CTBlankenship, on 21 Aug 2013 - 11:45 PM, said:

 

Whoaa ... back up there champ ... all I had was FTX Global installed ... after installing PNW my frames went to 15 to 16 ... that bloody test had all 8 cores at almost 100% ... where in God's name did you come up with THAT one!

 

C. T.

C.T.,

I assembled a new Sandy Bridge-E processor based system a few months ago and I'm seeing 13 -14 w/o HT enabled but w/ all maxed in full screen mode. I haven't tried HT enabled so that...I will try a little later tonight. Your reported performance at 15-16 frames is right in there with what one would predict for the superior single-core performance in the Haswell line. I opted for more cores & doable 6-core overclockability, plus upgradeability to IB-E on the same platform to enable PCIe 3.0 and a bit more performance. It will be interesting to see if intel uses fluxless solder on their IB-E 6 or 8 core processors. I do love this SB-E though. Fabulous smoothness and rock-solid performance. I haven't stressed it at all and it moves along at 4.4Ghz w/o breaking a sweat. If IB-E comes out w/ fluxless solder TIM I will take on more risk w/ higher over clocks on my 3930K.

 

 

But this makes me question the statement that FSX doesn't use those four virtual cores ... they were at 100% with all the others.  Maybe I'm doing something weird hum?

 

That's not how hyperthreading works. If the system is only in need of running four threads, then each physical core will run a thread. Once there are more active threads than physical cores, individual physical cores start to run parallel threads, until double the number of physical cores is reached.

I ran the test again with HT disabled and what appears to be a rock solid 4.4GHz overclock ... bouncing between 13 to 15 ... all stress tests for 1.29V @ 4.4GHz only pushes max temps to around 72.

 

The one I got from CyberPowerPC went straight to ~95C at a 4.2GHz overclock ... and a custom water cooling loop ... which can be expected with no special work being done other than to increase the multiplier to 42x ... which is what I'm sure they did.

@UlfB - HT really doesn't seem to impact FSX in a positive or negative way at all on modern hardware....but Windows itself can benefit from having HT on. If you use addons or have other stuff running in the background then why not leave it enabled so they can benefit from it?  You can force FSX to only run on certain "cores" if you will by using affinity masks in the config file but if I recall correctly Phil Taylor said it wasn't needed after SP2/acceleration( I could be wrong on this , couldn't locate the article)

There were many of us here at AVSIM that did som extensive testing with FSX regarding HT and using the affinity mask a couple of years ago after the release of SP2 and Accpack. The conslusion was that enabling HT had no effect on FSX. It only raised your cpu temps about 10C when stress testing. Better to disable HT and go for a higher OC.

 

Sure I know that Win7 and some applications will benefit from enabling HT. As I use Adobe PS and LR I have an OC profile with HT enabled.

 

I guess I misunderstood your answer on my post:

Sorry this is misinformation

FSX can benefit form HT. It's very easy to test. Disable two cores, or leave 2 enabled, enable HT and set AF=15.

 

Fire up FSX and slew fast & far away. As textures load, CPU usage will ramp up to 100% and stay there for a few seconds until textures are fully loaded, and you can easily see they load faster than without HT.

Lars & I did these tests too, actually he posted some extensive tests here proving all this. 

That's not to say that with four cores and "normal" use there's more than enough to have crisp textures unless you fly fighter jets at low altitude maybe. We fly FSX not slew party after all, but for dual cores HT is definitely a good thing and will somewhat enhance the experience. 

Also Efrain (BimmerCop) said once he clearly saw better performance in TileProxy with HT on his quad core.

 

FSX can certainly "see" virtual cores. It better can because with HT enabled all 8 cores exposed to the OS (on a quad) are virtual. Many people still think you get 4 virtual cores plus 4 physical cores with HT which is wrong. Each physical core is exposed as two virtual cores to the OS, and no special support is needed on the software side to take advantage of HT, other than placing a workload that involves running more than one thread per physical core, of course. 

 

I'm doing Mallorca photoreal right now and want to try different compression quality levels to see if I can maximize multicore performance and maybe have a use for HT.

 

Here's a pic, BTW, welcome to Orient

 

Dario,

 

Thanks for your reply. How much (some benchmark or so) will performance improve with HT enabled compared to an increasd OC with HT disabled?

 

I have an OC profile 4.2GHz with HT enabled and a profile 4.5GHz with TH disabled. I'll test the difference and see if I see any difference in smoothness and avg fps. But i doubt it. The placebo effect is very strong on the majority of people.

I ran the test again with HT disabled and what appears to be a rock solid 4.4GHz overclock ... bouncing between 13 to 15 ... all stress tests for 1.29V @ 4.4GHz only pushes max temps to around 72.

 

Interesting, then we're pretty much seeing the same w/ your Haswell outperforming the SB-E but only by 6-7% or so.  I would have expected a little more.   I have not had a chance to test w/ HT enabled on my hexacore so will get to that I hope later today.  I'm seeing core temps at 58-60C 4.4 using the Noctua-D14 cooler.

Dario,

 

With OC at 4.5Ghz HT disabled I get an increase of avg fps with about 20% compared to an OC at 4.2GHz HT enabled. 3DMark06 score with 4.5GHz is 37 800 and with 4.2GHz 36 400.

 

I'll guess that a higher clock will with HT disabled will allways beat a lower clock with TH enabled. Better keep HT disabled and go for a higher clock.

Dario,

 

Thanks for your reply. How much (some benchmark or so) will performance improve with HT enabled compared to an increasd OC with HT disabled?

 

I have an OC profile 4.2GHz with HT enabled and a profile 4.5GHz with TH disabled. I'll test the difference and see if I see any difference in smoothness and avg fps. But i doubt it. The placebo effect is very strong on the majority of people.

 

 

Dario,

 

With OC at 4.5Ghz HT disabled I get an increase of avg fps with about 20% compared to an OC at 4.2GHz HT enabled. 3DMark06 score with 4.5GHz is 37 800 and with 4.2GHz 36 400.

 

I'll guess that a higher clock will with HT disabled will allways beat a lower clock with TH enabled. Better keep HT disabled and go for a higher clock.

 

I haven't run any FPS comparisons to be honest. My HT profile runs @ 4.7GHz and even though it's just 100MHz, I still prefer the faster no HT 4.8GHz for my FSX.

Wanna make it very clear I wasn't not recommending turning on HT, just saying that it works and in some cases it might be worth considering to give it a try.

If you see a 20% drop in FPS with HT on, that must be that the main scheduler is now sharing a physical core with a texture loader I guess. You could try an AF = 254 to avoid that.

 

and why on earth 3dmark06? 

If you see a 20% drop in FPS with HT on, that must be that the main scheduler is now sharing a physical core with a texture loader I guess. You could try an AF = 254 to avoid that.

I'll give it a test.

 

and why on earth 3dmark06?

IMO 3DMark06 is a good test on how fast your system and graphics card are with DX9. A higher score with 3DMark06 will translate to a better FSX performance.

If you see a 20% drop in FPS with HT on, that must be that the main scheduler is now sharing a physical core with a texture loader I guess. You could try an AF = 254 to avoid that.

Nah, no change in avg fps. Would maybe have an effect if i would have a lot of active background processes.

 

No need for my to fiddle with the affinity mask. FSX is very smooth and nice with my system. I did a lot of tweak testing with my previous system a couple of years ago, but nowadays I don't spend time tweaking. Have my system and FSX setup following Nicks excellent advice.

 

The only future change to my system would maybe be to replace my monitor.

 

Thanks for all your good posts and good advice here at AVSIM, Dario!

 

 

Starting full AIDA64 now ... finishing up with an hour of Prime95 and saving the best for last ... Intel Burn Test V2.54.

 

Charles, for what it's worth I'd recommend running Prime95 for longer than an hour.

 

I've got mt overclock stable at 4.5 now (will push for higher still), but when playing around I was able to pass everything but Prime95 at up to 4.7Ghz (temps were a little high due to no delid). Take for example at 4.6Ghz I would pass everything else but Prime95 would fail after 4-5 hours. I'd recommend running atleast 12 hours.

Ant:

 

Will do ... have it running right now ... temps are staying in the mid 60s.

Ah ... it crapped out between 5 and 7 hours ... (went to watch a movie with my son ... World War Z) ... came back and the computer had a white screen.  It didn't blue screen ... it just "went away".

 

I've also had Prime95 itself crash on occasion with the message "Prime95 stopped functioning" ... or some such message as that.

 

I appears that I'm having difficulty establishing even a mid-level steady overclock.

 

I can't figure out what the problem might be ... the temps are good.  Maybe when NickN finishes up his Chapter 103 it'll shed some light on the problem.

 

 

I appears that I'm having difficulty establishing even a mid-level steady overclock.

 

Don't let it get you down CT just revel in the fact you can run FSX well enough to really enjoy it!