GeForce vs Quadro...
-
...once and for all!
Okay, days (hours) away from ordering the new 'chine. Found the Google rec for 512meg (3D class Video Card with 512MB of dedicated memory...?) but need to know if the quadro is worth the extra bucks...is it $4-500 better? I'd rather spend it than wish I had, but I'm not much into wasting money either, thanks.
Lay it on me please! (Sure hope you helpful Google folks way in on this too!?!)
-
i opted for the new Geforce 8800 GTX... at the moment i think it has the best price to performance ratio... u could save the extra bucks to get more ram power or more processing power per CPU..
-
I got two 8800 GTX's in SLI (had to up my PSU to 1000 however) and I have not looked back as far as performance goes, I cannot speak for the Quadro as i have never had one.
-
Tom,
If you decide to buy a Quadro, select one of the more expensive ones. Their price/performance ratio is much worse than that of the "consumer" cards. I consider buying one as a sort of future-proofing. The only advantage is if you want to run several OpenGL-using applications at the same time. For instance, on my GeForce-equipped laptop it is not possible to view a 3D PDF if SU is open at the same time, but otherwise the display performance is not much worse than with the Quadro 1400 in my work computer.
Anssi
-
I used a midrange geforce 6600 on my previous workstation and I thought it was pretty good. I now use worstation class laptop (hp compaq 8710w) that has a quadro fx1600 card which blows the socks off the previous machine.
If I was you I would find a waranteed refurb workstation with a quadro fx3500 / fx4600 card and save the money that way.
Here is a link with some technical info:
http://www.nvidia.com/object/quadro_geforce.html -
Hey, first of all thanks a bunch to all responding...really helped!
The article you provided, tv, though mostly Greek to me, put into one place other info I've read, but my question now is whether SU actually uses the enhancements provided by the Quadro cards and drivers. If not, it would seem to the bearly-understanding-all-this like me that a 728 GTX at $50 cheaper would be better than a 512 FX...? If so, then the quadro...?
It's a pity the helpful folks at Google haven't piped in...you'd think they'd have a file drawer or so of benchmark tests they could share...? (I can't get a response from their help site either...are answering such questions a corporate no-no? :`)
Speaking of ram:-) I read recently about the "3gig switch" which will allow XP to dedicate it's maximum(?) 2gig usage fully to a program. So what's the deal on ram? Since I rarely have more than one program running at a time, any reason to have more than 3gig? Beyond that, wouldn't my money be better spent on cache...and how much is too much cache memory?
Also, my rendering program of choice says: "though it will run on multi-core processors, PSP does not specifically divide processes between cores"...any reason I shouldn't find the fastest Pentium 4 I can find laying around rather than go with a slower duel core?
Hope you'll keep the answers coming, guys, this is great thanks!
-
Okay, finally got an answer from SU help...this link from 2003 :`):
Anyway, Yasser's last post there was quite helpful: if things haven't changed much from rel(what?)4...3? So...I repeat my last questions, thanks? But this would seem to say the 8800 GTX with 712meg would be better than the FX1700 with 512meg, the 3ghz dual w/ 4meg cache would be better than the 2.33ghz quad w/ 8meg, and doesn't answer the ram question...?
And now I'm wanting to read the "SU Concepts" help topic to find such good answers to whatever else is in there, but couldn't find it with several searches...can anyone help me with that? See quote from same below:
*I'm afraid for SU projects with a large number of polygons, the bottleneck is not the video card but the CPU. From the "SU Concepts" help topic:
"Transform
This second bottleneck has to do with how much geometric detail your CPU can handle. SketchUp is designed for design exploration, and as described above, each of the face and edge entities are a lot more "aware" of the other elements around them than in 3D modeling or CAD application. This means that your CPU has to do more work to process the geometry of a SketchUp model. This also means that if you have a lot of edges and geometric detail in your model, it's likely that your video card may not be where the slowdown is happening. A good way to test this is to resize your window to a very small size. This cuts down the fill rate requirement quite a bit thereby bypassing most of the video fill rate issues. A heavy model will still run slowly. The only way to improve the transform performance is to upgrade your CPU: The faster, the better."The rule of thumb is to get the fastest CPU you can afford, then get the fastest GPU you can afford of the same generation as your CPU. (They should be released and available for sale at roughly the same time.)
After that, the best results come from project management, I'e. splitting up the SU model into components that you can swap with lower poly "proxy" versions during modeling and swap back in as high poly during rendering or final output.
SU is exploration-oriented, and unfortunately we can't optimize OpenGL in exactly the same way as other 3D/CAD apps.*
-
bump :`)
-
As far as I know the old @last advice (mostly from Jim Holman -- may he be enjoying his post acquisition retirement to the hilt) is still true. Get the fastest processor you can buy (say an Intel P4 extreme at 3.8 Ghz -- good luck finding a ready made computer with one), have a decent amount of RAM (2 or maybe 3 Gb) and a decent video card, with video ram dependent on how big your display is. 256 Mb Video ram should be okay for a 1920 x 1200 single monitor. I must add, however, that since I switched to Quadro cards I have never seen an artifact. But the newer Geforce cards may very well be okay.
If you are only running SU (and not other OpenGL 3D programs simutaneously) you might want to look at the Quadro 1500 or 1600 series. And remember to use the model management techniques....
-
(Slightly unrelated..)
I had some concerns about using a slower dual core machine with Sketchup but I find that despite the lower processor speed working with multiple programs at the same time is a lot more responsive than on a faster pentium. This appears to be true of working with SU alone. Combined with the FX1600 card the 2.2GH dual core is a lot faster than 3GH+ pentium with geForce 6600 configuration of equal memory (2 Gig).Also be aware that since the Santa Roza version (2007) the dualcore centrino processor can speed up single thread processes:
http://en.wikipedia.org/wiki/Centrino#Santa_Rosa_platform_.282007.29@unknownuser said:
The Santa Rosa platform comes with dynamic acceleration technology. It allows single threaded applications to execute faster. When a single threaded application is running the CPU can turn off one of the CPU cores and overclock the active core. In this way the CPU maintains the same Thermal Profile as it would when both cores are active. Many expect Santa Rosa to perform well as a mobile gaming platform due to its ability to switch between single threaded and multithreaded tasks
We really need to develop a Sketchup benchmark script to be able to test hardware in a more scientific way. When you read the hardware tests on workstations you can see that benchmark performance on machines with similar configurations from different suppliers can be dramatically different. To throw components together arbitrarily does not necessarily guarantee a quick machine.
-
tv, paul...you have reinforced where I'm leaning: my head says go with the old knowledge (to a point), my gut says go with the new technology (to a point). Please also respond as you can (all who have been so kind here too) to my new thread on notebook vs desktop and the systems I have flagged for purchase this year.
My very best, Tom.
Advertisement