I have already had extensive talks with Tom Goddard about why the jaggies are there and there is alreadys an entry in the gnats database. Normally, I try to not burden the email list with these details and keep them in the gnats database, but to summerize: Without hardware multisampling/supersampling (which chimera uses if it is available), you'll always have jaggies for cylinders, spheres, and surfaces. It is just more noticable if you have a white background because of how the primitives are shaded (dark on the sides). The solution for printing, is to supersample, that is to generate an image that is 16 times larger and average 16 pixels into one (9x might be sufficient). Regular grid supersampling will work for molecular modeling images (instead of the jittered sampling that ribbonjr uses) because there are extremely few right angles (so there is almost no chance that the "picket fence" effect of primitives appearing and disappearing from one frame to the next). Greg On Wed, 14 May 2003, Tom Ferrin wrote:
Okay, but please respond to Tom Goddard's email (w/ cc to chimera-dev) about jaggies in Chimera images.
--tom
--------
Forgot I had a dentist appointment today. First in many, many years. Marriage is a good thing.
Greg
Greg, Which of the mid-to-high end graphics cards support hardware multisampling/supersampling? --tom --------
I have already had extensive talks with Tom Goddard about why the jaggies are there and there is alreadys an entry in the gnats database. Normally, I try to not burden the email list with these details and keep them in the gnats database, but to summerize:
Without hardware multisampling/supersampling (which chimera uses if it is available), you'll always have jaggies for cylinders, spheres, and surfaces. It is just more noticable if you have a white background because of how the primitives are shaded (dark on the sides). The solution for printing, is to supersample, that is to generate an image that is 16 times larger and average 16 pixels into one (9x might be sufficient). Regular grid supersampling will work for molecular modeling images (instead of the jittered sampling that ribbonjr uses) because there are extremely few right angles (so there is almost no chance that the "picket fence" effect of primitives appearing and disappearing from one frame to the next).
Greg
On Wed, 14 May 2003, Tom Ferrin wrote:
Okay, but please respond to Tom Goddard's email (w/ cc to chimera-dev) about jaggies in Chimera images.
--tom
--------
Forgot I had a dentist appointment today. First in many, many years. Marriage is a good thing.
Greg
I believe that all of the currently mid-to-high end shipping workstation class PC graphics cards support multisampling. It's not clear if it is enabled by default though, you might need to configure the driver. For instance, with the NVidia Quadro4 900 XGL graphics card on my desk, I changed the default settings to be "highest" quality in XP and for Linux I had to set some environment varialbes before starting the X server. Those cards are the NVidia Quadro4 and Quadro FX series, the ATI FireGL Z1 and X1, the 3dLabs Wildcat3, Wildcat4 and Wildcat VP series. The cost of these cards range from $400 to $3000. Any workstation graphics card that costs less than $400 is probably an entry-level card and should be avoided (i.e., some Quadro4 variations). As for the mid-to-high end consumer graphics cards (otherwise known as gaming graphics cards), they have multisampling support, but it is unclear if it only applies to full-screen applications (games) or if it works within a single window. The online literature is not clear. The literature seems to be missing any euphemisms for full screen only, so it should work. Those cards are the Nvidia GeForce4 Ti and GeForce FX series, and the various ATI Radeon 9000 series (the 9500, 9600, 9700 and 9800 are best). And just like the workstation cards, you probably have to tell the graphics driver that you're more interested in quality than speed. (Oh, and there's the Matrox Parhelia series, it should work too.) Unfortunately, we won't be able to tell for sure without getting our hands on more of these cards. The cost of the mid range should be $200 to $300. The high end should be $400 to $600. Greg On Wed, 14 May 2003, Tom Ferrin wrote:
Greg, Which of the mid-to-high end graphics cards support hardware multisampling/supersampling?
--tom
--------
I have already had extensive talks with Tom Goddard about why the jaggies are there and there is alreadys an entry in the gnats database. Normally, I try to not burden the email list with these details and keep them in the gnats database, but to summerize:
Without hardware multisampling/supersampling (which chimera uses if it is available), you'll always have jaggies for cylinders, spheres, and surfaces. It is just more noticable if you have a white background because of how the primitives are shaded (dark on the sides). The solution for printing, is to supersample, that is to generate an image that is 16 times larger and average 16 pixels into one (9x might be sufficient). Regular grid supersampling will work for molecular modeling images (instead of the jittered sampling that ribbonjr uses) because there are extremely few right angles (so there is almost no chance that the "picket fence" effect of primitives appearing and disappearing from one frame to the next).
Greg
On Wed, 14 May 2003, Tom Ferrin wrote:
Okay, but please respond to Tom Goddard's email (w/ cc to chimera-dev) about jaggies in Chimera images.
--tom
--------
Forgot I had a dentist appointment today. First in many, many years. Marriage is a good thing.
Greg
With the current Chimera web page setup, the only link to anything dealing with graphics cards is TG's "Hardware Benchmarks for Volume Data Rendering" page, and this is focused completely on performance. There are a couple of older pages that I used to regularily maintain before TG put together the volume benchmarks page that also included other info, like stereo viewing support. (These are still on the CGL web site, but with the new Chimera home page there's no direct link to my pages anymore.) We need to expand the current TG page or create another page that discusses other graphics card issues, like stereo viewing and multisampling. And we should call user's attention to graphics driver settings in general, so that they know that one setting may not always be the best. I'm not thinking of details of settings for inidividual cards, but rather just the fact that at times one might be focused on good performance and at other times highest quality may be more appropriate. Can you discuss this at today's Chimera's developers meeting? Unfortunately I can't attend as I'm in Washington D.C. Let me know what gets decided. --tom --------
I believe that all of the currently mid-to-high end shipping workstation class PC graphics cards support multisampling. It's not clear if it is enabled by default though, you might need to configure the driver. For instance, with the NVidia Quadro4 900 XGL graphics card on my desk, I changed the default settings to be "highest" quality in XP and for Linux I had to set some environment varialbes before starting the X server. Those cards are the NVidia Quadro4 and Quadro FX series, the ATI FireGL Z1 and X1, the 3dLabs Wildcat3, Wildcat4 and Wildcat VP series. The cost of these cards range from $400 to $3000. Any workstation graphics card that costs less than $400 is probably an entry-level card and should be avoided (i.e., some Quadro4 variations).
As for the mid-to-high end consumer graphics cards (otherwise known as gaming graphics cards), they have multisampling support, but it is unclear if it only applies to full-screen applications (games) or if it works within a single window. The online literature is not clear. The literature seems to be missing any euphemisms for full screen only, so it should work. Those cards are the Nvidia GeForce4 Ti and GeForce FX series, and the various ATI Radeon 9000 series (the 9500, 9600, 9700 and 9800 are best). And just like the workstation cards, you probably have to tell the graphics driver that you're more interested in quality than speed. (Oh, and there's the Matrox Parhelia series, it should work too.) Unfortunately, we won't be able to tell for sure without getting our hands on more of these cards. The cost of the mid range should be $200 to $300. The high end should be $400 to $600.
Greg
On Wed, 14 May 2003, Tom Ferrin wrote:
Greg, Which of the mid-to-high end graphics cards support hardware multisampling/supersampling?
--tom
--------
I have already had extensive talks with Tom Goddard about why the jaggies are there and there is alreadys an entry in the gnats database. Normally, I try to not burden the email list with these details and keep them in the gnats database, but to summerize:
Without hardware multisampling/supersampling (which chimera uses if it is available), you'll always have jaggies for cylinders, spheres, and surfaces. It is just more noticable if you have a white background because of how the primitives are shaded (dark on the sides). The solution for printing, is to supersample, that is to generate an image that is 16 times larger and average 16 pixels into one (9x might be sufficient). Regular grid supersampling will work for molecular modeling images (instead of the jittered sampling that ribbonjr uses) because ther e are extremely few right angles (so there is almost no chance that the "picket fence" effect of primitives appearing and disappearing from one frame to the next).
Greg
On Wed, 14 May 2003, Tom Ferrin wrote:
Okay, but please respond to Tom Goddard's email (w/ cc to chimera-dev) about jaggies in Chimera images.
--tom
--------
Forgot I had a dentist appointment today. First in many, many years. Marriage is a good thing.
Greg
participants (2)
-
Greg Couch -
Tom Ferrin