More specifically, it's the AMB on the MP's FB-DIMMs.

What does AMB stand for?

[Simon]

Originally Posted by angelmb 

What does AMB stand for?

Advanced Memory Buffer

From http://en.wikipedia.org/wiki/Fully_Buffered_DIMM
Fully Buffered DIMM architecture introduces an Advanced Memory Buffer (AMB) between the memory controller and the memory module. Unlike the parallel bus architecture of traditional DRAMs, an FB-DIMM has a serial interface between the memory controller and the AMB. This enables an increase to the width of the memory without increasing the pin count of the memory controller beyond a feasible level. With this architecture, the memory controller does not write to the memory module directly, rather it is done via the AMB. The AMB can thus compensate for signal deterioration by buffering and resending the signal. In addition, the AMB can also offer error correction, without posing any overhead on the processor or the memory controller. It can also use the Bit Lane Failover Correction feature to identify bad data paths and remove them from operation, which dramatically reduces command/address errors. Also, since reads and writes are buffered, they can be done in parallel by the memory controller. This allows simpler interconnects, more memory bandwidth, and (in theory) hardware-agnostic memory controller chips (such as DDR2 and DDR3) which can be used interchangeably. The downside to this approach is that it introduces latency to the memory request.

The AMB has to do a lot of work to do. You can check yourself that it gets very hot in the process. To overclock these MPs you have to increase the FSB clock because of the Xeon's fixed multiplier. That in turn means a whole lot more load on the AMB. And as it turns out, in this whole chain the AMB is the weakest link.

This is also the reason why overclocking limits on these MPs depend on the memory configuration. If you overclock with only two DIMMs installed, you find a max stable setting, and you then decide to add the rest of your DIMMs, you might end up having to go to a lower setting to get stable operation.

[mduell]

I thought the FSB speed was independent of the memory bus speed on the Intel 5000/5400 chipsets.

[revMedia]

Originally Posted by mduell 

I thought the FSB speed was independent of the memory bus speed on the Intel 5000/5400 chipsets.

I was under the same impression.

[angelmb]

Originally Posted by Simon 

Advanced Memory Buffer

Thanks, AMB happens to be my name+surnames first letters so it was mostly a curiosity… it is also my OS X account's name.

in this whole chain the AMB is the weakest link.

d'oh !!