I've always built my own computers and right now I have a C2D e6550 overclocked to about 3 Ghz, running stable at about 28 degrees.
But with CS4/LR2 and their multi-core support, would I see a noticeable difference in speed when going to a quad core processor? I'm looking at the Q6600.
Especially the rendering progress bars. You won't see much or any different for Brightness/Contrast, Levels, Color Balancing, Hue & Saturation, curves, and other tools that don't present a progress bar tho. So if in your typical editing day you see many progress bars then the quad core would help ya out. If not and you want more speed then look to memory sub-system specifications of MB and parts, HDD I/O buss and units, and Graphic card specs (MB buss and card specs) 1st.
Yea right now I have 6gb of DDR2 1066 RAM running at ~900 mhz, an e6550 running at around 3ghz, and a 8800GT 512mb vid card, it all runs pretty quick but I'd like it if lightroom ran a wee quicker and that it would export raw's to jpeg's quicker.
Only if you can find a good Q6600 (G0 stepping). If you can't get it up to around the 3ghz your better off just saving your money. The cores that are newer than the conroes also have SSE4 extensions that can provide a 15% boost imagine applications and up to a 40% performance boost in video encoding apps when enabled in the software. So you may also want to consider the E8500.
dan727 wrote:
Only if you can find a good Q6600 (G0 stepping). If you can't get it up to around the 3ghz your better off just saving your money. The cores that are newer than the conroes also have SSE4 extensions that can provide a 15% boost imagine applications and up to a 40% performance boost in video encoding apps when enabled in the software. So you may also want to consider the E8500.
The e8500 was another one I was considering. I've seen performance tests online between the C2D and quad core on CS3 and the C2D blew the quad core away, mostly because of the difference in processor speed (3.3 ghz vs. 2.7). I'm wondering though if CS4 makes more use of those other cores.
I can give you some cold hard facts since I have both Quad and Dual Cores.
I run my Q6600 at 3.25ghz air cooled, I also have a Dual Core E6750 2.66ghz running 3.5ghz and the Dual Core is not visibly faster than the Quad Core. Doing intensive things is PS CS4 like certain filters and such and timing the results the Dual Core is slightly faster.
This leads me to believe that Clock Speed rules over the number of cores. I also did some tests where Bridge was building thumbnails while in Photoshop, and again the Dual Core finished a bit quicker but I had to time it to see the difference.
Things that make PS CS4 run noticeably faster:
Ram, to a degree. 2 gig will choke it, 4 or 8 seems to be a sweet spot editing a one or two images in CS4 at one time.
64bit CS4 running on Vista 64bit, way faster than the 32bit version in XP.
10,000 RPM Hard Drives.
Separate Hard drives for Photos and Scratch files.
Raid level zero and ten with a REAL raid card, not motherboard raid!
I am unsure of the Video Card hardware acceleration since both systems the video card supports the CS4 GPU use.
Things that the Quad Core is faster at: Running VMWare with multiple Virtual Machines turned on, Encoding Video in my case taking DV to MPEG2 for DVD authoring.
So there you have it, given how I do things and what gear I have this should give you a pretty good idea.
Thanks for the great info dave. As I originally intended, I'll be keeping an eye out for an e8500. My e6550 has been a great work horse and does everything else without a hickup including a few games I play.
I do run two baracuda's at RAID 0, but I'm thinking about getting 2 more to make a 4 drive RAID 0 setup. Is there that much of a difference with a dedicated RAID card? I'm getting 130 MB/s constant transfer rate as is with the two drives working together, but it could of course always be faster. 10,000 and 15,000 rpm drives still to me do not have the value that current SATA drives do.
Yeah, I read your system specs... I don't think a quad would improve things much. And now I'm also noticing that not ALL of CS4 is multithreaded. Only some stuff... I did a few Lens Blurs today and wow... SLOW! Same as CS3... and system proc average held at about 10% for all 6 or 8 of the 30sec. progress bars... So, while some of it is better I'm discovering a few places where it still is not.
I think the CS line is somewhat limited in the use of 'multiple' cores beyond two - and with processor speeds so high, things happen so fast, that maybe it doesn't need to.
However, if you're doing more than just CS at the time, more cores will help. Also, I have ProShow Producer slideshow program that will use all 4 cores- as well as a couple other programs that I'm surprised will spike all 4 cores to 100% while rendering, processing etc...
So the quest for efficiency will be based on what needs you have now coupled with the money spent on your current solution VERSUS future needs with faster gear and ever reducing hardware prices.
Will Patterson wrote:
Thanks for the great info dave. As I originally intended, I'll be keeping an eye out for an e8500. My e6550 has been a great work horse and does everything else without a hickup including a few games I play.
I do run two baracuda's at RAID 0, but I'm thinking about getting 2 more to make a 4 drive RAID 0 setup. Is there that much of a difference with a dedicated RAID card? I'm getting 130 MB/s constant transfer rate as is with the two drives working together, but it could of course always be faster. 10,000 and 15,000 rpm drives still to me do not have the value that current SATA drives do....Show more →
The 10,000 rpm drives I'm talking about are SATA, but SCSI is certainly the king of performance hands down. The price may not be worth it for some.
Transfer rate score is total BS, it is a matter of your application running faster and sometimes transfer rate helps but it is not everything.
I did not mention a REAL dedicated raid card for my health!
Texturizer:.............. 1.03 sec.
CYMK CC:................ 1.27 sec.
RGB CC:................... 1.54 sec.
Dust & Scratches:.... 1.63 sec.
Water Color:............. 11.32 sec.
Texturizer 2:............ 1.21 sec.
Stained Glass:.......... 3.36 sec.
Lighting Effects:...... 3.41 sec.
Mosiac Tiles:............ 10.56 sec.
Extrude:.................. 44.52 sec.
Smart Blur:.............. 62.85 sec. Underpainting:........ 13.92 sec.
The first test set resulted in a Total Score of: 138.81 seconds at 3.06 GHz as opposed to 156.62 seconds at 2.66 GHz which is 17.81 seconds faster.
And the second one was: 25.01 seconds at 3.06 GHz as opposed to 26.53 at 2.66 GHz- so, about 1.5 seconds different.
The machine sure seems a lot zippier in comparison tho!
It runs no hotter either and stays as cool as always.
Anyway, so what does this mean? Well if you look at the number of procs, the clock speeds and the overall increase in test results it either means that PS was being slow because of non-CPU related subsystems or that the PS tools used are indeed NOT multi-threaded at all. So my information was kinda bogus even though it came from an inside source. Bummer.
I currently run a Q6600 at stock speed. What I would like to know (don't mean to hijack the thread) is whether to upgrade to e8500 or Q9550. Both are running 1333 FSB, 45nm procs, and the same media instructions (SSE4, etc.). The dual core is 3.16 GHz while the quad core is 2.83 GHz.
While posters above have stated that it seems clock speed wins out over number of cores, my question is, what about the cache? Q9550 has 12MB cache while the e8500 has 6MB Cache. That's twice as much - so, in order to better understand the differences, where does all that extra cache come into play? This whole balance between number of cores, clock speed, bus speed, and cache gets so obscure to all of us who are not uber tech savvy.
therock wrote:
I currently use the motherboard RAID controller. Can you guys put me onto some "Real Raid Cards"?
Before investing a lot of money in a "real" raid card i would spend some time investigating the likely performance gains (google is your friend here). Its a while since i looked at it really closely, but there has never been very much performance gain in using dedicated hardware for raid levels which don't include checksumming. And with the advent of multicore processors even this is debatable. Also, will an increase from 200Mb/sec to 300Mb/sec really make that much difference to your work?
I just built a server for processing images - up to 100,000 per day (yes, per day)
Granted, the tasks that I tested with were not 'normal' operations for most users. I had 4 independent threads rotating and shrinking 5000 images while another thread was dumping master images out to another computer along with another set of master images being pumped into the box - for a total of 10Gb of images processing and 20-30+Gb of images being copied.
The server is a Q9550 with 4Gb memory and 6 Velociraptor drives. Two of the drives are RAID1 off the motherboard for the two partitions: OS and smaller 'web' sized files. The master image partition was set to a single Velociraptor, RAID10 off the motherboard and I also tested a 24port, 64bit Areca ($1000) SATA RAID controller (granted, the Areca sat in a 32bit slot - so was kind of crippled from the beginning)
The single Velociraptor drive could feed the CPUs enough to process about 550 images per minute while copying 12Gb off the drives.
The Areca controller with RAID1E, could only muster about 500 images per minute but got about 14Gb off the drive.
The onboard controller with RAID-10 was able to process nearly 750 images per minute and get 16Gb out through the network.
Normally, I use the Areca contollers in RAID6 with 12+ drives, so putting the 64 bit controller in a 32bit slot was unfair - but...
Again, different workflow than most - and there are lots of different scenarios that will benefit from different aspects of a variety of hardware. But I'm pretty happy to save $500 on a controller. (although I do hope to test another card this week from 3ware)
