CPU choice

[pirate60]

All,

I have a choice to make for my next cpu.  First of all, my current system specs:

Mobo: Intel DG33TL

CPU: Intel E6750 2.66ghz core duo

GPU: Nvidia 8800 GTS 512

RAM: 4Gb corsair 6500C5

Sound: Creative xtreme gamer x-fi

Monitor: Acer 24" wide-screen

OS: WinXP 32 bit

The above system is fine and gives me good performance, but as always I am looking for a little bit more  .  I can't overclock the DG33TL, so therefore my only option to get more horsepower is to upgrade the CPU.

I was thinking of either an E8500 3.16ghz (dual) or a Q9550 2.83ghz (quad).

Any opinions?

Cheers

Dave Q

[solareagle]

Go QUAD for sure.  I was of the opinion that a highly clocked dual is the way to go, but I found even with a 4Ghz dual I still suffered from blurries and poor terrain texture loading.  After installing a QUAD all blurries were gone and I finally had sharp texures in nearly all cases.  Those who fly at normal altitudes often don't notice blurries, but when you fly at tree top level like me its a real problem with anything other than a Quad. 

I'd recommend ditching that MBD you have and get one that overclocks.  Thats the best thing to do with your money.  You could get a Q6600 G0 stepping for about $250, and a MBD to OC it for about $150.  Thats a lot cheaper than a Q9550, AND you will be able to get 3.5GHz out of it, where the Q9550 on your current board will be stuck at 2.8Ghz. 

So for $550 you can have a 2.8Ghz quad, or for $400 you can have a 3.5Ghz quad.  For m ethe choice would be clear, get a new MBD and overclock the heck out of a Q6600.

[Shotgun3]

hi solareagle,

apart from the QUAD upgrade, were there any other compnent you changed as well?  based on what i know, fsx is basically still a single-core application that does not leverage on other cores.  (i could be wrong).

i thought blurries are not just caused by slower cpu.  there are many other factors (ram, disk fragmentation, etc)

i only run a C2 Duo E6420 clocked at 2.66GHz, 3GigRAM, MSI 8800GTS OC version 384MB, nHancer, 2 500Gig SATA2 drives... and never encounter blurries.  running an average 24-30fps here with above-average sliders at 1680x1024 .    GEXn, FEX, UTX, ASX, FTX, XG, FDC - all these running.

that is probably because

- i tweak my XP Pro based on a suggested tweak at AVSIM by Nick_N

- i installed FSX on a separate physical drive

- regular hd defrags

- i use AlacrityPC utility

I am planning to build another rig, one of m choice is the C2 Duo E8400 3GHz and clock it to 4GHz... but I am still open on other suggestions.  It is a cheaper choice thatn getting the more recent Quads that runs slower clock.

CHeers,

Santiago

[solareagle]

apart from the QUAD upgrade, were there any other compnent you changed as well?  based on what i know, fsx is basically still a single-core application that does not leverage on other cores.  (i could be wrong).

i thought blurries are not just caused by slower cpu.  there are many other factors (ram, disk fragmentation, etc)

No I just swapped the CPU, and the SPx multithreading code in FSX will support up to 256 cores, so there is no basis for a claim that its only single threaded.  The multithreading helps with terrain texture loading and to reduce stuttering, so you do not get a framerate gain from it. 

Obviously there's more than one variable that influences blurries, but the primary cause is the terrain engine unable to keep up with texture laoding.  A quad has two more cores devoted to texture laoding, so there is VAST difference in how fast and reliably the textures load, and in how well it can maitain the highest LOD.  Despite the 60% clock speed delta a 4GHz Duo can't even compare to a 2.4Ghz Quad when it comes to texture loading performance. 

You say you have never encountered blurries, so try flying 300knots at tree top level with very dense autogen through New York.  You will see the engine is unable to maintain the highest LOD, and the textures will get blurry. 

[Maurice_King]

I must be missing something here ,  the Processor has far less to do with blurries than many realise . besides you DONT have an eagles eyesight.

Things will be indistinct at a distance due to your rendering radious and the pixel capacity of your display/s.

FS isn't Quad core aware so you might be spending way more than you need too. Personal opinion based on my PC.

[solareagle]

Things will be indistinct at a distance due to your rendering radious 

Yes thats what the LOD system is for.

FS isn't Quad core aware

FSX sees and uses up to 256 cores, and thats a fact.

[JohnG]

solareagle is right in that FSX does u's all four cores.

In Vista's side bar you can get a CPU usage meter,and you can watch FSX use the 4 cores.

Cheers....

JohnG....

[mstyles]

If you over clock make sure you get a good CPU cooler. 

The Thermaright 120 Ultra Extreme is probably the best performing cooler and good value at approx $80.  It's very tall and so may not fit in your existing case, better check 1st.

I installed one last night on a Q6600 G0 and idle temp went from 43C down to 27C  (that's a 15C drop!).  That was with one 120mm fan installed on the cooler running at 900 RPM. So its silent!  You can add another 120mm fan to the other side for even better performance.

Next I'm going to over clock the CPU from 2.4ghz to 3.00ghz (25%).  With this cooler the temps should still be very good.  Who knows maybe I'll reach 3.6Ghz.  May aim would be to stay below 65C at load.

Here's the link

http://www.thermalright.com/new_a_page/product_page/cpu/u120ex/product_cpu_cooler_u120ex.htm

[Matthew Webb]

I'm confused (but you all knew that already  ).  My understanding is that FSX uses as many cores as in the PC, but FSX (even after SP1) is primarily still a 1 core application.  Certain parts of FSX is batched off to the other processes, but you will always see FSX's main process use 100% of the core it runs on and the other cores jump around, rarely reaching 100% (except maybe during parts of the "load flight" sequence).

Is this assumption right?  If so, I would have thought performance of the core that the FSX process runs in is more important that the number of cores FSX can batch other processes too, as long as you have at least 1 other core (ie a 2 core CPU).  Is that assumption right? If so a high performance 2 core is better than an average quad?  \   

Cheers,

Matt.

[Barry Meiklejohn]

Partially correct Matt. My understanding of the issue follows.

You need to remember that a program like FSX is actually a collective of different code performing different operations. Yes, CPU speed will make a huge difference. This is because it will speed up each of the threads that are occurring. FSX SP2 is multi-core aware and various components of the program can be split off to different threads. These threads have the ability to utilise different CPU cores.

As alluded to above, one of the biggest issues in FSX, the blurries, is a result of textures not having sufficient time to display at their highest LODs. Some people may never see the highest LOD of the textures so they assume what they are seeing is clear when others may say the textures appear blurry. This all comes down to available memory and processing power.

However, another aspect of FSX is 'stuttering'. This is where the sim appears to lock up for a fraction of a second, or exhibits jerkiness. Generally this is caused as textures are being called too slowly for the display engine. This is where multi-core CPU's really shine. They have the ability to process the textures and have them ready for the display engine to call. The more cores, the more textures can be processed. So whilst they will not overtly affect the frame rate of the sim, they will increase the 'fluidity' or 'smoothness' of the sim.

[Barry Meiklejohn]

Here is some information provided by Phil Taylor of Microsoft ACES studio about FSX Multicore

Multi-core Performance Work

During loading, we run the DEM loader on threads. You'll see good balanced usage across all cores; as well as about 1/3 faster load times on average.

During flight we spawn threads for Autogen batch rebuilds as well as the terrain texture synthesis. The terrain texture work tends to be a bit bursty; as an area gets generated the load reduces true. But as you fly forward, as you bank, and as the terrain is lighted ( once a minute ) threads are spawned. The terrain grid system is radial around the current viewpoint, and, depending on level of detail radius can be up to 4.5 tiles in either direction, something like 64 tiles. So there is plenty of work to go around. Autogen is more constant, with a 2km extent being batched.

Even given the bursty nature of the core usage when flying; when there is load, its pretty balanced across the cores. And we got rid of as much of the stutters as we could by going to a lock-free synchronization style. Its solid work that we are deservedly proud of.

As far as practical limits on number of usable cores; currently SetThreadAffinityMask only allows explicit scheduling of threads on 32 cores ( the mask is a dword ) on Win32. So thats our effective limit on number of cores. But as soon as there is a way to explicitly schedule them, we can handle 256 cores.

[solareagle]

So simply put CPU clock speed determines your framerate, and additional cores improves upon texture loading and stuttering.  If you are not seeing an issue with blurries, then a highly clocked Duo might be better, where if blurries are a concern, the Quad may be the way to go. 

In my case I'm running a Quad with a lower clock and I'm happier than running the highly clocked Duo, which is solely due to my style of flight, which is prone to blurries.  Low and fast is the name of the game for me.

Here's some benchmarks to look at.  The first chart compares Duo and Quads at an equal clock speed of 3.0GHz.  As you can see adding two more cores has no impact on framerate, though the graph obviously cant show the improvement you get in the texture loading department.  You can also see that adding an addition 2MB of L2 cache per core results in a 9% framerate gain, and I would imagine that gain scales nicely with clockspeed.

This second chart shows how your FSX framerate scales perfectly with CPU clock speed, which means the bottleneck in these bechmarks is in the CPU.

[solareagle]

Delete.

[Quinn]

Here's some benchmarks to look at. 

Is it just me? - I can see that you have embedded images but they dont show on my machine even after 2 refreshes

[Barry Meiklejohn]

Might be your system Quinn as I can see the graphs.

[Maurice_King]

Ahem I di not say it could not use them I said it wasn't Multi core aware.

[Shotgun3]

I mentioned blurries was never my concern because my style of flying is low and slow in the countrysides.  Until FTX came out, I dont like to fly over big cities... unless I am in the mood to fly the private jets or the big heavy irons.  FTX yields better frame rates on my system over urban areas, quite surprisingly even at Very Dense scenery and Dense Auto-gen , and I like the customized autogen buildings/trees.

Still, I prefer buzzing grass/dirt airstrips with a singles (love those Scout or Beaver tundra versions)... which is why I am so happy with the new AU GA airstrips freeware!

So, with my style of Flight Simming, my now-outdated E6420 Core 2 Duo is enough for my preferences - a good balance between looks and performance, with enuf eye candy 

mstyles,

Many of my friends are using Q6600 and they're running their clocks at 3.6GHz with decent cooling!!  It seems this chip comfortably runs at 3.6GHz with the right cooling.

[solareagle]

Is it just me? - I can see that you have embedded images but they dont show on my machine even after 2 refreshes

I should have cited the source of the benchmarks.  Take a look at the page they came from.  I rehosted them as jpegs, but they were originally gifs.

http://www.behardware.com/articles/691-7/intel-core-2-extreme-qx9650.html

http://www.behardware.com/articles/657-6/amd-athlon-64-6000.html

[Quinn]

Might be your system Quinn as I can see the graphs.

Must have been - I was doing all sorts of upgrades and things earlier - they are there now - odd, because I actually tried page refresh 2 or 3 times

Anyway - problem solvered