OTOY Forums

smicha wrote:

glimpse wrote:

smicha wrote:Marcus,

Titaz Z question - the score for 4x Titans Z is 330 - is it for 8 gpus? Or for 2x Titans Z, 4 gpus?

from results it's clearly 2x TitanZs - 4GPUs -
Would be nice to have "*" sign explaining this
(in any dualGPU case =)

+1

4x titans Z is misleading. Please correct it to 2x titans Z. Same for all dual gpu cards.

or at least mention somewhere on the bottom.

We will put a note on the results page explaining that the configuration is the GPUs used, not the video cards.

Another thing to mention is vRAM issue - not all will get this clear that TitanZ's 12GB of vRAM will be "split" as software will see 2x6GB instead of full 12GB (as in the case of TitanX =)

for majority this page might become the best purchase advice & those facts should be mention otherwise forum will be flooded with messages, "I haven't seen that.." "..why noOne mentioned.." - I mean it's their problem, but..I'm pretty sure that will happen, so it would be nice to have that info, next to one mentioned prior.

glimpse wrote:Another thing to mention is vRAM issue - not all will get this clear that TitanZ's 12GB of vRAM will be "split" as software will see 2x6GB instead of full 12GB (as in the case of TitanX =)

for majority this page might become the best purchase advice & those facts should be mention otherwise forum will be flooded with messages, "I haven't seen that.." "..why noOne mentioned.." - I mean it's their problem, but..I'm pretty sure that will happen, so it would be nice to have that info, next to one mentioned prior.

I think we will add an explanation how to interpret the results.

Do you guy know of any web page that covers the various NVIDIA GPUs on a level that is relevant for Octane (i.e. architecture, GPU count, core count, VRAM, etc.)? That would save us a lot of time.

abstrax wrote:Do you guy know of any web page that covers the various NVIDIA GPUs on a level that is relevant for Octane (i.e. architecture, GPU count, core count, VRAM, etc.)? That would save us a lot of time.

I'm working on such site, Marcus - tomGlimps.com - will be more or less on Octane Render software features (like batch rendering), GPU related hardware (like 1, 2, 3), full builds explaining holistic approaches (like 1 & 2) , free scenes (more to come), material breakdowns, etc.

(it's still a mess but I'll sort things out..-need to find more time put more content)
If You have any specific topics to be covered, I'd be more than happy to help =)

For basic specs you can look at nvidia website, but cards perform differently depending on brand and cooling solutions.

http://www.geforce.com/hardware/desktop ... ifications
http://www.geforce.com/hardware/desktop ... ifications
http://www.geforce.com/hardware/desktop ... ifications

linvanchene wrote:

stratified wrote:Have a look here: http://render.otoy.com/octanebench/results.php

cheers,
Thomas

Thanks for posting the direct link!

While browsing around I noticed two web page related stumble stones:

- If you toggle: "Single GPU results only" it displays only the results of the current page.

This means if you are on page one you will not see any single GPU at all and have to manually browse to page 2. But then you will still not see the single GPU of pages 3, 4, etc...

- The search field only displays the results of the current selected page as well.

Example: If you enter GTX Titan on page 2 you will not see any results on page 1, 3, 4 etc.

- - -

In any case thank you for providing a simple to use benchmark.

+1

A single page with all results would be more readable.
Sorting with respect to min speed for a single card does not work.
2xTitan Z is not counted as a single card. Same for 690, 590.

Yea, there's some hiccups between the database and the front-end. The results seem to be hard-coded to a page number, which is no good. So checking the box for "single GPU results only" on the front page yields no results. And when going to the next page, the "single GPU results" setting is reset to off. And I challenge someone to filter by average score and single GPU only, and yield a result: sorting by average score resets to the first page, which never has single GPUs results.

Your webdev has some work to do. But on the plus side, I really like the charts plotting out where multiples scores of a single setup fall.

On the Titan Z (and all multi-GPU cards) thing, I think the normal assumption for end-users (who this is supposedly for) is that (1) Titan Z means (2) GPUs. So when you read "2x Titan Z" one would logically conclude you have (2) Titan Z cards installed, equaling (4) GPUs. If someone is benchmarking just one of the GPUs, that should be the exception, not the rule. Label it "0.5x Titan Z" or something else, but by representing one physical Titan Z card as "2x Titan Z" is going to cause (and already has) confusion for many viewing the database now and as time goes forward.

riggles wrote:Yea, there's some hiccups between the database and the front-end. The results seem to be hard-coded to a page number, which is no good. So checking the box for "single GPU results only" on the front page yields no results. And when going to the next page, the "single GPU results" setting is reset to off. And I challenge someone to filter by average score and single GPU only, and yield a result: sorting by average score resets to the first page, which never has single GPUs results.

Your webdev has some work to do. But on the plus side, I really like the charts plotting out where multiples scores of a single setup fall.

On the Titan Z (and all multi-GPU cards) thing, I think the normal assumption for end-users (who this is supposedly for) is that (1) Titan Z means (2) GPUs. So when you read "2x Titan Z" one would logically conclude you have (2) Titan Z cards installed, equaling (4) GPUs. If someone is benchmarking just one of the GPUs, that should be the exception, not the rule. Label it "0.5x Titan Z" or something else, but by representing one physical Titan Z card as "2x Titan Z" is going to cause (and already has) confusion for many viewing the database now and as time goes forward.

Exactly.

GTX 780 Ti Classified@OC version 1020MHz core-stock running Win 8.1 x64

My result - screenshot

GTX 980 is faster than my videocard at PMC kernel!?????? But wrong because attention: GTX 980 has non-reference speeds...

riggles wrote:Yea, there's some hiccups between the database and the front-end. The results seem to be hard-coded to a page number, which is no good. So checking the box for "single GPU results only" on the front page yields no results. And when going to the next page, the "single GPU results" setting is reset to off. And I challenge someone to filter by average score and single GPU only, and yield a result: sorting by average score resets to the first page, which never has single GPUs results.

Your webdev has some work to do. But on the plus side, I really like the charts plotting out where multiples scores of a single setup fall.

On the Titan Z (and all multi-GPU cards) thing, I think the normal assumption for end-users (who this is supposedly for) is that (1) Titan Z means (2) GPUs. So when you read "2x Titan Z" one would logically conclude you have (2) Titan Z cards installed, equaling (4) GPUs. If someone is benchmarking just one of the GPUs, that should be the exception, not the rule. Label it "0.5x Titan Z" or something else, but by representing one physical Titan Z card as "2x Titan Z" is going to cause (and already has) confusion for many viewing the database now and as time goes forward.

Tim fixed a bunch of issues with the results page. Could you have a look and tell us if that's better now?