DeusExMachina
New Member
Read somewhere that Win98 and below has no performance gain for anything greater than 512 coz it can't address anything greater.
How much DRAM for optimal image editing?
- Thread starter khinmarn
- Start date
- Status
actually its true for all 32 Bit software. There is only a max speed at which all data is processed through the whole system and having a super fast system with even the operating system trying to catch up would be a waste of computing power... By the way this goes for winxp and all other windows operating systems... dunno about macsDeusExMachina said:
pureflow said:
Per context, not per system. No SINGLE process can use more than 3.5Gb, but doesn't mean more isn't used, by treating the OS, and what is being used by the applications as seperate 3.5Gb memory spaces.
The OS get's it's own 32bit memory space, and each application get's it's own seperate memory space as well.
All modern Unix's and MacOSX certainly works like this. I assume Windows NT derivatives do as well. (NT, 2k, XP)
I would be surprised if a high end application like photoshop couldn't do it's own paging in order to use more than 32bits of memory space internal to the application. Adobe love doing special optimizations to pull in the benchmark figures.
pureflow said:
That might be an OS limit, not MB hardware. I have a vague recollection of the memory map in Linux being changed recently as 32bit applications were limited to 3.5Gb of RAM instead of 4GB as the OS had mapped 1/2 gig in the process context for it's own use. They changed the layout of the memory management to stop 'wasting' this 1/2 gig.
I never really have the time or background knowledge to understand it properly.
I have spent a bit of time over the years playing with 64bit platforms where this isn't a problem.
Ironicly my modern Shuttle 2.2Ghz 1GB ram PC is sitting on top of a 64Bit Alpha system that's 7-8 years old. The Alpha was a 266Mhz fitted with 384mb of RAM. (I couldn't afford more!) The 'LCA' Alpha was 'crippled' by a mesly 64bit memory interface, the CPU core spent most of it's time waiting for RAM.
BTW, the Alpha machine I have, while I built the board into a server tower chassis, was intended to be an OEM board for 64bit desktop computers. Apple did not release the first 64bit desktop personal computer. DEC did, 10 years ago with the Multia. It was a complete flop. Amoung other things the Multia was slower than the Pentiums of the day at 'normal' tasks, and couldn't have enough RAM fitted to really make use of the memory space afforded to 64 bit applications. The Multia was a 'cute' machine really.
The DEC engineers apparently optimised an Alpha version of Windows NT Alpha Photoshop and got really impressive transform benchmarks with it... In 32 bit mode, Windows never ran the Alpha AXP in 64 bit mode, it would be another 10 years before Microsoft worked out how to make windows work on a 64 bit CPU, by then DEC had been bought by Compaq who was 'merged' into HP and the AXP line killed of infavour of Intel's Itanium, which is based on a technique that DEC had rejected years ealier when embarking on the Alpha project.
I just read somewhere a few days back that current Adobe CS has 2 GB memory limitation on x86 with XP platform [32 bit ones I think 
]
I can't exactly remembered whether is the OS or the application itself :think:
I'll dig up my resources and post back here.
Anyway for DRAM ... if you are a Home user .. just chuck in 2x 1 GB DDR and set it in Paired mode [only on the new mobo's]
and dun forget to have 2 phyiscal HDD for scratch drive.
having 4x 1 GB is just too much now .. as the RAM price just skyrocketed this past few weeks.
=bob=
]I can't exactly remembered whether is the OS or the application itself :think:
I'll dig up my resources and post back here.
Anyway for DRAM ... if you are a Home user .. just chuck in 2x 1 GB DDR and set it in Paired mode [only on the new mobo's]
and dun forget to have 2 phyiscal HDD for scratch drive.
having 4x 1 GB is just too much now .. as the RAM price just skyrocketed this past few weeks.
=bob=
NashVillian
New Member
khinmarn said:
not necessary, my take on that is since there is nothing running in the background, the comp might have used cpu more intensively to hasten the process
The CPU will always perform at 100% capacity, nothing less. In another words, the utilization figure is derived as a percentage over time.
Thus, if the sampling interval is 1 sec, then 5% utilization means that the computation the CPU performed average only that much in that period.
If the same task is measured simply by changing the polling interval from 1 sec to say 1 minute, the CPU utilization figures would have been quite different if the peak is short.
So if you do not see your CPU utilization peaking then you have not yet to utilize the full capability of your machine. A benchmark is something we could objectively do to compare the performance in real world application.
For fellow Nikonians, we can measure the translation / processing of NEF file editing.
1. Using Nikon Capture Editor, edit multiple images by "open multi image window"
2. Perform simple editing such as auto-contrast. This is to ensure the processing for each image do not differ too significantly.
3. Then go on to the next image (as many as you want) but do not save or close the editted image at this point
4. After editing at least ten images, then select "Close" from the File menu. This would in turn batch open and save all the edited items.
5. Clock the time taken to save each editted file. You can turn on the speaker and measure the interval of the click sound for each completed file save. Or better yet take the total time and average it out yourself.
The pertinent info at this juncture are only
a. Processor Type
b. Processor Speed
c. Memory Size
The average time for my PC setup: Celeron 1.7G with 786Mb RAM, 9~10 secs.
As we become more critical, we can use increasing complex criteria to benchmark different PC setups including number of hard disks, swap space etc.
Thus, if the sampling interval is 1 sec, then 5% utilization means that the computation the CPU performed average only that much in that period.
If the same task is measured simply by changing the polling interval from 1 sec to say 1 minute, the CPU utilization figures would have been quite different if the peak is short.
So if you do not see your CPU utilization peaking then you have not yet to utilize the full capability of your machine. A benchmark is something we could objectively do to compare the performance in real world application.
For fellow Nikonians, we can measure the translation / processing of NEF file editing.
1. Using Nikon Capture Editor, edit multiple images by "open multi image window"
2. Perform simple editing such as auto-contrast. This is to ensure the processing for each image do not differ too significantly.
3. Then go on to the next image (as many as you want) but do not save or close the editted image at this point
4. After editing at least ten images, then select "Close" from the File menu. This would in turn batch open and save all the edited items.
5. Clock the time taken to save each editted file. You can turn on the speaker and measure the interval of the click sound for each completed file save. Or better yet take the total time and average it out yourself.
The pertinent info at this juncture are only
a. Processor Type
b. Processor Speed
c. Memory Size
The average time for my PC setup: Celeron 1.7G with 786Mb RAM, 9~10 secs.
As we become more critical, we can use increasing complex criteria to benchmark different PC setups including number of hard disks, swap space etc.
pureflow said:
The Raptors are about the price of Cheetah 10.6's anyway. With SATA not currently supporting Tag Command Queqe's, it's better to just invest in a 29320R. At least you can get a decent 5 - 10.6 drives on it before hitting the bandwidth ceiling. With less CPU usage on the SCSI adaptor too!
One thing I always wanted were the Platypus RAM drives. Sweet.. But costly at $2K+ for just the basic model w/ 512MB space.
pureflow said:
Hrmm.. I doubt the 10.6K can sustain 80MB/s throughout. IMO, closer to 60+MB/s avg., max. 70MB/s.
The Platypus is a "non-voilatile" RAM drive. At least, it uses an external power adaptor to hold the information. Allows you to install & boot OS'es like Windows and applications on the RAM drive itself. Behaves just like a HDD w/ much lower seek times. Uses SDRAM modules, BTW.
Creating RAM drives on mainboard memory doesn't allow you to do that. Except for *nix where you load the kernel on diskette and the rest into memory. Then again, you'd be using GIMP during then.
Enough crap talk.. Hehe.. Need to upgrade my machine soon. Wonder if I should venture into the 15K RPM drives. Then use my current Cheetah for swapfile and misc small apps.
....how much u spent on this cool machine ?? :bigeyes:pureflow said:
The only thing that comes to mind is that when your system is confined to one I/O type, then you are constricted by the bandwidth of the bus ie 266Mbps for one SATA channel, 160Mbps for SCSI Ultra320
With multiple I/O controllers ie. combination of firewire, SATA and SCSI then you are increasing the channel throughput for I/O transfers. It may increase the CPU utilization, but CPU is always ahead of the rest of the components such as memory, VGA, hard disk anyway. And amongst the three, hard disk is the slowest/ bottleneck in any PC.
Wouldn't you agree this could improve the I/O if indeed I/O is so crucial. I personnally doubts so many hard disks are required, perhaps for database apps but for image processing? I am not sure.
I just upgraded my pc JUST for photoshop. (switched back from AMD to Intel. something about motherboard support for DDR400/fully utilising the FSB)
Specs:
Pentium 4 prescott 3.0 EE
Abit IC7-G
old Graphic card ATI Radeon 9200+ (don't think I need to change this)
2 x 1GB cosair PC3200 DDR400 CL3 (kind of upset they didn't have 1GB CL2/CL2.5 )
Western Digital SATA 7200RPM 120GB for my OS partition/program files.
using the older IDE Hitachi IC35L 7200RPM 120GB as my swap file/archives
*yeah, I know I should use scsi, but I doubt any bottleneck is disk IO right now. 8)
Specs:
Pentium 4 prescott 3.0 EE
Abit IC7-G
old Graphic card ATI Radeon 9200+ (don't think I need to change this)
2 x 1GB cosair PC3200 DDR400 CL3 (kind of upset they didn't have 1GB CL2/CL2.5 )
Western Digital SATA 7200RPM 120GB for my OS partition/program files.
using the older IDE Hitachi IC35L 7200RPM 120GB as my swap file/archives
*yeah, I know I should use scsi, but I doubt any bottleneck is disk IO right now. 8)
argh! made purchasing mistake. its a pentium 4 3.0E (should be a prescott), got to check the invoice and box...
3.0E thought extreme...grr.
anyway, don't think it would be a bottleneck at all..
anyway, my old pc (amd xp1800 1GB ram) was sufficient for photoshop, except when I had like 50 raw files open and doing batch processing on "open files", then it starts to slow down when it does disk swapping/undo buffers.
3.0E thought extreme...grr.
anyway, don't think it would be a bottleneck at all..
anyway, my old pc (amd xp1800 1GB ram) was sufficient for photoshop, except when I had like 50 raw files open and doing batch processing on "open files", then it starts to slow down when it does disk swapping/undo buffers.
IceCoolBeer said:For serious image editing, so call Digital Imaging. You need as much RAM as possible (1G). And a good high speed hard disk with at least 80G or more. You need a good graphic card, and a good monitor at least 21 inch. Try not to use LCD screen. You need some external storage medium to free up some data along the way, like a CD writer, DVD writer or another external hard disk. Hope that help a bit.
Why not LCD screen?
Too "watery" ???
eagles_creek said:
Colour balance to hard to get right, and is greatly affected by viewing angle. The constrast ratios were also a bit low.
Colour calibraters that can do LCDs are around now, view angle has greatly improved over the years as has the contrast ratio, and I expect that unless you are in to professional artwork, 'near enough is good enough' in the colour dept anyway.
Just make sure you screen isn't set at an inappropiate colour temp.
- Status
