2016 Hardware Benchmark - DCS World 1.5.x

[BitMaster]

Yes, but I finally settled @ 1.34 Volts for now as it seems more stable and in my kind of use cool enough as I dont do Prime95 for money or joy.

 

The series of pics shows a walk through prime95 with all monitors opened so you can see how the temp goes from mid 70s to 90s depending on what the cores have to crunch.

 

The system is in full cooling mode, Tower and ext Radiator are at max rpm.

 

The caps are air cooled, I have the wc kit but didnt dare to take those coolers off and the NB wc set wouldnt fit due to the 980GTX directly 2mm above it.

 

Still...I am amazed. This CPU is getting better with time.

 

It ran as a WinSRV in a Rack for 3 or 4 years 365 and overclocked directly to 4.2 with stock cooler when I took it out to build my gaming rig. So I though, wc it and see how far you can go.

 

I directly managed to Auto-OC to 4.84 GHz stable with 1.44V through the Asus tool, but it also raised FSB to 103, which I dont like. So I set back to 4.8/100x48 @ 1.35V which is very stable even when 100% load for hours, this setting is idiot proof.

 

The 5.0 aint idiot proof, I would not do 30 videos with handbrake on all cores and walk away, that software crunches my mac to its kness and it will make this one bow too over time/night. For DCS 5.0 is ok as an Overdrive for a few hours as you can see on the temps.

 

With 4.8 and 1.35 temps are almost 20°C lower in prime95.....and only for DCS I torture this beast, where frames are everything.

 

For rippin' videos or mp3, terrabytes of backup-sql noodeling I do 4.8 and are safe 24/7/365

 

5.0 is on the edge tho it runs stable enuff to fly DCS and prime95 if really needed and wanted.

 

 

I got them both for free, cpu & mobo :) that is the best of all

 

Bit

 

*edit:

 

you can see form 1 - 6 how I lowered voltage from 1.36 to 1.34.

 

This Voltage is not what is actually fed to the CPU, that goes up to 1.6x Volts in my case and as low as 1.02V when @ 1.6GHz idling or YouTubing.

 

Still lowering that value by from 1.36 to 1.34 saves you 8 °C at peak load around where it matters if it is 85-87 or 89-92

Edited February 13, 2016 by BitMaster

[nap0leonic]

more and more results proven that broadwell with l4 cache are the king of performance clock/clock

 

goddamnit intel, you should release broadwell-e with l4 cache, those things will be monsters for enthusiasts that play this kind of game and other cpu heavy games like elite, star citizen, even cities skylines... unfortunately it remain a dream :mad:

 

also poor AMD, their GPUs are slaughtered in this game, this cement the fact that if you are playing many non mainstream AAA games, stay away from AMD because their GPUs are likely to be poorly unoptimized, that and the inherent disease they have under DX11 and lower : CPU overhead problem.

Edited February 17, 2016 by nap0leonic

[BitMaster]

more and more results proven that broadwell with l4 cache are the king of performance clock/clock

 

goddamnit intel, you should release broadwell-e with l4 cache, those things will be monsters for enthusiasts that play this kind of game and other cpu heavy games like elite, star citizen, even cities skylines... unfortunately it remain a dream :mad:

 

also poor AMD, their GPUs are slaughtered in this game, this cement the fact that if you are playing many non mainstream AAA games, stay away from AMD because their GPUs are likely to be poorly unoptimized, that and the inherent disease they have under DX11 and lower : CPU overhead problem.

 

 

clock/clock comparison is perfect, just show me a CPU without extra UP-price tag that will run at least at 4.6GHz, better more, to outrun a 2600k @ 4.8-5 in performance.

 

Said in plain words, for 5-10% more performance I wont pay 1500€ for CPU-MOBO-RAM combo.

 

Intel, you gotta come with some better stuff, another milestone release like SandyBridge was, afaik the next round again is Milestone round, gotta look up Intel's roadmap, they list it somewhere.

 

Milestone releases have far more new features and gimmicks, releases in between milestones are what you know as anything past SandyBridge....maybe the 2011v3 are Milestones...could well be.

 

* This milestone talk is directly from Intel, that's how they describe their strategy in releasing

Edited February 17, 2016 by BitMaster

[*Rage*]

Intel Milesone CPUs

 

Core2Duo - August 2006

i7 920 (Nehalem) - November 2008

i7 2600K (Sandy Bridge) - January 2011

 

All superb generational increases in processing power, IPC and clockspeed. Everything else released has been incremental and not really worth the outlay (from a general use/gaming POV at least) over a 2600K system.

[BitMaster]

Intel Milesone CPUs

 

Core2Duo - August 2006

i7 920 (Nehalem) - November 2008

i7 2600K (Sandy Bridge) - January 2011

 

All superb generational increases in processing power, IPC and clockspeed. Everything else released has been incremental and not really worth the outlay (from a general use/gaming POV at least) over a 2600K system.

 

Thanks !

 

marketing aint it all

[Ganesh]

Ok, here are my 2 cents:

 

today i´ve bought a used 780ti and did some fast tests compared to my existing 390X

 

With the latest 1.5.3 i choose the A-10C Instant Action (hard) russian lakes spring mission.

 

Going through the F-keys after the obligatory zoom out

here my results:

 

390x/780ti

F1 - 116/124

F2 - 120/196

F3 - 105/174

F4 - 100/160

F5 - 130/200

 

The F11 views (Airports) are all near the double e.g.:

Sochi 90/190

Batumi 52/100

Gelendzhik 111/210

 

The main menu cloud-render is total contrary 170/28

Edited February 19, 2016 by Ganesh

[Ganesh]

Then i did the "same" with Nevada 2.0.1

 

I choose the Instant Action Vegas Tour & Nevada CAS again with A-10C

 

This Time the difference is not so much only between 10-20 FPS

390X/780ti

 

F1 - 80/91 & 136/145

F2 - 103/110 & 153/165

F3 - 40/55 & 170/200

F4 - 60/80 & 122/150

 

F11

McCa. 43/53

Nellis 81/114

C130 on Nellis 49/64

 

Edit: on F7 view (vehicles) Weapon Practice & Nevada CAS it´s not possible to figure it out

+/- 5 FPS randomly on each side, so it seems there is no Winner in 2.0

Edited February 19, 2016 by Ganesh

[Ganesh]

A few more tests... all as before 390X/780ti

 

A10C RedFlag Campaign:

 

F1 125/126

F2 145/162

F4 97/113

F5 180/220

 

F2 scroll through Aircrafts:

139/150 108/113 126/133 151/151 214/224 162/179 188/200 186/190 158/165 169/179 150/150 124/130 146/158 150/163 146/158 133/152 122/133 136/140 128/136 127/138 185/197 196/200 and so on...

 

Mig 19 Nevada Intercept:

 

F1 90/113

F2 114/130

F3 88/112

F4 78/98

F11: Creech 113/153 Groom 122/173 McC. 41/52 Nellis 84/117

 

Thats enough for today and i don´t know what i should think about this "fast-test-results"? :huh: No such things like particle effects (e.g. flaredrops, weapon releases) were made.

 

Important questions remains unfortunately open:

Are 3GB VRAM enough for DCS? My sold 390X had 8GB

Edit: Up until now it seems so :)

 

My confirmation is: In 1.5. there is a huge difference between AMD & Nvidia. In 2.0 it seems the gap disapears more and more...

Edit: And what the heck about this 140frames "main-menu-render-madness" difference of 170/28? Any conclusions out there?

 

^^ settings for all my tests / nv-driver version 347.52

Edited March 28, 2018 by Ganesh
found a typo

[EAKMotorsports]

I want to say Thank you very much for tiborrr to make an very complex benchmark to DCSw, I always have missed and ask to site and reviewers to make an bench with dcs with no sucess, Now I have and big thanks. great job and I hope the more updates. thank you very much.

[TAW_Impalor]

@tiborrr:

 

Thanks for your effort! I totally agree with your conclusion that for multiscreen and VR there is some bottleneck in the engine. In fact, this bottleneck was discussed half a year ago in relation to A-10C FLIR/MAV bug, and I ran some tests:

 

http://forums.eagle.ru/showpost.php?p=2552722&postcount=212

 

That was with 1.5.0 and 1.5.1 versions of EDGE. Today I ran a similar test to yours to see if my ad-hoc solution still works. I.e. "load DCS with MINIMAL settings and then change them to high in game and launch the mission". I used a 3-minute track of my A-10C flying and used FRAPS, like you did. Here are the results (Min/Max/Avg):

 

1. DCS launched on HIGH/Usual settings, then I set all to LOW/NO and ran the track: 72/159/107

2. DCS re-launched on LOW/NO and I ran the track with these settings: 138/228/187 (75% gain!)

3. Set all to HIGH/Ultimate, re-launched DCS (important!), ran the track with these settings: 17/56/32

4. Now, set all back to LOW/NO, re-launched DCS(!), set all to HIGH/Ultimate, ran : 39/82/52 (60% gain)

 

I know the settings took effect with no rebooting - based on texture quality, shadows, visibility range etc. So, some people's claim that you must always reboot after each change is groundless. I was not surprised to see these results again, as I did 6 months ago. I am surprised nobody else noticed this. I am now going to open a bug ticket and see what the devs have to say about it!

 

So, there is a BIG difference (on my PC) whether DCS is launched with HIGH System settings or with LOW. I can't explain why it matters, but please go ahead and check on your side!

 

Thanks!

 

PS. I use an "old" NVidia driver 332.21 for my MSI 760 OC 2Gb, as all newer ones give me 10x(!) lower FPS on the "clouds" main screen. I see many people have this problem too...

 

UPDATE: Turned out MFDs resolution only applies after reboot, so the high FPS was with 256. Well, I can live with 512...

Edited April 8, 2016 by impalor

[DieHard]

(re: Message #111)

 

Wow, that's worth trying out.

 

I usually fly the Ka-50 on the 104th's server or the WWII German planes on the few servers that exist in MP, though I have most of the DCSW inventory.

 

Thanks.

Edited April 6, 2016 by DieHard

[inSky_1911CFZS]

I've heard the performance issues for ATI GPUs in DCS has been fixed since 1.5 update, but this thread shows that ATIs still got awful performance in the new versions?

That's a shame for ATI card users

[FireFox500]

What settings were used for the 4K GPU architecture tests? 70 FPS + seems very high for 4K and I'm not sure what had to be turned down (if any) to attain that high of a frame rate.

[inSky_1911CFZS]

What settings were used for the 4K GPU architecture tests? 70 FPS + seems very high for 4K and I'm not sure what had to be turned down (if any) to attain that high of a frame rate.

 

Seems the Depth of field effect has been turned off in all testing results.

[FireFox500]

Seems the Depth of field effect has been turned off in all testing results.

Were the rest of settings on max? Or close?

 

Sent from my XT1254 using Tapatalk

[Malefic Rage]

Extent of CPU bottleneck

 

Hello. I know this has been discussed a few times, I just wanted to share the results of some experimentation.

 

Until recently I had an AMD FX-6300 CPU with a Radeon R9 290 and have consistently had issues with FPS. Particularly significant was choppiness and minimum FPS when using TrackIR. Since my video card is relatively high-end, especially for the time DCS was released, and I only use medium settings, I couldn't understand the issue. After my motherboard died I replaced my CPU with an Intel Core i5 6600K which I overclocked from 3.5 Ghz to 4.2 Ghz.

 

After running a test run on the same settings with the same video card and with no AI units my results are as follows. (I also tested with my AMD CPU overclocked to 4.2Ghz. I can't remember the exact results so I've listed approximations based on what I remember.

 

AMD FX-6300 (3.5 GHz, 6 core) MinFPS - 21, MaxFPS - 70, AvgFPS - 38

AMD FX-6300 (4.2 GHz, 6 core) MinFPS - 26, MaxFPS - 80, AvgFPS - 43

Intel Core i5 6600K (4.2 Ghz, 4 core) MinFPS - 60, MaxFPS 120, AvgFPS - 90

 

The difference is spectacular. Basically it means that DCS will not play smoothly on any current-generation AMD CPU unless very modest graphics settings are used (particularly view distance) and only small missions are played. After observing the rendering stats, there seems to be a direct correlation between number of objects rendered and FPS. In hindsight I should also have taken results during a large mission but since my old AM3+ motherboard is dead, I can't any more.

Edited May 20, 2016 by Malefic Rage
Added info re AMD overclock

[Sporg]

Thanks for sharing.

 

You can see the same phenomenon in this post:

http://forums.eagle.ru/showpost.php?p=2624490&postcount=2

(the whole thread is interesting)

 

What "hit" you, is a combination of a lesser CPU architecture, coupled with a lower CPU clock.

 

I am running DCS on an i7 2600K, stock clock 3.8 GHz boost, normally overclocked to 4.3 GHz.

This overclock give me a very noticeable increase in fps, something you would probably be able to see on your system as well, if you test a bit.

Edited May 20, 2016 by Sporg

[Malefic Rage]

Thank you very much for that link. I hadn't come across that before. As for frequency, I should have mention that I tested the AMD CPU with an overclock of 4.2 Ghz. I got about an extra 5 MinFPS but it wasn't that significant. I believe the issue lies with the AMD architecture itself. (I noticed on that link all tests were using an Intel CPU). It will be very interesting to see their new generation coming out later this year. It is supposed to be a brand new architecture focused on per-core performance.

[BitMaster]

Thanks for sharing.

 

You can see the same phenomenon in this post:

http://forums.eagle.ru/showpost.php?p=2624490&postcount=2

(the whole thread is interesting)

 

What "hit" you, is a combination of a lesser CPU architecture, coupled with a lower CPU clock.

 

I am running DCS on an i7 2600K, stock clock 3.6 GHz, normally overclocked to 4.3 GHz.

This overclock give me a very noticeable increase in fps, something you would probably be able to see on your system as well, if you test a bit.

 

 

The i7-2600k has a base frequency of 3400MHz and a Turboboost frequency of 3800MHz out-of-the-box btw, 34/38 Multiplier

 

Using Ka-50 as a reference...I jump from mid 40's to high 60's in fps when switching while in the cockpit. I can see the same scene being rendered almost 50% faster when adding that last 1000MHz on top of Turboboost frrequency. Seems like the first 3-4GHz are for the game to be calculated and anything past 4GHz can be used to actually render faster.

 

This, I guess is the true nature of DCS in my understanding...beyond 4.xGHz comes eyecandy, anything prior to that number is basic calculations you gotta do, if you have cycles to spare..well, thats your extra fps on top of those with 4GHz max CPU's.

 

 

I assume, changing 980GTX to 1080GTX would not bring many more fps but keeping the 980GTX and being able ( in dreams ) to OC that 2600k to 6GHz would gain far more fps ( 1440p based )

[Sporg]

Ah yes, 2600K is 3.4 - 3.8 GHz, that's right.

It's so long time since I ran it at stock frequency, I forgot. ;)

 

I don't overclock so high since it demands more cooling and fiddling with voltage etc. And since DCS runs quite nicely for me at 4.3.

[Bestandskraft]

Does anyone have any experience or speculation as to how

 

1) the new Broadwell-E CPUs, specifically the 6850k and 6900k, would run DCS, especially compared to the 6700k, with and without overclocking?

 

2) an ultra-wide resolution of 3440x1440 would affect performance on a GTX 1080, especially compared to 2560x1440?

 

Does anyone have any good/bad experiences with running DCS in 3440x1440 other than performance, e.g. with regard to advantages/disadvantages of a 21:9 aspect ratio?

[Malefic Rage]

Does anyone have any experience or speculation as to how

 

1) the new Broadwell-E CPUs, specifically the 6850k and 6900k, would run DCS, especially compared to the 6700k, with and without overclocking?

 

First off, if you wanted to change you would need a motherboard capable of handling socket 2011-3 since they don't use the standard 1150/1151 sockets.

 

Contrary to the impressions the marketing gives, the new group of high core count Core i7s are actually likely to perform substantially worse than a standard 6700K in DCS.

 

DCS is not multi-threaded so the main (generally) metric of performance is single core performance. Anything more than 2 cores will not get any use by DCS so that $2100 Core i7 6950X Extreme Edition with 10 Cores is actually doing nothing at all for the money. In fact, at 3.0GHz stock speed, it is only 3/4 as effective as a Core i7 6700K at stock speed. Even a plain Core i5 6600 would be faster at 3.5 GHz.

 

As for overclockability. I don't think you would get anything more than you would a standard Skylake CPU; furthermore you would have less headroom for temperature management due to the extremely high TDP rating which would require some pretty serious cooling.

 

All in all, I'm not sure if these CPUs are optimal for gaming since game developers will develop their products to run on the average consumer's CPU. You'll certainly get benefit from a high-end CPU like a Core i7 6700K but unless you have something with a heavily threaded design which specifically requires as many cores as possible then you probably won't get much out of it. I doubt I would ever buy one myself.

 

Hope it helps. :)

[KansasCS]

2016 Hardware Benchmark - DCS World 1.5.x

 

I currently run with DDR3-1330 memory.

Is it safe to say that an upgrade to DDR3-2400 will see a considerable performance increase?

 

(I deduce this from the 2133 to 2400 Performance gain with DDR4, but I don't Know of this can be applied to DDR3)

 

 

Gesendet von iPhone mit Tapatalk

[Bestandskraft]

@ Malefic Rage:

 

Makes sense, after reading a couple of reviews of the 6700k and the 6800k+ it seems that the 6700k can be overclocked to a higher frequency with a lower temperature much better, even with an AIO water cooling solution which I'm aiming at getting. Thanks.

[SSTerf]

This is some great info. Wondering if anyone can help me. I have not used PC in many years or done any simulation for a very long time. I came across dcs and I am dying to do this. For someone new what is best way for me to figure out what hardware to buy? I only want to buy it once. Can you help me build a rock solid system?

 

I used to be very tech savvy with PC and Windows years ago... Been using Mac so I need to get back up to speed.

Edited June 8, 2016 by SSTerf