Jump to content
Sign in to follow this  
Danny Nolan

A major project I'd like to see: optimization

Recommended Posts

Don't get me wrong. I enjoy tech upgrades - the material project is pretty awesome, and something we def need. As was mesh, shadows, etc.

But a major project whose soul purpose is purely just optimization, I think, is something we really need. All this tech is getting crammed into a client that's already crammed.

Things I'd like to see is more threads, option of using more vram, SLI/crossfire working/helping again, network system cleaned up/optimized, etc.

Share this post


Link to post
Share on other sites

More threads?  More than 512 mb of VRam, SLI/Crossfire support?

All these are viewer features, so what use is your link to SL server articles?

None of these things are supported by any  viewer for SL.

Do you use some secret viewer that nobody knows of?

J.

 

Share this post


Link to post
Share on other sites


Nalates Urriah wrote:

Dude, where have you been?

Most of the last part of 2012 was doing just that. Read through these:
.

Okay. Those are server changes. While they're appreciated, they're not in any way related to my original post lol.

I'm talking core engine optimizations.

Share this post


Link to post
Share on other sites

Things I'd like to see is more threads, option of using more vram, SLI/crossfire working/helping again, network system cleaned up/optimized, etc.

When I first read this, as Nal, I first thought of the recent server optimizations that involve "more threads" and "network system" clean-up, but yeah, that has only involved the viewer code for a bit of network streamlining.

I'm no viewer dev, so I don't know what would be involved in expanding vram use, but then I don't even know what all the viewer does with vram besides the texture memory buffer. It certainly seems odd that my 2GB card (and somebody else's 4GB) has the same slider as an old 512MB, but I don't know what's really happening behind the scenes.

From what I see of the nVidia roadmap, I'm not sure I'd focus much development on SLI. In a stacked DRAM ("Volta") architecture with GPU-memory bandwidth in terabytes-per-second, SLI seems destined to buggywhip status in a couple years. On the other hand, it's been a very successful marketing feature for nVidia, so they may well figure out something to keep doing with it.

More viewer threads, however, seems an unmitigated win for any future hardware developments I can imagine.

Share this post


Link to post
Share on other sites


Qie Niangao wrote:

I'm no viewer dev, so I don't know what would be involved in expanding vram use, but then I don't even know what all the viewer does with vram besides the texture memory buffer. It certainly
seems
odd that my 2GB card (and somebody else's 4GB) has the same slider as an old 512MB, but I don't know what's really happening behind the scenes.

There is an old JIRA about video memory size:

Build 102206 does not pick up texture memory correctly

Zen Linden (Inactive) added a comment - 19/Nov/08 1:32 PM:

We now cap the amount of texture memory used at 512 MB to prevent some driver bugs when using more than that. 512 MB should be adequate for now.

[unquote]

It appears that the max 512 MB memory cap is still on even in the latest viewers.

Cannot put the texture memory higher than 512 MB.

 

PS

From the link below we can read:

http://community.secondlife.com/t5/Technical/Max-Video-Ram-and-rendertexturememorymultiple/qaq-p/717013

the max settable via the SL client is 512, however that is NOT the entireity of the VIDEO memory used, but rather the portion reserved for textures only.(some is also allocated to draw geometry and other operations. as well as what appear to be persistent memory leaks in the official client)"

[unquote]

 

 

So, if I understand that correctly, then it means:

• We set the texture memory allocation in the viewer to 512 MB

• Then that 512 MB of video memory is reserved and used for textures

 

If we have more than 512 MB video memory in our GPU is it of any use?

 

Yes, it is.  Memory above that 512 MB is used for (quoting from the above linked text):

"to draw geometry and other operations, as well as what appear to be persistent memory leaks".

(Provided that I understood the matter right.) :matte-motes-wink:

Share this post


Link to post
Share on other sites

You're quoting Linden comments that are almost 4 years old and  LL V1.2 jiras.

So there has been no development in 'broken' video driversGPUs in general ever since?

Seems adding shiny new stuff is more important than  keeping up with technical progress....

By the way: My GTX Titan offers 6 GB of VRam, I am sure using more than 8.3% of my GPU's VRam will not slow down things.....

J.

Share this post


Link to post
Share on other sites


Jean Horten wrote:

 

So there has been no development in 'broken' video driversGPUs in general ever since?

That's what I was wondering too when reading that ancient JIRA and seeing that the limit is still 512 MB in the latest viewers.  Perhaps there is some other reason too besides those 'broken' video drivers?  I have no clue.

Share this post


Link to post
Share on other sites

I'm not a DEV either; my only coding gift to the world was boob physics for emerald.

But, I really don't feel like allowing more vram would be much of an issue. Prob just not something they've thought about or considered.

Enabling SLI, though I can see as being a bit more steep. It involved working with the renderer, which is def more difficult code. But, it'd help pretty substantially. Shadows + DoF + Materials + SSAO + Screen Space Reflections. We're actually pushing a lot of work into the GPU - and we're at a point where SLI would actully be benficial.

Even if, say increasing max vram, isn't just as simple as I'm thinking - then why is it still an issue? LL has programmers - those of which get paid. Difficult or not, if it's in the best interest of your product, then I feel like it should have a substantially higher priority. *shrug*.

Share this post


Link to post
Share on other sites

Using GPU-Z I have seen almost 2gb of VRam used in some cases. Used by what I assumed textures but possibly what the other posting mentioned Geometry. How is the performance of your Titan on Sl compared to other video cards? I'm using a GTX 670 and amazingly can get it ramped up to 98% GPU usuage with shadows on etc.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

×
×
  • Create New...