djohannesen Posted July 23, 2003 Share Posted July 23, 2003 I have a Pentium 4, 1.6GHz, 1GB Ram, 60GB 7400 HDD & a Jaton NVIDIA Geforce 2 MX/mx400 with 64MB RAM. I wish to optimise the redraw rate & overall, the advice was to change processor, increase RAM, (at the time only 512MB), & get another card. Whilst I have increased the RAM, I can't justify the expense of a new processor & motherboard & RAM & video card. (Next year yes). My system will only run 4X AGP. Has anyone out there have a similar system, & if so could you please advise whether I can expect much improvement by changing the video card. As a reference to Ironcad performance, if I drag a standard block in from the catalog, use the Triball to copy it 30 times at a distance of 100mm, it takes 13-14 seconds to perform this. Any advice greatly appreciated. Regards, David J I have posted this notice previously on the forum & have appreciated all the replies. Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 23, 2003 Share Posted July 23, 2003 Copying a part 30 times has nothing to do with the video card. RAW cpu power will improve the time it takes to create new geometry such as that. The latest CPU will perform that operation in 5-6 seconds. ROTATING those 30 blocks once they have been created is all video card. Quote Link to comment Share on other sites More sharing options...
Rick Dorrington Posted July 23, 2003 Share Posted July 23, 2003 P4 1.5 1Gig (RamDac) 18Gig Ultra160 10k Rpm Fire GL2 6 Seconds Rotating the blocks? A breeze. 407 frames/sec @ 1600 x 1200 with this video test. Fast computers have so much more underneath the hood than processor, memory and video card. Again, you get what you pay for. Rick EdgeVBM.exe Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 23, 2003 Share Posted July 23, 2003 I may have to recant what I just said. Raw cpu *should* be the factor that impacts you the most; but I'm finding that IronCAD might be doing something unexpected that brings the disk into play. These are very very preliminary findings that are hopefully wrong but what I'm seeing is that SCSI verses an IDE hard drive might have a wildly significant role in this test when it shouldn't come into play at all. Here's a test that I just ran. 3.06g xeon with 2 gigs of ram and a 7200 rpm IDE drive: 7 seconds to perform that test. P2/500 with 256 ram and a 10k rpm SCSI drive: 3 seconds to perform the exact same test. These are disturbing results in the least. Do not quote me on them yet and do not view them as an official statement; it's just something that I'm working on. According to the way IronCAD creates those blocks; the data should only persist in RAM; but it seems like the disk is involved where it potentially should not be. I'll let you know when I test this further and talk with our developers about it. Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 23, 2003 Share Posted July 23, 2003 It's been confirmed... Whenever you perform a copy operation, or when you shade a 2d view; ironcad must temporarily use the local disk to perform the operation. So copying a block uses not only the CPU, but the local disk as well; thus, copying a block is not an accurate gauge of CPU power. I'm going to come up with a test scene for everyone to run and report back on that will use only the cpu. Quote Link to comment Share on other sites More sharing options...
djohannesen Posted July 24, 2003 Author Share Posted July 24, 2003 Thank you all for your replies. Very much appreciated. Will be waiting to hear your final results Chris, as there has been some pondering on this subject from others here. Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 24, 2003 Share Posted July 24, 2003 Ok; here's a good test that taxes the cpu only. 1) Load this scene (3.5 megs) --Don't touch the part or rotate the scene 2) Hit "Select All" using either ctrl+a or from the edit menu 3) Pull down the Shape menu (from the top, not the catalog) and choose "Regenerate" (Don't time this regenerate) 4) Wait *5)Re-do the select all+regenerate and this time, take not of how many seconds it takes for you to regenerate. I tried to keep the part simple so that a P2/500 could regenerate it in about a minute (around 57 seconds). It takes our 2.0 ghz box 22 seconds to regenerate; and our 3ghz 17 seconds. If course this is very simple and is intended to quickly gauge the CPU speed only. The percentage difference between the different times will be greatly magnified when you add more data. So while the difference between a 2.0 and 3.0 ghz system is only a few seconds with this small data; it could easily be several minutes with larger machines. Note: If you rotate the scene and you can not render this part in full shading while rotating; then you need a stronger card. All cards should "studder" when you first attempt to rotate this but should quickly recover and display the faces in real time. facet_test.zip Quote Link to comment Share on other sites More sharing options...
Rick Dorrington Posted July 24, 2003 Share Posted July 24, 2003 P4 1.5 56 Seconds. 3 tries. Yes for stutter, but just a blink. Curious how a P2/500 can score 57. Must be a mutant. Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 24, 2003 Share Posted July 24, 2003 Nooooooo; I'll retest on our P2. Quote Link to comment Share on other sites More sharing options...
Guest mwalls Posted July 24, 2003 Share Posted July 24, 2003 Chris, This question goes back to the copy test you did, so here is an interesting one for you. I have a Dell with 2.66Ghz xeon, 1gig mem, 128meg Fire GL X1 video, Seagate ST336753LW 15K rpm SCSI drive. When I do the copy thing with the block, it takes around 7 seconds to perform the task. I see that you said you were able to do it in about 3secs. with a far less powerful system. Do you have any idea why my system would take so long with the 15K rpm drive???? This is curious to me. Any ideas would be helpful. Thanks. Mike Quote Link to comment Share on other sites More sharing options...
Rick Dorrington Posted July 24, 2003 Share Posted July 24, 2003 P2-450 147 seconds. ( 2 min 27 sec) Forget the video. That P2-500 must have the wrong sticker on it. Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 24, 2003 Share Posted July 24, 2003 Ok it wasn't a P2 and, the counting was off...so just ignore the times that I put in that post above for the P2 Here's what we just tested -P3/650: 72 seconds -P3/900: 50 seconds (retested this many times. This is a dell laptop. Can't explain why it's faster than your 1.5) -We don't have anything in the 1.xghz range Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 24, 2003 Share Posted July 24, 2003 Mike is your 15krpm disk connected to a built in SCSI Ultra controller or is it on a Perc Raid? Hmm...why am I asking...perhaps it could be a read/write cache/read-ahead/write-through setting in the controller. I ran the block copy test on a Dell Precision 410 with a 9 gig scsi drive running on the built in Adaptec 7890 scsi controller with no partitions and nothing on it other than Windows XP and IronCAD; with the single drive running the ntfs. Perhaps you have multiple partitions; perhaps you're running something other than XP; perhaps you're on FAT32 verses NTFS? Umm...other large files loaded...maybe it's a service pack in XP...the disk could be fragmented... Hm... Quote Link to comment Share on other sites More sharing options...
Joel Parrish Posted July 24, 2003 Share Posted July 24, 2003 Just to give you Intel guys something to chew on... Doing these tests for fun... Copy Test 1 to 2 sec Chris's regen facet_test. 18 sec AMD 2500+ Barton 1 Gig of 400Mhz DDR 7200 UATA-100 GeForce 4 Ti4800 8x AGP 128 Meg Joel Parrish www.pai-design.com Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 24, 2003 Share Posted July 24, 2003 OOoohhhh thanks Joel. Your 2500 cpu is the same speed as our 3.06ghz xeon but the copy test is VERY interesting. Quote Link to comment Share on other sites More sharing options...
Rick Dorrington Posted July 24, 2003 Share Posted July 24, 2003 I bid 40 seconds. I have discovered that the "surface smoothness" affects the facetting part of the test. Mine was set very smooth (120) If I set to 25 or less then we get 40 seconds. Is everyone on the same setting? Does the facetting section have anything to do with video? Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 24, 2003 Share Posted July 24, 2003 The surface smoothness is specific to the part (I set it up in the part properties); so it's the same for everyone that load my scene (even if their default is set to something else). I set the surface smoothness "on high" intentionally so that it would take longer for the part to regenerate. The smoothness most certainly has an impact on the ability to rotate the camera in real time; since a higher smoothness adds more facets to the part. Quote Link to comment Share on other sites More sharing options...
Guest mwalls Posted July 25, 2003 Share Posted July 25, 2003 Chris, I'm using an Ultra320 SCSI card, Win2K Pro, single partition, NTFS, 2% on fragmentation, and the only other thing running was MS Outlook???? I'll poke around to see what the read/write cache etc. is at. Thanks for the info. I've always been impressed with the few times I have seen something similar done on an AMD system vs. an Intel. Interesting....... Mike Quote Link to comment Share on other sites More sharing options...
Guest svangeldern Posted July 25, 2003 Share Posted July 25, 2003 Sony Laptop P4 2.4 512 RAM Regen facet test: 36 sec. Copy Block test: 1-2 sec. Home Built AMD 950 768 RAM Regen facet test: 56 sec. Copy Block test: 1-2 sec. Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 25, 2003 Share Posted July 25, 2003 Hmm, there's something more to this copy block test. Steve did you set your units to millimeters and then copy it 30 times at 100 mm distance? It should have been much slower on the laptop. Are you running an ATA drive on the AMD? Thanks for posting your results everyone. Quote Link to comment Share on other sites More sharing options...
B. Ludin Posted July 25, 2003 Share Posted July 25, 2003 I get almost the same results as Steve with my Dell Inspiron 2.4GHz (1GB RAM, nVidia 4200 64MB), i.e. approx. 2s for the block copy test 33s for the part regeneration and no problems whatsoever with rotating the part at 1920x1200 or while sharing the screen with Rick's swirling masks (rate drops from 65 to 55fps) Cheers, Beat Quote Link to comment Share on other sites More sharing options...
Guest svangeldern Posted July 28, 2003 Share Posted July 28, 2003 Chris, I ran the copy test again this morning (mm, 30 copies at 100mm) ...The laptop is still 1-2 sec, but the home built is about 15 sec. Steve Quote Link to comment Share on other sites More sharing options...
Chris Lohman Posted July 28, 2003 Share Posted July 28, 2003 Steve what laptop do you have and what type of disk drive is running in the clone? Quote Link to comment Share on other sites More sharing options...
Guest svangeldern Posted July 29, 2003 Share Posted July 29, 2003 Chris, Home Built HD is an ATA: Maxtor 52049U4 Maximum Capacity of 40.9 GB UltraDMA 66 Data Transfer Rate 2 MB SDRAM Cache Buffer 7,200 RPM Spin Speed < 9.0 ms average seek time Laptop SONY PCG-GRV550 Fujitsu MHR2030AT Maximum Capacity of 40.9 GB 2MB data buffer UDMA/100 Transfer Rate 4200RPM Spindle Speed Average Seek Time: 12ms Steve Quote Link to comment Share on other sites More sharing options...
Guest dlalonde Posted December 3, 2003 Share Posted December 3, 2003 Thanks for the EdgeVBM.exe openGL test Rick. I've found it useful to determine how effective my performance tweaks have been. Speaking of which, I recently upgraded my pc to a shuttle 'lunchbox' computer. AMD XP3000+ [2167Mhz] 512 Ram [DDR dual-channel] I replaced my beloved old Geforce256DDR graphics card with a Geforce3 200 Ti [made by MSI] which I purchased new on ebay for US$43.00 After some softquadro magic, it is detected as a Quadro DCC [Quadro 3] and it does 476 fps at 1600 x 1200 with Rick's test. Yeah baby! $43 bucks! Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You are posting as a guest. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.