You Gonna Hate Me, But...

[SUNDR1V3R 72]

I've been making Pc's as a bussiness since 2012 specially for real pilots.. I must say i built around 30 of this pc's.. With a profit between 100USD and 250USD by PC..

 

Anywaysthe hardware was like this:

CPU: i7 4770K
GPU: GTX 770 2gb ASUS Direct CU II

MOBO: Ga-Z87MX-D3H

RAM: 8GB @ 1333Mhz

HDD: 1TB SATA 3 7200RPM

PSU: 550Watts CM 80PLUS

CASE: N200, N400

 

An a ton of tweaks, from FSX.CFG to the SP1 and SP2 without the acceleration Pack, NVIDIA inspector, DX10, etc.. And i must say it runs amazing.. In any weather, in any airport, in any time, with all traffic .. I Use add-ons for almost averything..

So no complains on that PC..

 

But, the thing is that i've been doing some research lately and a I discover that the i5 4670K could be doing the work just fine and im just wasting money in the i4770k..

So thats why im here asking for your help.. I built a PC like a month ago with this specs:

CPU: i5 4670 (no K)

GPU: R9 270X Vapor X

RAM: 8Gb @ 1600Mhz

MOBO: H81M-E

 

And it actually runs very well, of course not as good like the one above, but in a scale from 1 to 10 thinking that the i7 4770K was 10 this was like 8 and customers are able to save  like 400USD..

So here are my questions:

What is the most important hardware when you think in FSX:

 

-CPU right? So:

 

• Does FSX takes advantage of the HT?

 

• Would you recommend the i7 4770K over the i5 4670K even with the HT turn off in the i7? (Thinking in the OC in both processors), im thinking doing 4.5Ghz OC in the i5.

 

• Is there any reason besides the HT its worth getting the i7 4770k in FSX? Probably better OC in the i7 or best quality in the Chips, anything..

-How good does GPU FSX need? Im actually and NVIDIA INTEL fan guy, but after using this 270X i saw really nice results ATI/Radeon GPU *i'll show photos with this GPU in the end of the post*

I know the better the best, in this case if you can afford the 780 stick with that but.. The point of this thread is to be able to max out the hardware of the CPU, in this case probably its better to get an 750Ti but working at 80% rather than a 770GTX working at 40%.. Im just saying dont actually know the facts.. 

 

-How important is to use a RAM at high frequency like 1800 against 1333Mhz?

I really apreciate your time for reading and answering this topic.. Thank advance friends..

Greetings from Mexico,
Have an excellent day!!
 

Screens with the i5 and R9 270X 2GB VaporX Sapphire

[mjrhealth 820]

There are many experts out 5there tah tsay HT does nothing for FSX, and in DX9 i may agree, but when using DX10 many have said itmcan make the whole sim smoother. Of course it does come with ectra heat, and you may have to drop the overclock speed a notch or oncrease cpu volts to use it.

[gboz 157]

 

 

Screens with the i5 and R9 270X 2GB VaporX Sapphire

 

Hey, really nice.

 

gb.

[martin-w 3,234]

•Does FSX takes advantage of the HT?

 

 

No, FSX isn't coded to take advantage of HT.

 

 

•Would you recommend the i7 4770K over the i5 4670K even with the HT turn off in the i7? (Thinking in the OC in both processors), im thinking doing 4.5Ghz OC in the i5.

The 4770K has a bigger cache, which is said to aid FSX. However, I have no idea if you would notice the difference in real world use.

 

"May" also be a better overclocker, depending on how Intel bin these chips.

 

 

 

 

 

-How good does GPU FSX need? Im actually and NVIDIA INTEL fan guy, but after using this 270X i saw really nice results ATI/Radeon GPU *i'll show photos with this GPU in the end of the post*

 

 

 

Theoretically, FSX is CPU bound. However, these days, as we are using higher res, higher AA and graphically demanding add-on's, it seems to me that the GPU is more advantageous than it was. I myself noticed a very nice improvement jumping from a GTX 580, to an EVGA GTX 770, 4GB, Duel BIOS card.

 

It still applies that FSX favours the CPU though.

 

 

-How important is to use a RAM at high frequency like 1800 against 1333Mhz?

 

 

 

Haswell loves fast RAM. However, how well an individual chip tolerates faster RAM is a variable. The integrated memory controller is now on the CPU, this variable can mean that individual chips vary in their memory frequency capabilities and the timings they will tolerate.

 

Again, "theoretically" faster RAM is an advantage, but you would have to carry out proper tests, a proper comparison to state definitely that it makes a "significant" difference in real world term.

[dazz 69]

Well, you don't code things to use HT. It happens in a lower level, and it's been already proved by Lars here that it sort of works in FSX under the right workload. Problem with FSX is that it's not properly threaded, and HT will only be truly beneficial at highly threaded loads. With high detailed photoscenery, you get (slightly?) better texture loading but can come at the expense of a bit lower FPS or stutters.

 

The L3 cache is a non factor. The notion that it helps comes from the same NickN guy that said HT doesn't work in FSX.

 

I would say, for the regular user, an I5 is fine. The benefits of an I7 are marginal at best, but I would still get the K version paired with a CM 212 to leave his overclocking options open.

 

For a GPU, a midrange one is fine at moderate AA levels, and RAM makes a difference, but not a huge one, if you can get cheap 2400Hz 8GB  kits, better than 1600MHz 16GB ones

[dave2013 1,123]

CPU clock speed has the most impact on FSX performance, so if you can clock an i5 to the same frequency as an i7 then I doubt you'd see any difference.  The i5 is just a slightly cheaper CPU with a little less cache, but otherwise identical to an i7, with the caveat that the i7 may be of higher quality and thus may overclock better.  

 

A powerful GPU is necessary for higher levels of antialiasing, especially if you want to utilize transparency sampling.  If you're using DX10 then the GPU plays an even more significant role in FSX graphics processing.

 

RAM speed very likely does make a difference, but the question is whether that very small performance increase you might get going from 1600MHz to 2400MHz is worth it, or even noticeable when using the sim.

 

This is all based on my experience and knowledge gained over many years of flightsimming, reading forums, and a good deal of technical knowledge and experience with computers, electronics, and programming.  However, I am no expert in any of the aforementioned fields, so I could be mistaken.

 

Dave

[martin-w 3,234]

Well, you don't code things to use HT.

 

Really!

 

Well it wasn't just NickN that told us HT doesn't benefit FSX... Microsoft told us that FSX does NOT support HT.

 

If you disagree with NickN fair enough, I too disagree with a lot of his comments. However... his source for this information was Phil Taylor and Adam from Aces, who actually coded FSX.

 

In my opinion, to disagree with Adam, the Microsoft expert that coded FSX, is quite a bold move.

 

As Nick Needham said...

 

#7. Intel Hyper Threading: FSX has no code to recognize or process logical core hyper thread. I don't know how many times I have read some 'guru' has come up with an assessment that FSX will make use of or runs better with hyper thread enabled. This person really needs a sign! The REAL terrain 'guru' Adam from Aces as well as Phil Taylor specified years ago that FSX does not support any logical core threading (hyper thread).

 

 

 

The reason you SEE logical core activity in the Windows CPU monitor window is because the physical cache is used for each logical (hyper thread) core. "logical" core means there is no 'physical' hardware core but the hardware cache in use is there, and, FSX is DUMB so a thread is spawned due to the physical cache in use, but the data is never processed and used by FSX!

http://www.simforums.com/forums/the-fsx-computer-system-the-bible-by-nickn_topic46211.html

 

 

you get (slightly?) better texture loading

 

 

So "slightly" you say. In my opinion "slightly" isn't something you can definitively prove. More down to subjective opinion that objective evidence perhaps.

 

If you can show me any definitive evidence, as always, I will change my mind.

[dazz 69]

If you can show me any definitive evidence, as always, I will change my mind

 

You just need to set the right conditions.

 

I quick test here with LOD=9.5, over max resolution photoscenery and AF=252 to run texture loading on 3 physical cores / 6 logical cores

 

https://www.dropbox.com/s/ml259rroqo4onvz/Captura%20de%20pantalla%202014-04-25%2018.19.00.jpg

 

All 3 texture loading cores maxed out, and if you think they just look to be doing something when their not, you can run a simple test and slew fast forward, then stop and mesure the time it takes for textures to get crisp. It's clearly faster with HT on. I'll look for Lars' tests that are extremely detailed.

 

That's not to say that it makes a huge difference in "normal" situations, but clearly proves to me that whatever NickN says, is to be taken with a grain of salt at best. It's up to you to believe what you want, but I'm pretty sure no one from ACES ever told him anything like that. You can code something to make better use of HT, but you don't need to do anything to allow threads to run on logical cores. That's ridiculous to be honest, since with HT enabled ALL of the cores exposed to the OS are logical anyway. Going by that logic, FSX wouldn't run at all with HT enabled

EDIT: Lars' (SAAB340) analisis using the technique described above to test texture loading

 

http://forum.avsim.net/topic/429204-a-further-look-at-hdds-and-ssds/#entry2880592

 

EDIT 2: And just to clarify things further, there's a big difference between seeing some activity in two logical threads as opposed to having both maxed out because that's when you make the most out of HT. Just about any workload will show some sort of activity in both logical cores of a given physical core.

[martin-w 3,234]

It's up to you to believe what you want, but I'm pretty sure no one from ACES ever told him anything like that.

So you are saying Nick Needham is deliberately lying to us?

 

What motivation could he have for such a thing?

 

 

[Gregg_Seipp 565]

I just got a new rig a month ago with an i7-4770K and GTX-770 4GB RAM.  I did all the optimizations to see what worked best:

• HT off seemed to get me 2 more FPS (don't know why) but no more smoothness so I turned it back on.
• I had serious jaggies so I spent a lot of time tweaking that and found that 8xSGSS did the best job.   (8xSGSS really smooths jaggies, cuts way back on flickering on trees and lights). Of course, the video card bogs down seriously with that setting.  With that setting, I can get 40-60 at less demanding airports but only 20 FPS at the more demanding airports  Overclocking only the video card I could get it up to 30 but with no more smoothness.  Turned off the overclocking.  I haven't overclocked the CPU yet at all.  So if you need/want 8xSGSS the video card is a bottleneck. 

Gregg

[martin-w 3,234]

Trouble is there are so many variables. You can conduct tests, think you are keeping all variables the same, and see a subtle difference the next time you test.

 

One or two frames here, one or two frames there, "maybe", "perhaps" slightly better texture loading with HT on... none of this is very scientific.

 

There are no doubt all kinds of variables having a subtle influence that we aren't aware of. Not forgetting the placebo effect of course. It's easy to convince ourselves that we see what we want to see.

[dazz 69]

So you are saying Nick Needham is deliberately lying to us?

 

What motivation could he have for such a thing?

 

 

 

My guess is he's just not up to date. HT implementation has improved over the years, new generation of CPU's have a much better memory bandwidth witch helps HT in some situations, and there have been improvements in the way newer OS's manage HT. But yeah, I pretty much ignore anything he has to say 

 

 

none of this is very scientific

[...]

It's easy to convince ourselves that we see what we want to see.

 

 

It's empirically scientific. I already told you how to test it so you don't need to convince yourself.

 

At any rate, once again the way I see it, HT off (or better, an I5) will do just fine for most simmers in most situations. But that's just me, I have it off myself. Guys like SAAB340 (Lars) or BimmerCop (Efrain) prefer the extra texture loading HT or even 6 cores can provide in photoscenery/TileProxy

[Gregg_Seipp 565]

Trouble is there are so many variables. You can conduct tests, think you are keeping all variables the same, and see a subtle difference the next time you test.

 

One or two frames here, one or two frames there, "maybe", "perhaps" slightly better texture loading with HT on... none of this is very scientific.

 

There are no doubt all kinds of variables having a subtle influence that we aren't aware of. Not forgetting the placebo effect of course. It's easy to convince ourselves that we see what we want to see.

 

I suppose I should have added the YMMV disclaimer.  I did do the same test I do for all of my optimization and only changed one thing:  PNW KSEA north departure to 1000 ft.  I got the average frame rate over time from LINDA console so it was less subjective.  It's good enough for my purposes, though, perhaps not for this.

 

Gregg

[martin-w 3,234]

My guess is he's just not up to date. HT implementation has improved over the years, new generation of CPU's have a much better memory bandwidth witch helps HT in some situations, and there have been improvements in the way newer OS's manage HT. But yeah, I pretty much ignore anything he has to say

 

So not lying then, and Phil Taylor and Adam from Aces did actually say that "FSX does not support any logical core threading (hyper thread)." Previously you said... " I'm pretty sure no one from ACES ever told him anything like that"

 

But yeah, I pretty much ignore anything he has to say

 

 

I wouldn't agree with that approach. By doing so you miss out on the stuff he tells us that actually is valid. It also tells me you are biased against anything he says.

 

It's empirically scientific. I already told you how to test it so you don't need to convince yourself.

 

 

I would say it isn't. There's nothing scientific about the tests at all. There are too many variables to make such a definitive judgement. Especially when we are dealing with such subtle effects. Even just rebooting the PC can add a variable.

 

When we "think" we see  "slightly" faster texture loading it's purely subjective, based on an individuals perception. You would need some kind of "objective" metric, to objectively prove that there was faster texture loading present. And then of course those results would need to be replicated by many others.

 

If you can't objectively measure the effect in such a way that it avoids a human beings subjective opinion, you have nothing.

 

I'm not trying to be argumentative, and you may well be right, but to be certain of any benefit at all, we would need to be a lot more scientific about it than we have been up to now.

 

I'm yet to be convinced I'm afraid.

[dazz 69]

I'll try to explain it again.

If you set the proper conditions (high LOD, high speed, enough texture loaders...) the effect of HT is so obvious that it just can't be ignored. Nothing subjective there.

When I said "slightly" better texture loading I was referring to situations where you don't speed up the sim or the flight hence reducing the texture load.

But HT works, and it increases throughput under the right circumstances. Not sure if you have HT there to test it yourself, it takes 10 minutes. (Fast Slew, stop, wait for textures to load)

HT only helps with performance with enough multithreading going on, obviously. You won't get better frame rates cause the main scheduler is still a single thread running in it's own core so you can only make things worse there by making it share resources. So the easiest way to mesure HT effectiveness is to increase texture loading demand, even if it's with settings you would never use, risking OOM's and the likes. It just to prove that:

 

"a thread is spawned due to the physical cache in use, but the data is never processed and used by FSX!"

 

really doesn't seem to be happening.

 

Other than False Sharing, I'm not sure what could have such an effect on threads to just sit there doing nothing

 

 

 

I certainly would like to see a link to something Adam Szofran said himself instead of what NickN says he told him