Clock speeds are certainly peaking and the majority doesnt have demand for anything much more. How fast do you want to open MS office? It also doesnt have much effect on surfing the internet. The ones who gain from faster GHZ are flight simmers (especially old code ones ) and video processing and CAD applications. 

Not even sure if mainstream gaming has hit a peak, I dont play much other games often, and the ones I see dont seem to be struggling much with FPS 60+. 

 

So since the market is drastically reduced, there doesnt seem much pushing Intel and AMD for that matter. We may have to buy super computers from IBM to run our simulators. lol

I looked on CAEs site but cant see what kind of hardware is running the highend simulators, probably cause its highly customizable. The simulator room I was in in 1996 for a DC10 had a huge room of mainframes. (I was trying to get into the simulator maintenance program for CDN) . 

 

In either case, flight simulation be it corporate or desktop simulation, pushes a computer like no other with so much going on and in our case, its not even having to send hyrdaulic and pneumatic info among other things. 

So it seems we will be in this lull for sometime overclocking like mad to squeeze every juice we can, unless some breakthrough software comes along to massively reduce the load of calculations. 

 

 

So since the market is drastically reduced, there doesnt seem much pushing Intel and AMD for that matter.

 

True, or I hate to say it, a console!!  PS4 and Xbox One Console designs really are pushing hardware towards complete optimized systems aimed at gaming/simulation ... maybe this will be our final destiny (I hope not).

 

I keep wondering what "could" be done if we kept the power up and optimized the CPUs for processing performance rather than less power.  Use the smaller size and less heat to increase CPU clock cycles and use a much larger CPU cache ... use 22nm or smaller, but keep the die size large optimized for heat dissipation and higher clock frequencies.  I just have a hard time believing Intel/AMD can't come up with 4 core or 6 core CPU that runs at 6 or 8 Ghz ... make it big, make it power hungry, make it expensive, but just make it!

To a certain extent making chips low power also helps to create affordable processors with good performance. When you increase the current in the chip, heat will become a huge problem and with ever smaller die sizes/transistor count it causes problems in the data circuits and transistors unless you have some special cooling solutions. And smaller manufacturing process is in itself a requirement to keep die size low and still retain high clocks. Die size is huge factor in manufacturing costs and keeping it in affordable level while performance, features and instruction sets increase. Smaller manufacturing process also allows the developement and application of new technologies such as quite recent 3D transistors. Actually Intel made a mistake with their NetBurst architecture, where they practically left power consumption and heat out of the primary design. At first, Intel thought that they could during the normal developement eventually rise the speed of the NetBurst all the way to the 10GHz range and the whole consept of high clock speeds was actually essential to make up with the performance loss of the deep pipeline. But in the end they bumped in to a wall at under 3.8GHz and last Prescotts/Cedar Mills were extremely hot without practically any headroom left for higher clocks. Also chip was much more difficult to implement for smaller manufacturing processes, than Intel at first thought. After that they went back and designed Core 2 which was actually based on older P3/Pentium-M architecture.

 

It is true that there is much thought put to low power applications with modern Sandy Bridge, Ivy Bridge and Haswell processors, though they are marketed for dektops and as Xeon line for servers with negligible or larger differences depending on the Xeon model. Still, especially Intel's 6 core high performance desktop processors are wonderful chips and are still powerful compared to many older designs, while there is much more emphasis in the low power features in the micro architecture than before, one example that low power is not conflicting with performance needs. Market requirements for consumers are still quite small with these kind of processors when you look at the volumes today: laptops rule over the desktop market and with smaller ultrabook designs that become more and more popular, heat is a problem and integreated graphic chips and low power technologies also help to keep the system simple to design, smaller and cheaper. Also all technologies developed to minimize power consumption are essential for server market too, which also in many cases has very different requirements for the processors compared to powerful desktop processor, like single thread performance doesn't have to be that high, but you need to process huge amount of threads simultaneously (that is up to dozens of processors in the server) and you need extremely high performance memory and disk operations with the system.

 

Things have changed and will change. In the future many tasks requiring high computational performance may be processed in the cloud making powerful desktop processors more or less useless as a form of PC we know today (mainframes, anyone? ). There are of course many things hindering this, mainly in the network infrastructure: you'd practically need fast low latency fiber Ethernet connections to be able to run realtime cloud based applications.

 

About next generation consoles they are an interesting example of computer platform design, although to me, it seems that larger steps were made with the last generation. There is really nothing special design charasteristics with new systems, perhaps the only one is the GDDR5 main memory of the PS4. Both use low power CPU which do have the capability to handle 8 threads, that is one per core, if I remember correctly. Cell processor used in PS3 for example could also deliver the same amount of threads and it is far more interesting consept than either of the new consoles and offer still really good performance after 8-9 years of the developement for certain applications, but game developement proved to be quite difficult because of the complicated dev platform (I've heard), the ring bus design of the Cell, which requires careful thread timing in the software design and because of the low performance of the PowerPC core of the Cell, which again practically demands effective use of the SPEs and thus multitasking. This makes Cell really a good processor, even today, for many scientific, encryption and multimedia tasks, but for game design where multithreading is often quite difficult or you at least you gain smaller and smaller performance benefits for threading it more and more and still every time complicating the software design unnecessarily. New consoles will bring the consoles to the 2010's when it comes to performance, especially from GPU side, but the performance estimations I've seen suggest that the CPUs itself can deliver the performance of less than one quarter per core of that of i5 3570K (http://www.rockpapershotgun.com/2013/05/27/week-in-tech-hands-on-with-those-new-games-consoles/). That is quite underwhelming and even though your desktop PC can't handle the 8 threads that are possible with new consoles, its raw power makes up most of that easily. Software optimization is of course wholly another thing as is the performance of the less complicated OSs of the consoles compared to complcated all capable, all around computer OSs.

 

Just my 2 cents about hardware at this point...

AMD just managed to get by the 5ghz barrier. but going beyond that and cost is prohibitive I would imagine.

I wonder what happened to dual CPU MBs. They just seem to have died out as quickly as they came in probably because single cpu multi core outperformed them. But if you were to take that technology and put two of them on the MB, you could essentially have a simulator written so that one CPU is dedicated to the aircraft itself, systems, FDE, and the second CPU would handle the outside world around. The problem is bandwidth to handle sending all that info I would imagine. 

 

That said, I dont think I'll be owning a CAE type simulator anytime soon. Just watched a video on 787 testing, and one building a 777 simulator which took 3 weeks, and costs 8-16 million. (Thats a steal compared to PMDGs recent listing of 165 million LOL)

I was looking for the computers that run the show, but everything is well integrated now, and far different from 15 years ago. 

 

Anyhow Im getting off topic from the OP, our hopes still have to lie in an eventual 64 bit, platform that is highly optimized with no bottlenecks. 

Like AMD products not overpriced like intel is. It's easier to upgrade and CPU A10-6800k clocked at 4.1 and combo APU setup with equalivant 6670 type intergrated graphics card your good to go at mid settings for under $500. Amd get 5.0 fx 9590 will become enthusiast CPU in 2020 for $300 dollars.

 

 

You'll need to apply Steve's DX10 adjustments (files available in AVSIM library) and/or wait for Steve's soon to be released retail package that makes the process easier.

 

Sorry for digressing back to this topic, but will Steve's DX10 retail package be fully legal under MS EULA? I would hate for this project to be thwarted in any way. I am no expert, so I am just throwing this question out there. 

It's good to see FSX CPU variance test data but I wouldn't use that as definitive for a few reasons:

 

1.  Chipsets and motherboards are not the same in all cases (can't be) so not really apples to apples.

2.  Affinity setting, 6 core vs 4 core, again not apples to apples. 

3.  Chart is confusing, were all CPUs tested at 4Ghz only?  Would have thought max stable Ghz for each would have been more appropriate (either air or water)?

4.  Trying to do CPU loading test (aka use a lower performing graphics card) doesn't really represent FSX hardware.  FSX does push more on the CPU but it still benefits from improved GPUs (Titan comes to mind when working thru 4096 clouds with high levels of AA).

5.  Haswell 1150 dual channel vs. 2011 quad channel.

 

The data is interesting, but just as FSX is balancing game, so is the hardware ... the two are not independent of each other.  RAM timing improvements on a quad channel setup produce overall better performance than the same timing improvements over a dual channel motherboard ... but again, not really apples to apples and synthetic tests don't really add-up because they aren't emulating what real world applications/games might do.  And then there is the issue of having 2 additional CPUs available to process other OS activity and/or Add-Ons.  So again, not really Apples to Apples and it removes real world context from the equation.

 

But for the sake of debate (because I know this might generate an endless threads of people defending a purchase/CPU choice), lets assume the out-of-context 10% improvement value can be applied to any system with a Haswell ... 10% of 30 fps is 3 fps (difference is even lower in problematic areas where you need it the most).  At the current rate (about 18-24 months) Intel is "increasing" performance, it'll take over 10 years to go from 30 fps to maybe 45 fps in FSX -- possibly longer given Intel's shift in focus to less power and mobile computing where significant desktop updates are moving further and further apart.

 

What I would like to see Intel/AMD focus on for the desktop computer is much improved bus bandwidth, 8 channel (or higher) RAM (= even more expensive motherboards), and much better chipsets that can leverage the abilities of multi-cores and multi-GPUs.  Sadly Intel appear to be getting out of the desktop/motherboard business: http://www.zdnet.com/intel-to-leave-desktop-motherboard-business-by-2016-7000010269/ ... door open for AMD?

 

Rob

 

I think the goal has always been to achieve a balanced FSX install (good basic add-ons across the board, decent terrain and 3d objects loaded etc) when running benches there really is no need to even use the Affinity settings unless you are trying to do something very specific.

 

As for the hardware I'm pretty sure that all this has been considered in the benching process already, kind of basic 101 stuff. Yes we could start an endless debate about all things that effect the outcome but don't miss the forest for the tress Rob, really the performance of the Haswell is clearly evident.

 

A performance advantage is what it is.

 

 

but will Steve's DX10 retail package be fully legal under MS EULA

 

If so, then the AVSIM library is also in violation because much of his work is already in the library.

 

Microsoft no longer lists FSX in their licensing and terms site here: http://www.microsoft.com/en-us/legal/intellectualproperty/UseTerms/default.aspx

 

But I do wonder why you would want to bring back the EULA debate?  Just be aware that whether something is given away for free or is sold for money, it has absolutely NO bearing on the EULA ... reason I say this is because you used the word "retail".

Argh more EULA debates, all it does it creates artificial roadblocks for us all

 

Guys like Steve are only helping us!

In fact, it will probably run slower than now.

 

Because, instead of crashing with an OOM after 4-5 complex scenery+airplane are used together in the same area, users will be free to stuff it at ungodly unnatural levels, and will obviously complain about the collapsed fps...

 

Of course, Virtuali is 100% right!

 

To convert FSX to 64-bit would be a herculean task, not for the faint of heart or light of wallet.  Then all the hang-on 3PD codebuilders would be compelled to follow suit.

 

And like X-Plane, which is ALREADY 64-bit, the end-user could push the sliders and autogen through the roof and bring even a 64-bit system to a stuttering mess - I know, I've DONE IT.

 

While 64-bits has some advantage in being able to address huge memory resources, it takes more than RAM to make a flight sim, my dear Virginia!

 

You have other bottlenecks that come into play, not the least of which is you are still talking about quad-core processors in most cases, and  your VRAM on your gaming card can easily be flooded out in XPlane unless you spring for the GTX Titan, and even THAT card can have its 6GB (6 billion bits of VIDEO ram) flooded by insane high-def graphics and clouds.

 

What's the rub in Flight Sim world is that the simulation is not a finite limited game, like GTA  (Grand Theft Auto) for example.  The pilot can plug all sorts of EXTRAS into the gaming engine, all of which have a NEGATIVE IMPACT on performance.

Microsoft (and now the Devs of GTA) have seen the light, and are building online virtual worlds for multiplayer games, and that is where I ultimately see flight simulation shining.  The better ISPS can give amazing throughput now, and if you build a supercomputer that is running your simulator, chances are your graphics and overall performance will likely blow away anything you can build at HOME.  Part of that is the motherboard builders' fault, and the Operating Systems as well. A system with 4 quad core processors is of course theoretically possible for a WORKSTATION, we need workstation computing power with like 4 video cards in SLI running Direct X 11 (as an example).

 

Then you begin to envision something in terms of CGI graphics like they obtain at rendering farms at Pixar Studios (for example).  Something cinematic in quality, much more than the graphics we now have - and aren't we spoiled?  Look at the earliest video games (like PONG, or even Microsoft Flight Simulator in its first iteration.  Caveman-like graphics.  So today we are incredibly spoiled, and like a spoiled child, we want so much MORE than what we have already.

 

The Ground Services X alone is a stellar add on, that steps up the level of immersion immensely in FSX.  It's developer (Virtuali) knows of what he speaks.  It's all about optimization, and you have to give a little on this to uptick on that...

 

all because it's not just ONE unified game with NO hang ons.  The hang ons make it appealing, but carry a resources price as well.

GTA flight sim do traffic pattern and see 100 police cars at the end of the runway and retract flaps and push throttle to full to find another airfield. That GTA flight sim lol. Its safe to land that cessna, yes, really want to want to, no been bush flying items from the jungle that snow white color and dropping them off in the waters of Miami or liberty city lol.

Just be aware that whether something is given away for free or is sold for money, it has absolutely NO bearing on the EULA

Correct. all that matters is that the terms of the EULA aren't broken.

.
.

Correct. all that matters is that the terms of the EULA aren't broken.

 

Eulas are pretty difficult to break.  Consider the WHAMMO brand "Eula-Hoop".  Millions were sold, and most are still intact.  Now that's a strong EULA!

 Robert!

I thought you had given up our FSx for X-Plane!

 

If that's not the case did you fly the PMDG tutorial #2? If so, since you and I have virtually identical setups , CPU, Graphics processors and TM HOTAS Warthog, did you experience extreme frame rate slow down on your approach to LOWI?

 

Mine slowed down to what felt like 1FPS(ONE FRAME PER SECOND)  after Rattenberg(RTT) in P3d.

 

Thanks,

 

Chas