LDBecker wrote:
I don't want to talk about it... it was painful... Larry

By NewEgg prices and then the case:
- Core i7 Extreme 965: $1000
- i7 mobo: $250
- Geforce 280: $250
- 12 gb: $500
- 2x Intel SSD's: $1300
- 2x WD 1TB: $150
- Zalman case: $700
- Vista Ultimate 64: $200
- CS4: assume upgrade at $150
TOTAL: $4500 assuming you kept your keyboard, mouse and monitor
This is today pricing - not a month or two ago.

LDBecker wrote:
I just got tired of having pcs go obsolete every 2 years.

Hate to break it to you Larry:
http://en.wikipedia.org/wiki/Moore's_law
Two years is the going rate for bleeding edge technology - or maybe less?

So how long do you think that that box will be fast enough when in two years, your secretary has a 8Ghz, 16core laptop with 1TB of solid state that acts as hard drive and memory?

BTW - I do envy that case!!! I am on a quest for a quiet PC myself - replacing fan after fan with water or Noctua fans! (Ok, so I'm jealous of the whole box!)

Yes, Hammy... I'm bleeding... thanks for poking a stick in my wounds!
My point in building it wasn't to build the cheapest, or to have a state of the art pc for several years... I just wanted to build the best pc I could build. And, three months ago, even if I could even find someone to build it for me, it would have been double the cost - but such things weren't really available in the mainstream. Dell's Alienware came close, but at nearly $9k it wasn't really feasable, and not up to the spec I wanted. I built it for about half, so your numbers aren't far off.
Another reason for doing it myself was I could put the components in it that I wanted, not whatever was available. It was especially an issue on drives and graphics cards. I also liked the challenge - I hadn't built a pc from the ground up in over 15 years - things changed just a little bit since then...
So now, when a better significantly improved processor/mb combo comes along, I can replace just that. Or graphics card (not easy to find them watercooled a few months ago...), or drives... ORRRRR - an operating system - Win7 is coming soon... The parts that are going to really get obsolete quickly are the SSD drives. Bigger and faster drives are coming quickly, I'm sure. Then I can retask these to other uses -- still WAY faster than any harddrive, and much more reliabale long term.
In 2 years I can replace bits of it, or OC the system - don't really need to do that now (although my graphics card is OC'd from the factory).

Larry (bleeding but having fun)
Edit - the case IS really nice... one of the best parts of the build. It will last for several more builds - it's that good... completely quiet, very cool, no distractions, weighs a TON

LDBecker wrote:
My point in building it wasn't to build the cheapest, or to have a state of the art pc for several years... I just wanted to build the best pc I could build.

Larry,

I totally agree, I've been building boxes for over 20 years. Used to work for some computer stores, and now have my own business. I did buy some Dells a few years ago to get things started - complete packages with warranty did the job for close to what I could build them for - but they had no specific task.
In the past few months, I've built about 18 boxes:
- 12x Quad 2.5Ghz, 4Gb, 640Gb boxes for image download and simple image processsing
- 3x Servers with Quad 3Ghz, 4Gb mem, 7x Velociraptors in RAID10 with dedicated Areca controller for massive image handling and web serving
- a few more boxes inclucing my CS4 box: Q6600 oc'd to 3.4Ghz, 8Gb mem.

So I understand being able to build your own and getting what you want. My old Q6600 oc'd to 3ghz and realizing its quite loud in my office so I've looked at that Zalman case in the past few months.

For servers especially - every time we look at building a high end server, a name brand costs 3x what we can build one for - and we can tune ours for our needs better.

Sorry about keeping the wound open - I really do like the components of your build! It'll be a few month from recovery of the above systems (and 36 DVD duplication towers) before I can build my own monster box.

Congrats for doing it right and not getting an overpriced name brand.

LDBecker, I second the congrats for doing it right and not getting an overpriced name brand.

I am curious now. Do you notice a difference between letting photoshop use your graphics card to help versus with that feature turned off?

Second curiosity is wow come using 64bit was faster than 32bit? I could swear this process doesn't take more than 3.2gb of ram (which is the max that 32bit can address) so I would have thought 64bit doesn't have any advantages. Did you maximize the amount of ram your CS4 32bit can use under performance options?

okay, nvm, just checking memory usage and with 32bit, it is maxed out already so it probably has to use scratch disk for this operation and take way longer than 64bit.

Edit:
Oh WTF! I'm comparing what happens under 32bit when using the C drive as the scratch disk versus using a separate physical drive and now I'm getting that photoshop only needs 2.6Gigs... gr.... This might be because I saved the post noise added image so photoshop doesn't need to remember as much in it's history states.

Strangely, I am getting faster times on the 32bit than my original 64bit time so something might be wrong

under 1 minute!!!

small disclaimer: that is, under 1 minute for it to fail and tell me there's not enough RAM

2 GHz cpu, 2 GB of ram, laptop

Okay, I just did a bunch of tests and here are the results:

Hammy wrote:
- Core i7 Extreme 965: $1000

- 12 gb: $500


Yeah, that $1,000 vs a 920 at $265 is pricey. I usually hang back at least taht amount - to the 920 - or a quad.

Unemployed now though, so I just let it go ...


Go price 24 Gb of DDR3 RAM though - about $2,800! Steep premium for the 4Gb sticks over the 2Gb.


I too find that I usually can't build a box with an OS that I have to buy for less than Dell prices.

Have fun!

Cheers,
M.

Weiyang Liu wrote:
I'll start off:
PC - i7 920 @ 4.33 Ghz w/ 6GB Ram @ 1648Mhz - Benchmark: 3:02

Ten days later - after somebody with a better system that is NOT overclocked, but better components overall that posts a legit time of 2:42 using 64bit software, and almost a minute LESS using 32bit software

Weiyang Liu wrote:
Okay, I just did a bunch of tests and here are the results:

...all these trials showed approximately 2:38-2:39
(and 3:12 for 64bit)


So what got upgraded in the past 10 days to get 25 seconds FASTER in 32bit?
or what got degraded to get 10 seconds SLOWER in 64bit?

Or is it just convenient that re-running the tests put you back as the fastest in this FM poll?

I'm really actually impressed with the OC results that the OP stated. I haven't seriously tried to do that - made a couple of attempts and, well, thank goodness EVGA motherboards have reset button - that's all that needs be saiid about that...
Anyway, that's pretty impressive. One of the reasons I went with thfe Intel 965 was it was significantly overclockable, where I had read that the 920/940 processors were "locked." Turns out that lots of people are having success doing that. Cool. I actually don't have a lot of time to play, so I wanted something that just worked and was really fast...
I also had NO experience with water cooling and wasn't sure how effective it would be. My last (air cooled) Dell P4 was so noisey and ran so hot that I was determined not to have that problem again. Turns out that this case is really, um, cool - pun intended. I've got 2 ssd's, 2 1TB drives, a water cooled graphics card, a 1KW Zalman power supply (that is itself very quiet), and a sound card - all running dead quiet, very acceptable temps (36 degrees C as I write this here in Los Angeles), and I'm encouraged that WHEN I feel the need for more speed, I can take the system up to the 4ghz region and be ok. If I sort out the OC thing.
I will probably be driven to do that because of HD video editing. I have a Sony HD camera, and the 5DII -- those files are just plain huge, and the overhead of the AVCHD format is obnoxious. Even on my (yes, insane) system, I get a little stuttering of video from time to time. Grrrr. Probably will go to something like Cineform Neoscene to manage the files... not a need today, but soon.
So --- CUDOS to the successful overclockers! I'm envious and in awe! I may be asking for help soon to OC my system!

Larry

Hammy wrote:
Weiyang Liu wrote:
I'll start off:
PC - i7 920 @ 4.33 Ghz w/ 6GB Ram @ 1648Mhz - Benchmark: 3:02

Ten days later - after somebody with a better system that is NOT overclocked, but better components overall that posts a legit time of 2:42 using 64bit software, and almost a minute LESS using 32bit software

Weiyang Liu wrote:
Okay, I just did a bunch of tests and here are the results:

...all these trials showed approximately 2:38-2:39
(and 3:12 for 64bit)


So what got upgraded in the past 10 days to get 25 seconds FASTER in 32bit?
or what got degraded to get 10 seconds SLOWER in 64bit?

Or is it just convenient that re-running the tests put you back as the fastest in this FM poll?



You don't have to believe me but if you want a proof, I can post a video if you are willing to make it worth while for me. As for the 10 second gain from 64 bit, well I found that different add noise trials changes the run time. Since add noise is a randomized command, each time u do it, it is slightly different. There is another possible reason. The 4.3Ghz that I am getting is at bclk=206 with multiplier set to 20 with turbo on. This means the multiplier goes to 21x in almost all cases unless your cpu is being overstressed (eg. using Prime95, it goes down to 20x in 2 seconds). With this photoshop benchmark that gained 10 seconds, it could be because the ambient temperature in my room increased enough to cause throttling from 21x down to 20x (although I doubt it cuz i monitored that before and it seemed to stay at 21x).

LDbecker >> There are ways around the "locked" cpu's and they usually come from the motherboard being good. CPU multipliers have been locked for quite some years now except when you pay for the highest end ones. The way to overclock is to overclock the bus clock going into the cpu. So say a 2.0Ghz computer is made from a processor with 20x multiplier and 100Mhz bus from the motherboard. On a high end processor that is "unlocked" you can increase the multiplier. On a "locked" CPU, what you do is increase the 100Mhz bus to say 150Mhz and the cpu till multiplies it by 20x giving you 3.0Ghz overclocked speed. OCforums.com is an excellent resource. I highly suggest it before you overclock.

Weiyang Liu wrote:
You don't have to believe me but if you want a proof, I can post a video if you are willing to make it worth while for me. As for the 10 second gain from 64 bit, well I found that different add noise trials changes the run time. Since add noise is a randomized command, each time u do it, it is slightly different. There is another possible reason. The 4.3Ghz that I am getting is at bclk=206 with multiplier set to 20 with turbo on. This means the multiplier goes to 21x in almost all cases unless your cpu is being overstressed (eg. using Prime95, it goes down to 20x in 2 seconds). With this photoshop benchmark that gained 10 seconds, it could be because the ambient temperature in my room increased enough to cause throttling from 21x down to 20x (although I doubt it cuz i monitored that before and it seemed to stay at 21x).



I certainly don't need proof. But it is obvious that this 'benchmark' is seriously flawed:
A: you've chosen a 'randomized' filter that has varying degrees of time
B: somehow, you're 32bit CS4 runs quite a bit faster than 64bit but not on other boxes.

For a "future proof" bench, it should be 64bit tuned - that is the future of computers and OS and software.
For you to get SLOWER times on a box that is FASTER than Larry's, then either you need to contact Adobe for a crippled version of 64bit software, or your benchmark is memory limited, or the capability of the Extreme processor (cache?) simply walks away on your version of a benchmark.
In ALL other benchmarks that are realistic, there is a VERY LINEAR scale per processor family based clock speed.

So obviously, there is no static benchmark comparison to be had here. It is hardware dependent, software dependent and variable in pattern.

Hammy >>

I agree with what you are saying. I just don't see how in the world the 32 bit version runs faster than the 64 bit with a modern system running 64bit operating system. The things that comes to mind are that (a) I screwed up in the install or (b) windows7 64bit is still very much not 64bit optimized yet since it is still a release candidate that I am using.

In terms of using a randomizing function, I thought 64,000,000 pixels were enough to create a distribution that's unique AND about the same difficulty to process with filters. I was expecting a variation of <1% but instead got a variation of a few %.

Hammy wrote:
In ALL other benchmarks that are realistic, there is a VERY LINEAR scale per processor family based clock speed.

So obviously, there is no static benchmark comparison to be had here. It is hardware dependent, software dependent and variable in pattern.


I wouldn't say that is true at all. Harddrive and memory usage is pretty important. For example, in this benchmark, I scored faster times than the 4.0ghz i7 extreme on all processes that lasted a few seconds or longer, but scored slower times for processes that were short. There was a lack of linearity compared to their times. You have to realized that if you want "realistic" benchmarks as you are saying, then it is indeed supposed to be not just cpu dependent. It needs to be hardware/software dependent and depends on multiple facets of software/hardware. These realistic benchmarks are meant to test what it takes to do a process on a computer that you might actually use. For me, as I've stated above, is for testing noise reduction but since not everyone has noise reduction software, I used this. Unrealistic benchmarks are the ones often referred to as the synthetic benchmarks that are specifically tailored to test a specific hardware or software component (or a specific set of hw or sw components), and be highly independent of the other components. Maybe this is what you were looking for?

Now I kind of feel like going back the the driverheaven benchmark to see how my 32bit times compare with my 64bit times. I'm hoping it's because win7 64 is not optimized yet.

Weiyang,

I too am confused or something isn't running right - like you suspect.

When I see benchmarks like:
http://www.anandtech.com/bench/default.aspx?b=25
we see a very linear transgression of improvement.

There is no doubt that Larry has a very nice system, but I would think that your 20% higher clock speed should offer more merit on the 64bit side - unless your test really is RAM limited and Larry's 12Gb takes it in stride. But that makes no sense that your 32bit times are faster

But yes, I would think that to get a general basis of how fast a box runs a test, would be more than just one filter. Yes, there are always factors to consider and some tested are tuned to take advantage of certain components. But broadening the spectrum running a realistic set of filters over static images would result in truer results.

Model Name: MacBook Pro OS 10.5.7
Model Identifier: MacBookPro4,1
Processor Name: Intel Core 2 Duo
Processor Speed: 2.4 GHz
Number Of Processors: 1
Total Number Of Cores: 2
L2 Cache: 3 MB
Memory: 4 GB
Bus Speed: 800 MHz

CS4 32 bit

Total time 20:00 no scratch disk

Jeff >> How do you set it up to have no scratch disk? Under performance preferences, it forces me to select at least one hard drive for scratch disk.

AMD 9850 Quad overclocked to 2.8g/core 8gigs ddr2 800
Running Windows 7 64 bit
10.08

Weiyang Liu wrote:
For example, in this benchmark, I scored faster times than the 4.0ghz i7 extreme on all processes that lasted a few seconds or longer, but scored slower times for processes that were short. There was a lack of linearity compared to their times.

Okay, I redid that benchmark using 64bit getting 159 seconds which is perfectly linear this time around. Using 32bit, I got 180 seconds. This makes it even stranger how for the radial blur filter, I was able to achieve a faster time with 32bit CS4.

Jeff Donald wrote:
Model Name: MacBook Pro OS 10.5.7
Model Identifier: MacBookPro4,1
Processor Name: Intel Core 2 Duo
Processor Speed: 2.4 GHz
Number Of Processors: 1
Total Number Of Cores: 2
L2 Cache: 3 MB
Memory: 4 GB
Bus Speed: 800 MHz

CS4 32 bit

Total time 20:00 no scratch disk

Model Name: MacBook Pro
Model Identifier: MacBookPro3,1
Processor Name: Intel Core 2 Duo
Processor Speed: 2.4 GHz
Number Of Processors: 1
Total Number Of Cores: 2
L2 Cache (per processor): 4 MB
Memory: 4 GB
Bus Speed: 800 MHz

I had about 16 minutes, maybe 10.5 runs slower or needs more ram?

^ wow, that's a huge difference on basically the same computer. Could the L2 cache make that big of a difference?

Core2Duo 2.2, 8GB RAM, Xp 64bit, CS4 64bit.

31.33

Ouch. Was still browsing the internet with a couple of browser windows plus has the usual anitvirus, backup software, etc running in the background if that helps. I didn't see that PS used more than 3.5gb of RAM, certainly didn't go to scratch...

18:20

15'' - Macbook pro OSX 10.5.7

Photoshop CS4

Video Card: GeForce 8600M GT - 128 Meg RAM

2.2 GHz Core 2 Duo, 3 gigs ram. ~ 3 years old

not too bad for a 3 year old LAPTOP

8:35
Dell Studio 540
Intel Core2Quad Q9400 2.67GHz x4, Bus speed = 1333MHz
Microsoft Vista Home Premium x64
RAM: 8GB DDR2 SDRAM at 800MHz
Video Card: ATI Radeon HD 3450, 256MB
Photoshop CS4 64-bit

I'm pretty satisfied considering I paid $450 for the computer.