Chriscool

Member
  • Content Count

    329
  • Joined

  • Last visited

Community Reputation

0 Neutral

About Chriscool

  • Rank
    Member
  1. Chriscool

    Computer a bad audiction

    HisMajesty, upgrading your girlfriend (a new breast and alike) can be far more expensive. you'd better had stuck with hardware However, we upgrade our hardware so frequently because : 1- it is a passion 2- we can affoard it People are used to spend a lot in their passion. Because, beside passion, computers can be so useful, it ain't that bad finally.
  2. Chriscool

    Software Development

    Chewy, i understand you'd like to use your different CPUs. However, i'm not sure that running multiple slots (the -j 4 parameter) will really perform better. If i were you, i'd do some benchmark here (unless you already did). Olaf, i agree that compiling is very cpu sensitive. After auditing my cpu, it happens to be mostly at 100% when compiling. BTW, what do you mean by "Object files can be written with delay" ? Could anyone confirm that cpu is still the bottleneck with recent cpus (mine is only 2.2 Ghz) ? Chewy, could you confirm than cpus are still the bottleneck with multiple threads (i expect performance to degrade quicker on hard drives than cpu as the threads go more and more numerous), if you were to test ?
  3. Chriscool

    Software Development

    I don't have any benchmark, but i am pretty confident that the raptor solution would perform better. You usually have not multiple threads compiling (how your makefile is suppose to dispatch the files to the different threads ?), so this is a typical workstation pattern. With Tenths Mo of source, i would bet on an independant drive to write the object files. My system uses scsi, but more because i use virtual machines to simulate an entire environment (server+client).
  4. Chriscool

    3ware 7506-4LP perf in WinXP Pro env

    First, benchmarks only give comparison values between different systems using the exact same benchmark. If you want to know how your system compare to the one tested by tom's hardware, you must use the same benchmark and compare values. Second, drives performances slow down dramatically when multitasking like you are when you read and write from and to the same drive. This will always give you values far far from the values any benchmark could give. Third, tom's benchmark uses raptors, which are obviously better performer than your drives. I will repeat myself : if you want to know how your system can compare to another one, use the same benchmark. Values by themselves are pretty useless (real values are commonly far below benchmarks results).
  5. Chriscool

    SATA raid for cluster computing

    However, you should check different possible bottlenecks : * cpu usage, network interface Just to be sure, try auditing the cpu(s) usage. Also, how far your i/o are from your network limits ? * filseystem With many little files, check out what filesystem would possibly do the best job. The new reiser file system is supposed to be quite good for this, but is still a little bit experimental. * raid subsystem If you cannot go for SCSI, at least try devices with TQ abilities (both for your controller and your drives). As it has already be said, raid 5 is not quite adequat if write performances are an issue. If possible, you should try raid 10 or 50 just to know if it helps or not (actually, for exhaustive test purposes, you could as well test raid 0, just to know precisely what is the precise impact of your raid configuration on your performance issue).
  6. Chriscool

    3ware 7506-4LP perf in WinXP Pro env

    The performances seem ok to me : raid 5 is not the fastest regarding writes, and 40Mo is quite good if you read and write files from the same (raid) drive.
  7. Chriscool

    DVD Codec

    Try to download ac3filter here http://ac3filter.sourceforge.net/
  8. Chriscool

    DVD Codec

    Maybe an AC3 codec is missing. Check this out...
  9. Chriscool

    Basic (?) SCSI question

    Connecting SE and LVD devices on the same cable will slow down to SE the scsi bus. The adaptec cards 29xxx in LVD had a technology to connect two cables on the same bus (one for SE and one for LVD). They used to call it SpeedFlex IIRC. However, i couldn't tell for the 19xxx...
  10. What could be added is that those recommandations don't apply for just mas series, and are relevant for any drive
  11. Chriscool

    Raid O vs single drive for gaming

    It is likely to get a few improvements when using scsi RAID, as those drives are not optimized for workstation patterns. However, what you will get using raid with ide drives is completely random and you cannot assume anything. It is not unlikely that the raid layer conflicts with a drive's firmware optimized for workstation loads. That is why, in this case, my advice had been just to not bother.
  12. Chriscool

    Raid O vs single drive for gaming

    There is no definite answer to your question about single drive and raid-0 performances. The raid-0 has a better potantial, but involve one more mail, and as such, another possible point of weakness. With raid-0, your overall performances will rely on firmware of your disks, of your raid controller, and OS drivers. For large data, the increased transfert rates is a major improvement. However, real performance for program loading would rather be completely random (ie, you cannot be sure of anything before you tried). In the worst case, you shouldn't be able to notice real differences between your two configurations, so don't bother...
  13. Chriscool

    DVD media standards

    DVD +R are supposed to have less compatibilty issues with stand alone players (despite i had no problems with other standards, even RW on my old stand alone pionneer DVD player).
  14. Chriscool

    Who here likes noisy HDDs?

    Solid is the opposite of mobile. Solid State Drive implies a stuff without any mobile part
  15. I completely agree with Honold. This is FUD. Porting software or rewritting it implies the same thing : potential bugs. Software, OS, procedures and users are parts of the whole system. Tha main failure here is more about management than OS failure. About the comparisons abut unix not going down when the software is broken, it is clearly off topic in a dedicated system like this one. In this case, software down implies the system cannot provide its services anymore, no matter if you get a segmentation fault or a BSOD. It is FUD because you cannot assume what is responsible of the failure. We could as well look at the hardware they were using and say things like "see, they were using Dell (or Compaq or IBM or whatever) and they got a breakdown". There are just enough known facts for superstitions and FUD. The most certain responsability here is clearly a poor management : the system was known to have weaknesses, the procedure to overcome them was clearly weak and not applied, and there was no backup solution.