i7 3770 vs 3820 for photoshop?

1       2      

Registered: Oct 22, 2008
Total Posts: 9300
Country: Japan

There's no difference. One just takes longer so I can get a better idea of how the code execution is scaling across the cores. It still does everything one at a time just like if you came home and wanted to process only 10 images. The trouble with processing only 10 or something is that it's so lightning fast on this Xeon system that I can't tell if the procs are even being used at all.

That said, I've come home with nearly 2,000 RAW images a few times!

sjms wrote:
yep I run out and shoot 6000 raw images and then batch process them a few times a week. I'm sorry I forgot about that.

And I didn't batch process all 6,000. Just created a large catalog to work with and then processed a few at a time in a few various ways.

Bifurcator wrote:
I don't think it's at all irrelevant.

I thought this was one of the better discussions on FM about buying computers! I'd like to continue a little with the Workstation thing tho.

Let the truth be known, PS CS6 even more than PS CS5.5 scales really well across multiple processors. People saying it doesn't are probably still using CS4 or maybe early CS5. I rather dislike LR so I dunno exactly but the versions I have tried scaled fairly well and Adobe has said they are making extra efforts to improve MP scaling in LR and PS. The CS video editing apps especially, just scream the more physical procs ya give them. And I wipe the floor with home built overclocked boxes when it comes to apps like Handbreak. One of my 5-year-old Xeon machines at 2.66 x 8 still keeps up or overtakes the OC Gammer boxes of 2012 for video processing and is right there at shoulder height from image editing too. Virtual (logical extra) procs as I explained earlier do almost noting for typical single app performance. So it's quite easy to calculate directly:

3.8 x 12cores = 45.6
3.6 x 12cores = 43.2
3.4 x 12cores = 40.8
3.0 x 12cores = 36.0
2.6 x 12cores = 31.2

3.8 x 8 cores = 30.4
3.6 x 8 cores = 28.8
3.2 x 8 cores = 25.6
3.0 x 8 cores = 24.0
2.6 x 8 cores = 20.8 (shown in the screen-shot below - actually 2.67 x 8 so the potential is 21.36GHz)

3.8 x 6 cores = 22.8
3.6 x 6 cores = 21.6
3.2 x 6 cores = 19.2
3.0 x 6 cores = 18.0

4.0 x 4 cores = 16.0
3.6 x 4 cores = 14.4
3.2 x 4 cores = 12.8
3.0 x 4 cores = 12.0

And yes, I do see the full potential of all 8 (PHYSICAL) cores in PS CS6 about 80% of the time when PS is actually processing something. Even just opening a RAW file uses all processors. Just like an Corei7 does. And I'm constantly checking the CPU graph too so I know of what I speak.

Just finishing opening twelve 28mpx images - notice the CPU temperature and fan-speed besides of course the CPU usage.

And if you just HAVE TO overclock your system there are ways to overclock some Xeon setups:
And I've seen setups that mask socket connections in order to do similar things on yet other systems.
They don't typically interest me tho as I don't need to overclock in order to go fast. Most of the apps I use for photography, video, and animation scale just fine and as the numbers above show, that's plenty good enough for me.

Just because there was one CPU named here which was ridiculously priced doesn't mean they all are - as indeed the Dell system prices show.

IMO what we have here are some guys who love to tinker with their hardware advocating the most tinkerable systems. What was it Swoop said: "...for 10 years now (about 8 different systems)..." whereas someone like me doesn't touch anything for 5 or 6 years or longer. I still think all 4 ways have good merit!

1 Build your own PC
2 Buy a prebuilt PC
3 Build your own WS
4 Buy a prebuilt WS

And each will appeal to different individuals differently. In my particular case the appeal is in the reverse order of the list above there (4, 3, 2, 1) mainly because of the simple math shown in the MC potential horse power calculations above and the support packages which are available. I have both here (DIY PCs, and Prebuilt WSs) and there's no question in my mind which I prefer. Others here have other preferences - of course - but a WS system still remains IMO anyway, a viable solution to consider for any photographer thinking about getting a new system.

---------- EDIT ----------
I reinstalled LR and updated it to v4.3 and it scales excellently! I created a catalog with 6,000+ RAW files from the GH2 and applied various modification, ran a few plug-ins, and just generally had another look around under the hood so-to-speak. This is very typical of what it looked like when it was doing something computationally demanding:
Other times the graphs were just many little peaks of between 20% and 90% depending - but I didn't see almost any operations where only a single core or two peaked without the others. So that's an excellent indication that LR is very capable of best utilizing MP/MC systems! And also again even after an hour of hard use trying to push the application the CPU temperatures never went much over 30˚C and the system remained virtually silent.

Registered: Jan 27, 2008
Total Posts: 3909
Country: Norway

Batch exporting large numbers of images from LR is one of the places that multiple cores really do get well utilised. In general use LR also scales pretty well, but does seem to have a bottleneck when it comes to noise reduction.

Bifurcator - if you feel like getting back into geeking for a day, your input to the recent LR performance scaling thread would be more than welcome - http://www.fredmiranda.com/forum/topic/1165920

Registered: Oct 22, 2008
Total Posts: 9300
Country: Japan

I will a little later for sure! Thanks for the invite.

Just to note tho, the CPU usage being shown was from displaying a before and after window with a buttload of adjustments applied - including NR.


1       2