#: 90570 S16/Photoshop 11-Nov-93 20:15:31 Sb: #Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: all Sorry about the sensationalistic subject name! This, as many of you know, will hopefully be a useful thread in comparing (albeit unscientifically) the Mac and PC performance with PS under different conditions and on different systems -- some with very tricked out setups, others not so (hey! I remember when my 486dx\33 was considered fast!). The comparison goes like this: anyone, with any system, provided it can run PS at any speed, is welcome to join in. Lurkers are encouraged to join in and be immortalized if this thread goes into the Library. When you post, it is only useful if you post your results along with system info. Minimums would be processor type, speed (in Mhz), memory, OS, and bus type. Please note anything else going on in your system that would affect the performance (caches, DSP, etc). (Special note to DSP users -- to get an idea of the benefits of a DSP, I think it would be useful to post performance with and without the DSP enabled, if this is possible -- thanks!). All tests should be performed on a 1 meg file, and on a 15 meg file. The test so far: Gaussian Blur at .5 Unsharp Mask at 5%, .5 radius, and 0 tolerance. Resample image to 1/2 original dpi (which results in a file 1/4 the original size). Rotate image 1 degree. Feel free to post test suggestions!... Nat There are 2 Replies. #: 90573 S16/Photoshop 11-Nov-93 20:29:12 Sb: #90570-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: Nat Merriam 72202,142 My system: Intel 486DX\33 w/8M ram. Running Windows 3.1 as a shell over MS-DOS 5.0. ISA. Have 1M PC-Kwik disk cache, 256k hardware cache. My swap disk is a compressed drive, ala Stacker 3.0. My results: 1M | 15M | Blur 6 sec | 2:05 | Sharpen 9 sec | 2:31 | Resize 6 sec | 1:26 | rotate 11 sec | 2:57 Nat #: 90726 S16/Photoshop 12-Nov-93 19:56:16 Sb: #90573-Pc/Mac PS Shootout Times Fm: Paul Serrano 76336,563 To: Nat Merriam 72202,142 Good work. Macintosh IIsi (68030, 20 MHz) w/17M ram (14M for Photoshop). Running System 7.1.1. 32KB ram cache. 96KB software disk cache. My swap disk is a Maxtor 202 MB SCSI drive (which I think has a transfer rate of about 1.2 or 1.3 MB/sec). My results: 1M | 15M | Blur 26 sec | 5:02 | Sharpen 25 sec | 5:50 | Resize 8.5 sec | 1:46 | rotate 25 sec | 5:13 (Let's all use bicubic interpolation.) #: 90838 S16/Photoshop 13-Nov-93 21:30:01 Sb: #90726-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: Paul Serrano 76336,563 (X) Oh, thanks! I just assumed everyone would use bicubic interpolation, but should have stated it... Nat #: 90900 S16/Photoshop 14-Nov-93 12:25:40 Sb: #90838-Pc/Mac PS Shootout Times Fm: J.C. Bourque 71600,2730 To: Nat Merriam 72202,142 (X) Nat, I inadvertently started a new thread with my times, so here they are again. Also discovered an error, so toss the old report. Have you volunteered to compile this data, or should we all be collecting it? Seems like it would be neat to see this in a scatter graph when it's done. Since the timings are not done under controlled circumstances, we should only expect to see trends, I figger'. Mac ci, 80mb RAM, 60 assigned to PS, Daystar Turbo 040/33mhz w/128k static RAM cache. 670mb Micropolis HD 1mb File: Blur :06.4 USM :08.7 Rotate :06.0 Res down :04.0 Res up :03.8 15mb File: Blur 1:24 USM 1:40 Rotate 1:42 Res down :44 Res up :44 Mac Centris 650, 20mb RAM, 16 assigned to PS, Apple 230mb HD 1mb File: Blur :06.7 USM :08.2 Rotate :07.5 Res down :04.9 Res up :04.5 15mb File: Blur 1:40 USM 1:58 Rotate 1:58 Res down 1:07 Res up :59 #: 90931 S16/Photoshop 14-Nov-93 18:09:06 Sb: #90900-Pc/Mac PS Shootout Times Fm: robert phillips 76711,1337 To: J.C. Bourque 71600,2730 JC--- I'm saving both these here threads, so not to worry! When they're done, we'll confab here about how to preserve them in lib 16. %%robert #: 91039 S16/Photoshop 15-Nov-93 10:23:46 Sb: #90931-Pc/Mac PS Shootout Times Fm: Phil Williams 71121,2054 To: robert phillips 76711,1337 I ran into this message over on the DTPFORUM. Thought I'd add it to the soup . . . a little forum cross-pollination. vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv Sb: #408442-Why PC's for DTP Fm: Jonathan Hirschman 70274,2526 To: Calvert Dayton 72760,3163 Calvert, Your message to Kyle strikes a chord. Our shop recently kicked the Macs out. We were using some pretty old color Macs, and it came down to a choice between replacing them, or getting the Art Department Mac users to convert to the rest of the company standard, PCs. We let them decide... When we told them we could buy 4 PCs on steroids for the price of two Fat Macs, they opted for the PCs. Best of all, they are now all telling me that it was the best decision they could have made! They really like the extra horsepower of a DX2-66 with a SuperMatch Spectrum card running on VL-Bus...and this sentiment is echoed by one artist that was running a top shelf Quadra setup at her last job. --- Jonathan Hirschman ...via AutoPilot... #: 91043 S16/Photoshop 15-Nov-93 11:08:02 Sb: #91039-Pc/Mac PS Shootout Times Fm: robert phillips 76711,1337 To: Phil Williams 71121,2054 Appreciated, Phil! Thanks! OpEd is a *good* thing for this thread! %%robert #: 90948 S16/Photoshop 14-Nov-93 18:51:16 Sb: #90900-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: J.C. Bourque 71600,2730 Compile? I knew I was forgetting something! Actually, I have missed a lot of it, so if anyone has been getting this whole thing, that would be great. A scatter graph would be nice, but I don't have a spreadsheet (whew! Talk about getting lazy because of computers!), so it would take me a while. Anyobody out there want to volunteer their Quattro sevices if we can get together the data? Nat #: 90979 S16/Photoshop 14-Nov-93 22:27:22 Sb: #90948-Pc/Mac PS Shootout Times Fm: robert phillips 76711,1337 To: Nat Merriam 72202,142 Nat--- I'm saving the messages! If there's to be a graph, it needs to be readable by both platforms. Acrobat would solve that, but not everyone can read Acrobat files (yet!). %%robert #: 91025 S16/Photoshop 15-Nov-93 08:57:15 Sb: #90979-Pc/Mac PS Shootout Times Fm: J.C. Bourque 71600,2730 To: robert phillips 76711,1337 <> How bout GIF? JC |-(:-) #: 91044 S16/Photoshop 15-Nov-93 11:08:04 Sb: #91025-Pc/Mac PS Shootout Times Fm: robert phillips 76711,1337 To: J.C. Bourque 71600,2730 GIF sounds plenty comfy to me, JC. %%robert #: 91189 S16/Photoshop 16-Nov-93 00:43:21 Sb: #90979-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: robert phillips 76711,1337 Well, this is the photoshop forum -- I should hope we would have the capability to share a graphic file in JPEG or TIFF format! (g) Of course, that would require that the spreadsheet be able to export a graph that was readable by PS (Hey -- something that would probably work better on Windows than a Mac, if you're good with OLE)... Nat #: 91216 S16/Photoshop 16-Nov-93 08:13:36 Sb: #91189-Pc/Mac PS Shootout Times Fm: robert phillips 76711,1337 To: Nat Merriam 72202,142 Well, yes, Nat! OLE and Publish/Subscribe -- both give me headaches. JPEG might not be such a bad idea, but I'll leave it to the active participants to decide. At this stage, I'm the order-taking troll. %%robert #: 91078 S16/Photoshop 15-Nov-93 16:05:15 Sb: #90900-Pc/Mac PS Shootout Times Fm: Tom Kirby 76645,112 To: J.C. Bourque 71600,2730 Here goes, Quadra 950, 40 meg RAM, 37 to PS, 1.2Gig Maxtor, Daystar Charger 1 meg 491x491 file W/Charger W/O Charger Blur :04 :07 Sharpen :04 :08 Rotate :06 :08 Resample :04 :07 15 meg 1620x3240 file W/Charger W/O Charger Blur :38 1:16 Sharpen :48 1:36 Rotate 1:16 1:39 Resample :23 :56 Tom #: 90575 S16/Photoshop 11-Nov-93 20:48:01 Sb: #90570-Pc/Mac PS Shootout Times Fm: Andrew Rodney 71511,3205 To: Nat Merriam 72202,142 -->All tests should be performaed on a 1 meg file, and on a 15 meg file. Sounds good. I'm game to try but a suggestion first. I don't know of any way to tell PhotoShop "create a 1mb file". If you could give us an exact pixel dimension we would all be using an EXACT same file size (assuming it's RGB). For example, a file that's 2048x2048 in RGB is 12mb. However a file that's 2048x2040 (or something like that) may be rounded to indicate "12mb". So for the test to be fair, we should all be using the same RGB file with an exact pixel count. We could then have an exacting way to do the resample (ie, go from 2048x2048 to 1024x1024 etc). Andrew #: 90703 S16/Photoshop 12-Nov-93 18:36:39 Sb: #90575-#Pc/Mac PS Shootout Times Fm: J.C. Bourque 71600,2730 To: Andrew Rodney 71511,3205 Also, let's define when to start and stop the clock. JC |-(:-) There is 1 Reply. #: 90728 S16/Photoshop 12-Nov-93 20:10:42 Sb: #90703-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: J.C. Bourque 71600,2730 Start the clock when you hit "enter" to begin the process. Stop the clock when the progress indicator goes away (you don't have to wait for the screen to redraw)... Nat #: 90753 S16/Photoshop 12-Nov-93 22:18:48 Sb: #90703-#Pc/Mac PS Shootout Times Fm: Andrew Rodney 71511,3205 To: J.C. Bourque 71600,2730 (X) -->Also, let's define when to start and stop the clock. Yes! I would say from the time you hit the button to the time it beeps. What do you think? Andrew There is 1 Reply. #: 90790 S16/Photoshop 13-Nov-93 11:06:10 Sb: #90753-Pc/Mac PS Shootout Times Fm: J.C. Bourque 71600,2730 To: Andrew Rodney 71511,3205 (X) <> I agree. Otherwise, you're timing your video card. JC |-(:-) #: 90711 S16/Photoshop 12-Nov-93 19:31:59 Sb: #90575-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: Andrew Rodney 71511,3205 Hmm. I just fiddled with the numbers when I created the new file. Whatever ppi you use, if you start adjusting the dimensions pixel by pixel, you'll come across the exact dimension. Unfortunately, I didn't write down the pixel size I used. I don't think the proportions would matter as long as the same number of pixels is present. I'll try to get an exact pixel size if someone doesn't post one by my next visit... Nat #: 90754 S16/Photoshop 12-Nov-93 22:18:51 Sb: #90711-#Pc/Mac PS Shootout Times Fm: Andrew Rodney 71511,3205 To: Nat Merriam 72202,142 (X) -->Whatever ppi you use, if you start adjusting the dimensions pixel by pixel, you'll come across the exact dimension. We should make this something simple like a 2000x2000 pixel file (or any round number). As far as I know, it makes no difference if the file is brand new and doesn't have anything other than white in it. That way everyone could just create a new file at an exact pixel size. PS doesn't care if the 2000x2000 pixel's are all different colors or all white. As soon as the rule's are set, I'm ready to rock. Andrew There is 1 Reply. #: 90839 S16/Photoshop 13-Nov-93 21:30:04 Sb: #90754-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: Andrew Rodney 71511,3205 Okay, here's what I have gotten as dimensions. 1 meg file: 1in x 2in x 413ppi (826x416 pixels). 15 meg file: 4in x 8in x 405ppi (1620x3240 pixels). I've noticed that these sometimes would be off by as much as 300k in the 15 meg file, but I don't see that such a small variation would throw a wrench in the works... Nat #: 90906 S16/Photoshop 14-Nov-93 13:24:27 Sb: #90839-#Pc/Mac PS Shootout Times Fm: Andrew Rodney 71511,3205 To: Nat Merriam 72202,142 (X) -->1 meg file: 1in x 2in x 413ppi (826x416 pixels). 15 meg file: 4in x 8in x 405ppi (1620x3240 pixels). OK so it's official right? 1mb file is an RGB file that's 826x416, 15mb file is 1620x3240. -->I've noticed that these sometimes would be off by as much as 300k in the 15 meg file, but I don't see that such a small variation would throw a wrench in the works... That's why we need to set this up with pixels. It's the smallest and most accurate way to describe a file. If everyone creates a new file with these exact pixel dimensions, there should be no variance introduced by the file itself; down to the single pixel that has to be pushed about. Let's also agree how the timing should be done. Several have suggested that timing start from the second the button/menu is clicked to the second one hear's the beep. That means that redraw is not a factor since the beep goes off just before the screen starts to redraw. I've seen several folks start to post results and I plan to do the test tomorrow. Let's make sure that those that have posted results are using the same criteria mentioned above. Sound good? Andrew There is 1 Reply. #: 90926 S16/Photoshop 14-Nov-93 17:26:58 Sb: #90906-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Andrew Rodney 71511,3205 Andrew/Nat, Just to clarify. My 1MB file was 591 x 591 pixels. My 15MB file was 2290 x 2290 pixels. Timing was from pressing the button to when the progress d'log disappeared or when the screen started to redraw -- i.e., it excludes redraw time. (We don't get beeps on the Mac when a filter finishes doing its thing). If you feel it's important, I'll try redoing the tests with an 826 x 416 and a 1620 x 3240, but I suspect that the results will be very similar. Bruce #: 90956 S16/Photoshop 14-Nov-93 19:33:45 Sb: #90926-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: Bruce Fraser 72511,131 I don't think it's necessary to redo the test. The difference in the # of pixels used was less than a percent, and this isn't brain surgery (g)... Nat #: 91008 S16/Photoshop 15-Nov-93 04:50:55 Sb: #90956-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Nat Merriam 72202,142 <> I was hoping you'd say that. Of course, it would be interesting to see if it made a difference... Bruce #: 91190 S16/Photoshop 16-Nov-93 00:43:23 Sb: #91008-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: Bruce Fraser 72511,131 >><>pixels used was less than a percent, and this isn't brain surgery (g)...>> >>I was hoping you'd say that. Of course, it would be interesting to see if it >>made a difference... Hey, don't let me stop you! (g)... Nat #: 90949 S16/Photoshop 14-Nov-93 18:51:18 Sb: #90906-Pc/Mac PS Shootout Times Fm: Nat Merriam 72202,142 To: Andrew Rodney 71511,3205 I think we established that the timing was from the time you hit "enter" to the time the progress indicator goes away (or beeps). As you said, that'll eliminate display as a factor... Nat #: 90725 S16/Photoshop 12-Nov-93 19:56:13 Sb: #90575-Pc/Mac PS Shootout Times Fm: Paul Serrano 76336,563 To: Andrew Rodney 71511,3205 You're right. 1 MB file is 591 pixels square. 15 MB file is 2290 pixels square. (I've noticed in the past that images with an extreme aspect ratio take longer to rotate.) #: 90866 S16/Photoshop 14-Nov-93 06:45:40 Sb: Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: All Equipment: Mac IIfx, Fusion Data TokaMac 33MHz 68040 accelerator, SuperMac Thunder II GX 1360 w/built-in DSPs, Seagate ST11200N 1.2 Gig drive, System 7.1, 32k disk cache, 80MB total RAM, with 64MB devoted to Photoshop. ~~~~~~~~~~~~~~~~1MB/DSP ON 1MB/DSP OFF 15MB/DSP ON 15MB/DSP OFF Gaussian blur 0:01.67 0:07.12 0:19.61 1:30.55 Unsharp mask 0:02.42 0:08.29 0:26.09 2:01.50 Resample 50% 0:03.07 0:04.27 0:25.97 0:40.99 Rotate 1 degree 0:05.87 0:07.40 1:07.65 1:32.67 I also reduced Photoshop's RAM allocation to 32MB, to bring virtual memory into play. The results for the 1MB image were unchanged, but it's easier to cut and paste, so I include them anyway. ~~~~~~~~~~~~~~~~1MB/DSP ON 1MB/DSP OFF 15MB/DSP ON 15MB/DSP OFF Gaussian blur 0:01.67 0:07.12 0:35.36 1:50.44 Unsharp mask 0:02.42 0:08.29 0:34.70 2:10.93 Resample 50% 0:03.07 0:04.27 0:26.38 0:51.33 Rotate 1 degree 0:05.87 0:07.40 1:21.03 1:45.31 The difference with the smaller memory partition is considerably less dramatic than I would have expected, and certainly less dramatic than it would have been in Photoshop 2.0.1. Clearly the Knollmeister has done major work on optimizing the virtual memory performance. It also leads me to the conclusion that a SCSI array might be a better investment than more RAM.... Bruce #: 90908 S16/Photoshop 14-Nov-93 13:24:37 Sb: #90866-#Pc/Mac PS Shootout Times Fm: Andrew Rodney 71511,3205 To: Bruce Fraser 72511,131 (X) -->It also leads me to the conclusion that a SCSI array might be a better investment than more RAM.... That's good news. Man that DSP of your's really is cutting the times down on some operations! Very impressive. Andrew There is 1 Reply. #: 90925 S16/Photoshop 14-Nov-93 17:07:11 Sb: #90908-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Andrew Rodney 71511,3205 The DSP seems to have the biggest effect on Guassian blur, USM, and upsampling -- the gains on the other operations are much more modest. But considering how vital gaussian blur and USM are, it seems worth it. It also speeds up feathering quite a bit. I just wish someone would figure out how to speed up the *^&^ magic wand! Bruce #: 90968 S16/Photoshop 14-Nov-93 21:34:19 Sb: #90925-Pc/Mac PS Shootout Times Fm: Roberts Howard 71514,3624 To: Bruce Fraser 72511,131 >> But considering how vital gaussian blur and USM are, Bruce, please explain in what context they are vital...Rob #: 91009 S16/Photoshop 15-Nov-93 04:50:58 Sb: #90968-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Roberts Howard 71514,3624 (X) Rob, USM is needed on just about every image I deal with. Most of the scanners I use don't do their own sharpening on the fly, and scans pretty much always need sharpening. The other sharpening algorithms don't give anything like as good a result. Guassian blurring is something I mostly do to alpha channels to soften the edges of selections -- it's indispensible if you want to make a soft-edged, realistic shadow, for example. I've also been known to use it to get rid of obvious film grain in skies, but don't tell anyone. Bruce #: 91035 S16/Photoshop 15-Nov-93 09:47:01 Sb: #91009-Pc/Mac PS Shootout Times Fm: Roberts Howard 71514,3624 To: Bruce Fraser 72511,131 Aha! I see how you're using those tools. I agree with your uses especially with photography. My priorities are a bit different with illustration and I spend more time mucking about with the Brightness/Contrast tools...Rob #: 91092 S16/Photoshop 15-Nov-93 17:05:19 Sb: #91035-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Roberts Howard 71514,3624 <> In that case, you may want to check out Levels and Curves, both of which are on the same menu as Brightness/Contrast. B&C is something of a blunt instrument compared to those two... Just a thought. Bruce #: 91230 S16/Photoshop 16-Nov-93 09:53:22 Sb: #91092-Pc/Mac PS Shootout Times Fm: Roberts Howard 71514,3624 To: Bruce Fraser 72511,131 >> B&C is something of a blunt instrument compared to those two... Rather than use them as an allover tool like Levels and Curves, I use them for selected areas. I prefer to manipulate Value without effecting Hue and Chroma, thus a simple tool like Brightness/Contrast does it with no chromatic shifting. #: 91070 S16/Photoshop 15-Nov-93 15:05:50 Sb: #91009-Pc/Mac PS Shootout Times Fm: Alan Boucek 72411,3405 To: Bruce Fraser 72511,131 Bruce- have you tested any of the new sharpening plug-ins- I'm looking for a workable adaptive sharpening system- something a little smarter than Daystar or Adobe USM that can smooth flat areas, and strenthen edges. alan #: 91152 S16/Photoshop 15-Nov-93 21:12:17 Sb: #91070-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Alan Boucek 72411,3405 Alan, I like Color Access' sharpening -- it's very controllable, and does just what you're asking for. The only third-party sharpening plug-in I know of is the Sharpen Intensity plug-in in Kai's Power Tools, which is much improved in v.2. I'd love to try some others if I knew where to look. You have anything specific in mind? Bruce #: 91276 S16/Photoshop 16-Nov-93 13:35:53 Sb: #91152-Pc/Mac PS Shootout Times Fm: Alan Boucek 72411,3405 To: Bruce Fraser 72511,131 >>You have anything specific in mind?<< I saw something listed in the PrePress direct catalog (which I trashed unfortunately)- a Photoshop plugin. Has KPT 2.0 shipped- I haven't received mine. I'll probably have to get ColorAccess, I've thought about it, but never was fully convinced that I need it- this might convince me. Does CA deal with large files well? ie 50-100mb? I have to find out what kind of USM is available with Eclipse (on SGI) and whether its a part of the main package or part of the separation package. alan #: 91341 S16/Photoshop 16-Nov-93 20:09:30 Sb: #91276-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Alan Boucek 72411,3405 KPT 2.0 has indeed shipped, though it may not be in the stores yet. Color Access is really a tool for converting RGB images to printable form. As such, it produces excellent results. Your question, does it deal with large files well, has no easy answer. It isn't a speed demon, but you can run batches of files to separate overnight. It isn't an interactive, arty kind of program like Photoshop -- it essentially duplicates the controls you'd find on a drum scanner. Bruce #: 91389 S16/Photoshop 17-Nov-93 11:03:40 Sb: #91341-Pc/Mac PS Shootout Times Fm: Alan Boucek 72411,3405 To: Bruce Fraser 72511,131 >> It isn't an interactive, arty kind of program like Photoshop -- it essentially duplicates the controls you'd find on a drum scanner.<< That's why we have Photoshop- my remaining question- does CA have a good facility for previewing before doing the sharpening and color correction? It sounds like something that I should get. alan #: 91489 S16/Photoshop 18-Nov-93 04:30:41 Sb: #91389-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Alan Boucek 72411,3405 (X) <> I'd say so. It has a lockable on-screen densitometer too, which is probably the most important tool.... Bruce #: 91026 S16/Photoshop 15-Nov-93 08:57:20 Sb: #90925-Pc/Mac PS Shootout Times Fm: J.C. Bourque 71600,2730 To: Bruce Fraser 72511,131 <> I have. Quit using it. JC |-(:-) #: 91091 S16/Photoshop 15-Nov-93 17:05:17 Sb: #91026-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: J.C. Bourque 71600,2730 I've pretty much quit using the magic wand too. But wouldn't it be nice if it worked, and did so quickly enough that you didn't have to go to lunch while it did its thing? Or do you think it's intrinsically useless? Bruce #: 91196 S16/Photoshop 16-Nov-93 01:50:38 Sb: #91091-Pc/Mac PS Shootout Times Fm: Steve Pollock 70244,1436 To: Bruce Fraser 72511,131 --> Or do you think it's intrinsically useless? Close to it in its present form. The concept of density masking is SO powerful, it's a shame a *really* interactive front-end hasn't been designed for PS. Just the idea that this might be remotely workable w/ input in the form of a single tolerance value...c'mon! Steve #: 91340 S16/Photoshop 16-Nov-93 20:09:26 Sb: #91196-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Steve Pollock 70244,1436 (X) <> I have to agree. Bruce #: 90950 S16/Photoshop 14-Nov-93 18:56:37 Sb: #90866-Pc/Mac PS Shootout Times Fm: Pawel Bodytko 70733,644 To: Bruce Fraser 72511,131 Bruce, I've been looking at those numbers ( 0:03.07 0:04.27 0:26.38 0:51.33) and am not sure what they mean; are those minutes, seconds an 1/100s of a second? If so, how do you measure a 1/100 of a second? For myself I have trouble to measure a single second since I am not sure when should I call it off... Do you do that after the screen redraw or after the black bar on the bottom is gone. (If you wait for the screen redraw, there will be a diferance depending on enlargement and the actual % size of your image on the screen... The first test with the 8 meg file I did with the screen redraw which took about 20 seconds because of my slow display card and preview showing the whole file... Let's clarify those things as well since it is great to actually see how your setup compares with the others. Pawel #: 91007 S16/Photoshop 15-Nov-93 04:50:53 Sb: #90950-Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Pawel Bodytko 70733,644 Pawel, The numbers are minutes, seconds and 1/00ths of a second -- it's what my stopwatch gives me! I've done enough of this sort of thing to know that my reaction time is around 1/10th of a second, so you can take that as the margin of error. The times don't include screen redraw. I pressed the stop button either when the progress indicator disappeared, or, in the cases where there was no progress indicator, at the instant that the screen started to redraw. Hope this helps. Bruce #: 91027 S16/Photoshop 15-Nov-93 08:57:26 Sb: #90950-Pc/Mac PS Shootout Times Fm: J.C. Bourque 71600,2730 To: Pawel Bodytko 70733,644 Run the test on two files: 591x591 pixels and 2290x2290 pixels, both RGB. On real short times, i run the test several times to try to average out my reaction times. JC |-(:-) #: 91085 S16/Photoshop 15-Nov-93 16:32:48 Sb: #90866-Pc/Mac PS Shootout Times Fm: Andrew Rodney 71511,3205 To: Bruce Fraser 72511,131 (X) Bruce, I've just checked out your speed tests and compared them to mine. When you reduce your memory to 32mb (close to the 35mb I have assigned), it looks like your IIfx and my accelerated Quadra rank pretty close with DSP operations. This is what I would expect since all the work is going to the DSP. We have different boards but the results are interesting and pretty close. I've placed my times for DSP operations on the 1mb &15mb file below yours. Another interesting thing is how much faster your times were with that 64mb devoted to PS. Seems it's true that ram is needed for max DSP speed. -->I also reduced Photoshop's RAM allocation to 32MB, to bring virtual memory into play. The results for the 1MB image were unchanged, but it's easier to cut and paste, so I include them anyway. ~~~~~~~~~~~~~~~~1MB/DSP ON 1MB/DSP OFF 15MB/DSP ON 15MB/DSP OFF Gaussian blur 0:01.67 0:07.12 0:35.36 1:50.44 0:02.91 (AR's) 0:34.96 (AR's) Unsharp mask 0:02.42 0:08.29 0:34.70 2:10.93 0:03.49 (AR's) 0:02.91 (AR's) Resample 50% 0:03.07 0:04.27 0:26.38 0:51.33 0:02.65 (AR's) 0:20.?? (AR's) Rotate 1 degree 0:05.87 0:07.40 1:21.03 1:45.31 0:05.59 (AR's) 1:06.?? (AR's) Andrew #: 91153 S16/Photoshop 15-Nov-93 21:12:19 Sb: #91085-#Pc/Mac PS Shootout Times Fm: Bruce Fraser 72511,131 To: Andrew Rodney 71511,3205 (X) Andrew, Did your system really do USM on the 15MB image in 0:02.91? Or is that a typo? If it isn't, I want what you've got! Note that my IIfx actually has a 33MHz '040, which is what brings it so close to Quadra levels. I still suffer from slow SCSI, but the '040 plugs into the PDS slot so it isn't limited by the NuBus bandwidth. Yes, RAM seems to make much more difference with the DSP than without. Bruce There is 1 Reply. #: 91169 S16/Photoshop 15-Nov-93 22:34:04 Sb: #91153-Pc/Mac PS Shootout Times Fm: Andrew Rodney 71511,3205 To: Bruce Fraser 72511,131 -->Did your system really do USM on the 15MB image in 0:02.91? Or is that a typo? Oh sorry. That was that PowerPC I'm playing with . Actually it was a major typo (and wishful thinking on my part). Should have been USM on 15mb file with Quadra 43 seconds and change with the DSP and 31 seconds with the KEPS USM filter. I know where I got that 02.91 figure, it was Gaussian Blur on a 1mb file. Andrew #: 91086 S16/Photoshop 15-Nov-93 16:33:25 Sb: Pc/Mac PS Shootout Times Fm: Andrew Rodney 71511,3205 To: All I just realized I posted my speed times under a different thread heading so here we go again (sorry for the rerun's). OK here are my results on two systems: (Did all tests several times. Times posted are average of all tests. The 10th of a seconds are hard to time especially when working with the smaller files.) System #1 is a IIci with a DayStar 33mz 040 accelerator w/cache. 20mb of ram (16mb to PS).DayStar Charger DSP and a MicroNet 303 mb drive with SCSI-2 board. 1mb file with above pixel size created new: Gaussian Blur -----2.80 seconds Rotate 1 degree ---5.55 seconds Resample Down -----2.18 seconds Resample Up -------2.92 seconds USM ---------------4.22 Seconds 15mb file: Gaussian Blur -----52 seconds Rotate 1 degree ---1 minute, 33 seconds Resample Down -----28.71 seconds Resample Up -------29.21 seconds USM ---------------1 minute on the nose System #2/. Quadra 900 accelerated to 34mz with a Newer VSO. 40mb of ram with 35mb assigned to PhotoShop. MicroNet Raven gig array drive. Charger PRO DSP. The Pro is a newer and faster DSP that uses KEPS precision filters. Both USM and Gaussian blur are available as KEPS filters so I timed these as well (they are faster and said to be better too). A rotate and resize filter is available but I didn't do time tests since this seems unfair. This filter does both operations at once! 1mb file with above pixel size created new: Gaussian Blur -----2.91 seconds** (2.21 seconds with KEPS) Rotate 1 degree ---5.59 seconds** Resample Down -----2.65 seconds** Resample Up -------2.92 seconds USM ---------------3.49 Seconds (2.21 seconds with KEPS) **Interesting that the IIci seems to be a tad faster than the Quadra. However we are only dealing with 10th's of a second difference which could be human error in timing. It also shows that small files totally in ram going to the DSP don't make much difference with native processors since the DSP is doing all the work. Same is true of the HD's. The Quadra should be a little faster but with files in Ram and the DSP, both machines are about the same. PLEASE NOTE that all these operations are going to the DSP; none are using the native processor in either machine. One should see bigger differences with no DSP running. Also note that when KEPS filters are used, the difference between the IIci times with a 15mb file and the Quadra running KEPS is about 100% (ie, USM went from 1 minute to 31 seconds). With the 15mb file, I believe that VM is taking place. The gig array on the Quadra is most likely the reason why the times are so different between each machine. 15mb file: Gaussian Blur -----34.96 seconds (31 seconds with KEPS) Rotate 1 degree ---1 minute, 6 seconds Resample Down -----20 seconds Resample Up -------28.71 seconds USM ---------------43.74 seconds (31 seconds with KEPS) Andrew