ZDNet: New Utlity to Overclock Mac Pro

One comment, based on my MacPro1,1

*Are there any risks involved in overclocking by a great deal?*
As with any system crash, you lose unsaved files. Since HFS+ is a pretty robust filesystem, it recovers well at the next reboot.

I was surprised because I had a clean system and had run DU and Disk Warrior earlier, and had not had any problems (for months, since last fresh install). In fact, I had cloned my system, run DW etc all in getting prepped for 10.5.4.

No other programs were loaded or running except System Profile (I wanted to watch for memory or parity errors). Probably should have restarted the Finder or done a sync command to flush changes to files and directory before running. My directory had some orphan files and needed to be repaired with both Disk Utility and Disk Warrior.

I don't consider OS X filesystem to be all that robust, nor does Disk Utility look to see all the possible problems, which is why you still must use 3rd party tools (MicroMat, Alsoft, etc). It is also possible for the journal to become corrupt and need to turn off, then turn it back on (booted from another drive) to get it back to working.

Of course the best course of action is to restart before using any utility like this, and do your testing on a drive other than your normal boot drive that you use for regular work and use.

So my own feeling is that HFS+ is a little less robust, a little more fragile.

Pete, what you can try is to remove all non-Apple-RAM and try again. Just for a test, if memory is the problem here.

Also, see our FAQ http://www.zdnet.de/enterprise/mac/hardware/0,39038647,39192217-5,00.htm (last question) for troubleshooting memory related problems.

-Christoph

Yep, poor memory sticks... ...I thought better of OWC... ...never mind.

OK. With just the basic 2GB Apple memory allowed me to achieve a stable speed of 3.361GHz before it panicked and asked to be restarted. Shame about the memory!

Nice utility, very impressive!

Christoph - Thanks! I understand much better, and the FAQ is really a good idea.

Psst: I wish the numerical readouts would update as you move the slider. Right now, one moves the slider blindly, stopping to see where you've gotten. And if I close the main window, there seems to be no way to reopen it, short of quitting and re-running the app.

*****
TV Pete - Your inability to speed up your Mac pro may not reflect on OWC's RAM. I also have 6 OWC 1-GB chips in my 2008 2.8GHz Mac Pro, plus the original two from Apple. I easily got my clock up to 3101 MHx (so far), ran mprime for half an hour or so, and tinkered with several applications and fan-rpm experiments vs temps. Neither Console nor System Profiler reported any RAM errors.

I suspect another bottleneck may exist in your system, as there are processor dependencies other than the RAM. Luck of the draw, I guess.

It is a pity, that it don't work for machines that could use a speed bump much more than the latest, and fastest gerneration... like the old G4s... THAT would be nice...! 😉

The only thing I'm a bit worried about is the temperature of the Northbridge...

Under normal conditions the heat is close to a common car cigarette lighter... I don't even wanna think about what it would be under OC conditions...:-(

OK. You've got me fascinated! What else would cause such a problem, that would be affected by the addition of the memory, that would be absent when the memory was removed?

I'm somewhat ignorant about complex modern computer hardware. I'm afraid I cut my eye teeth on main frame and minis, and my last real playing about with hardware was my Amiga. Nowadays, I stick to just adding drives, memory or graphics cards!

Here are some sample data points on operating temps while overclocking.

While running my 2.8GHz Mac Pro at 3101 MHz, and while running mprime, I also set CPU + RAM fans to 900rpm minimum (default 500) using smcFanControl. After temps stabilized, I took a picture of all of +iStat Menus+ temp readouts. They were in ºF:

Ambient = 76
CPUs = 128, 130 (heatsinks at 110)
Northbridge = 113 (normally a bit above the CPUs, e.g. 103-106)
RAM modules = 155, 164, 162, 162, 171, 169, 167, 158
Power supply = 121, 139

At 900rpm minimum, the fans are barely audible, a distant rushing noise. I don't know what a "safe" max temp is for these RAM modules or at what temp the fans would ramp up on heir own, as mine never have. I like smcFanControl for the control it gives me over the cooling vs noise issue.

That's why I'm so anxious...!

Here are my numbers... running nothing but OSX (in a AC cooled Room) NOT Overclocked.

Ambiend = 77
CPUs = 117, 101, 124, 103
Northbridge = 175...!
RAM modules = 157, 149, 140, 137
Power supply = 108, 124

Fans running with 600, 500, 600, 500... as usual.

OK. I've been playing a little more, and I will continue to do so, given some of the curious results I have now encountered!

But, first off, I would like to ask a question no-one seems to have voiced, and that is, how could memory that is sold for use in a machine that could be running at 2.8, 3.0 or 3.2GHz - the 3 Jan 2008 models - when no question is ever asked which speed of machine it is required for, refuse to run at different speeds up to 3.2Ghz? Yes, above 3.2GHz, but why it would cause a problem at lower speeds?

I've been moving the memory modules around the slots, but still 8GB, and one new position allows me to run up to 2.92GHz before it just freezes, another caused the machine to instantly restart at 2.84GHz. Very strange.

My machine seems to run cooler than some other readers' machines; with 79ºF ambient, and the fans running at 7-900rpm, I get CPUs @ 81, Northbridge @ 86, memory modules @ 101 to 124, power supply @ 101, Boot Disk @ 103, and Data Disk @ 90

Or is it the FSB speed increases that causes the problem with the memory?

I ran the „mprime-stress-test“ yesterday... and it turns out, that the 0.4GHz speed bump, which was looking good during „normal“ work, is to much form my system – it quit after about a minute. So I went down a few steps to 0.32GHz OC and it ran for about 15 minutes... before the system quit working. Funny – the CPUs wasn’t used up to 99% during the tests (like they did under normal conditions). They was used about 90%... at least iStatPro said so...

I stoped the testing session right there...

So... before you start cheering up your OC success, put it under the stress-test for about a day, or so, if it still ran you can start the party... not before...! ...and i guess you will find out that the stable... ! OC-Speed for your machine is not quite as impressive as you thought...

A stress-test is just what the manufacturer put their produced hardware through, before they put it into the „good“, „better“, or „best“ basket... the components for the different systems (2.8 / 3.0 / 3.2) are all the same... but their speed ability is not.

It’s the same old story...

The German manufactury „Radiotelefon- und Apparatefabrik Ideal“ produced headphones. After a headphone was finished, they checked it... if it was solid it could be sold... but if it was working better than expected, they put a little „blue dot“ on it and sold it for a higher price. After a while the customers ony bought the models with the „blue dot“ on it – that’s how the brand „Blaupunkt“ was created (Blaupunkt = Blue Dot).

So... if you own a 2.8 MacPro, one... a few... or all of your main components wasn’t able to work in a higher system under stress conditions. Or do you realy believe Intel, Apple, etc. sell their stuff cheaper than they could...?!

The only way to get a system that realy is able to run in a higher class, is that they ran out of a few components during production and have to use better components to fulfill your order. But that will happen 1 out of 10.000 times... at least we all hope we are the lucky ones 😉

If it turns out that EVERY 2.8 machine could run stable overclocked to 3.0... and THEY will run stable at 3.2... and so on, than Apple is holding back the real potential of their systems for an upcoming speed bump.

Matthias, your Northbridge at 175 seems way out of line. Also, that your RAM chips are so different from each other makes me wonder if you are getting the airflow you need. Just now I am coasting along at 2.8GHz, browsing the Internet, with CPU/MEM fans at 600 instead of 500. Ambient is 74º. CPUs are 90º, Northbridge 97º, RAM chips (not the heatsinks) in the mid-130s. Has your Mac gotten dusty?

Further, the slower the fans, the more uneven the internal temps, I have found. There may be air-flow "dead spots" if the fans aren't aggressive enough in forcing air where it must go. So I add a mere 100rpm to the defaults and see a difference without any perceivable increase in noise.

I really do suggest running smcFanControl and boosting the base speed of CPU_MEM fan to, say, 700rpm to see what happens. I have found the temps are remarkable responsive to fan-speed changes, and a little boost seems to go a long way without causing much noise.

http://homepage.mac.com/holtmann/eidac/

Speculation: Makes me wonder if a heatsink on the Northbridge chip is misaligned or something.

TV Pete: "how could memory that is sold for use in a machine that could be running at 2.8, 3.0 or 3.2GHz - the 3 Jan 2008 models - when no question is ever asked which speed of machine it is required for, refuse to run at different speeds up to 3.2Ghz? Yes, above 3.2GHz, but why it would cause a problem at lower speeds?"

Suggestion: It's NOT the RAM but one of its closely-tied controller/manager chips that is freaking out. Data gets in and out of RAM via highly-managed pathways with synchronization considerations, etc.

+„Matthias, your Northbridge at 175 seems way out of line.“+

Exactly...! That’s what makes me nervously...

+„Also, that your RAM chips are so different from each other makes me wonder if you are getting the airflow you need.“+

I think, that’s why they changed the layout of the RAM-Modules in the early 2008 models... turning the upper card „up-side-down“... remember – I have the first (2007) 8-Core-Model (MacPro 2.1)... both riser cards are looking into the same direction.

+„Has your Mac gotten dusty?“+

I don’t think so. First of all, I get the same results the day it arrived, and on the other hand i’m cleaning my system at least twice a year with an air compressor an a vacuum cleaner. I opened the machine about a week or so... erverything was fine.

+„Speculation: Makes me wonder if a heatsink on the Northbridge chip is misaligned or something.“+

That is my fear as well...!
Have someone a link to a picture of the board-layout, so I can figure out the northbidge and check it...?

I ran the stress-test again for about 10 minutes... just watching what happens to the fan speeds an the temperatures. Guess what... while the fan for the Exhaust and the CPUs are speeding up to 1.000 – 1.200 RPM, the heat on my Northbridge sunk to 150...! ...while the whole system is working at its limit... remember, without any work at all, and usual fanspeed, the temperature was higher...!

Strange... isn’t it...?!

I’m now installing smcFanControl... let’s see what happens...

I found this on the net:
http://www.hatsuon.net/macstuff/Northbridge/NorthbridgeTooHot.html

So it looks like a well known problem of the late 2007 models 😟

At least I'm not the only one out there...