I have recently built a new pc and I am using Autocad Architecture 2009 (x64). -Windows XP x64 -intel q6600 2.4Ghz @3.0Ghz -8GB RAM -ATI Radeon HD3870 @FireGL v7700 etc... I have "softmodded" my HD3870 to a FireGL v7700, which happens to be a certified graphics card listed on the autodesk website. I am confused as autocad still tells me that the card is not certified. I have applied the xml certification database update so something seems wrong. Does this matter? Here is my 'TUNE LOG' (set at default settings) : Performance Tuner Results Log Version: 17.2.0.4 Date of Last Tune: 9/6/2008 Machine Configuration --------------------- Processor Speed : 3.0 GHz RAM : 8190 MB 3D Device --------- Name : ATI FireGL V7700 Manufacturer : ATI Chip set : FireGL V7700 Memory : 512.0 MB Driver : 6.14.0010.6764 Your machine contains a 3D Device that is not certified. As more graphics cards and 3D display drivers are tested, they are added to a list on the Autodesk website. You can check for updates and download the latest certification list at any time. Run the 3DCONFIG command and click the Check for Updates button. Current application driver: Software Current Effect Status: ---------------------- Enhanced 3D Performance: Not recommended and on Smooth line display: Not available Gooch shader: Software emulation in use Per-pixel lighting: Not available Full-shadow display: Not available Texture compression: Not available Available application drivers and effect support: ------------------------------------------------- Software driver Effect support: Enhanced 3D Performance: Not applicable Smooth line display: Not applicable Gooch shader: Software emulation only Per-pixel lighting: Not applicable Full-shadow display: Not applicable Texture compression: Not applicable OpenGL driver: Not Certified Effect support: Enhanced 3D Performance: Available Smooth line display: Not available Gooch shader: Available Per-pixel lighting: Not applicable Full-shadow display: Available Texture compression: Available Direct3D driver: Not Certified Effect support: Enhanced 3D Performance: Available Smooth line display: Not applicable Gooch shader: Available Per-pixel lighting: Available Full-shadow display: Available Texture compression: Available ...anyway, this is my main question; I have been using a low-end onboard nvidia graphics card on my work computer, and i can't say i notice a great difference in performance between the two.(although i am working with a 40MB+ file) I'm very new to autocad so my 3d work isn't that complex. I plan to start learning 3DSmax when i get the chance, but on a glance I have noticed that the performance is alot better when orbitting compared to autocad. I would just like to know if I am having any benefit from this mod or should i just revert back to the original hd3870 drivers. Is the firegl mod worth it? opinions please..
I've never used autocad for anything that required enough 3D power that a FireGL would make a difference... anyway, your needs may be different, and the mod is definitely useful in maya if you want to confirm that your mod was successful, run specviewperf: a result around 30-50 for maya means it didn't work, above 200 means it worked
what driver do you use? d3d or opengl on autocad2009? as i use adt2006 (for the best 3d performance) it uses opengl and it improves performance noticable. and i also use 3smax9 with firegl maximum drivers, that improves performance also very much compared to default d3d or opengl drivers. it's a pitty that there is no firegl maximum drivers for 3dsmax2009 and from adt2007 and up autocad versions, 3d performance decreased, so i stick to adt2006 and max9
Well I have it on direct3D for both Acad2009 and 3DSmax as it sets this as default. Should i change it to opengl? ...if i am able to confirm that my softmod was a success, this would be a good start.
as I said, to see if the mod was successful, launch specviewperf (and look at the maya result); for this you don't need to have maya, or know how it works
as i know there is a the moment no x64 mod for the 3870. maybe youre the first who reached it.. or i cannot use google... i dont want to except this if you can paste your mod script i will tell you if it works.
Softmod is for OpenGL mode only. If the application is running in D3D mode then softmod is useless. You might want to change to OpenGL mode and test again.
You just have to look around in this forum and you will find the patchscript for xp64 mod for 3870 I did mine and here's my result specviewperf 10.0 Q6600 @ 3.6 GHZ 3dsmax-04 50.91 no result catia-02 47.58 no result ensight-03 58.56 no result maya-02 267.59 no result proe-04 42.94 no result sw-01 99.50 no result tcvis-01 37.16 no result ugnx-01 56.07 no result @ 3.75 GHZ (**I didn't prime95 this one but it managed to finish off the benchmarks**) 3dsmax-04 53.84 no result catia-02 50.44 no result ensight-03 59.80 no result maya-02 272.90 no result proe-04 45.63 no result sw-01 106.16 no result tcvis-01 39.31 no result ugnx-01 56.39 no result I use solidworks alot so just anxious to see if i can break the 100 mark.. cheers
I followed a 'HOWTO' from cgsociety.org and i used the patchscript found in kilema's posts. i'll download specviewperf and see how i get on, i'm now in opengl.
I use proengineer, so i applied the patch on a radeon HD 3870 to a firegl V7700. Running the specviewperf 10 the proe-04 scores are over 40, but for work the card is the same or slow as a quadro fx1400. This made me think if there is some driver trick that identifies if there are running benchamarks or specific applications. If someone coul confirm this I would appreciate.
Pro-E scores and forth... I ran tests on a lot of configurations. I ran them on on different configurations of : CPU s : Pentium 630, Pentium E2180, Core2 Duo E4600, Core2 duo E6300, Core2 Duo E 6750, core2 duo E8400-E8500, Core2 quad Q6600. All running on stock clock speeds. I OC ded E8400 and E8500 to analyze sensitivity of SPEC tests. RAM : 2 x 1 GB DDR2 667 and DDR2 800. HDD : 7200 rpm SATAII and 15K rpm SAS disks VGA :Un modded Quadro FX1700, Quadro FX 3500, Quadro FX 4600, 8800 GTX, HD 4850, HD 4870 Modded : HD 3650, HD 3850, HD 3870 : run on stcok clock speeds and OC ded settings. Furthermore I run SPECapc tests for 3DS Max, UG NX 4, Solidworks 2008. And I analyzed lots and lots SPECviewperf and SPECapc scores. My conclusions are : Given CPU and RAM are the same : 1 ) Maya and UG NX prefer FireGL architecture; any low level FireGL beats even Quadro FX 5600. 2 ) On other tests ; any low level Quadro FX beats even FireGL V7700. There are Quadro NVS and/or Mobile Quadro scores that beat FireGL v7700! Given that RAM and VGA is the same : 1 ) 2.4 GHz Core2 Duo E4600(2MB cache/800MHz) beats 2GHz Pentium E2180(1MB) by nearly 10 percent or so. 2 ) 1.86 GHz E6300 ( 2MB /1066MHz) beats Core2 Duo E4600 by 10 percent or so. 3 ) 2.66 GHz Core 2 duo E6750 ( 4MB, 1333 MHz) beats Core2 duo E6300 by 10 percent or so 4 ) 3.16 GHz Core2 duo E8500 (6 MB,1333 MHz ) beats Core2 Duo E6750 by 10 percent or so. 5 ) Faster clock speeded Core2 Duo E6750 and Core2 Duo E8400/E8500 beats Core2 Quad by 10 and 20 percent respectively. SPEC tests heavily prefer higher cache and higher speed over more cores. Given that CPU and VGA is the same : 1 ) 800 MHz RAM and 667MHz RAM does not effect test scores. 2 ) 7200 rpm SATAII and 15K rpm SAS disks does not effect test scores. Given that CPU, RAM are the same and we use modded HD series cards : 1 ) The applications that favor Quadro FX architecture benefit from CPU overclocking ; 20 percent overclock is roughly equal to 20 percent increase in scores; however Maya and UG NX scores do not change at all. 2 ) The applications that favor FireGL architecture ( Maya and UG NX ) benefit from VGA overclocking; 15 percent increase in VGa speed roughly equals 15 percent increase in scores for these tests; however other 6 test scores do not benefit at all. 3 ) If we overclock both CPU and VGA; NVidia favoring tests suffer a huge blow; FiregL preferring test scores suffer less. Probably that is due to the fact that I used 400W and 500W PSU s which had max 27-28A capacity on 12V rails and PSU was not able to cope with these loads. So : a ) every application is written to perform a minumum score on their preferred VGA; b ) If you need more score, you need to upgrade to a higher scoring CPU or overclock. c ) To get good performance on HD 3850/3870 mod you need at least E8xxx series Core2 Duos which can supply the VGA card with sufficient data. Even E6750 is not able to utilize 3870 s processing power. I also have an engineering sample 3 GHz Core2 Quad Q9650 ( 12 MB, 1333 MHz ), but could not test the under SPECviewperf or other SPEC apc tests. I run SPEC apc UG NX 4 on both E6750 + Quadro FX 3500 and E8500+HD 3870 runnign on 8.44. The SPEC apc scores are : 4.92 for Quadro FX and 9.98 for HD 3870. Today I changed the VGA of HP xw4600 ( E6750+2GB 667MHz+15K rpm SAS+Quadro FX 3500) to HD3870 running on 8.453.1.6. The user opened a very heavy UG NX 4 model and told me it was at least as good as FX 3500.since he also uses Pro-E, he will try it out for a few days more. My conclusion for you as a Pro-E user : You could have easily bought yourself a Quadro Fx370 or 570 and could get the more or less the same performance as a HD 3870. I hope you can analyze your components according to these informations and reach some conclusions.
