Hi David,
"Very Large" can mean many things to different people, so if you could quantify that a bit better, it may yield some better suggestions for you.
Why the X6800? I did the same, and I’m really thinking it was overkill. Why did I go with it? Well, I’ve never built a system for overclocking and the design of the CPU facilitates a broader range of overclocking options. Combine that with fast DRAM and I thought I might be able to push the performance to the max. Guess what? That’s not the case, and I could’ve saved a good bit of money by stepping down a notch to the 6700 at least, not to mention DDR2-866 DRAM rather than the 1066 variety that I’m unable to get 1066 performance out of. Admittedly, I am running my system overclocked at 3.65GHz and getting very good performance out of it, but I’ve seen others get similar results with cheaper CPU/DRAM combinations.
I’ve yet to do much image editing, but a few tests suggest that PC CS2 performance on this system I’ve built will be very good and easily meet my needs when working with files typically no more than 100MB or so. As for Premiere? It is great! Where my old Athlon 3000 Compaq system was taking as much as 15 hours for some video renders, it appears as this system is going to be paring that down to 2 hours or less. Even there I’ve only played around with a few videos along the way of building this system up and getting some problem troubleshooting done, but I’m very optimistic.
In a nutshell, what I’ve built is a WinXP Pro system based around an ASUS P5WDH Deluxe motherboard, Intel Core 2 Extreme (X6800) CPU, 4GB Corsair DDR2-1066 DRAM, a pair of 150GB Western Digital Raptor SATA hard drives, and a 750GB Seagate Barracuda SATA hard drive. The drives are independent (no RAID), with my O/S and apps on one Raptor, the Windows pagefile and other data space on the 2nd Raptor, and the PS scratch disk plus additional data space and backup storage on the Barracuda drive. My system is at <
http://ambress.com/pc> if your’e interested in seeing it. As for the Samsung LCD monitor, that’s all I’ve got connected at the moment, but PS work is done on a calibrated Iiyama CRT, not this LCD monitor.
Regards,
Daryl
I should have specified — large in my case means– between 1 & 2 gigabyte. David
ps Thanks for replying — I had a look at yr set up – it seems interesting. Any chance of you testing 2G photoshop files on it?
Stitching multiple images together to create a large panoramic and trying an architectural crop on all four sides of the whole image would be an interesting test.
My files come from hi resolution scans of large format sheet film..so they can be very large!1
David
Yep, I’d call that large! 🙂
What I’ve typically seen folks here using, where dealing with files of such size, are RAIDed drives. I don’t know what your budget allows for, or what drive configurations that Gigabyte board supports, but I’d probably be looking at running at least two RAID pairs, one for a primary data workspace, the other for the PS scratch disk and extra storage. For video, one RAID pair would be your project drive while the other would be to render the output. Whether or not the O/S and apps gain much from being on RAID array, I’m not sure, but I think activation of Adobe apps is easier to deal with when you stay with a single drive. I think a RAID of WD Raptor drives is probably the fastest setup you can go with, short of higher-cost Ultra320 SCSI drives, and I’m not even sure if a SCSI RAID would be faster or not.
How much of a performance penalty there would be with the Windows pagefile sharing the same physical hard drive as the primary data workspace I’m not sure either, but it may still be worth doing. As I think about it, one thing I learned with WinXP Pro 32-bit is that the pagefile cannot exceed 4GB on a single partition, so you may wind up splitting that across multiple partitions if you feel you need a larger pagefile. The X6800 CPU may well be a better choice since normal clock speeds are still higher, and you may not want to push your luck with overclocking on system for professional use; stability may be more the concern. I’d probably do something like this:
CPU: Intel Core 2 Duo 6700 or Extreme 6800.
DRAM: 4GB DDR2-866 (possibly more later if you move to a 64-bit O/S)
Drive C: 150GB Raptor – O/S & Apps, 2GB fixed pagefile (to minimize fragmentation). Drive D: Two 75GB WD Raptors in RAID 0 – 10GB partition for pagefile, Windows TEMP space Drive E: Remaining 65GB for data, video capture
Drive F: Two 75GB WD Raptors in RAID 0 – 10GB partition for PS scratch disk (Would a larger partition benefit PS? I don’t know.)
Drive G: Remainining 65GB for data, video render
Drive H: Seagate Barracuda 750GB drive for data storage, backup, etc.
I chose 75GB Raptors simply for the cost savings over the larger 150GB variety, but the latter may be preferable for the added storage, given the size of files you’re working with.
A full tower case with good airflow and cooling will be highly desirable too, given the number drives involved. Raptors, at 10,000rpm, run on the warm side. Mine seem determined to hover around 40-41°C, which is a bit warmer than I like but well within spec. I attribute that to the mid-tower chassis and rather packed interior. With another pair of Raptors installed, I’d move to a tower chassis with more airflow.
For a graphics card, you needn’t spend much if high-end gaming isn’t also a requirement. I’d just let your monitor choice(s) drive your requirements there….single or dual-head, analog or DVI out, etc. If you plan on having a large flat-panel monitor, just check to see first if it has any specific card requirements.
I think this is a reasonable overall recommendation, but I’ve never dealt with a system for processing such large image files. My suggestions on the partitions, pagefile, scratch disk, etc. are all just an educated guess really, more than the application of practical experience, so I’ll be interested in seeing what others comment on.
Hmmm…a 2GB test? I’d be hard pressed to have any images suitable for that, so I’d be having to do some serious upsampling of some 15MB NEF raw files from my Nikon D200. I might be game to try something though if you could elaborate a bit more. What is an architectural crop? How many images are being stitched? Is the stitching being done automatically with the Photomerge (I assume not)? Are layers being retained or flattened? What part of your process requires the most time to perform? I’m not sure how quickly I could do this, but it would be interesting to see what my system can handle in a reasonable amount of time.
Regards,
Daryl
Thanks for your thoughts and input – itis v. much appreciated and seems sensible. I do wish MS$ had not made such a mess of their 64 bit O/S!!
In regard to testing if you wanted to do it I would suggest you took any raw image on your Nikon which has some vertical and horizontal lines. Shoot the image against a fairly neutral background (It will make the stitching a breeze).
Load one copy into CS2 and note the file size. Using the layer menu create layer from background- save file as name01.psd but do not close the file.
Now repeat the process until you have a total of around 1.4G of files open in PS. Now you can use the stitch facility in CS2 to stitch them together horizontally to create a wide panoramic (just drop the images in next to one another using the manual option).
Save the new file.
For the sake of reality add some adjustment layers before cropping.
Use the crop tool from the tools palette. Tick the perspective option. Make a crop let the crop tool cover the whole image and then, using the handles at the corners adjust the crop area to take a triangular section off each edge. You would for example take nothing off the bottom at the left edge but a small amount at the bottom right edge. O the top reverse — zero off the right top and a little bit off the left top. Do something similar for the sides.
Thenclick the green tick to start the crop – I would be intereste to know how long the redraw takes.
Thanks again
David
David,
If there is time when I get home tonight (I’m working a bit late), I may try this to see what I can do. I may even cheat a little and use an image from which I cut out a center horizontal strip, as if it was a panorama, and then upsize that and slice 10 overlapping layers from it. Each layer could be saved to a file and each file then resized to 140MB or so and pulled back in for stitching via Photomerge. Saving the assembly as layers, I could then apply a few adjustment layers and the crop as you suggest, and see how long the screen redraw takes upon applying the crop. Does that sound viable to you?
Daryl
David, why not send Daryl your file via <
http://www.sendthisfile.com/> or some such transfer service? Seems that would be a simpler way to level the playing field, and you could include other users (likely with more bourgeois machines) in the test.
Joe,
Thanks for the suggestion….I’d thought at first to ask David if he had a set of good images for this test, but given the potential size of them, I think I can just as easily work from something of my own. Even if I upsample some smallish image, the image quality itself isn’t so much as concern as just the capability of handling large files. I believe the last idea I mentioned should work reasonably well, although I’m open to alternative ideas. Taking an approach as I mentioned, I might could even "actionize" part of it from some base image in JPEG format that would be small enough to share, should others want to try a similar test. The action would take care of increasing the image/file sizes into the range of interet, plus this might allow the inclusion of all steps up to the point of preparing the perspective crop.
Daryl
Sending a large file would be difficult atm as I am 8,000 miles away from my office!!
Daryl’s idea should work fine But I am not certain about using a Jpeg –(due to the way PS handles jpegs) a rawfile would give more consistent results.
As Daryl says it is the file size that is the most critical issue. I have suggested the image has areasonable number of vertical & horizontal lines within it so you can visually check that the perspective adjustment has happened. The perspective crop of very large panoramics files seems to generate heavy loads in PS– and provide a good test of performance.
David
Daryl, since sending the huge file is geographically unworkable, your idea seems a good choice. A smaller file, whether JPG, TIF, PSD or whatever, accompanied by an action that would grow it into a large file for the test, should provide a decent benchmark. A second action to perform the "bit grinding" would make timing the process easy.
Ooops…major goof in my test last night. I’ll repost with updated results later.
I suggest the action starts with a raw or tif not jpegs. David
OK guys, here is some info:
Firstly, I didn’t try creating an action as I thought I might, but rather just found an image I thought might work for this task and took a stab at things. I didn’t have a good image from my Nikon D200 that I thought would lend itself well, so I found an older photo shot with my D70. I opened the 6.4MB NEF image using ACR upsampling to a 25MB, 6155×4086 px file. From that image, I cropped out a horizontal slice that I then further resized up to 11,000×4770 px for an indicated 150MB size.
So, working from that 150MB image…
1. I cut it into 11 overlapping layers that were enlarged to 3632×14,437 px (150MB)
2. The layers were exported as files that ranged from 90-116MB in size for a total of 1.09GB
3. With all images closed, I opened the folder of 11 files using Photomerge with "Attempt to Automatically Arrange Source Images" selected. The time elapsed from when I selected OK to invoke Photomerge until all images were assembled and displayed was 1:58(mm:ss)
4. With "Keep as Layers" selected and then selecting "OK" in Photomerge, the time elapsed until the fully assembled image was drawn completely was 09:30 with an image size of 33,248×14,490 px (1.35GB).
5. With all layers selected, I then added Layers, Photo Filter, and Channel Mixer adjustment Layers, with the time to do so being negligible.
6. With all layers selected, the Crop tool was chosen and the full image selected with the crop marquee, a perspective correction to the top and bottom of the image was made by bringing down the top left and moving up the bottom right anchors about 3 inches each (image was about 60 inches high). The crop was then applied and was completed with the image fully redrawn in a time of 02:36. Final resulting image was 33,257×13,867 px (1.29 GB).
During this exercise, the scratch file peaked at 13.3GB in size. The source images resided on my 150GB Raptor boot drive while the windows pagefile and PS scratch disk were allocated to a 25GB partition of my second 150GB Raptor. The pagefile was fixed at 4GB. I did not use the Seagate drive since, although a high performance drive, it is not quite as quick as the WD Raptors. Also, the RAM allocation in PS was set to 90% for a 2.2GB allocation.
The before and after images reduced to 700px wide are at <
http://ambress.com/gallery/gsm/mingus_mill.jpg>.
If anyone wants to play with the image slices I created for reassembly, I saved them as Quality 10 JPEGs and zipped them together as a 53MB file found at <
http://ambress.com/gallery/gsm/mingusmill.zip>. When unzipped and opened in PS CS2, they should still decompress to being 150MB images, so I’d expect them to be usable for the discussions here.
Hope this is informative…even if not served up as a benchmark.
Daryl
Hi Daryl
Thank you so much.
That is pretty impressive redraw result compared with my existing 3 yr old system that takes over 20 mins for +1G files!
I am curious whether making the crop for all four sides would make any difference?
I am however curious about the time the photomerge operation took which seemed overly long by comparison with the crop. The time taken here suggests that the routine had some difficulty with the image placing rather than the stitching routine.
My guess is that that had something to do with the images.
As a matter of curiosity if you get a chance maybe you could try the photomerge with the manual option and see how long the stitching takes.
Thanks again for a great effort.
Now I need to decide whether to wait for the QX6700 or bite thebullet now!!
David
Hmmm – an interesting effect Daryl!
I wonder if processing by slices is also making the image brighter and more well-defined?
Deebs,
No, the changes you’re referring to are just the results of some rather arbitrary edits I applied through creation of Adjustment Layers for Levels, Photo Filter (Warming 60%), and Channel Mixer (Never used this before!). I wasn’t so much try to improve the image as just throw some stuff onto it for the sake of this exercise. The annotations on the photos don’t include my performing the adjustments.
David,
I’m not sure how much difficulty Photomerge had placing the files, as the assembled result required no further editing. But, as the initial slices were created from a full image, I somewhat expected a clean reassembly should be doable. I did notice that there was a series of 3 passes through all the files before the assembly was presented in Photomerge, and the processing window indicated "Transform" in at least one of the passes. I can certainly try opening Photomerge without the auto-arrange of images selected, just to see how long it takes before the Photomerge screen is presented. I’ll do that later tonight since I’m now at work.
Yes, it is pretty amazing to see how much performance is delivered in current PCs vs. those from a few years ago. I had a test action I devised as my own sort of benchmark that, upon applying to a 22MB test image, took about 3.5 minutes on an old dual 550MHz P3 system. The new PC completes that same action in about 9 seconds! 🙂
Now, what I didn’t mention was that my posting replaced one from last night where I must’ve been too sleepy to recognize the error of my ways. I have only 3 drives yet was thinking 4, and the PS scratch was on my Seagate drive where I also first had the image files. So, for all the processing that was done, that drive was being exercised quite a bit. Further, in creating my slices of the original image, I did that rather quickly and overlooked that each slice was still a layered image, so I had roughly 200MB images instead of 100MB, and the assembled image was on the order of 2.3GB prior to the cropping. That stretched out the initial Photomerge processing to about 13 minutes, while the final perspective crop took about 5.5 minutes. I also saw a Windows warning of low drive space on the scratch disk, which was "only" 10GB. Fortunately, I had allocated the 10GB Windows page file partition on a Raptor as my secondary scratch drive and as the pagefile was only 4GB, there was enough room for the PS scratch to use up what it needed…another 5.3 GB. So, for whatever that’s worth, I’d just call that more of a worse case scenario since the data files and scratch disk shared the same drive.
Other things were going on too…I’d overlooked shutting down background processes (firewall, antivirus, spyware blocker, etc.), all of which I disabled this morning before my updated test run.
If I ever run into a situation of working with such large files myself, which is unlikely, I may add on a 4th small, high-performance drive and dedicate it solely to PS use. Where I thought 10GB would be plenty of space for a scratch disk, I see now how that can quickly be eaten up when working with images that are 1GB or larger.
Daryl
The relative placing of images seems to be a pretty intensive process — (incidentally stitching of jpeg images adds another layer of complication for processing because reconstruction of lossy compressions means that the adjacent reconstructions are always chancy matches!! .. that is why raw shots are best if subsequent stitching is envisaged e.g when creating panoramas where the aspect ration is usually at least 3.5:1 and in my case often around 5.5:1) -usually there is an overlap between adjacent originals and the stitch algorythm works to blend the images through the overlap. My guess is that the placing routine took a lot longer with the files you used than in a real life situation.
I also think the results of the previous interpolation might have made the image placing more demanding.. I have noticed that phenomenon when I have done something similar.
Thanks again
david
David,
Although I provided the JPEGs for others to play with if so desired, I should have clarified that I don’t really suggest the files be kept in JPEG format for the testing. This is done simply to keep the files to a manageable size for downloading and, once opened in PS, they could be resaved to TIFF or PSD format.
With regard to the overlap, I did provide for perhaps a 10-15% overlap in the images I created so that they couldn’t simply be placed edge-to-edge for reassembly. So, the positioning algorithm will still have to work its magic. The fact that the images already line up horizontally and have no perspective distortions relative to one another might shorten the processing time, but that’s simply a guess on my part. This does all have me a bit curious to try a "real life" case…perhaps this weekend.
Daryl
If you do I strongly recomend you use a really solid tripod, use at least f22 and give meticulous attention to the spirit levels!! It really does speed up the matching process quite apart from reducing the amount of PS work when the images have been stitched.
David
Spirit levels???
Ha! Just kidding…I know what they are but there are none on my old Bogen 3011, so I’d have to resort to my shorter Bogen-clone from Amvona with built-in levels. My Really Right Stuff BH-40 ball head is pricey enough to deserve a better base beneath it, but these two tripods work well enough for most of my needs. Ideally, it would be good too if I could align the camera & lens to the nodal point or whatever it is called, but that’s all more than any image I’d be shooting would justify the trouble of doing. I’ll see what I can get and if the results are satsifying, I’ll share more details later.
Daryl
Here’s the latest update:
With Auto Arrange deselected when launching Photomerge, there was no appreciable difference in the time required before the Photomerge editor was presented: 2:02 vs. 1:58 this morning when Auto Arrange was selected.
Repeating the Photomerge open with Auto Arrange and again applying the result with Preserve Layers selected, tonight’s run was 15 seconds quicker at 9:15
Finally, applying a 4-sided perspective crop took 21 seconds longer at 2:57 vs. the prior 2:36 observed this morning.
Regards,
Daryl
Thanks again
You have been diligent.. the results seem to be consistent — the results seem to indicate that your configuration offers a real advance.. I am now wondering about whether to use the X6800 or to wait u8ntil the QX6700 becomes available (which should be quite soon).
I have been talking to Gigabyte about their board. There seems to be some availability problem for sets of 2x2G DDR800 at the moment. I do not want to buy 2x2x1G sets and then have to pull them later to move the system up to a full 8G.
I have also been looking at power supplies. It seems I ought to think of having two raids – each 2x150G Raptors plus two 750G Seagates. When I put the power requirement for that lot together with two high end graphics cards each needing 22A there are some serious power and cooling issues.
I cannot find a suitable power supply (it would be better to have separate power supply for the graphics cards butno one seems to have them available).
The closest seems to be the Enermax 650 – that has two 22A rails but that leaves no margin.
Anyway thanks again for your contribution – it is pushing me on — I will need to do something soon. The other thing is I wish Vista was available..but that looks like late January — all in all it seems that availability issues are pushing me backwards while processing requirements are piling up into a backlog!!
Indecision rules OK!!
David
You are going for a rather "ultra" system, aren’t you? Is this strictly for working with the large images of the sort you’ve mentioned here? I’m curious about the two high-end graphics cards and why they would be needed, unless you have demanding display requirements due to monitor sizes and/or the number of monitors being driven? Also, does each card actually need 12V/22A power or is that the combination of the two? I see that the nVidia dual-SLI spec is for two cards combined to have a 12V/22A power suppply…or at least that’s how I interpret it.
If that p/s you’re looking at is the 620W (rather than 650) model Enermax Liberty, priced at $165 from Newegg, and you’re really concerned about having the headroom you need, then another $115 over that will buy the new Thermaltake Toughpower 850W W0131RU which is $280. That’s a healthy jump in price but might be a worthwhile investment for a high-end system where you’re even questioning going quad-core. The specs on the Thermaltake certainly look nice…quad-SLI approved with two 12V/30A and two 12V/18A rails.
As far as the graphics are concerned I have had extreme difficulty working with panoramic images where the ratios are at miniumum 3.5:1 and more often 5.5:1 and sometimes even 7.5:1. The difficulty is being able to examine the whole image in sufficent detail.
When the final output is going to be printed at least 13 inches high and 6 feet long at high resolution you can see why multiple monitors and rapid redrawing are desirable!!
Subsequently when I have created such images they may also be used for some clients as a multi frame background against which other images are panned in premier — it all gets very image intensive!!
Thanks for the suggestion about the Thermaltake I will look at that one. I must say I prefer to have a power supply coasting rather than close to its limits.
The cards I am looking is the 6950 which needs 22 at 12V for each card and I am reckoning on driving 4 monitors.
Have to go will come back
David
The video card I am considering is :
EVGA GeForce 7950 GT 512MB GDDR3 SLI PCI Express Dual DVI HDTV
Digging a bit if I assume 2.5A per drive that is 15A PEAK 2 GeForce 7950’s @ 22A 44A PEAK
That Thermaltake model has a total PEAK of 62A which means that the 12V rail is left with only 3A for fans, cooling etc etc so it looks as though I will need something even more beafy. So if really decide to go for such a high component spec it looks as though the 1000 Watt 132RU will be required and that is marked as "Coming soon"!!!
I only started doing such projects quite recently and seem to have found my niche. They are a challenge artistically, photographically and technically. From a design point of view they are uniquely interesting but the proceesing time has been a real drawback. It can take 4 hours work on my existing system to process one image – with mmost of the time spent waiting for redraws. With higher processing speeds I reckon that could be brought down to 20 mins or less- so the incentive is there!
David
David,
You know a heck of a lot more about your needs than I do, but if you are strictly talking about doing 2D imaging and maybe some video processing, I wonder if you aren’t going way into overkill for your video cards? I’m not certain what benefits are realized from having an amount of video memory far exceeding the display requirements, but it has always been my understanding that large amounts of video memory are more to satisfy 3D graphics as with gaming. I didn’t ask what resolution your 4 monitors run at, but let’s just assume they are 30-inch Dell widescreen models running at 2560×1600 and 32-bit color.
Video memory requirements = 2560 * 1600 * 32 (1 byte/8 bits) = 16,384,000 bytes or approximately 16MB.
So a dual-head card driving 2 such monitors would require 32MB minimum and I’ll bet 4 times that amount would be more than enough overhead for whatever is to be gained. That is a 128MB card should satisfy any 2D needs and as long as the card supports a fast enough refresh rate at the desired resolution, then I think that’s all you’d need. I would nearly bet you could easily get away with a much less power-hungry pair of cards, saving some money on both the cards as well as the power supply and electric bills. Unless you’re doing 3D rendering or high-end 3D gaming, do you really need the cards you’re looking at? Also, I assume you’d not be running in SLI mode which I understand as two cards working together as one, so power needs for cards should be based upon independent use of each card.
Looking at the eVGA 256-P2-N554-AX Geforce 7600GT 256MB 128-bit GDDR3 video card at NewEgg which is $143, I see that even it still quotes a 12V/18A power requirement in non-SLI mode, but that does buy you some extra power supply headroom. A pair of those would still leave some change vs. the price of one of the 7950 cards you’re looking at.
I just went to the nVidia site and found that for a dual 7950GTX SLI setup, one of the power supplies they recommend is Thermaltake I cited earlier, but also several others including a PC Power & Cooling Silencer 750 EPS12V. That P/S is spec’ed with a single 12v rail up to 60A. I know their products generally get very good reviews but also tend to be very pricey in contrast to other highly-regarding p/s units. Their website also cites this p/s as one of two recommeneded for the Gigabyte mobo you’re looking at.
One final thought….while these cards are spec’ing 12V/18A+ power requirements, I’m sure that’s based upon peak loading rather than constant delivery. And, I’ll bet regardless of how many monitors your’e driving, the power demands for 2D graphics handling will not tax the cards the way 3D would.
I know you don’t want to buy something short of providing you a solid, stable work platform, so just take all of this as food for thought.
Daryl
Hi daryl
Some very useful observations there and you clearly have some well thought out approaches.
Here is my problem I need to spread over probably three monitors the video output from a single application whilst running a number of other applications in the background. i.e the output from PS needs to be much wider than the output from two monitors. Are yopu saying I( can achieve that over three monitors without using SLI? If so your approach has a lot to commend it.
The fourth moonitor would be used for running premier where I do not need multi monitor output.
The othr application is rendering of large panoramic images. This requirement does not represent a large proportion of my work but has been seen as a bonus!!!
Thanks for giving so much of your time – it is really appreciated. David
David,
Guess who’s still up! 2:15am here. I really don’t know much about SLI other than the bits and pieces I’ve absorbed through what I’ve seen regarding it in spite of my lack of interest or need for it. I’m not a gamer. But, I am pretty certain the idea behind both SLI and ATI’s Crossfire technology, is that two video cards work together as one, considerably increasing the video bandwidth and power for intensive 3D graphics, but NOT increasing the available video outputs. I’m sure you could do a bit of online research to confirm that. For more video outputs, you simply need more cards…no special technology beyond an O/S that supports multiple displays.
At work, we bought an IBM Intellistation A Pro workstation which came with a PNY nVidia Quadro 1100 dual-head card, but for which we also bought a unique 4-head Colorgraphic Xentera 4 graphics card for, so that we had 4-monitor support. The Colorgraphic (ATI graphics chipset) card worked fine with a Dell workstation the recommendation was based upon, but it was not compatible with the IBM, so I went out and bought a $100 GeForce FX5600 dual-head card. Even the Quadro card was overkill price-wise. But, those two cards provided 4 outputs total, to support 4 1280×1024 LCD panels and yes….I’ll double-check this tomorrow (today) if I can…. that was with a single desktop spanning all 4 screens and hence, an application window should be able to as well. And this was with two different models of graphics cards no less, although I’m sure two identical models would be preferable. Although these were nVidia cards, the IBM workstation did not have SLI support, nor was it required.
Buying a current production motherboard that is SLI-capable does ensure you that you’ve got two card slots of the same interface type, so at least you can buy a pair of matching graphics cards. But, you don’t have to make use of the SLI functionality. You might actually have to disable it…in the BIOS I suspect.
If you should think about going for a specialty 4-head graphics card, then I’d suggest you review the following for some insight:
Matrox: QID PCIe @ $800
<
http://shopmatrox.com/usa/products/datasheet.asp?ID=801>
ATI: FireMV 2400 PCI Express @ $430
<
http://ati.amd.com/products/firemvseries/specs.html> <
http://www.provantage.com/colorgraphic-100-505115~7COLC00Q.h tm>
nVidia: Quadro NVS 440 @ $412
<
http://www.nvidia.com/object/IO_14605.html>
<
http://www.pagecomputers.com/store/product.asp?catalog%5Fnam e=Video+Cards+%26+Adapters&category%5Fname=33g33c200s220 4&product%5Fid=1010346>
Then, once over price shock…I’d go out and buy a pair of GeForce FX 7600GT cards or similar, and save a chunk of money!
Daryl
David,
As I thought, you can have one window span all four displays of a PC configured with a pair of dual-head video cards, with no special requirements other than setting those displays up properly in Windows. The IBM workstations we have at work with 4 monitors do not have any Adobe applications installed on them, but an Excel spreadsheet spanned all four monitors without any problem. Unless there is something inherent in the GUI design of Photoshop that would not permit such a broad window span, then I’m pretty certain you will be OK. For that matter, if it didn’t work, then I’d not expect it to work with any kind of grahics card configuration unless perhaps a single, quad-head card was used.
The only exception to this that I’m aware of, and this may vary with the video application used, is that video will not span the displays. At least that was the case using Windows Media Player where the video was only shown on the display from where Media Player was initially opened.
Regards,
Daryl
Hi Daryl
Thanks again for your input and thoughtfulness — it is much appreciated. I think my ideas are beginning to settle. Iwill probably go for the Intel Board and an E6300 for the short term and then, when the rprices come down maybe in a year or so replace that with the quad QX6800. So I am going to configure the system so power supplies, coooling etc will be suitable for the upgrade. At the time I upgade the processor I will then consioder installing Vista — so following pretty much your own thoughts of waiting a while for that. Going Vista also opens the dialogue about then shifting to DX10. By that time DX1o should have shaken down and cards available at a reasonable price.
As I am planning on OC for the E6300 I am revisiting the cooling and wondering whether it would be sensible to spend some of the money I save buying the cheaper processor on water cooling. If so do you have any recomendations?
Thanks again
David
Hi David,
I have no experience with water cooling and this system I’ve built is the first I’ve ever overclocked. But, for all I’ve read, the Core 2 Duo overclocks well even with a good air cooler. In fact, as I was reading earlier at NewEgg of someone easily overclocking their 2.13GHz 6400 up to 3 GHz using only the Zalman CP9500 cooler as I have in my system. That’s consistent with what I’ve read elsewhere too. Watercooling has the advantage of higher efficiency and being quieter I believe, and it might be fun to try, but I like the simplicity of a standard heatsink/fan combo and don’t find my Zalman cooler to be that loud. It is running at 2500 rpm this moment and may be the loudest fan that I’m hearing, yet still is very acceptable.
Daryl
