Re-tested on an edgy model.

V2 = 220 ms
V3 = 120 ms

Much better. my bad.

Hi Burr, yup that's more like it... Also an edgy model with Hidden line display enabled will probably see the biggest difference.

With this edgy model, do you see any difference if setting DisplayThreadLimit=6 or DisplayThreadLimit=8 on your dual-quad-core ? I'm still not so sure that one thread per "virtual" core is the best way to go on a hyperthreaded machine or not.

- Michael

Hi Mike,

> I have a Pentium 4 (2 core), 1 g. ram
> NVIDIA GeForce4 Ti 4800 SE

Not really too surprising there isn't much gain there, because the GeForce4 Ti 4800 SE is basically an antique at this point - it comes from the 4th generation nVidia wave, they are on the 15th generation right now...

Especially if it's in an AGP 1x or 2x slot probably the card is the bottleneck. Multi-core CPU use can only help things out if the video card and bus is speedy enough that the card can handle more stuff than what one CPU can chug out. If the card is not able to keep much ahead of just 1 CPU then having more than 1 CPU also going won't really do much.

However recent trends have been that video cards have become faster at a greater pace than CPUs - for various reasons CPUs are basically stuck at around their current speed and the focus is more on having several cores available, so this new system in MoI is pretty good for the direction that things are going.

Your GeForce4 card though is nearly 10 years old at this point, as far as GPUs go that's like an eternity, it's basically a nice feature that MoI is able to run at all on these kinds of old generation cards (MoI actually does run on the 1st generation GeForce and Radeons!), but you're probably not going to get much benefit for things that are tuned more for current generation stuff.

I would have expected a better result on your work machine though definitely... From what you were describing it kind of sounds like the multi-core use of the driver is still on despite changing that setting, I don't know if possibly that is a driver problem that has been fixed in a newer driver version, but that could be possible.

- Michael

"""""""this edgy model, do you see any difference if setting DisplayThreadLimit=6 or DisplayThreadLimit=8 on your dual-quad-core ? I'm still not so sure that one thread per "virtual" core is the best way to go on a hyperthreaded machine or not.""""""""""

Hi Michael,

Video card defaults
no entry in MoI ini thread limit = 130
thread limit 6 = 140
thread limit 4 =150
thread limit 1 =180

Video card threaded optimazation off in global, no entry in ini file = 150

Seems about right.

Hi Burr, thanks for those tests!

So it does not seem that one separate thread per virtual core (which is what you should be getting with no entry set, or with it set to 0) is hurting anything there, that is good to know.

- Michael

What I need is $2k after taxes. ;-)

The card was a quick replacement after lightning took all kinds of stuff out in my home PC, including the better card.
The motherboard is the only thing left that wasn't (too) effected, and it has evident charring around the power strip connector.

I'm still going to poke around some next week to see what the deal is with my work PC...
My brother put it together a year ago, and he tweaks gaming systems as a hobby.
I believe for the sake of keeping my boss from having anymore headaches, he clocked everything down to "reliable" and "stable" values.
And I remember a lot of stuff done in the fancy BIOS setup.

Ya gotta admit, you can still do a lot with MoI on an old system... the watch, the tank. ;-) Now that's a product.

260GT - i3930K

April

Frames: 76 - Time: 10000ms - Avg: 7.600 - Min: 7 - Max: 8

July

Frames: 116 - Time: 10000ms - Avg: 11.600 - Min: 8 - Max: 13 No DisplayThreadLimit - Nvidia Thread opt: Auto Hyper Thread ON

Frames: 119 - Time: 10000ms - Avg: 11.900 - Min: 11 - Max: 13 No DisplayThreadLimit - Nvidia Thread opt: Off Hyper Thread ON

Frames: 97 - Time: 10000ms - Avg: 9.700 - Min: 7 - Max: 11 No DisplayThreadLimit - Nvidia Thread opt: On Hyper Thread ON



Frames: 104 - Time: 10000ms - Avg: 10.400 - Min: 9 - Max: 12 DisplayThreadLimit=2 - Nvidia Thread opt: Auto Hyper Thread ON

Frames: 105 - Time: 10000ms - Avg: 10.500 - Min: 9 - Max: 12 DisplayThreadLimit=2 - Nvidia Thread opt: off Hyper Thread ON

Frames: 104 - Time: 10000ms - Avg: 10.400 - Min: 9 - Max: 11 DisplayThreadLimit=2 - Nvidia Thread opt: on Hyper Thread ON



Frames: 120 - Time: 10000ms - Avg: 12.000 - Min: 11 - Max: 13 DisplayThreadLimit=4 - Nvidia Thread opt: Auto Hyper Thread ON

Frames: 120 - Time: 10000ms - Avg: 12.000 - Min: 11 - Max: 13 DisplayThreadLimit=4 - Nvidia Thread opt: off Hyper Thread ON

Frames: 119 - Time: 10000ms - Avg: 11.900 - Min: 10 - Max: 13 DisplayThreadLimit=4 - Nvidia Thread opt: on Hyper Thread ON



Frames: 120 - Time: 10000ms - Avg: 12.000 - Min: 11 - Max: 13 DisplayThreadLimit=6 - Nvidia Thread opt: Auto Hyper Thread ON

Frames: 120 - Time: 10000ms - Avg: 12.000 - Min: 11 - Max: 13 DisplayThreadLimit=6 - Nvidia Thread opt: off Hyper Thread ON

Frames: 120 - Time: 10000ms - Avg: 12.000 - Min: 11 - Max: 13 DisplayThreadLimit=6 - Nvidia Thread opt: on Hyper Thread ON


Will do an other batch with Hyper Treading OFF ... (does the result make any sence for you Michael Oo)

So my conclusion, the nvidia thread optimisation 'auto' is in fact turned off for MoI.
Forcing in 'on' slow down the result.

So for the next test I keep nvidia thread optimisation on 'auto', but I disable the intel hypertreading on my mobo.



Frames: 120 - Time: 10000ms - Avg: 12.000 - Min: 11 - Max: 13 No DisplayThreadLimit - Nvidia Thread opt: Auto Hyper Thread OFF

Frames: 107 - Time: 10000ms - Avg: 10.700 - Min: 9 - Max: 12 DisplayThreadLimit=2 - Nvidia Thread opt: Auto Hyper Thread OFF

Frames: 120 - Time: 10000ms - Avg: 12.000 - Min: 11 - Max: 13 DisplayThreadLimit=4 - Nvidia Thread opt: Auto Hyper Thread OFF

Frames: 120 - Time: 10000ms - Avg: 12.000 - Min: 11 - Max: 13 DisplayThreadLimit=6 - Nvidia Thread opt: Auto Hyper Thread OFF

Hi PaQ, thanks for collecting that data!


> (does the result make any sence for you Michael Oo)

Sort of... That data does seem to suggest that maybe it's better to use only half as many threads on a hyper-threading machine. Seems to be the opposite conclusion from some of Burr's previous data though.

I guess it's fairly likely that the optimal settings for different things (there are other internal variable factors too like number of buffers used per thread and vertex/index buffer size) is different on different machines depending on all kinds of factors like bus speed and video card speed, etc... and also probably even different on the same machine with different model structures as well.

- Michael

Hey PaQ,
Do you have another display script with the more detailed output from something previous?

"frames" + time

Thanks.

Hi PaQ,

> So for the next test I keep nvidia thread optimisation on 'auto', but I
> disable the intel hypertreading on my mobo.

So it looks there that hyperthreading being turned on in the bios by itself doesn't harm anything but that there's not much to gain there by one thread per "virtual core", and that the increased contention ends up as a slight net negative.

This probably has to do with the whole process being memory bandwidth intensive instead of only calculation intensive which would suit hyperthreading somewhat better I think.

I think it will probably be best to go with a default of one thread per "real core" instead of one per "virtual core" for this particular thing.

- Michael

Hi Michael,

You're welcome, glad if it helps.


Hi Burr,

Not sure if I understand what you're asking.

I'm using fraps to benchmark the viewport (http://www.fraps.com/ ), with the autorotate2 script here: http://moi3d.com/forum/index.php?webtag=MOI&msg=1001.1 .

Using time display script in MoI....

1000 Spheres in Array (cube arrangement), mesh=5 deg.

2000 ms in prior version.
1069 ms in new version.

2x increase.

So it's good for larger shear quantity of surfaces. - when its more about large amounts of poly data being processed.

Not as much noticeable with simpler models. - when it's more an issue of the bottleneck with the PC communicating with the vid card.
In one case, when I was using a smaller number of objects. I noticed that the newer version was slower than the older.

Hi Mike,

> In one case, when I was using a smaller number of objects. I noticed
> that the newer version was slower than the older.

By very much or were they pretty close? But also with a small number of objects things are probably going at a fast enough pace that you probably would not be able to tell the difference if the numeric readout was not there, or was that not the case?

- Michael

Hey PaQ,

"""""""""Not sure if I understand what you're asking. I'm using fraps to benchmark the viewport """"""""""""""""""""

Ok, my bad... I thought you were doing it with the script Michael posted. we only got "ms" with the script. I didnt know how you were getting "frames per second."

Thanks.

Very cool... Thanks Michael... I'll announce this new beta an my forum as well.

Peace,
Rich_Art. ;-)

Great job Michael G. on the tapered extrude! Good for folks like me that do injection molded plastic part designs!

Michael T.

Is this new cpu optimization working on the mac version??

Hi ed,

> Is this new cpu optimization working on the mac version??

Yup, the same thing is working on the Mac too. I only have dual-core macs over here, and I saw about a 30% improvement, a quad-core should probably see better than that.

- Michael

Michael, I'm seeing a possible bug with the new version:

==========================
FreeForm curves (control point, Through Points, and Sketch) are not allowing you to highlight them or select them in one ortho view if you created them in another one.
==========================
Regardless of how they were drawn in one view, it seems to happen when they appear "planar" in another view.
It appears to be intermittent and is random in occurrence. (So if you try, try a few different times)
Scrolling in and out seems to bring this problem on and off.

I've drawn some FreeForm curves while in "Front" view, I go to "Top" view and Moi is not highlighting them for selection.
Nor am I able to move them.
This is also happening in the same view I draw it in. But it will sometimes work again, then not work if I zoom.

It doesn't seem to be a problem with non-planar curve objects.
Also strange, I don't seem to be able to snap things to it's ends or other parts... almost like it is invisible when seen in planar.
Even if I select it with a window selection, I still can't move them.
Oh, and some (objects planar in view) are highlightable/selectable and some will not.
And those invisible objects do not have the preview-to-snap "x"'s within the tolerance circle, not show up either.

The same effect is happening with shapes, lines and anything else planar, like a Planar box. I've rebooted just to clear up that possibility.
I also set the thread limit to 1, just to rule that out.

Just for comparison I've opened up the older version and pasted the same things into it...
Of course, No problem in the older one. The older version seems to have less lag when moving things, and everything highlights with a mouseover just fine.

Anyone else have this issue? Draw a FreeForm curve with a few points in one flat view, (in a straight fashion (no curving)) and go to another flat view and see if it is not selecting.

 
I just noted that this doesn't seem to be a problem between views in Split View, it's just the full version single views of each viewport.

===================================
Also, there seems to be a little lag time when I try to move control points over the previous version.
Also a lag when I am using box selection. The box is somewhat behind in catching up to the cursor.