SATech.com is offering a 5% discount to diglloyd.com readers on Mac Pro memory, or anything on their web site. See All About Mac Pro Memory for details.
with 2% reward and free expedited shipping
World’s fastest wide angle zoom for full frame DSLRs!
- Discount on Mac Pro memory 03/29/07
- New Canon EF 16-35 f/2.8L II tested 03/29/07
- Mac Pro article updated, Mac Pro memory 03/28/07
- Memory for sale: 4 X 512MB FB-DIMM for Mac Pro 03/27/07
- Canon rebate scam— Caveat Emptor with a modern twist 03/26/07
- SanDisk Extreme III 16GB for US$238 03/26/07
- Stable memory 03/23/07
- Dream Machine 03/21/07
- Exercise and the brain 03/20/07
- Inkjet paper for everyday printing 03/18/07
- SeriTek 2SE2-E two-port SATA card 03/17/07
- Six hard drives in a Mac pro 03/17/07
- 8-core Mac Pro rumored soon 03/13/07
- Dead dinosaurs, dead cats, and dead yoga instructors 03/13/07
- IbexWear clothing 03/09/07
- Sandbox 03/09/07
- Followup on Mac Pro/X1900video problems 03/08/07
- ECC memory errors followup 03/07/07
- ECC memory errors again 03/07/07
- How much memory to buy 03/07/07
- ECC memory errors 03/06/07
- Mac Pro memory performance 03/05/07
- Severe FireWire performance bug on Mac Pro 03/04/07
- Mac Pro video problems (ATI X1900 video card) update 03/03/07
- “Goodbye Nikon?” article updated 03/03/07
- Self-cleaning sensors 03/03/07
- Evolution of digital camera features 03/03/07
- Mac Pro memory temperature 03/03/07
- Mac Pro memory—stress testing 03/02/07
- Mac Pro memory—installation 03/02/07
- Mac Pro memory comments 03/02/07
- Tricking Apple Mail into storing its data elsewhere 03/02/07
- Backup strategy 03/02/07
- Memory stress-testing OK 03/02/07
- Mac Pro power usage 03/02/07
- Mac Pro video problems (ATI X1900 video card) 03/01/07
Favorites: 15/2.8 Distagon • 25/2.8 Distagon • 35/1.4 Distagon • 50/2 Makro • 135/2 APO-Sonnar
I’ve received the new Canon EF 16-35 f/2.8L II lens. I’ve not yet had an opportunity to shoot it enough to give a detailed performance assessment, but use so far is suggestive of the following:
- Flare control is outstanding. Outstanding!
- Images show superior sharpness compared to the original 16-35, at least at the focal lengths I’ve tried so far.
- Consistent with the Canon “Wide” MTF chart, at 16mm corners beyond a 39mm image circle are mush wide open, and f/8 or f/11 is required to produce acceptable results. (43.2mm is the absolute corner of a 36mm X 24mm frame). However, image sharpness over the rest of the frame is very impressive even wide open, noticeably superior to my first-generation 16-35 f/2.8L. UPDATE: See the April 4 blog entry. Further shooting suggests that the improved sharpness might be due to focus error with the original 16-35, making an accurate test with the “II” problematic.
- Chromatic aberration (red/cyan) color fringing remains prominent at the 16mm end (on the EOS 5D at least), repudiating Canon’s press release that “Two UD lens elements virtually eliminate chromatic aberration”.
It seems likely that this new 16-35 will be a stellar performer wide open on the new EOS 1D Mark III. With its cropped-frame sensor (28.1 X 18.7mm = 1.3X), all the mushy areas will be well outside the sensor’s 33.7mm image circle. A terrific 21-45mm zoom in effect.
I’m selling the Apple OEM 4 X 512MB modules (2GB total) that came with my recently-purchased Mac Pro. These modules cost $600 (plus tax) at the Apple Store when purchased separately. Asking $450, you pay shipping. Email for-sale-2007 at this domain (diglloyd.com) [email not spelled out to avoid spammers]. I’ll respond to offers in the evening, PST.
My rebate postcard from Canon, claiming missing items
I quoted the awful experience of an acquaintance of mine with Canon rebates in my Feb 11 blog entry. About six weeks ago, I submitted my own rebate form for three Canon items, including an EOS 5D. The rebate totaled $700, and I pre-qualified all the items online. I took a picture of the entire rebate submission, well aware of the difficulties reported by many people on the web.
Today I received a postcard from Canon (shown above) claiming that the receipt and warranty card were not included in my rebate submission! That’s rather odd, since they’re in the picture I took just before I stuffed them into the envelope! There is no mention of which of the 3 item(s) did not have the required receipts/cards. And I’m given until May 3 to produce them (again).
That is the way the scam works—it’s a percentage game—some postcards will be lost, it won’t be worth the hassle for some people, some people will have sent in the originals without a record, etc. No matter what, the rebate company makes extra money—my $700 could pay the wages of a claims representative (phone bouncer) for several days.
When I phoned, I was told that I must send in the “missing” items. I asked why I had to send them in again, since I had sent them in once already. I asked them to fax me their records of what they claimed I had sent in. I was told they didn’t do that. I asked how they could claim that I hadn’t sent the items in if they had no records. I was told again I must send in the “missing” items.
Finally, the bouncer (aka “claims representative”) asked if I wanted to speak to a supervisor. I said yes, and within a few minutes a new voice came on the line and said my claim was approved.
Incredulous, I asked how that could be, since I had just been told 2 minutes prior that items were missing from my rebate submission and that I would be required to send them in again. I was told that it was an “internal error in our systems”.
Quite interesting--an internal error. Their systems magically worked once I called to complain and asked to speak to a supervisor. As a software engineer, I’d be pleased as punch if system bugs just magically fixed themselves this way.
If I had time on my hands, I’d file a class action lawsuit against Canon, which bears the real responsibility in this matter.
Looking for a really large compact flash card? Look no further than mydigitaldiscount.com, which has SanDisk cards available at very low prices. I ordered a SanDisk 16GB Extreme III card a few weeks ago, and while they were very slow to ship (10 day delay), they did send a tracking number once it shipped. B&H Photo is selling the 16GB card for $299. Amazon.com is selling it for $249, with free Super Saver shipping.
I appears legitimate in every way—I compared it closely to another 16GB card that I had previously purchased from B&H Photo in New York. Having a bad past experience with a counterfeit SanDisk card, I called SanDisk to verify that the edge code is genuine, and I hope to hear back within 24 hours. It formats fine and has worked well so far in my Canon EOS 5D.
Like its sibling, it formats to only 8GB in my Nikon D200 and Canon EOS 5D. I hope Nikon and Canon provide firmware updates soon to support 16GB cards. And that Sandisk would feature compatibility information more prominently (or at all) on its packaging.
I’ve been testing 8GB of Mac Pro memory from two vendors: SATech and Other World Computing “Netlist Apple Certified”, for a total of 16GB—4X2GB + 4X2GB. Under full load with multiple copies of diglloydTools testing about 15GB of memory simultaneously, thus generating maximal heat, I’ve observed no ECC errors, though a test at a high ambient temperature is yet to be done. I’m not ready to endorse either vendor’s product, but I might do so when I’m certain my results are dependable. To be clear: diglloyd.com gets no commission from SATech or any other memory vendor.
If you are interested in the SATech memory, be sure to call them and specify “modules of the same quality that Lloyd Chambers of diglloyd.com is using”. Pricing on the SATech website might not be reflective of this style and quality of memory—my perspicacity in testing resulted in SATech providing me with what appear to be very high quality modules. I’ll clarify the part number and pricing with SATech sometime soon.
By swapping the two sets of 4 modules, I’ve determined that memory in slots 1 and 2 on both riser cards always runs about 10°F hotter than if the same memory is placed into slots 3 and 4. Keep this in mind if you do any comparisons of different brands.
Replacement memory from SATech looks great, runs cool and shows no errors so far. More on this another day.
I’ve long felt that the only strongly efficacious remedy for prevention and/or mitigation of a wide variety of health issues is regular vigorous exercise. I make it a point to burn (on average) about six thousand (6000) calories (Kcal) per week. Shown below is a portion of an 800 calorie ascent (estimated).
Now, there is a fascinating feature in the March 26, 2007 Newsweek magazine about how regular exercise is not only correlated to better physical health, but better mental health as well, including recent findings of neurogenesis—the creation of new nerve cells and their interconnects. There is also speculation that the path to avoiding degenerative diseases like Alzheimer’s might be as simple as exercising throughout life. But since exercise doesn’t come in a pill or with pepperoni, the findings aren’t likely to have much impact on the general population, and it will be difficult to get the drug companies excited.
The state of America’s health
These findings should come as no surprise to those of us who are both self-aware and who already enjoy regular vigorous exercise—it’s like hearing a newsflash that the sky is blue most of the time. And certainly, establishing the scientific basis for the “obvious” is required in order to engage public policy makers. And like any pill, there is no cure-all—health is influenced by many mental, physical, genetic, environmental and dietary factors.
I print many more pages of my blog, other web pages, receipts and miscellaneous stuff than I ever print fine art images. For everyday content, I’m looking for a paper that gives me pleasing text and decent image quality—spending $5.00 on paper alone to print 10 pages of a web site with perfect color makes no sense at all.
My paper of choice has been Epson Photo Quality Ink Jet Paper, a relatively lightweight paper with one printable side yielding the aforementioned “decent” results. When I ran out of paper yesterday, I drove to my local camera store only to find that they did not have it in stock. I drove to another store which I usually try to avoid like yesterday’s smelly socks (Fry’s Electronics) and the selection was much better. I purchased four inkjet papers for comparison to see if the premium-priced Epson paper was worth the roughly 40% higher price.
I printed 3 pages on each paper with the Epson R2400 that were representative of the types of general printing I typically do (mostly text and some images, such as this blog). The black ink used in the R2400 was the matte black. I examined the results under a 4700°K Solux desk lamp—typical reading conditions for me. The results are summarized below.
|Inkjet Paper Prices, Performance and Comments|
27lb, brightness = 98, 100 sheets
Slightly yellow look, very similar to the HP paper, but a slightly dingy off-white compared to the HP paper.
??lb, brightness = 100, 100 sheets
A close match for the Epson paper, but at 71% of the price. A shade more yellow than the Epson paper. My least favorite.
28lb, brightness = unspecified, 100 sheets
Very bright white—brightest of the four papers, text is very high contrast and easy to read.
Note: HR-101N might be different than HR-101 (not sure). HR-101 is more expensive.
32lb, brightness = 98, 150 sheets, double-sided
Very similar to the Epson paper, but printable on both sides, and brighter white. And 62% of the price of the Epson paper.
The upshot is that any of these papers will produce pleasing results for the casual printing of text, web pages, etc.
My favorite was the Canon High Resolution Paper for its high-contrast easy-reading text, with images at least as good as the other papers. My 2nd choice would be the HP Premium Presentation Paper Matte; printed single-sided, it’s the cheapest of the bunch, and printed double-sided, it’s 1/3 the price of the Epson paper, and 1/2 the price of the other papers. I did print on both sides of the HP paper with very good results, even with mixed text and dark graphics.
The price savings are a bit silly to worry about unless you do a great deal of printing. But I go through about 2000 sheets per year. Saving $0.04 per sheet is worth $80—not bad if you actually prefer the cheaper paper! And if the brand one prefers is out of stock, then it’s good to know that all these papers provide good results. Neither store had the 500 sheet ream of Epson paper, which is what I’ve bought in the past; prices can be much lower in quantity or on the web.
No doubt the archival Ultrachrome K3 inks of the Epson R2400 are “wasted” on these papers, but my R2400 does double-duty for both fine-art and everyday printing.
I have no desire to repeat the interrupt storm problems I’ve had with the Sonnet Tempo SATA cards, so I purchased three FirmTek SeriTek 2SE2-E two-port external SATA cards, filling up the slots of my Mac Pro.
FirmTek SeriTek/2SE2-E two-port external SATA card
Each diminutive card offers two external SATA ports, so this allows connecting up to 6 external SATA cables. I’ve connected mine to the quiet and extremely well-built SeriTek/2EN2 and SeriTek/2eEN4 enclosures—six cables to six ports. (I particularly like the SeriTek/2EN2 for its quiet operation and compact size). I’m waiting for a few hard drives and then I’ll post some performance figures.
FirmTek also states that the 2SE2-E also support Port Multiplication with a compatible enclosure. Like all my experiences with FirmTek products, these cards are performing very well, except for one erroneous dialog which FirmTek tells me will be rectified with a firmware update due out shortly. Firmtek will have several more storage cards and enclosures available soon.
The Mac Pro includes four built-in drive bays, but it has six SATA ports on its logic board (see diglloyd Dec 26, 2006 blog entry Eight Internal Hard drives in a Mac Pro). MaxUpgrades offers two optical bay assemblies which allow the installation of either two additional hard drives or four additional hard drives. The 4-drive assembly requires externalizing (removing) the optical drive, which I decided against.
I ordered the US$129 SZ-MCMPOPT01 2-drive kit. The kit is well-designed, with a sturdy aluminum construction. Unfortunately, it was delivered with no instructions whatsoever, leading to a frustrating experience trying to figure out how it should be installed. I emailed MaxUpgrades, and the support person responded with multiple pictures of the installation and cable routing. But there’s no valid excuse for shipping this product without installation instructions; very few users will figure it out (I couldn’t).
There are several helpful things to know about the installation:
- The original metal “cage” containing the optical drive does not get reinstalled.
- To allow access, remove the leftmost two built-in hard drives.
- Route the SATA cables from the optical bay down to the SATA ports. I recommend a needle-nose pliers to assist with this, but be careful and make absolutely sure you’ve unplugged the power cord from the Mac Pro.
- Disconnect the power cable from the video card (if you have the power-hungry X1900 video card, which requires its own power cable).
- Attaching the SATA cables: Remove the single screw holding the plastic baffle and wiggle the whole assembly up about 1/2" out of the Mac Pro enclosure so that the SATA cables can be attached to the ports on the logic board. Then push the plastic baffle back and screw it down (a magnetized screwdriver is very helpful). Without doing this, I found it impossible for my fingers to fit into the gap and attach the cables, though MaxUpgrades support insisted it was possible. Not for me.
I added two additional Maxtor 7h500F0 500GB hard drives (see diglloyd.com Hard Drives), bringing the internal storage to 3 terabytes. Using Temperature Monitor, I found that the newly-installed hard drives run about 5-10°F cooler than the ones in the standard drive bays. This is what it looks like after installation:
Optical bay with MaxUpgrades kit and 2 additional hard drives [Mac Pro]
My “Master” drive is now a six-drive striped volume, offering nearly 400MB/sec sustained transfer speed. The remainder of the space was used to create two 3-drive striped volumes. Failure of one hard drive means loss of Master and one of the backups, but not both. This setup is designed for speed and convenience; I maintain several other external backups.
Rumor has it that a dual quad-core Mac Pro (8 cores) is imminent. But the disappointing fact is that except for the very rare compute-bound application which accesses relatively little memory, an 8-core Mac Pro running at 2.66 GHz will almost always be slower than a 4-core Mac Pro running at 3.0 GHz. This is a fact of life with multi-core machines, and a key reason I proceeded with my Mac Pro 3.0 GHz purchase. Photoshop users should be content if even two cores can be fully utilized a modest percentage of the time.
There are three key factors at work inhibiting the performance of an 8-core machine:
- Software designed for up to eight-way concurrency that scales linearly up to 8 processor cores is rare, especially since memory bandwidth inherently throttles such scalability.
- Memory bandwidth is inadequate for 8 cores. It’s already a limiting factor with the current quad-core
3.0 GHz Mac Pro. Memory copy speed is at best 2.9GB/sec on the Mac
Pro, in spite of Apple’s highly misleading claims of
21.3 GB/sec (“maximum processor bandwidth of up to 21.3 GB/s”—bandwidth is a bit more than double the memory
That’s a measly 700MB/sec per core on a quad-core machine, and only 350MB/sec per core on an octa-core machine. By comparison, a 6-drive hard disk RAID array can easily perform at over 400MB/sec!
- Clock speed is slightly lower for the Intel quad-core Xeon, topping out at 2.66 GHz vs 3.0 GHz for the dual core chips (although a 3.0 GHz quad-core chip has been announced, it will not be available for some time). That’s a 12.8% advantage (at best).
Of course, if Apple devises a higher bandwidth memory bus for an 8-core machine, higher performance is possible. But this is not likely, since the quad-core chips share the same pin-out and in fact can be swapped for the dual-core chips currently present in the Mac Pro. That latter point is good news for hard-core techies who could obtain the low-end 2.0GHz Mac Pro, and swap the chips out for faster quad-core ones.
In short, the 8-core Mac Pro will be of value only if your workflow is compute intensive with modest memory bandwidth requirements. Server applications that are highly multi-threaded such as a web server might also benefit substantially. Don’t pay extra to run more slowly until you see some test results!
similar savings with 18-135mm or 28-135mm, half off body only
If you’re a customer of Amazon.com, you know that new recommendations are suggested to you based on what you previously purchased. Usually this is a fairly useful process, and benign when Amazon is doing it (but not so amusing when one day it’s acknowledged that the feds are doing it too, and produce a more nefarious mismatch than Amazon did—see below).
Especially if you have kids, you might have already viewed the various Walking With… DVDs, produced by the BBC. They’re all good (and interesting for adults too), but my favorites are Walking With Dinosaurs and the superbly-narrated When Dinosaurs Roamed America( Discovery Channel). The best deal is The Complete Walking with… Collection, but it does not include When Dinosaurs Roamed America.
So what did Amazon recommend for me? How about One Foot In the Grave, a TV series by BBC. Maybe dead dinosaurs are similar to dead cats and dead yoga instructors!
As a cyclist, photographer and general outdoor type, I really enjoy the exceptional quality and comfort of IbexWear clothing—I have quite a collection of shirts, pants, tops, jackets and gloves to cover a variety of conditions. Most items are wool, but not the scratchy, uncomfortable stuff, but rather wool that can be worn directly against the skin all day long. You’ll be hard-pressed to find anything of comparable quality and comfort anywhere else, regardless of price.
Best of all, they’re having a sale, and with free shipping, too (though selection is somewhat limited now as Spring approaches). My favorite items include the Adirondack shirt (I frequently get compliments wearing these), the Shak jersey and full-zip Shak jersey, the Woolster pant, and the Loden gloves. If you spend time outdoors in cool and/or windy/damp conditions, the Shak jersey and Woolster pants are the most comfortable, versatile combination I’ve found over a range of temperatures, yet can be worn inside and even when temperatures warm to 70° or so. Cyclists in particular should take note of the Shak jersey, a daily-wear item for me during winter (cycling and otherwise)—and it doesn’t accumulate “stink” after a few uses, unlike synthetic fabrics.
When I head to the mountains and/or cool temperatures for photography, IbexWear clothing is what I wear, because I know it won’t let me down, even if the weather changes. I wore the Woolster pants and Shak jersey and Loden jacket all week for my Death Valley trip in February.
If you sometimes feel like a big kid, this place is for you. Click for larger image.
For the past 3 mornings, the Mac Pro has woken up from sleep mode with no video trouble, as described in my March 1 entry. The past two nights have been with the replacement card, but the night before that was with the original card. In short, the problem just stopped happening on its own.
Regarding yesterday’s entry about ECC memory errors under heat stress—a diglloydTools run-stress-test test run this morning of 90 minutes showed no ECC errors. However, the ambient air temperature (as reported by the Mac Pro itself) was 73° F, and the memory modules reported temperatures about 10° F cooler than yesterday’s test run. This suggests that adequate cooling is indeed an important factor with Mac Pro memory, and leads me to the following recommendation—
How to run a memory stress test
To test that your Mac Pro memory is robust under duress, place the Mac Pro in a location in which the ambient air temperature can be allowed to rise at least to 85° F, but not more than 95° F.
Use diglloydTools run-stress-test to yield a maximal load on the machine, using enough copies (one per Terminal window) to test all but about 1GB of memory. Use of 'nice' is suggested so that you can perform basic tasks like reading mail while testing, but don’t expect to get any Photoshop work done while running this test.
For my 10GB Mac Pro configuration, I use four (4) copies of 'dlt' as follows:
nice dlt run-stress-test -l verbose --memory-per-thread 560M --duration 8h
(See the diglloydTools documentation for details). Four copies of the above will test about 9GB of memory (4 X 4 X 560MB). You can adjust the amount of memory, so long as it does not exceed the available physical memory. The output will look something like this:
While the test runs (and afterwards), be sure to check for ECC memory errors periodically (see yesterday’s entry).
I wrote the entry below about 9:30 am. By 10am, I observed “ECC Correctable Errors: 48464” with module A2 (same module as before) immediately after starting a test run, and that was at an ambient temperature of 75° F—relatively cool. So it seems that the module in slot A2 is flaky.
SATech is going to exchange the modules for me, replacing them with ones using heat sinks meeting Apple’s specifications. I’ll follow up in a few days when I have the new modules.
Still relevant, readers contemplating a Mac Pro might want to review my August 20, 2006 blog entry How Much Memory To Buy.
When Mac OS X 10.5 Leopard becomes available, a trickle of true 64-bit applications will begin, but don’t hold your breath for Photoshop CS3 to be 64-bit anytime soon. For my part, I do plan on revising diglloydTools to be 64-bit when Leopard arrives, as this will simplify stress-testing large amounts of memory.
The Mac Pro uses Error Correcting Code (ECC) memory modules. ECC memory modules contain an extra bit for every byte (8 bits). This extra bit allows single-bit-per-byte errors to be detected and corrected. Multi-bit errors can also be detected, but cannot be corrected, and cause a kernel panic (crash). See my Jan 24, 2006 blog entry for more on ECC memory.
Memory errors can be caused by excessively high operating temperatures and/or cosmic rays (cosmic rays are considerably more of a problem at high altitudes, making one wonder why laptops don’t use ECC memory; they often are used on airplanes at 30,000'). Poor-quality memory chips are also more subject to random errors than high-quality memory.
Today, with diglloydTools run-stress-test running on 9GB of memory, I left for the office. I had intended an 8-hour test, but I inadvertently started the test at the default run time of 1 hour. So it ran for an hour, then the Mac Pro idled (without sleeping) for another 9 hours in a closed room. The ambient temperature upon my return, as measured by the Mac Pro itself and displayed by Temperature Monitor , was 84° F. Apple alludes to ambient temperatures up to 95° F for FB-DIMM modules in Technical Note TN2156, and my test was 12° F below that temperature—hardly an extreme test.
Upon my return, I checked the memory status (=> => ). This is what you don’t want to see—ECC memory errors (but at least they were all correctable).
Rebooting did not make the errors go away; after reboot DIMM A2 still reported 88339 “ECC Correctable Errors”. However, cooling the room down by about 10° F was sufficient to make all ECC Errors vanish. The wonderful thing about ECC memory is that instead of a mysterious system crash, the system just keeps working. No data corruption, and no system crash. Good stuff.
The 2GB module in question is one of the SATech modules previously discussed. I will be asking for a replacement of that module. It’s hard to say if the non-compliant heat sinks on the 2GB modules are part of the problem, but I don’t want to use memory that has any ECC errors.
I’m not a fan of theoretical technical discussions about memory bandwidth (transfer speed) , , , except as a starting point for investigation—which I’ve now done*! Also, measured results are more persuasive than theoretical ones. While performance should ideally be measured for actual usage scenario(s), such results lack broad applicability, and are tedious to perform.
The number of memory modules, their “rank”, and their proper installation all affect memory bandwidth (speed) on the Mac Pro. See Recommendations.
For some programs, memory bandwidth will have little or no effect on performance. But for programs that make extensive access to memory, especially large amounts of it, performance could vary substantially between optimal and non-optimal memory configurations. Of course, an adequate amount of memory of any speed has far more impact; this discussion is largely irrelevant if Mac OS X must swap memory contents to and from the disk (“virtual memory”).
I had 8 FB-DIMMS available: four (4) Apple 512MB modules (single-rank), and four (4) SATech 2GB modules (dual rank), for a total of 10GB. I tested the speed of various combinations. “Speed” here means the speed of large-block memory copying.
|Mac Pro Memory Bandwidth|
|Configuration||Installed In||Memory Copy Speed
(aggregate, 4 threads)
|2 X 512MB||A1, A2||1440 MB/sec||49|
|4 X 512MB||A1, A2
|2 X 2GB||A1, A2||1724 MB/sec||59|
|4 X 2GB||A1, A2
|2 X 2GB +
2 X 512MB
|4 X 2GB +
4 X 512MB
|A1, A2, B1, B2
A3, A4, B3, B4
Modules must be installed in pairs of matched size (see Apple’s Replacement Instructions and RAM Expansion Product-Specific Details). So there can be 2, 4, 6 or 8 modules installed. The riser boards are nicely labeled as to which slot is which (slots 1/2/3/4 on each riser board).
- Aim for a configuration of either 4 or 8 modules, not 2 or 6. Each group of 4 modules should be the same size, and occupy the same slots on each riser board.
- Avoid single-rank modules. Apple’s 512MB modules are single-rank. However, 4 of them do offer 90% of the performance of the dual-rank modules, so they’re still speedy in groups of 4.
- If you have mixed sizes, install the fastest modules in the lowest-numbered slots. Mac OS X allocates memory from the lowest-numbered slots first.
Keep in mind that the performance of any specific program might or might not benefit from the optimal configuration. On the other hand, there is no reason not to purchase the optimal configuration, assuming your budget allows it.
As I explained in my Feb 28 blog entry, I reserve internal drives for a striped RAID array containing data only (no system software or applications). I boot the Mac Pro from a FireWire 800 drive on which I’ve installed Mac OS X, plugged into the FireWire 800 port on the rear of the machine.
To my puzzlement, I had observed poor performance on the boot drive while copying files, and now I’ve determined the cause. Below are the performance numbers from DiskTester on a partly-full boot drive ('disktester run-sequential -c 128M Boot', speed would be a bit faster on an empty drive):
Write rate: 63.0MB/sec
Read rate: 69.4MB/sec
Plug in any of the following peripherals (it doesn’t matter whether the peripheral is FireWire 400 or FireWire 800):
- Epson R2400 printer using a FireWire 400 cable to the front of the Mac Pro (turn printer on after plugging in);
- Lexar FireWire 400 card reader to the Cinema Display port;
- SanDisk FireWire 800 card reader to the front of the Mac Pro. (Yes, FireWire 800).
- A 2nd FireWire 800 hard drive, connected to the first one, or directly to the Mac Pro.
Testing again with DiskTester yields the following greatly-impaired performance figures:
Write rate: 23.4MB/sec
Read rate: 51.2MB/sec
I verified this multiple times, plugging and unplugging with the same results every time—write speed is cut to nearly 1/3 of the speed without the peripheral! Read speed drops by 26%. While disappointing, one might expect that plugging in a FireWire 400 peripheral would cut performance to FireWire 400 speeds, but plugging in a FireWire 800 peripheral ought to have no effect!
The PowerMac G5 line suffered from a broken implementation of FireWire with impaired write performance (see various benchmarks at barefeats.com). These results show that the fancy new Mac Pro suffers similar problem—FireWire is worthless as a means of extending storage capacity when performance is a goal, should you want to connect even a single other peripheral.
What good are additional FireWire ports if connecting more than one device of any kind kills hard disk performance?
Note that connecting a 2nd FireWire 800 hard drive produces the same impaired performance. Don’t be suckered into using FireWire 800 for RAID, or any other multi-drive setup unless these low levels of performance are acceptable.
The Mac Pro/ATI Radeon X1900GT video bug after overnight system sleep discussed in the March 1 entry occurred two nights in a row with two monitors attached. Leaving only a single monitor attached (Apple 30" Cinema Display), the problem did not occur this morning. (So far I have only three nights testing, and overnight sleep is needed to induce the problem).
Of course, those are small sample sizes, and it’s hard to be sure that’s the key factor. Next I’ll try using two Apple-brand displays, instead of the 30" Apple Cinema Display and the 24" Sun Microsystems display.
3.0 GHz, 8GB memory, 512GB flash drive
I’ve added some more thoughts on the switch from Nikon to Canon in my Goodbye Nikon? article, especially in consideration of the recently announced Canon EOS 1D Mark III. I have a EOS 1D Mark III on order, and will be reporting on it, though I’m as yet unsure whether I will do a full-blown review of it, as I’ve done with other Canon and Nikon digital bodies.
I predict that self-cleaning digital camera sensors will become a standard feature soon, one that will be required to be competitive in the DSLR market. The recently-announced Canon EOS 1D Mark III is the first professional camera to include this feature. Perhaps it will be a bit longer in coming to full-frame sensors, but it will come.
We’ve now passed an inflection point. The race used to be about increasing the number of megapixels, and to some extent this is still seen in the idiotic megapixel increases in the consumer digicam market, where merely zooming the lens can approach or even exceed the ability of the lens to supply adequate resolution to the sensor (smaller aperture = diffraction).
The new race becomes one in which usability and overall image quality become the market differentiators. This includes things like self-cleaning sensors, lighter weight, longer battery life, WI-FI, frame-rate, buffer size, color accuracy, dynamic range, high ISO performance and improved optics.
There will be continued incremental increases in megapixel count; expect Canon to introduce a EOS 1Ds Mark III at 22 megapixels or so, a realistic optical limit with a full-frame sensor. Such increases are likely to stop at around 28 megapixels for full-frame sensors, unless optics improve dramatically. Nikon already has nowhere to go with its smallish DX sensor; the photosites of the D2X are already so tiny that low micro contrast is an issue, even if the optics can supply adequate resolution—and critically-accurate focus is a very real problem.
As I noted yesterday, the heat sinks on the SATech memory have greatly reduced surface area compared to that of the Apple modules. I downloaded Temperature Monitor to check out the temperature of the memory in my Mac Pro. I’m a bit concerned that the 2GB modules seems to be running very hot as compared with the 512MB modules. Do keep in mind that the 2GB modules are “dual rank” which means they have twice the chip count (chips on both sides of the module), and thus each chip has twice the memory density of the chips on the 512MB module, so greater heat should be expected. Or perhaps the chips are the same density, which would require 4 times as many chips—I’m not sure.
Shown below are the temperatures of the 8 modules (ambient temperature about 72° F). Modules A1/A2/B1/B2 are the 2GB modules, and A3/A4/B3/B4 are the 512MB modules.
Temperatures of the 8 FB-DIMMS
Fully stressing the machine, using diglloydTools to run all 4 cores at 100% and access 8GB of memory for 30 minutes or so results in a modest temperature rise of no more than 10° F, but that pushes the temperature close to 200° F at an ambient temperature of 81° F.
I want to emphasize that I’ve had no glitches with the memory, and the ECC memory status has never shown any errors. But memory bandwidth is apparently a concern in the Mac Pro if the modules are running too hot—see Apple Tech Note Thermal Considerations for Mac Pro FB-DIMMS.
I asked SATech.com to investigate the temperature issue for me. They responded:
The memory module temperatures will be definitely higher than the 512mb due to the density. 2gb FB DIMMS are designed to work in a server environment only, so the heat spreaders are designed to work in that environment. The heat spreaders on these modules are Jedec compliant and the modules are Super Micro and Mac certified means tested under extreme operating conditions. So no cause to worry even if the outside temp increases.
A reader requested pictures and several are shown below. Click for larger images.
Mac Pro riser board, fully populated with four FB-DIMMS
Slots 3 and 4 (top) contain Apple 512MB modules
Slots 1 and 2 (bottom) contain 2GB SATech modules
Apple 512MB FB-DIMM [side view]
How to determine that the memory is reliable, even under stressful (hot) conditions? Run the diglloydTools run-stress-test command, designed to tax a Mac OS X system to its limits (excepting the video card).
The system needs memory to run of course, so diglloydTools can’t test all the memory. But that’s not critical for its intended purpose. For systems with more than 2GB of memory, running multiple copies of diglloydTools is the answer to stressing most of the memory. Test all but about 512MB; it is counterproductive to attempt to test all of it as this will provoke system “paging”, or swapping of memory contents back and forth from the hard disk.
With 10GB memory, running 4 copies of diglloydTools (in 4 Terminal windows), each with 2GB of memory (512MB per thread), is a good way to stress most of the memory in the system (8GB). (A fifth copy could be run to soak up another 1.5GB or so, or each of the four could test 2.4GB or so).
I used the following command in Terminal for each of the 4 copies:
nice dlt run-stress-test --num-threads 4 --memory-per-thread 512M --duration 4h
No ordinary use of the system could ever stress the system (and memory) nearly as much. Even under this incredible load, the fans run only slightly louder, even though all four CPU cores are running at 100% and accessing most of the system memory.
Installation is a breeze—the easiest I’ve yet seen in a Mac. Pull the riser cards, insert the modules, and re-insert the riser cards. Don’t mix sizes and/or brands, and fill DIMM slots 1 and 2 first on each card eg slot 1 on riser A and slot 1 on riser B, etc.
The first step after installing memory is to verify that it’s recognized by the system. Use the “About This Mac” command in the Apple menu to bring up this dialog:
Mac OS X “About This Mac” dialog
Mac OS X has a nice status display for memory (shown below), which can be viewed by clicking “More Info…” in the “About This Mac”. All modules should read “OK” even after heavy testing for multiple hours.
similar savings with 18-135mm or 28-135mm, half off body only
I purchased additional memory for the Mac Pro yesterday from satech.com [pricing page]. As of this writing, SATech (aka RamFinder) charges $1990 for 8GB (4X2GB), or $249/GB. By comparison, Apple charges $2699 to upgrade to 8GB, or $337/GB. Going the SATech route, one keeps whatever DIMMS already came installed—4X512MB = 2GB in the case of my refurbished Mac Pro.
I’ve had multiple good experiences with SATech over the years, and they stand behind their product. I recommend sticking with “factory approved memory”—the highest quality. However, the FB-DIMMS I purchased do not have the same heat sinks as the Apple-installed 512MB modules, so it’s a bit misleading for satech to claim “This is the same memory product you would buy from your system manufacturer (Apple, Compaq, Dell, HP, IBM, Gateway, Sony and others) directly when you buy the system”. I have forwarded this concern to satech, as I believe they should retract this claim.
The Apple heat sinks on the four 512MB modules do appear that they would be more effective. But see the stress-testing discussion below.
As I mentioned in my Feb 28 entry, one can “trick” Apple Mail (or nearly any other program) into thinking its data is in one location, but actually in another location, even on a different volume/disk.
Suppose you have a volume “Master” on which you wish to store all your important data. Begin by being disciplined enough to store your data/files on that volume only, organizing things as you normally would within folders, subfolders, etc. Next, “trick” programs like Apple Mail into storing mail there too. In this manner, all data is conveniently contained with the “Master” volume.
The steps below explain how to relocate your Apple Mail folder. Please don’t follow the steps below unless you are comfortable with them, and are sure you’ve backed-up up your mail first!
- Quit Apple Mail.
- Copy the “Mail” folder to your “Master” drive. While you’re at it, make a backup somewhere else, too. The “Mail” folder is found within “Library”, which is in your home directory ( in the Finder).
- Throw the original “Mail” folder you just copied in step(2) into the trash and empty the trash.
- Open a Terminal window. (In the steps below, “llcMP” is the name of my computer, and “Master” is the name of the master data volume). Commands to be typed are shown in blue.
llcMP:~ lloyd$ cd ~/Library llcMP:~/Library lloyd$ ln -s /Volumes/Master/Mail Mail llcMP:~/Library lloyd$ file Mail Mail: symbolic link to `/Volumes/Master/Mail'
You’re done. Launching Apple Mail should result in exactly the same mail as you had before, but the folder actually resides on the volume “Master” now.The “Mail” folder within “Library” will now display as a shortcut like this: .
Locating all data within/underneath a single “Master” folder enables a backup strategy in which a full backup can be performed by a single drag-copy of the “Master” folder to other volume(s). It also allows erasing the boot drive at any time, without fear of losing anything more than one’s system preferences. And it enables the transport of one’s data between similar or dissimilar computers via an external hard drive.
I prefer manual backups. Backup software needs to be paid for, upgraded regularly, and learned. In my view, a backup procedure should be brain-dead simple, and should not require special programs, which can contain bugs, not to mention confusing features that can be misused. My strategy involves backing up my Master data folder to four different backup volumes, cycling through them on a week basis.
When I backup, I follow these steps:
- (optional) Erase the backup volume using Apple’s Disk Utility. This ensures a problem-free file system. Of course, if the backup volume holds multiple previous backups, then you would not perform this step. And don’t erase the wrong volume!
- Create a new folder on the backup volume using the notation eg “20070302” to represent March 02, 2007 (this notation sorts nicely by date when multiple backups are viewed).
- Drag-copy the “Master” folder to the folder created in step (2).
- (optional) Disconnect the backup disk and store it in a safe place, preferably away from the computer or at another site.
There’s only one problem with this strategy: what about things like Apple Mail, which stubbornly insist on storing mail in your home directory on the boot volume? Remembering to backup your “Master” data folder and a variety of other data folders is error-prone.
I also ran Apple Hardware Test (boot from the Mac Pro system software DVD while holding down the “D” key). With 10GB of memory, it takes a long time, even for the short test, so long that I thought it had crashed or hung. But after somewhere between 7 and 15 minutes (guesstimate), it passed just fine. The fact that the mouse cursor responded 20 seconds or so after moving the mouse led me to believe that the test had crashed or hung. I never was successful in canceling the test, reinforcing my concern that it had crashed or hung. Fortunately, it is just really, really sloooooow. Apple should clarify this point for those of us with gobs of memory.
The wonderful thing about the Mac Pro is that even under full load for most of an hour, it remains barely audible, at least under my desk. The Intel Xeon “Woodcrest” processors must indeed generate less heat than the PowerPC G5 processors.
The Mac Pro is reasonably power efficient, though its power usage at idle seems rather wasteful. I used my Watts’s Up power meter to measure power usage as shown below.
|Mac Pro 3GHz Power Consumption
10GB memory, ATI Radeon X1900 GT, 4 X 500 GB Maxtor 7H500F0 hard drive
(400% CPU usage, 9.2GB continual memory access, no disk activity)
The ATI Radeon X1900 GT video card contributes significantly to the power usage; it has its own dedicated fan and hot air vent out the rear of the chassis. The internal hard drives draw only about 8 watts while idle, so the figures won’t change much with only one internal hard drive.
Add another 90 watts or so for the Apple 30" Cinema Display, and about 30 watts for an external boot drive.
Disabling one or more CPU cores using the developer “Processor” control panel did not alter the power usage at all.
I put the Mac Pro to sleep around midnight, and woke it up around 7am to a headache-inducing display.
The problem is not unlike the problems I experienced with an ATI X1900 on the G5 Quad, as described in my Jan 31, 2007 entry. The problem is provoked in exactly the same way: everything is fine after a short system sleep, but things go haywire after sleeping overnight. As this has never happened with other brands of video cards I’ve used, it seems reasonable to conclude that it’s a problem with the ATI X1900, possibly induced by having two monitors attached.
What to do now? I’m keeping the Mac Pro. So perhaps the X1900 has to go.