Benchmark here (DCS World 1.2.4 A-10C module reqired)

[Presing]

...

 

Plenty of power left in the Titans, but the CPU is a bottleneck.

 

I7 3820 bottleneck?? You are joking right? Can you say how much processor usage you have when you are playing?

It's a game limitation it's a very ... engine using one core for hole game except sound and sound use approximately 5-10% of second core. That I7 can not be bottleneck. Also what is method for AA (MultiSample, Adaptive MultiSample or Supersample) that is very hard hitter for FPS in the all ED products.

 

P.S. No-one from the top do not like to talk about game engine that's why ... are there.

Edited June 14, 2013 by Presing

[Waxi]

Hey Waxi, could you try what you get with stock speed and if it can handle it, a higher clock on the CPU and tell what you get?

 

These are the results using the stock speed of my i7-920 2.6GHz:

 

AVG FPS : 19.15

GPU MEMORY USED : 2157 MB

GRAPHICS CARD : 4096MB Gigabyte GeForce GTX 670 OC (GV-N670OC-4GD)

CPU : Intel Core i7-920 2.6GHz @ 2.6GHz

MAIN MEMORY SIZE : 12 GB

 

 

For your reference, these are the results at 3.2GHz:

 

AVG FPS : 21.70

GPU MEMORY USED : 2139 MB

GRAPHICS CARD : 4096MB Gigabyte GeForce GTX 670 OC (GV-N670OC-4GD)

CPU : Intel Core i7-920 2.6GHz @ 3.2GHz

MAIN MEMORY SIZE : 12 GB

 

COMMENTS: 3200x1200 dual screen (viewport 1920x1200; second screen 1280x1024) with both MFCDs exported to second screen

 

My system gets unstable when going beyond 3.2GHz. Hence, I cannot provide you with results at higher clock speeds.

 

 

I made another test running a single screen at 1920x1200 for better reproducibility:

 

AVG FPS : 19.77

GPU MEMORY USED : 2396 MB

GRAPHICS CARD : 4096MB Gigabyte GeForce GTX 670 OC (GV-N670OC-4GD)

CPU : Intel Core i7-920 2.6GHz @ 2.6GHz

MAIN MEMORY SIZE : 12 GB

 

In all tests I used the graphics settings shown in the attached screenshot.

Edited June 14, 2013 by Waxi

[GKOver]

AVG FPS : 11.5 FPS

GRAPHICS CARD: SLI 2 x GTX-295

CPU : i7-2600

CPU Utilization: Only 2 of the 8 threads have only 50%

 

MAIN MEMORY SIZE: 16GB

MEMORY USED : 4700 ( 11500 still available)

GPU Memory: 896

GPU MEM USED: 98,9% / 886 MB

GPU Utilization: 50-70%

COMMENTS: It seems that SLI is not really working here or the insufficient GPU memory is the bottleneck. I will change some parameters and the drivers. But it is for sure, that I will never get more than 20 FPS with these two cards....

Edited June 19, 2013 by GKOver

[Nealius]

AVG FPS: 16.73

MEMORY USED: 597mb

GRAPHICS CARD: HD7850 2GB

CPU: AMD Phenom II X4 965BE @ 3.68GHz stock

MAIN MEMORY SIZE: 8GB

GPU MEM USED: unknown

 

Why is my memory used so much lower than everyone else's? Also 16fps with only 9 jets is really sad.

Edited June 14, 2013 by Nealius

[GKOver]

Why is my memory used so much lower than everyone else's? Also 16fps with only 9 jets is really sad.

 

Because your settings are very low....

[Nealius]

Well that doesn't make sense. If performance is dependent on CPU/GPU yet the settings only affect memory, then why do people say to use lower settings when they have poor performance?

[Erforce]

Probably because if you're experiencing such a performance with only 9 A10 around you, the game would be unplayable in a high populated scene, or getting through a dynamic cloud (dynamic weather was an issue on my xeons). (just my 2 cents)

anyway, i'm posting screens for my i7 3770 / 670dcu2

 

setting for 26fps

 

If i put MSAA 16xQ or such, i'm down to 23fps.

anyway, while in flight, i still fly @ 40fps, so not a problem.

GPU VRAM rise to 1500Mb, and still @97% load.

 

major problem, is this scene is only 750 objects.

I saw scene with 2000+ objects, and these hurt a lot.

 

Note : i took this screen after the FPS Hit. (first 10 secs i'm at 36FPS)

[Nealius]

Weather hasn't impacted me much. Heavy scenes like the campaigns are about the same.

 

It must be more than clock speed, because my clock speed is the same as some of the i7's out there. Is it the cache maybe?

[pacotito]

The Intel processors are more efficient at processing(higher ipc). Meaning they do more work in a clock cycle than the phenoms. Say the i7 has a 20% ipc than your phenom and yours can do 100 calculations per cycle. The i7 does 120. So the same clock would be faster with the i7

[xracer]

These are the results using the stock speed of my i7-920 2.6GHz:

 

AVG FPS : 19.15

GPU MEMORY USED : 2157 MB

GRAPHICS CARD : 4096MB Gigabyte GeForce GTX 670 OC (GV-N670OC-4GD)

CPU : Intel Core i7-920 2.6GHz @ 2.6GHz

MAIN MEMORY SIZE : 12 GB

 

 

For your reference, these are the results at 3.2GHz:

 

AVG FPS : 21.70

GPU MEMORY USED : 2139 MB

GRAPHICS CARD : 4096MB Gigabyte GeForce GTX 670 OC (GV-N670OC-4GD)

CPU : Intel Core i7-920 2.6GHz @ 3.2GHz

MAIN MEMORY SIZE : 12 GB

 

COMMENTS: 3200x1200 dual screen (viewport 1920x1200; second screen 1280x1024) with both MFCDs exported to second screen

 

My system gets unstable when going beyond 3.2GHz. Hence, I cannot provide you with results at higher clock speeds.

 

 

I made another test running a single screen at 1920x1200 for better reproducibility:

 

AVG FPS : 19.77

GPU MEMORY USED : 2396 MB

GRAPHICS CARD : 4096MB Gigabyte GeForce GTX 670 OC (GV-N670OC-4GD)

CPU : Intel Core i7-920 2.6GHz @ 2.6GHz

MAIN MEMORY SIZE : 12 GB

 

In all tests I used the graphics settings shown in the attached screenshot.

 

The benchtest is really hard on GPU and looking at the below maybe there isn't much more fps to get for you by doing any work on getting yout CPU stable at higher speed. What you have (around 20 pfs) may be what to expect with the i7 920 with that card. Im in the process to install my GTX-780 so then we may see if there is anything left to get in the i7 920. Some scary stories about ppl installing the GTX-780 with the newest drivers 320.18, saying it has KILLED their CPU's!! Gotta read a little here before i do something. Also im going from AMD to Nvida so there is some drivers/regedits that need to be cleaned. Thanks for supplying the testdata. Its good for us to know what is wise to do when upgrading regarding DCS. No point wasting money if there is no useful gaining in it. The way i looked at it performance wise was that GTX-670, 680 and 770 more or less in the same group and GTX-780 and Titan in another. Then i thought ok i'll put in the money ( converted was $910 USD!!) to hopefully be a little ahead of time. The Titan for me would be only for those doing the combo with high resolution, high settings and multiple screens. I just hope that 3 GB mem is gonna be enough for a while.

 

AVG FPS: 26

MEMORY USED: 2863 MB @2400mhz

GRAPHICS CARD: NVidia GTX670 2GB

CPU: Intel i7-3770K @ 3.9 GHz (stock)

VRAM : 1150MB

AVG GPU LOAD : 98%

 

Notes : 1920x1200 WHQL 320.18, all high. no exception.

 

2nd pass OC to 4.8Ghz : AVG FPS : 26.

Edited June 15, 2013 by xracer

[xracer]

Well that doesn't make sense. If performance is dependent on CPU/GPU yet the settings only affect memory, then why do people say to use lower settings when they have poor performance?

 

As the other guys say, your system has a problem with the AMD CPU (note! that is if your system in general is optimized and thats alot to look at in that department too). But i think also that the 7850 could be holding things back GPU wise. In this test you need serious GPU upgrade to gain alot. What you should do is to do a small test. Increase your visuals in options slowly one by one until you start to get close to 2 GB used GPU. If you still get same fps its the 7850 which is maxed out at around 16 fps with that AMD u have. Do the tests to see if we can find out anything usefull. What i think is happening is that DCS shuffle all it can into GPU mem, then its mostly up to the 7850 to keep the fps up. Not so much the CPU in this benchtest.

Edited June 15, 2013 by xracer

[TZeer]

Obviously this benchmark was just made for you to get some kind of idea of what your current setup with its current settings would achive if the missions contain something like this in close proximity. Scary stuff..hehe. Of course to get a strict benchmark on this all should use the same settings, but i guess its relevant anyway as most have the settings the way they want for everyday use. Anyway from the results delivered so far, there is 2 which sticks out:

 

AVG FPS : 43

MEMORY USED : 3,6 GB system memory, 1,8 GB Video Memory

GRAPHICS CARD : 2x GTX Titan

CPU : i7 3820 @ 4,6 GHz

MAIN MEMORY SIZE : 16 GB

 

AVG FPS: 42

MEMORY USED: 1523 MB

GRAPHICS CARD: NVidia GTX580 3072 MB

CPU: Intel i7-2600K @ 4.6 GHz

MAIN MEMORY SIZE: 16 GB

GPU MEM USED: N/A

 

Something doesnt add up here. The 3820 is a better CPU than the 2600K. Same speed here, but 3820 has 10MB L3 cache and that usually is the main thing which makes it better with the same speed. Looking at the Titan SLI even a 3820 is not good enough to utilise a SLI setup with these cards on the DCS engine, it could even be that performance is lower with SLI for him than using a single Titan. Maybe he could comment on that? It looks like to use a SLI setup you have to be 100% sure that the CPU is more than capable for the spesic cards you want to use. For GKOver's setup it's different cause his SLI cards are alot weaker and he gets alot better performance cause his CPU is more than capable for his spesific cards. SLI is more to raise the low end of fps and stabilize fps overall and in most cases if the application supports it, a more than capable CPU will give you generally stable good fps with less peaks doesnt matter what resolution, ultra settings etc. As of that a strong reliable system fps wise.

 

Obviously if the DCS on the 2600K is tweaked alot with LOD changes then its not even point in comparing. Btw. mention LODS. After all i've messed with this DCS sim I have a single conclusion for myself. There is a single main factor which has importance in DCS and that is:

 

LOD AND COMPLEXITY OF SINGLE UNITS. The variation is HUGE on many units, and some like these in this benchmark will bring all but the best systems down. This is the cause of most big variations in fps. You cant really say its the drawdistance or the number of objects. An example would be the infantery man. Small and neat, but a huge resource hog cause of how its designed. There are units in the system which are way to detailed for current general systems.

 

2600K and the 3820 are built upon the same architecture, so they will give similar performance. Yes, the 3820 might be a better CPU on paper. But DCS isn't capable of utilizing those benefits.

 

Regarding SLI, will not matter if I turn it off. Only thing that could have happened is performance going down if the GPU was maxed out.

 

 

I7 3820 bottleneck?? You are joking right? Can you say how much processor usage you have when you are playing?

It's a game limitation it's a very ... engine using one core for hole game except sound and sound use approximately 5-10% of second core. That I7 can not be bottleneck. Also what is method for AA (MultiSample, Adaptive MultiSample or Supersample) that is very hard hitter for FPS in the all ED products.

 

Yes, it's the bottleneck. DCS uses 2 cores, 1 for sound, 1 for the rest. 1 core goes up to about 90+ % when the FPS goes below 60 FPS.

 

It's quite simple really. My SLI setup was moving along at comfy 50-60% usage, so it's not the bottleneck. If you look at the number of objects on some of the screenshots, you can see it goes above 1000. 1000-1200 is the magical number for my system, depending on the complexity of the scene, that is the max amount of objects I can have, before FPS starts taking a dump.

 

IPC and speed > number of cores, for DCS.

[xracer]

2600K and the 3820 are built upon the same architecture, so they will give similar performance. Yes, the 3820 might be a better CPU on paper. But DCS isn't capable of utilizing those benefits.

Regarding SLI, will not matter if I turn it off. Only thing that could have happened is performance going down if the GPU was maxed out.

Yes, it's the bottleneck. DCS uses 2 cores, 1 for sound, 1 for the rest. 1 core goes up to about 90+ % when the FPS goes below 60 FPS.

It's quite simple really. My SLI setup was moving along at comfy 50-60% usage, so it's not the bottleneck. If you look at the number of objects on some of the screenshots, you can see it goes above 1000. 1000-1200 is the magical number for my system, depending on the complexity of the scene, that is the max amount of objects I can have, before FPS starts taking a dump. IPC and speed > number of cores, for DCS.

Hi TZeer, probably alot of truth in what you say here, but i have a little new here.

 

After a tivolitrip with the kids today i've just now installed the GTX-780 and it was a breeze to install. 25 min from start until i was into DCS. Absoultely fantastic card is all i can say! Even on my system. What i get with good settings on the benchmark is 37 fps. No tweaks either now in DCS. After putting on one after the other max setting until all on, full distances and whatever, i only dropped to 32. Love this stuff. Monster card! Guess am not used to high end stuff probably. Am using the 320.11 drivers now until they've sorted out their driverhickups. Haven't tweaked any yet either. All i can say to anyone who are looking to upgrade. If you have the funds and already have a CPU like mine or better and are going for a NEW card. Dont even think about a 770 or less for DCS! While i remember, I do have a Corsair 1000 W PSU in my comp so remember that part too. Quite some money for those too. It does look like i can do a bit on the CPU side, but i also know that i have some very slow settings on mem in BIOS (1046 MHz), plus some other things. I have had the 920 definately at the limit with my EK HFX-240 watercooling which was 4.3 GHz, but stable is around 4.1 only. So i will try to see what i get then on the 780.

 

I just did a little testing regarding memory use on the GPU. I think at least i haven't got that right at all. The memory seems like its allocated and deallocated on the fly. That what it looks like let say if i press F11 all around. Anyway i did get one indication on 3990 MB used, but mostly its below 1 GB even with all settings to full. So this brings me to this.

 

I never use the AERO DESKTOP. My desktop is as simple as it can be. I also have AERO disabled in DCS. Could it be that which play us a trick in the readings in the benchmark when ppl get 2-3...GB in use?? If you dont switch that one off in DCS maybe you waste alot of GPU mem which could have been used in DCS and even some lost performance. Just my 2 cents...

 

AVG FPS: 34 (39 if OC CPU to ~4.0 GHz which is as far as it goes stable)

MEMORY USED: 2808 MB

GRAPHICS CARD: EVGA GTX-780 3GB

CPU: Intel i7-920@ 3760 MHz

MAIN MEMORY SIZE: 6 GB

GPU MEM USED: 870 MB

 

COMMENTS: All settings full except: No Heatblur, 512 MFD, 16x AA, No Civil Traffic, HDR Normal, 1920x1080 res. Low GPU mem use could be NO AERO, And just direct read in cockpit after planes are lined up (no 360 degree lookaround or other function buttons pressed).

Edited June 15, 2013 by xracer

[Nealius]

As the other guys say, your system has a problem with the AMD CPU (note! that is if your system in general is optimized and thats alot to look at in that department too). But i think also that the 7850 could be holding things back. In this test you need serious GPU upgrade to gain alot. What you should do is to do a small test. Increase your visuals in options slowly one by one until you start to get close to 2 GB used GPU. If you still get same fps its the 7850 which is maxed out at around 16 fps with that AMD u have. Do the tests to see if we can find out anything usefull. What i think is happening is that DCS shuffle all it can into GPU mem, then its mostly up to the 7850 to keep the fps up. Not so much the CPU in this benchtest.

 

Surely a 2GB card should be plenty? What is a good program that monitors and saves in graph form GPU usage? I only have one screen so I can't monitor GPU usage as I'm playing, and MSI Afterburner's log files are a cluttered mess that I cannot read/interpret.

 

 

I'm thinking about doing a full reinstall at this point. Vsync off doesn't appear to actually work, as my framerates never go above my refresh rate (60) with Vsync on or off.

Edited June 15, 2013 by Nealius

[xracer]

Surely a 2GB card should be plenty? What is a good program that monitors and saves in graph form GPU usage? I only have one screen so I can't monitor GPU usage as I'm playing, and MSI Afterburner's log files are a cluttered mess that I cannot read/interpret.

 

I'm thinking about doing a full reinstall at this point. Vsync off doesn't appear to actually work, as my framerates never go above my refresh rate (60) with Vsync on or off.

 

Hmm..I did set up some monitoring onscreen as i played using Rivatuner Statistics together with HWinfo64 and Afterburner to control my fan/temp on GPU, but setting up HWinfo64 with that is a bit of work. Can't explain that here, but i think there is a Youtube video on the setup for that. But that was ARMA 3 and DCS of course had to be special there too as i couldn't get the data as onscreen overlay there?? In fact its only Fraps i get to do that in DCS, but only FPS of course. HWInfo64 is the top of the line for logging though. You can fill you entire screen as overlay with all possible kinds of data with HWInfo64 and log it to a file at the same time.

 

Can't you just run Afterburner logging to graph in background then just ALT-TAB to it and pause graph to check after? DeAttach the graph and full screen to see better?

 

And yeah 2 GB should be plenty. Have you checked that VSync is not controlled (locked) by Catalyst Control Panel? Then its no use to set it off in DCS.

 

Thats all i've used so can't help more.

Edited June 15, 2013 by xracer

[pacotito]

I use gpu z. http://www.techpowerup.com/gpuz/

 

It does do a graph although on a short time period. You can log all the readings to a text file though

[Nealius]

I can try pausing it but there's no way for me to view the graph in full screen.

 

And yeah 2 GB should be plenty. Have you checked that VSync is not controlled (locked) by Catalyst Control Panel? Then its no use to set it off in DCS.

 

Catalyst Vsync is set to "On, unless application specifies."

Edited June 15, 2013 by Nealius

[xracer]

I can try pausing it but there's no way for me to view the graph in full screen.

 

Catalyst Vsync is set to "On, unless application specifies."

 

VSync was working on my 5850 earlier today, but cant get to that setting now with the 780 installed, but in registry i have:

 

DalAllowHSyncVSyncAdjustment = 1 (hex) as a leftover

 

Drivers are 12.6

 

But generally your setting should be fine...

[Waxi]

AVG FPS: 34

MEMORY USED: 2808 MB

GRAPHICS CARD: EVGA GTX-780 3GB

CPU: Intel i7-920@ 3760 MHz

MAIN MEMORY SIZE: 6 GB

GPU MEM USED: 870 MB

 

So it seems that the i7-920 is still quite capable (at least in this benchmark). And I was already considering to replace it by Haswell. Maybe I keep my good old i7 for some more time. This is really the CPU with the best price/performance ratio I ever had so far, in particular when considering the time I have it (I bought it when it hit the market). :thumbup:

[xracer]

So it seems that the i7-920 is still quite capable (at least in this benchmark). And I was already considering to replace it by Haswell. Maybe I keep my good old i7 for some more time. This is really the CPU with the best price/performance ratio I ever had so far, in particular when considering the time I have it (I bought it when it hit the market). :thumbup:

 

Yeah, also many games use HT also and i gain quite alot on that compared to the i5's plus it has VT-x and VT-d. An i7 is an i7...almost ..hehe. Now am looking for a cheap i7 950. They are great as they can go higher maybe 4.5 GHz, The 990X beats most modern CPU's still, but still VERY expensive. So this X58 board could still have some life left. Stable as a rock apart from that i overclocked the board so hard that one of the USB hubs is halfway gone.

Edited June 15, 2013 by xracer

[HiJack]

So it seems that the i7-920 is still quite capable (at least in this benchmark). And I was already considering to replace it by Haswell. Maybe I keep my good old i7 for some more time. This is really the CPU with the best price/performance ratio I ever had so far, in particular when considering the time I have it (I bought it when it hit the market). :thumbup:

I got the Haswell i7-4770K and if you are on i7-3770K then there is no point in upgrading to Haswell. If you are coming from older CPU's then go straight to Haswell and pick up the extra 10%'s and its a 4th generation CPU with added features over Ivy and Sandy.

[Nealius]

I redid the test twice. It doesn't make a whole lot of sense to me. Same system:

 

AMD Phenom II X4 965BE @ 3.68GHz, HD7850 2GB, 8GB RAM

 

(LOW settings per first picture)

 

AVG FPS: 18.00

MEMORY USED: 1272mb

GPU MEM USED: 242mb

GPU % USED: 68%

 

(HIGHER settings per second picture)

 

AVG FPS: 16.69

MEMORY USED: 653mb

GPU MEM USED: 992mb

GPU % USED: 68%

 

(EVEN HIGHER settings per third picture)

 

AVG FPS: 15.11

MEMORY USED: 533mb

GPU MEM USED: 1246mb

GPU % USED: 78%

 

I don't get it.

 

1. Why does my GPU not go above 78%? I thought that with Vsync off it was supposed to max out the GPU.

2. My GPU memory is hardly used at all, except for when I used the higher settings and even then it was only about half of what my card has available

3. Why is less memory being used on higher settings?

 

Judging by how little the framerates change, it's obviously not my graphics card. But looking at Performance Monitor my CPU stays around 30-40%. If neither my GPU nor CPU are being maxed out why are the frames so bad?

Edited June 16, 2013 by Nealius

[pacotito]

Dcs only uses 2 cores one for sound and 1 for the rest. So 30-40% CPU usage is maxed out for the sim engine. Check the individual core usage and you should see one maxed.

[xracer]

I redid the test twice. It doesn't make a whole lot of sense to me. Same system:

 

AMD Phenom II X4 965BE @ 3.68GHz, HD7850 2GB, 8GB RAM

 

(LOW settings per first picture)

 

AVG FPS: 18.00

MEMORY USED: 1272mb

GPU MEM USED: 242mb

GPU % USED: 68%

 

(HIGHER settings per second picture)

 

AVG FPS: 16.69

MEMORY USED: 653mb

GPU MEM USED: 992mb

GPU % USED: 68%

 

(EVEN HIGHER settings per third picture)

 

AVG FPS: 15.11

MEMORY USED: 533mb

GPU MEM USED: 1246mb

GPU % USED: 78%

 

I don't get it.

 

1. Why does my GPU not go above 78%? I thought that with Vsync off it was supposed to max out the GPU.

2. My GPU memory is hardly used at all, except for when I used the higher settings and even then it was only about half of what my card has available

3. Why is less memory being used on higher settings?

 

Judging by how little the framerates change, it's obviously not my graphics card. But looking at Performance Monitor my CPU stays around 30-40%. If neither my GPU nor CPU are being maxed out why are the frames so bad?

 

We got to understand a couple of things here. The field "MEMORY USED" is supposed to be the total main memory used (DCS+Windows w/apps) which you have at around 1.2 GB max. That is impossible since DCS alone take that amount. This should be above 2 GB unless something really strange is goin on. Also talking about CPU/GPU usage. There is Core usage and CPU usage. CPU usage when reading bottom of let say performance tab in task manager, when it shows 50% on a 4 core when running DCS then 2 cores is almost at their peak which is all that CPU will be able to deliver to DCS, Note that is in an ideal situation when no other apps get timeshare. So in general you wont see more than 50% there unless there is some other app hogging the system too. The value will most often be a a quite low percentage unless running things like SuperPI or IntelBurner stuff. The point is how maxed is the exact core in use. Also the reason you may see low GPU usage is that maybe the CPU cant supply enough data for the GPU (Benchmark.miz should be demanding) OR the GPU can push on (view straight up) should give GPU load when Vsync off. But someone else needs to verify that. There are some factors involved in trying to get a concise understanding of GPU usage, since u have GPU load and GPU utilization etc. "View up" will give load but little utilization. Try and learn and use HWInfo64 together with MSIA and Rivatuner SS if you havent (they can be connected all together). They will give you most answers about your systems use. Check attached values for me when running benchmark.miz. Values seems reasonable, with my less powerfull CPU. Remember this is also on a default DCS install. No mods :thumbup:

Edited June 16, 2013 by xracer

[GKOver]

@Nealius: I had the same behaviour and now I have COMPLETELY reinstalled the graphics driver. The SLI was not (really) working, so only one graphic card was really used.

 

So, may be in your system SLI is not working as designed....

Or the CPU is the bottleneck in your system. To find out if your CPU is the problem you have to check if there is at minimum 1 core with a CPU utilization of 100%.

 

The total CPU utilization is not important for being monitored: there can be a total CPU utilization of about 12 % (1/8 of you CPU-cores including the virtual cores) and your CPU is already the bottleneck.

Edited June 18, 2013 by GKOver