art taylor

After a false start due to my own ignorance followed by a butt-saving eBay turnaround, I have finally obtained what I can call moderate success in upgrading the CPU in my Mac Mini from a 1.66GHz Core Duo to a 2.0GHz Core 2 Duo (T7200).

First, the sexy part:

Then, the slightly puzzling (at times) part:

The new configuration is clearly faster than the previous level, which was in turn moderately faster than the earliest version. My Mac Mini now sports a 7200RPM SATA drive and a T7200 2.0GHz Core 2 Duo CPU. I guess I like the number 7200.

I fully expected xbench results similar to those displayed above. It's a straightforward benchmark, and it tests synthetic operations that can only be roughly mapped on the real world.

What vexes me slightly are the application results. Admittedly, I didn't do anything especially time-consuming, such as encoding a DVD, which would spread out processing differences over a longer sample period. I was anxious to start _using_ this thing, but I'll try an encode tonight. However, I ran through a simple series of applications that I generally run that don't have any involvement with the network. Hence, I avoided my 200+ tab Firefox startup, Mail.app, etc.

The installation was rather quick and painless, apart from a moderately unsettling moment when I broke one of the plastic heat-sink retaining pins. However, I improvised something with a couple tie-wraps (man, I'm white trash) that kept the heat sink seated firmly against the CPU.

That previous paragraph is what you're going to be pointing to in, oh, about ten seconds.

I followed the procedures that have been posted to the web and they worked quite nicely. I tried using an expired credit card first (instead of a thick business card), but that loaded up the opposite side with too much pressure. Instead, I used 3x5 cards in pairs and the case was cracked in no time.

Having been through the exercise before, it was short work removing the top chassis and motherboard, and the CPU was exposed, released, replaced, re-Arctic Silvered, and sealed back up in about twenty careful minutes, other than the swearing when the spring-loaded clip flew across the room. The smoke test passed, and I ran through the "benchmarks" mentioned above. I did the same for the pre-7200 RPM drive upgrade, thinking I'd be making a huge performance leap with HD and CPU upgrades, but so many things have changed in the interim that the comparison wouldn't be meaningful.

After running through the above sequence and restarting the machine, I started up my usual coterie of permanently-resident applications -- Firefox (with its multiple windows with 30-70 tabs each), NetNewsWire and its 600 feeds, Mail.app, iCal, iTerm (the application with the weirdest memory usage on earth), and a few other tiny proglets.

For some reason, the fan comes on with a disturbing increase in frequency and volume compared to the old CPU. Go ahead and re-read that paragraph I pointed out for you up above. I'll wait.

For those others of you who have upgraded from a T2300 to a T7200, and didn't bollix up things as I have, are you seeing the same thing?

To confound the straightforward issue of "2.0 = 120% 1.66 == teh h0t" and any 1:1 speed/watt comparisons between Core Duo and Core 2 Duo, and of course my foolish "snap the sprogs on the retaining pin", I upgraded to 10.4.8 last night, which has a reputation for increasing the fan usage on other Intel-based Macs. I noticed this last night while playing bzflag, but the fan level playing the game last night is the baseline level today. However, the computer hasn't shut itself down yet, which is a good sign.

I am interested in other people's experiences with this upgrade, precisely because of the fan issue. I've had to relocate my previously-silent Mini underneath my desk for the time being, as I hate any sort of computing noise, whether it be drive or fan spinning, or just weird scritching or GSM cell phone synching.

If I did the math, which I always end up doing in these situations, I have about $1,204.35 + $174.26 + $337.20 = $1,715.81 in this Mac, which is pretty close to the low-end configuration for a Mac Pro. I know this, and this happens to me time and again due to my sporadic demand for instant gratification.

If you have any clever suggestions for additional benchmarks or tests to run to compare to your own Core Solo/Duo Macs, please leave a comment. Also, if you want my old T2300, send me an email.

[edit]

I got sick of the fan noise, so I opened it back up, popped the heatsink off, and followed the Apple recommended practice of putting on about 900cc of thermal paste. After closing it up and a disconcertingly long restart, the fan is back to normal.

That doesn't mean everything is working properly, just that the fan is quieter.

9:30 PM: Start Parallels and create a pre-configured instance for Windows XP, modifying RAM allotment to 512MB, relocating the data files to an external FireWire Drive, and changing the CDROM from an ISO file to the Mac's SuperDrive DL.

9:35 PM: Put Windows XP disk in drive, start VM.

9:46 PM: Windows XP installer says,"39 minutes left to install," as it copies files, probes hardware (probably especially easy with a virtual machine with constrained virtualised hardware), and tells me how much better life will be with XP. On my last install (on a PC), 39 became 23 which lasted for FOUR HOURS.

10:01 PM: Installer asks for configuration information.

10:02 PM: Installer finishes and reboots.

10:04 PM: I'm logged in and ready to go, looking at what I swear is Palouse country.

I installed the "Parallels Tools", (similar to the VMWare tools and the Virtual PC tools) which make the mousing smoother if nothing else. However, window movement is still a little flaky (I resized the window to 1200x800 or so, so it might be fine at 10x7 or another "natural" resolution.), but everything worked as it should. Surprisingly so, and so, so much faster than Virtual PC did on my G4, which is no surprise.

Again, much to my surprise, it used about 510M RSS and 1.25GB VSS to manage this, much less than I expected. Of course, I'm running IE right now and not Outlook, Office, Visio, and streaming video.

On the other hand, I thought I'd try a simple test tonight, and went to http://www.dslreports.com to try the speed tests, using the SF-based Megapath server. I expected there to be some cost to the virtual bridging in Parallels.

Captive Windows using IE and the latest Java 5 JRE plugin:

dslreports.com speed test result on 2006-04-07 01:42:14 EST:
4887 / 503
Your download speed : 4887 kbps or 610.9 KB/sec.
That is 197% better than an average user on pacbell.net

Your upload speed : 503 kbps or 62.9 KB/sec.
That is 38.5% better than an average user on pacbell.net

Host Mac OS X using Camino 1.0 and the craptacular Apple Java 5 JRE plugin:

dslreports.com speed test result on 2006-04-07 01:54:00 EST:
4896 / 499
Your download speed : 4896 kbps or 612 KB/sec.
That is 197% better than an average user on pacbell.net

Your upload speed : 499 kbps or 62.4 KB/sec.
That is 37.4% better than an average user on pacbell.net

Not much difference, and what difference there was, the Mac came out behind. So much for BSD networking! (Yes, that's a joke.)

I'll continue testing later this weekend with two of the PC games I keep my T40 around for: Scrabble and Age of Kings II: Age of Empires: Conquerors Expansion: Can we add more colons?

Regardless, at $49.95, I can't wait until they get out of Beta so I can buy the Intel Mac version. Based on what I've seen so far, I'd buy it in its present state.

(This is what is known as "foreshadowing". Be prepared for a raft of "Bloody hell this sucks!" posts in the near future.)

Edit: Wow, that was quick. Attempting to eject the Windows XP CD from inside the Windows VM caused a nice kernel panic in OS X. As my pappy used to say, when life gives you poop, make poop juice! So I used the forced reboot as an opportunity to install 10.4.6.

Oh, and I can't upgrade to SP2 because I don't have enough disk space (I went with the default disk size) and I can't find the resizing tool they talk about in the control panel.

Beta. Now I get it.