AT Mac Section: Taking Requests

65 Comments

View All Comments

• Steve - Tuesday, June 15, 2004 - link

  Hi Anand et al,
  
  Some years back the warrantees on HDDs were cut from 5 to 1 year for 2MB cache drives and 3 for 8MB cache drives.
  
  That clearly indicated that OEMs knew HDDs were no longer as reliable as they had been.
  
  It is easy to find latest/greatest HDD reviews all over the net that dwell on the easy to test things like access time & write speed. Most of us, however, would gladly swap some of that performance for drives that don't let us down!
  
  When are we going to see reliability assessed?
  
  Some drives are undoubtedly bought with performance requirements in mind but I BET the majority are bought to ensure our digital photos/family videos don't disappear into the digital ether.
  
  So please try to assess reliability or at least suggest a database people can contribute to regarding HDD models/purchase and failure dates/ DOD rates/ heat at idle 20C ambient.
  
  Thanks for your time, Steve.

• Jason Clark - Tuesday, June 15, 2004 - link

  Why would it matter how loud an XServe is? Since it will most always be in a server room, or in a NOC facility? I can see noise tests for desktops or notebooks, but a server?
  
  L8r

• Eug - Tuesday, June 15, 2004 - link

  Quote: "Just remember to make sure the Mac is made to look great otherwise hordes of raving MacZealots will be after you, and Eug Wanker won't link to this site 100 times a day anymore :D"
  
  Heheh. I'm famous! ;) Can't say I link here very often though, although now I have a good reason to... :) P.S. This is being posted from my XP box. ;)
  
  Anyways, it would be interesting to see a good analysis (technology preview?) of the the G5 2.5's liquid cooling system.
  
  Also, it would be good to go a bit into the differences of the GUI design and implementation in OS X. (eg. 3D acceleration in Quartz Extreme, Exposé, etc.)
  
  There are a lot of cross platform apps for benchmarketing, but that's gonna be a difficult analysis.
  
  Premiere? Nobody uses it on the Mac side. (Final Cut Express and Pro.)
  After Effects?
  Cinebench?
  Lightwave?
  Maya?
  BLAST? Apple has their own version - A/G BLAST, which is much faster than NCBI BLAST, but A/G BLAST is Altivec'd.
  Photoshop? Certain filters favour x86, while others favour Macs.
  Video encoding?
  Video decoding? (I think WMV9 is faster on x86, of course, but H.264 is faster on G5s.)
  
  And, there are a bazillion *nix apps that could be used on the Mac side, vs. *nix on x86.
  
  Of course, what would be nice would be a summary of your preferences and dislikes about your Power Mac for everyday use.
  
  Cheers,
  
  Eug

• David Smith - Monday, June 14, 2004 - link

  One place to look for performance tools is Apple's CHUD tools (search around on developer.apple.com if you haven't seen them). The benchmarks included are quite simplistic, but the performance analysis tools are very nice.
  
  Another thing to check out would be how well SMP and Altivec are used in common apps or the OS. SMP is easy (CHUD allows you to turn one proc off), not sure how one would go about changing the availability of Altivec. sysctl lists it, but I don't think it's changeable there.

• ViRGE - Monday, June 14, 2004 - link

  While G4s are on their way out, I'd like to see some sort of comparison on the processor, similar to how you do traditional CPU benchmarks. How are the G4's affected by cache size? FSB speed? It's somewhat of an ambiguous task, since acquring G4's with similar clock speeds(but different features) is hard, but the results would be worth it, however you figure it out. And if you can't do that, at the very least, even a simple comparison between an iBook/PowerBook at similar clock speeds would be useful.
  
  PS Dave, this isn't the place to be making personal attacks

• miranjan - Monday, June 14, 2004 - link

  The thing most interesting to me is MPEG2 and MPEG4 encoding tests. Use whatever tools are best for that platform, then compare speed and quality.

• Anonymous - Monday, June 14, 2004 - link

  Many people would like the RAPTOR guy to kiss our collective ass. (ahem, not you specifically, John)
  
  I'd like to see how to overclock our G5's, for those of us like me with mil spec soldering certificates and tools that can desolder and resolder resistors the size of a head of a pin. I would like to turn my 1.8 dual into a 2.0 dual, or conversely I'd like to know it cant be done and here's why.....
  
  I'd like to see some common game performance tests between platforms, with the same game available on both platforms.
  
  I'd like to see ease of adding upgrades and software compared between the two platforms with real world examples of the same solution applied to each platform and how well it turned out or each.
  
  I'd like to see the G5 get equal billing and consideration on Anandtech when events warrant it.
  
  
  
  

• John - Monday, June 14, 2004 - link

  Many DBAs would like to see the performance of the Raptors in a RAID 5 configuration, particularly while doing a large UPDATE or INSERT where pages on disk muct be modified to complete the transaction.

• Guy - Monday, June 14, 2004 - link

  I suggest a whole new type of benchmark: taking inspiration from Let1kWindowsBloom, I propose 100TaskTime, a test of how long it takes to perform 100 common tasks eg.
  
  1. Connect a digital camera
  2. Reconfigure from scratch your printer settings
  3. Mount your external hard drive and transfer 100 Mb of files
  4. Burn the files to CD.
  5. Change screen resolution twice
  ....
  
  Of course the test would be colored by many variables at first, but over time effort could be made to standardize tasks (open source this - invite readers to submit and vote on tasks?), and I suppose that in fact, with 100 tasks, discrepancies would to some extent average out, even operator skill as users become familiar with performing the tasks. Even just getting through the tasks successfully may be revealing. Extra points could be alloted or deducted for eg. no. of popup windows encountered, no. of restarts required etc.
  
  The idea is to test for basic usability. Comparisons among users' experiences would be qualitatively valuable ("I crapped out at step 37," "I finished at the faster end of the 90-minute quintile, but I rate my run-time operator skill as only a 7"), and, given work, quantitative discrepancies could be minimized.

• Dalis - Monday, June 14, 2004 - link

  Another vote for looking at HD and/or RAID options. What is better price/performance; a 10kRPM SATA drive? A modified case with several extra large 7200rpm SATA drives and PCI-X RAID in 0+1 or 5 config? How much would either option improve performance? There are many variables and it is something an individual would have a hard time evaluating.