For general purpose computing (what I'll be doing also) you can use all 12GB ... for SLi 3D work then yes 6GB each.

 

Cheers, Rob.

 

I have no stake in what you decide, as I personally won't make any moves until 20/16nm (20nm FinFET), but out of pure curiosity, would you be able to do the same with 2 Blacks (use all 12GB for general work) ? 

 

Whatever you decide I can't wait for the vids/results  :lol:

 

 

ould you be able to do the same with 2 Blacks (use all 12GB for general work) ? 

 

Yes.

 

 

 

Whatever you decide I can't wait for the vids/results

 

Hopefully I'll not be needing Fraps any more and will record on a separate PC using 4K capture card ... that's this years plan anyway.

 

Cheers, Rob.

NVIDIA market it as a 12GB GPU. If it's 6GB per GPU though, you might be right that only 6GB is usable. I'll look into the matter more, wouldn't be surprised if NVIDIA call it a 12GB card just to trick buyers in. If that's so though, even less of a reason to get a GTX Titan Z over two GTX Titan Blacks.

 

Rob, if a GTX 790 gets released, it will be aimed at the consumer market, which probably means crippled double precision performance (like the GTX 780 and the GTX 780 Ti). I don't think you'd want that.

NVIDIA market it as a 12GB GPU. If it's 6GB per GPU though, you might be right that only 6GB is usable. I'll look into the matter more, wouldn't be surprised if NVIDIA call it a 12GB card just to trick buyers in. If that's so though, even less of a reason to get a GTX Titan Z over two GTX Titan Blacks.

 

NVIDIA, AMD, retailers, etc. are all guilty of it unfortunately. 

 

The 295X2 is marketed as an 8GB card, but it's the same deal, with only 4GB available to each GPU when running crossfire.

 

Strictly speaking, it is an accurate spec, but yeah, it's a bit misleading for anyone not familiar with multi-GPU tech.

Hopefully I'll not be needing Fraps any more and will record on a separate PC using 4K capture card ... that's this years plan anyway.

 

Cheers, Rob.

 

I think I'm most interested in just how high you can get with VRAM utilization.

 

My experience with other big ticket DX11 titles is that they will at times reserve quite a bit more VRAM than what would actually be required for smooth operation, which is fine, but makes it a bit difficult to make a clear determination of what is actually required. Going to 6GB x2, it will be interesting to see just how far it will go.

NVIDIA, AMD, retailers, etc. are all guilty of it unfortunately. 

 

The 295X2 is marketed as an 8GB card, but it's the same deal, with only 4GB available to each GPU when running crossfire.

 

Strictly speaking, it is an accurate spec, but yeah, it's a bit misleading for anyone not familiar with multi-GPU tech.

 

Yes, multi-GPU tech is very confusing and complicated (when it comes to driver support, scaling and microstuttering). Hopefully with NVIDIA Pascal, single GPUs will be able to handle 4K.

Yes, multi-GPU tech is very confusing and complicated (when it comes to driver support, scaling and microstuttering). 

 

It can be at times, no doubt about it, but it has come a long way, even just over the past two years or so, and is a much more credible (and easy to manage) solution than it once was.

 

It isn't without caveats and exceptions even today, but with even minimal application support, and, at times, the use of tools like NI or RadeonPro, the results can be pretty darn good. 

 

I guess all I mean is, and I think I said this in a post to J van E not long ago, that for the most part, the difficult issues have been mostly solved, in both camps. I really can't wait to see P3Dv2 in multi-GPU action with proper support. My cards are aging fast, but should still be able to pull off some impressive performance, especially at my tiny resolution.

 

That said, if the recently leaked 390X specs are of any merit, it won't matter, as that will be in my machine regardless.

 

Spinning wheel, etc...

So essentially, a doubled GPU yet on one card is equivalent, limitations-wise, to two cards in SLI as far as FSX/P3D are concerned.??

So essentially, a doubled GPU yet on one card is equivalent, limitations-wise, to two cards in SLI as far as FSX/P3D are concerned.??

 

For the most part, yes, as far as any app is concerned that can use them.

 

There was a time when it was much more difficult to fit two GPUs to one PCB/card without having to make significant compromises to clocks, etc., to keep within power/temp limits. At that time, separate cards would of course then have an advantage.

 

Node efficiency improvements, more advanced GPU monitoring, dynamic clocks (Boost, etc), and even things like BIOS switching via the Uber Mode switch (I think the 6990 had an AWSUM switch...lol) have mostly worked around that problem, and now, the only really unattractive features of dual-GPU cards are price, and in some cases (literally and figuratively), heat. Titan Z will employ (it seems) a revised version of their excellent vapor chamber cooler from the 590/690, and as you may know, AMD has gone with an AIO type water cooler for the 295X2.

 

It may just be bias on my part, maybe I'm remembering the days of the sandwich-PCB GTX 295 and the 4870 X2 (shudders), but I just don't really see the benefit in dual-GPU cards, except in situations like I mentioned before, where space is an issue. They are definitely better now than they were, but the price, plus decreased flexibility (for me anyway), are kind of major points against the idea.

 

(I think maybe Rob is leaving the door open for quad-SLI with two Titan Zs...*drools*)

 

 

Rob, if a GTX 790 gets released, it will be aimed at the consumer market, which probably means crippled double precision performance (like the GTX 780 and the GTX 780 Ti). I don't think you'd want that.

 

Agree, I may not want that ... was just tossing it out there for others to consider.  I've heard conflicting reports of how much VRAM will be on the GTX 790 ... 10GB (5X5) or 16GB (8X8) ... or maybe two version?

 

Yeah, door is open for two Titan Z's ... moving to 4K monitors may require two Z's regardless of what nVidia suggest (they seem to focus on 3D shooters and we all know that P3DV2.x is far more performance demanding than 3D shooters).  But start with one and see how it goes.

 

I've gone back and forth over single/dual GPUs over the years, early multi-GPU solutions were difficult to keep running reliably.  However, multi-GPU support has improved over the years -- it's far more stable that it used to be.

 

Cheers, Rob.

Rob, being retired I can't justify 2 video cards worth more than my 3770K based PC.  I have a 680 and so far have been amazed at what it can do

with P3D.  Could you list the top 3 or 4 settings I should back off?  Would cloud shadows be #1 on that list? Or is there some other setting harder

on single GPU frame rates?

 

Thanks

Ken

I'm surprised that anyone's surprised about the shadows eating performance at dawn/dusk.

 

Shadows are very demanding so it's surely obvious that at dawn/dusk shadows give a larger performance hit as the shadow drawn is much longer/larger due to the angle of the Sun compared to day shadows where the shadow is beneath the cloud.

 

Id like to know you guys with your 780 titan cards how much this affects tour performance ie is it stuttering what are tour actual framerates and what is thé load on thé gpu? Is it still visually ok with shadows at dawn/dusk?

 

I haven't really noticed a FPS hit at dusk/dawn when the sun is in view.  It's when it goes behind an actual cloud (relative to your point of view) that the FPS virtually halve, in my test 44 to 23 and then back to 44 again as soon as the sun clears the cloud.  It is cloud shadow related, but I think it is an issue that has crept in since the beta.

 

The above is on a titan black and a 4770K at 4.3 ghz.

 

Regards, Daz

Thanks daz

 

I've not noticed that myself.

 

I guess I should have read properly a few people mentioned this properly

 

I'll test this out myself on my lowly 570

 

So to clarify this can be at any time of day? Even midday?

 

 

Could you list the top 3 or 4 settings I should back off?

 

How about my top 5.

 

Shadow Map Count

Terrain Shadow Cast Distance = 0m  (aka OFF)

Cloud Draw Distance 80mi

Cloud Shadow Cast Distance

Water Detail

 

Do not use SGSS AA and stay with standard 4X MSAA.

 

Cheers, Rob.

Thanks daz

I've not noticed that myself.

I guess I should have read properly a few people mentioned this properly

I'll test this out myself on my lowly 570

So to clarify this can be at any time of day? Even midday?

I have only noticed it at dusk/dawn. Have not experienced any obvious issues during what P3D considers to be the daytime.

 

Daz

I got be honest guys...don't hate me.

 

So I ran FSX and P3Dv2 through the same flights over the weekend and, I hate to say it, but, my FSX is still out performing at this point. Running comparable settings except for Interior/Exterior/Simobjects shadowing enabled, terrain,cloud shadowing disabled, volumetric clouds disabled in P3D, and no traffic.

 

Don't get me wrong, P3D is performing better than it ever has, and is definately enjoyable for the most part, but, cloud shadowing and terrain shadowing definately needs to be disabled to get even relatively close performance on my system. Frame rate drops in  urban areas, in particular, are killing me in P3D, yet the same area, FSX doesn't even flinch an inch.

 

I know all will have different experiences, yet my system is no slouch and has been tuned to run FSX pretty darned well.  And that is with AA and SGSS settings in NI and cfg tweaks we all know and love, of course.

 

P3D AF settings set to all logical cores, improved considerably in v2.2, with no other cfg tweaks configured or necessary, as far as we know. NI not used and internal AA set to 4x. Shimmerfest in cockpit.  

 

I am sort of back to FSX at this point, despite the better exterior shadowing and excellent interior shadowing in P3D, which imo are the only advantages as of now. P3D is certainly more of an experimentation for me and I do like where it's going, just needs more time and more work I think.