I use a Gigabyte GAX79 UD3. I read in the manual that the first and the (third?) slot are fully SLI-supported. I also use an Nvidiatool to enable (force) PCI-E 3.0 after installing a new driver. PCIe operates with 100Mhz, only the multiplicator for the CPU is used for overclocking and for RAM I use Corsair-RAM with an XMP-profile. Okay ,the hardware is not the latest but like you said it should not built a bottleneck. In 2.5 SLI works great for me, no comparsion to 3.0.

 

cheers, Kai

[. In 2.5 SLI works great for me, no comparsion to 3.0.

 

cheers, Kai

 

 

 

 

So you are on 3.0? Might be confusing then since i am on 3.2 WHQL 364.51

 

Michael Moe

 

 

EDIT: some motherboards also don't support SLI very well, only one x16 slot and the other slot is really an x4.

 

Hmmm, how do I know this already, Rob?  :mad:  :rolleyes: 

 

Jeff

 

 

Hmmm, how do I know this already, Rob?

 

Haha, I wasn't actually thinking of your situation Jeff, but it's fairly common issues for SLI/Crossfire users to encounter a motherboard limitation.

 

Looks like nVidia released new drivers today also 364.72 ... haven't tested them yet -- it seems nVidia have been releasing drivers at a RAPID pace recently.

 

 

 

In 2.5 SLI works great for me, no comparsion to 3.0.

 

Forcing PCIe 3.0 or leaving it at 2.0 will not make much of a difference, however the slots will if you mix a 16x with an 4x ... just tossing out possibilities - sounds like that's not the issue in your case Kai.

 

SLI in v2.5 has artifacts (boat wakes, smoke fx, and a few others) ... in V3.2 I'm finding pretty close to the same performance benefit as in V2.5 less the artifacts ... but also need to keep in mind V3.2 support Dynamic Reflections (feature not in V2.5) and wakes now blend with water, shoreline waves, etc. etc.  I HIGHLY recommend you move to a fresh V3.2 install and before you do any adjustments or 3rd party installs, but a based SLI test with and without.

 

Cheers, Rob.

I seriously thought we would never see those shoreline waves again. Pass on to LM thanks for bringing them back  

 

The water in 3.2 is amazing with a floatplane. I think the water(waves, effects etc.) is one case where SLI will really shine.

Thanks Steve, that reminds me I need to fire up the Twotter for a test spin.

 

Cheers, Rob.

Good to see the waves back but they are far too fast and need a greater LOD.

 

gb.

Looks like nVidia released new drivers today also 364.72 ... haven't tested them yet -- it seems nVidia have been releasing drivers at a RAPID pace recently.

 

Hi Rob,

 

Superb post Rob - well up to your usual high standards!

 

Bit the bullet last night. Updated my UEFI BIOS to 2001 and thought, what the heck, went ahead and also updated from 361.91 to 364.72 via GeForce Experience. Been using GFE since day one of this completed build and have had no issues whatsoever. Mind you, I always opt for the custom route and perform a clean install. The only downside is having to spend a little time re-establishing all my profiles in NVIDIA Control Panel. NVIDIA Inspector is easier as I only use it for SGSSAA in P3D.

 

The 364.72 driver set seem to be performing very well so far. In passing I note that the DSR option has now appeared under the Global tab for those of us using SLI. When I 'optimised' Skyrim in GFE it was applied automatically. However, I didn't like the mouse input lag so quickly reverted back. Also didn't feel the visuals were significantly better than at 2560x1440 where they are superb

 

Had a great flight last night using quite high settings and the experience was pretty much flawless.

 

Incidentally, after exhaustive testing, and with Steve's very patient and helpful input, I have discovered that an Affinity Mask of 21760 (01,01,01,01,00,00,00,00) applied to the 5960X with Hyperthreading ON is providing the best performance in Prepar3D v3.2.

 

I am using batch files* to keep any Addons like 'Air Traffic Manager', 'SimLauncherX' and 'FS Global Real Weather' off those 4 cores as per Steve's detailed instructions at codelegend.com. 'Ideal Flight Pro 10' is a special case and can be configured internally to assign its own affinity mask. It too is an excellent front-end from which P3D can be configured and launched. I used it all the time while testing and collecting data.

 

We have the luxury of being able to use any combination of the first 4 Cores. Currently I'm assigning affinity masks of 'D' (11,01) or 'E' (11,10) which works well and leaves 2 Cores/4 LPs free for anything else, including Windows. The [start /B "" /affinity (value in hex) "drive:\path to the addon executable"] is definitely the way to go as it eliminates the need to eliminate spaces (where exist) in the name of an executable.

 

*Steve has his reservations about using 'Process Lasso', but I must say my observations confirm that it assigns affinity masks accurately. So, it's quite possible I will move over to using it for any Addons running concurrently with the sim. Having said that, I do NOT use it to assign an affinity mask for Prepar3D as it's best to do that via the [JOBSCHEDULER] section of Prepar3D.cfg. However, I do use Process Lasso to ensure P3D is running at max performance all the time and is excluded from any 'Restraint' activity. Process Lasso seems to do a good job of Restraining other background process from interfering with P3D activity. I like it!

 

Cheers,

Mike

 

 

 I HIGHLY recommend you move to a fresh V3.2 install and before you do any adjustments or 3rd party installs, but a based SLI test with and without.

 

I think I will do that. Maybe a fresh complete 3.2 will help more than testing with the updated 3.0.

 

cheers, Kai

 

 

Incidentally, after exhaustive testing, and with Steve's very patient and helpful input, I have discovered that an Affinity Mask of 21760 (01,01,01,01,00,00,00,00) applied to the 5960X with Hyperthreading ON is providing the best performance in Prepar3D v3.2.

 

Interesting Mike, I'm always up for test, will try that out with one of my scenarios in my table and see how it compares.  Thanks for sharing your results.

 

I don't use "process lasso" ... I have tried it in the past but it does nothing more than what I can do manually via Process Explorer ... what really raised my eye was it's memory management claim (reducing VAS usage) -- 30+ years of software engineering from the birth of the PC and I can assure you their claims to memory management are completely FALSE.  Because of that, it does leave me doubting the validity of the developer.

 

Cheers, Rob.

what really raised my eye was it's memory management claim (reducing VAS usage) -- 30+ years of software engineering from the birth of the PC and I can assure you their claims to memory management are completely FALSE.  Because of that, it does leave me doubting the validity of the developer.

Hi Rob,

 

Can't really comment about this aspect as it's never been an issue for me so I just leave Windows 7 to do its thing. However, I've just had a quick look at the relevant section concerning Virtual Memory and it does seem to me that the developer is being upfront about it:

 

https://bitsum.com/virtual_memory_facts_and_memory_optimizer_scams.php

 

"SmartTrim is disabled by default because we still do not believe RAM optimization is necssary for most users. However, we accept the fact that some advanced users can use it to good effect."

 

"Virtual Memory management at a low level is *much* more complex than I've laid out here. I mean, you would be shocked at just how sophisticated the management of it is. That is why it is best left alone, in most all cases. Besides, these days most applications themselves will page their own inactive memory out before the virtual memory manager even has a chance to do it, so that their perceived memory footprint is lessened. Yes, it is a trick many developers use to hide their real virtual memory utilisation."

 

I don't think he is trying to pull the wool over our eyes nor does he offer this feature as a wonder cure, but the option is there for those who wish to experiment.

 

Cheers.

Mike

So Sli here 2x gtx970 4790k 4.7ghz HT off p3d v3.2.

I've been running process lasso for a while now , but don't think I have it right what cores should I dedicate for trackir,vpilot, & other utility's I use for fs. & then also what cores for p3d .

Cheers

Riche

Hi Riche,

 

I'm afraid I have no experience with 4-Core CPUs. However, if I followed the same logic as used with my 8-Core then I would be trying an Affinity Mask value of 240 (11,11,00,00). Don't use Process Lasso for this; instead, enter this entry under the [JOBSCHEDULER] heading added to Prepar3D.cfg. I am assuming you know how to do that. This will give P3D exclusive access to 4 logical processors and leave cores 1 and 2 (4 LPs) free for your Addons, just as I have achieved.

 

Remember to turn Hyperthreading back ON.

 

You could then assign Affinity Masks to your individual Addons using either Process Lasso or batch files (see the penultimate paragraph in post #38 of this thread).

 

Note: values "D" and "E" are hexadecimal values for decimal 13 and 14 respectively.

 

For instructions about using batch (.bat) files go here: http://www.codelegend.com/forums/viewtopic.php?f=2&t=632

 

I prefer using the 'start' command as it is just a single line and you don't need to worry about having to eliminate spaces in folder names or an Addon's executable - very useful when an Addon is updated and those spaces (if existed) return.

 

From Steve Waite: Always try to give 2 or more LPs to each app. Remember the main thread of almost all apps is the first one, the LP it first comes across unmasked (a one in mask).

 

Let us know how you get on.

 

Regards,

Mike

My Fs runs smoother with 32 bit clouds than DXT5 clouds...

Don't know why but it does...

 

Me too.. In fact, after select 32bit clouds in REX soft cloud´s setting, the sttuters went way... Thank you!

 

 

My Fs runs smoother with 32 bit clouds than DXT5 clouds...
Don't know why but it does...

 

Are you running the REX waves at 32 bit as well? I think the other option is DTX1.