Photo processing benchmarks?

[PeterAit]

My photo PC is getting on to 3 years old. It works fine but I am starting to wonder if hardware has advanced to where I could get a meaningful speed boost with a new system. It's hard to tell, though - just how fast is my current system compared with today's fastest? If I could double speed I might consider an upgrade, but a smaller increment would likely not be worth it. Thus, I am looking for valid benchmark(s) so I can compare my system with published results for current hardware. Thanks for any suggestions.

[ErikKaffehr]

Hi,

Try geekbench, http://www.primatelabs.ca/geekbench/

It's available for Mac, Windows and Linux.

Regarding poto processing capabilities, it is hard to know. Adding more memory or an SSD disk may improve an existing system. Or switching to a 64 bit OS, as 32 bit OS-es are limited to around 3GB of RAM.

General idea:

If you see a lot of hour glasses and spinning beach bolls it means that system is working mutch against disk. Adding RAM or SSD for swap would help a lot. You need 64 bit OS if you add more than 3 GByte RAM. Geekbench would measure upper bound on performance if the system is not bottlenecked by RAM or disk.

Best regards
Erik

My photo PC is getting on to 3 years old. It works fine but I am starting to wonder if hardware has advanced to where I could get a meaningful speed boost with a new system. It's hard to tell, though - just how fast is my current system compared with today's fastest? If I could double speed I might consider an upgrade, but a smaller increment would likely not be worth it. Thus, I am looking for valid benchmark(s) so I can compare my system with published results for current hardware. Thanks for any suggestions.

[Farmer]

3 years so you're probably on a Core i series?  I'd wait for IvyBridge - updating to SandyBridgeE is going to be less impressive than waiting a couple of months for SandyBridge.

[hjulenissen]

How many cores can photoshop, lightroom etc use in a meaningful manner? If they scale anywhere near linear up to e.g. 16 cores, that is a simple (although expensive) route to performance beyond the typical PC at any point in time.

Using your GPU for photoshop seems like a feature that is talked a lot about in the media and marketing but is it really cost-efficient?

I have a 64-bit system and 12 GB of ram.

I do recommend SSD as that makes the computer experience generally more snappy, it may or may not affect your photoshop activities (I have 20GB og lightroom cache on my SSD).

You might check up things like this:

To measure performance under Photoshop CS4 we turn to the Retouch Artists’ Speed Test. The test does basic photo editing; there are a couple of color space conversions, many layer creations, color curve adjustment, image and canvas size adjustment, unsharp mask, and finally a gaussian blur performed on the entire image.

The whole process is timed and thanks to the use of Intel's X25-M SSD as our test bed hard drive, performance is far more predictable than back when we used to test on mechanical disks.

Time is reported in seconds and the lower numbers mean better performance. The test is multithreaded and can hit all four cores in a quad-core machine.

-h

[Rhossydd]

I'm also interest in how more modern systems compare to a couple of generations back. CPU clock speeds or numbers of cores haven't changed that much, but trying to understand how increases in ram speed/bandwidth, larger on chip caches and new architectures effect performance on real world tasks is difficult to find, most reviews (as above) only seem to compare current models or just back one generation, if that.

Thus, I am looking for valid benchmark(s) so I can compare my system with published results for current hardware. Thanks for any suggestions.

I think the only way to do this is first find a site with some current benchmarks you'd trust, then try to get hold of the same software as they use.
Unfortunately this may prove quite difficult.

The second issue is what use will benchmark results be ? You'll see different performance with different software, so what might work well and give benefits for Photoshop might not give as great benefits with, say, Lightroom.

Comparing to three years back, I think the arrival of affordable SSDs can make a big difference to performance, plus the latest boards can take lots of, currently cheap, ram. I'm highly tempted to upgrade to a 3820 with 32gb ram and another SSD. It's difficult to know if the forthcoming ivy-bridge offering will be significantly faster at an affordable price in the near future.

[hjulenissen]

I'm also interest in how more modern systems compare to a couple of generations back. CPU clock speeds or numbers of cores haven't changed that much, but trying to understand how increases in ram speed/bandwidth, larger on chip caches and new architectures effect performance on real world tasks is difficult to find, most reviews (as above) only seem to compare current models or just back one generation, if that.

Actually, Anandtech lets you compare a bunch of cpus in terms of this photoshop CS4 test, see link below:
http://www.anandtech.com/bench/CPU/25

This link includes simple Core2 duo, AMD Sempron and recent Intel i7s.

Of course, the processing time for a given set of actions in Photoshop CS4 may or may not correlate well with the perceived "snappyness" of your actions in some photo app. But use with care, I think such benchmarks have their place.
-h

[Rhossydd]

Actually, Anandtech lets you compare a bunch of cpus in terms of this photoshop CS4 test, see link below:
http://www.anandtech.com/bench/CPU/25

Interesting data, I guess that's about as useful as it gets.
The top two give a fine example of being careful not to waste money on kit that doesn't deliver value. Just .3s difference (3% improvement, could you spot that ?) but the difference in price is over 300% !!! £233 against £779

The added complication is that those tests are probably done at default settings, whereas optimising bios settings for speed rather than economy can make a big difference to perceived performance, then some were sold specifically for over clocking so in real world use often deliver more than the stock figures might suggest.

[Farmer]

It's difficult to know if the forthcoming ivy-bridge offering will be significantly faster at an affordable price in the near future.

My system is a similar age - perhaps slightly older, but with a mix of a few new parts.  I'm deliberately waiting for Ivy-Bridge at this point.  It brings with it better USB3 intergration, potential thunderbolt, PCI-E3 (and the potential for OCZ to deliver a PCI-E3 Revodrive for even more peformance), quad-channel memory architecture and the 24nm processor architecture.  Some discussions on this board with Steve Waldon and John Murray have highlighted a number of likely benefits and speed improvements from these various updates.

Given the proximity of release compared to what's available now, to me it makes sense to wait and then upgrade to the new architecture rather than take the latest of the current.  It's likely that the upgrade path will be easier for the next few years based on an Ivy-Bridge system rather than continuing with a Sandy-BridgeE (let alone continuing wiht the level of machines that we likely both have at this point).  Incremental updates during the lifecycle are often quite cost effective and worthwhile.

[Rhossydd]

Given the proximity of release

Maybe the end of June ? http://www.tomshardware.co.uk/Intel-Ivy-Bridge-Delay-Confirmed-cpu-processor,news-37861.html
Then add the premium of the early adopter and a wait to make sure it's a safe platform (remember the Sandy bridge debacle of having to recall every board sold?).
I doubt it'll be a sensible proposition for most of us until nearly Christmas.

Meanwhile LR4 runs like treacle......

[Farmer]

Yeah, delayed, but I'm looking at August to buy.  The Sandy-Bridge issue isn't likely to be repeated and you do have the option of continuing with LR3.

However, a question about your LR4 - did you rebuild your catalog with new previews etc?

EDIT: That's August at latest - if it's only delayed by a few weeks to a month as others have suggested then it could be late June or early July that I'll be jumping in.

[Rhossydd]

The Sandy-Bridge issue isn't likely to be repeated

I admire your optimism ;-/

you do have the option of continuing with LR3.
However, a question about your LR4 - did you rebuild your catalog with new previews etc?

WRT to keeping using LR3; On the portable I have no choice as it runs XP and I don't want to spend lots upgrading that. The core location jobs of checking for dust or other issues and keywording will still be fine.
LR4 does have some useful new features I'd like to use, so the main desktop is using it. Hopefully Adobe will pull their finger out and crank up the responsiveness in a future update.

Rebuilding the previews ? no I haven't knowingly done that, but the LR3 preview folder is about 2k files @ 1gb whereas the updated LR4 preview folder is 20k files @ 50gb
what's going on there then ?

[Farmer]

Well, yeah, it might be an aspect that needs to be looked at - you can imagine that things will be slower with that kind of difference.

I'm sure Adobe is aware of the issue - it's affecting some and not others - so I'd expect an update at some point!

[149113]

I just ran this on my i7-920 (OC at 3.9) with 24GB and got 13.5. Hand timed so not exact.

[Rhossydd]

it's affecting some and not others - so I'd expect an update at some point!

People are obsessing about ram & cpu speed, but I suspect a lot has to do with screen resolution. On LR3 I noticed a significant slow down when I upgraded my main monitor to a 27" unit, suddenly we're talking a very big stretched desktop to refresh every time one of those sliders is moved. All the people just using laptops with little screens seem not to be shouting much.

I doubt the first update will sort the speed issues out based on past LR development.