Topic Author's Original Post - May 10, 2007 - 05:52pm PT
Download the following image, open it in Photoshop, and then select:
Filter > Blur > Radial Blur.
On the Amount, put 100, and the Method, put Spin, and for Quality select Best.
Time how many seconds it takes for the Radial Blur filter to render image.
Please note the machine you are running and the processor type and speed.
The RAM and video card are irrelevant, this is pure processor action!
Here are some of the results from my machines - or those of friends:
Ok, on the cheapest Dell Optiplex you could buy a year ago with a 3ghz Pentium 4 with a gig of crappy ram ($450 with XP Pro) it takes 1:11 to render in PS Elements 4. I will try on my Core2Duo next week sometime.
I'll have to try this when I get home... I have a dual boot setup on my PC. I run XP as well as OS X on it. I guess I can also run it on my AMD box as well, which also runs OS X and XP. It will be interesting to see the comparison between PC hardware running OS X versus XP.
one of the top background painters in the motion picture business today (Fifth Element, etc.) runs Photoshop 4.5 no lie, w/ his custom brushes on a monster dual processor PC FYI
Any current 32 bit OS (MS Windows, earlier versions of OSX, etc) can use 4 gigs. My workstation runs three gigs PC 3200 on a single channel platform (pretty old school), so, basically anyone with a modern PC can blow my benchmarks away.
I'd build a new ultra crusher but, hey, my current box works great.
some (perhaps even several) of Photoshops plug-ins are multi-threaded but, the application itself is not.
Photoshop is a single-threaded app.
Raydog - please explain why my friend's Dual 2 Ghz G5 (48 sec) kicks the snot out of
my 2 Ghz iMac G5 (1:40). As well as the 2.5 Ghz G5 Quad Core (19 sec) smoking the
2.3 Ghz Dual core (41 sec)?
Are you sure Photoshop isn't multi-treaded? What accounts for the speed differences?
one other fact about CS2 -
it can see (use)2 gigs of RAM before it goes to scratch.
regarding cores- maybe if one supports the OS, then one runs the apps - you have a much better performance -
however I will say this, the area of benchmarking your rig is an interesting and complex one, even just benchmarking a HDD (64megs per second max data transfer rates for conventional 7200 RPM storage technology - pretty slow) - gamers do it, I'm not that into it.
The system runs, grinds thru it's duties with out a problem or a delay - I'm happy.
Enjoy your Mac!
Benchmarking AND over-clocking is way interesting though...
rendered in exactly 1:10 sec on a 2004 Gigabyte K8U board w/ 3 gig single channel memory (high latency) an Athlon 3700+ single-core w/ a clock frequency of 2.4 gigahertz running Photoshop CS2 off a 28 gig partition on a 7200 RPM (conventional technology, not perpendicular) drive.
My XP2 installation is fairly tweaked as well.
Now I'm going to try it after dragging the image into an appropriately sized Photoshop art-board as the second (not the background) layer and see what it does.
Ok, had to chime in here...
didn't try it yet but first I want to clarify a couple things...
If you put 4 gigs of ram in a PC,
the system bios will report 4 gigs but the OS will report
anywhere from 2.75 and 3.5 gigs. No way around that.
Any more than 2 gigs is really a waste and can result in a decrease in performance BTW.
Just because an app can virtually address 4 gigs doesn't mean it can use that much physical ram.
It's complicated
but basically it's limited by the Intel X86 design architecture.
Physically addressable memory and virtual memory are totally different.
Edit: The RAM and video card are irrelevant, this is pure processor action!
This is only half true. The video card doesn't enter the equasion for the actual calculations the cpu is doing but the SPEED of the ram is a big part of the overall performance.
Faster Ram = faster time.
Also the reason a multicore procesor will always outperform a single core, no matter what kind of app your running,
even if it's a single thread app, is because you are always running multiple threads! You're never just running one program.
Open task manager and you'll get the idea.
right Shack but are you sure the need for scratch in Photoshop is not decreased by running more than two? I mean since the app itself can use 2? What do you think?
ok I re-did the test and...
as layer one it took 1:17 w/ my old school tank of a PC.
my bottom line sentiment on my own (personal) computer usage is that the thing is a tool, like a shovel, and as long as it digs the holes I need in a manner suitable to my style the shovel is working and I do not need a bigger one.
If my cave man PC breaks, I can fix it and, with overnight FedEx from Newegg - have it back up and running quickly and for minimum bucks - this is why I choose build, not for some theoretical reason, but because all systems fail: it's a when not an if.
The question about memory useage etc. is a really interesting one.
Just woke up from a nap and checked the thread, still waiting for a PC (Piece of Crap) to beat my time... Lots of dogs barking, but no bite! Ha ha ha...
OH and yes 667Mhz Ram is very common right now.
800Mhz is the max unless you overclock or get the EVGA 680i MB...it was developed in parallel
with the Corsair Dominator RAM.
The MB recognizes the ram when you put it in and automatically
sets the CAS latency times and the speed to 1142Mhz!!
And corsair just came out with an even faster one that runs at 1250Mhz!!
For true power in a PC,
you need to jump to 2 or more physical procesors.
It used to be more common but now you see it only in servers.
(mostly because you need to run NT)
Ray are you talking about something like this? hehehe
My Powermac Dual G5 2ghz has a dead logic board that I can't afford to fix just yet so I'm falling back on a 800 mhz G4 and I'm not going to worsen my envy by testing this, cause I'll be waiting for photoshop enough in the next few months.
I ain't testing my laptop either cause It would be like racing a donkey
Ray, look at the specs of the system I posted about earlier.
The MB supports upto 1333Mhz Front side bus!
Supports 1200Mhz DDR2 Ram!
Put a Dual or Quad core in there with the fast ram and a good video card...
It will Kick Ass all over a comparable AMD. Guaranteed!
Don't get me wrong, I think AMD is really good too...
but you can't build an AMD that fast.
All parts are available at Newegg too!
Put them in a Cooler Master Stacker 830 case and lookout.
Shack, I know you're right, another PC friend of mine says the same thing - maybe I will go that route. Honest I'm more concerned about the next system having bomber Quadro GPU support, massive screen real estate and either Raptor HDD's or Seagate Perpendicular stuff.
Alright,
this one kicks ass on everything...
badass!
and yes it really is a dumpster PC and yes I really did start out building 'em out of pure and total junk.
smokin!!!
(that HDD that's hanging out is running the show, of course)
The Raptor X has a clear cover so you can see the platters and the heads!
Run 2 of them in RAID 0 and you'll get burst transfer rates of 200MB/second! That's Smokin'!
Stay away from RAID 0 if you care about your data.... either disk fails and your data is lost, so double the likelihood of data loss with no redundancy.
If you need that much throughput, better to go with a RAID 1+0 approach (but requires more disks). Or go with RAID5 if you want to skimp a bit on resiliency but still be better than RAID 0.
"Stay away from RAID 0 if you care about your data.... either disk fails and your data is lost, so double the likelihood of data loss with no redundancy. "
So are you saying that everyone who only runs one hard drive is better off?
If 2 drives doubles your chances of failure, then RAID 5 is 4 or 5 times as likely to fail?
Actually, with more drives, they are LESS likely to fail...
With 2 disks, each disk gets "used" half as much!
Just keep your data backed up and you'll be at no greater risk of data loss than any standard 1 drive PC.
I didn't want to have to do this, but here goes... it's sort of like the equalization vs. no extension debates in building anchors... what do you want to optimize for?
say you have 4 disks. There are several ways you can logically configure them to optimize for speed, resiliency to disk failures, or both:
JBOD:
"Just a Bunch of Disks" where you have different drive letters in a PC, don't know how MACs do it. But you manually copy your data around to back it up. It's not real-time and transparent to the operating system like the RAID versions are.
RAID 0
stripe some data across each of the 4 disks, so each file actually exists in bits spread across each disk. Lose any one of the 4 disks, and the file/data is corrupt. But it's very fast, because your PC rotates between writing to all disks rather than waiting for any one of them.
RAID 1
mirrors each bit of data on each disk. It's totally redundant, no single point of failure. Any disk dies, and you have perfect copies of your data on the other disks. But it's just as slow (if not slower) than having a single disk.
RAID 1+0
Have two groups of two disks. You stripe data (RAID 0) within a group to increase your throughput, and then you mirror this group for resilience. If you lose a disk, then the other member disk within that group contains corrupt data. But you have the whole group mirrored in another 2 disks, so those other 2 disks together contain your complete set of data.
There are more setups, but we'll keep it simple for now.
I believe he's referring to the fact that there is no mirroring with RAID 0. You have data spread across 2 drives - if one fails, neither is accessible. With later versions of RAID, 1+0, or 5 for example, the data can be spread across a set of volumes, displayed as one, but is also mirrored for redundancy.
edit: sorry, nutjob. your post wasn't up when I started typing. And you know you actually "enjoyed" doing that. =)
I understand what all the different RAID levels do.
My point is:
So what if there is no redundancy in RAID 0!
There is no redundancy in a single drive setup either.
Raid 0 gives roughly double the throughput of a single drive.
We are talking about performance...not data security.
WD Raptors cost 4 times as much as a standard drive.
Too expensive to use in a RAID 5 setup.
4 times as much as a standard drive isn't much if you remember the days of drives that were actually expensive AND ran SCSI drives.
I remember those days. I ran a local BBS (talk about geek alert) that had 15 Gigz online, which was HUGE at the time. Half the space was Quantum Fireball 1080's, (1gig) whihc were fast, but also about a grand each. Seagate Barracudas were the only thing faster at the time, but since those were SO expensive, and since I had a friend working at Quantum...
Good discussion on RAID setups, gentlemen. Very informative.
In my initial post, I related that my 2 Ghz Macbook rendered the horse in 29 seconds.
Well, truth be told, that was knott accurate. It was actually 32 seconds.
Here's a cool video I made of the action (you'll be on the edge of your seat, I promise)...
Interesting that the 2.33 Ghz Macbook Pro did it twice as fast in only 16 seconds,
since the processor isn't all that much faster - at least knott on paper...
I'm posting this as a followup to my last post (5 up).
I went to an Apple store yesterday to try the test on the new 3 Ghz, 8 core Mac Pro.
Alas, there was knott one on display. However, since I found the results of the
Macbook Pro claimed above to be astonishing, I thought I'd run the test on the
same configuration at the store (2.33 Ghz). I did the test twice, and both times
got 29 seconds (3 seconds faster than my 2 Ghz Macbook). This seems more
like it, given the 10% faster processor...
In the meantime, prepare to bow down to the mighty Mac Pro 8 core!
Ok, to shut the crickets up for a moment, 22 seconds on my Core2Duo with 2 gig of ram. No overclocking running pc800 memory. But then this doesn't let me use my Nvidia 8800GTX video card to it's fullest advantage.
In that the only reason I have a PC at home is to play games, I don't really care how fast it will do a radial blur. I care about my frames per second when playing Crysis.
Early 2011 2.2 GHz MacBook Pro Intel Core i7
Mac OS 10.7.3
8 GB RAM (1333 MHz)
Photoshop CS5.1
256 GB SSD
6.8 sec
Quite impressive that the MacBook Pro beats Jerry's 3.2 GHz quad-core tower.
Interesting that the 5 year-old (2007) 3 GHz dual quad core still has the best time of 5.5 sec
(as seen in video a few posts above). Still a lot of performance from that ancient machine...
Let us know!