Benchmarking Arma with new i5-6600K and 3000mhz DDR4

[forteh 11]

I've upgraded my machine and performance has increased by over 100% since having arma3 so any perceived fps loss has not been apparent :)

[ceeeb 147]

Thanks for your reply. I doubt it's due to my windows install, I recently reinstalled windows 7 and I have it on SSD. I'm running a very bare bones setup in terms of software installed right now, might be about 2 months since last reinstall? I did move ARMA to my SSD though, but it didn't help in terms of frames (as it shouldn't).

But since you are replying in the first place, I could swear that I have lost about 10 frames over maybe the last year or so in ARMA. Has there been a performance drop over time on your part? Back in alpha, I had allmost better performance in ARMA III compared to ARMA II (albeit it was on Stratis of course).

 

Hmm okay, I suggest a clean start as I have also done that recently. It used to help a lot in XP days but windows 7 seems more stable over time. Up until six months ago, I had a GTX 570 and an older SSD, which I replaced when the video card died. I don't think the upgrade helped Arma 3 very much.

Alpha did feel surprisingly smooth, but I was away from home with only a laptop for most of 2013, so I really can't compare either. You can load legacy builds of Arma 3 through Steam, but AFAIK you can only go back to the last milestone build.

[fabio_chavez 103]

meanwhile, i overclocked to 4,5ghz (+600mhz), on the halo benchmarks i get the exact same results (With quality preset "High" in 1920x1080) as with stock speed and a ~2 fps increase on YAAB (now 41fps in avg as opposed to 38 before)...
so far the only significant fps gain i had was due to activating the XMP profiles to run the RAM on 3000mhz 15-15-15-35​.

Maybe i should do a rerun on Ultra e.g. since Halo Altis seems to be maxed out at 60

p.s. considering to revert the oc and undervolt instead since the lame piece of silicone i got seems to undervolt better than to overclock anyway

[forteh 11]

You should be much higher than 60 in the helo benchmark, you haven't got vsync turned on have you?

Assuming you don't have vsync on the overclocked results aren't as high as expected, jumping from 3.5 to 4.5 gives almost 30% more for me compared to less than 10% for you.

[fabio_chavez 103]

You should be much higher than 60 in the helo benchmark, you haven't got vsync turned on have you?

Assuming you don't have vsync on the overclocked results aren't as high as expected, jumping from 3.5 to 4.5 gives almost 30% more for me compared to less than 10% for you.

Stock turbo is 3,9 ghz, so the oc is 600mhz, generally i do experience higher fps ingame, however with these benchmarks and videosettings on the "high" preset, i get these results... GPU load is kind of maxed most of the time, the question would be if other people would get different results with same conditions of one or a few different hardware parts, i think what mostly kills fps for me now are the fast camera movements through towncenters, due to the nature of that issue, i would expect 600mhz zu help a BIT but not much, that would align with the 2 fps in avg i get on yaab

[forteh 11]

Don't forget that turbo speeds are only attained when certain numbers of cores are utilised. On the 6600k when only one core is used it runs at 3.9ghz, two cores active it's 3.8ghz and all four cores active it's only 3.6ghz.  As arma3 does utilise quad cores then your processor will only by running at 4 x 3.6ghz.  When you overclock you should have turbo disabled so you get 4 x 4.5ghz naturally.

[fabio_chavez 103]

The clocks are permanently on 4,5ghz with ~50% load across all 4 cores according to hwinfo overlay

[forteh 11]

Yes, when overclocked the clocks will all be at 4.5ghz, but at stock speeds the processor will only be running at 3.6ghz.  In my opinion turbo is a marketing gimmick because windows will unload cycles to unused cores, as soon as the core is used it drops the turbo speed even if it is only 1% usage on the spare core.

[I give up 152]

Unfortunately there is no significant performance increase with the 6th Intel generation when compared with previous one.
And DDR4 is total flop, to have the same performance of DDR3 1866 CL7 we need to have DDR4 clocked up 3800 Mhz CL17 also the second and third timings need to go to insanely high values.
Unless you are building a new machine because you really need, there is no reason for a upgrade.
 

[fabio_chavez 103]

The thing about ram timings is not true as far as i understand but whatever... Skylakes are also not too bad but much too expensive... Offcourse i wouldnt have bought it if my old pc wasnt broken

[heavygunner 179]

Ofcourse it's not true, since Bratwurste is known for spreading misinformation.

[fabio_chavez 103]

In terms of hardware evaluation, i prefer my gipsy aunt and her crystal ball over random opinions from the internet anyway :)

[St. Jimmy 272]

The thing about ram timings is not true as far as i understand but whatever... Skylakes are also not too bad but much too expensive... Offcourse i wouldnt have bought it if my old pc wasnt broken

Timings do matter this much http://www.hardware.fr/articles/940-5/cpu-ddr4-vs-ddr3-pratique.html (change the 7-zip test to Arma 3)

Not as much as Bratwurste said but still some effect. CL7 1866 would be likely match CL15 or CL13 2400MHz on that list. DDR3 and DDR4 are equally fast if their clocks and timings match.

[I give up 152]

The only applications that are having benefits from the higher frequencies of DDR4 are the ones that are not dependent of graphics rendering. Maybe this may change with DX12.
There is no single game having benefits with DDR4 when compared with DDR3.
Like i have said, DDR3 1866 CL7 performs better than DDR4 3200 CL 16 (this with games in general and ARMA 3 is included).
This may also change when memory modules manufacturers start to provide modules with higher clocks and lower latency, right now all of them are stuck at 3000/3200 CL 15/16.
With a few specific modules is possible to achieve 3800 Mhz but, also like I have said, CAS (in particular second and third timing) needs to be insanely increased and because of that they are losing the advantage provided by the higher frequency.

[heavygunner 179]

Would something like these 19-25-25-45 4000mhz modules be in the right direction?: 

http://www.gskill.com/en/press/view/g-skill-conquers-4000mhz-speed-barrier-and-releases-new-trident-z-and-ripjaws-v-series-ddr4-memory

[I give up 152]

They have dropped it, The max that you can get from G Skill is 3200 and these ones do not go at those speeds, not even close.
Corsair has ones at 3600 MHz CL 18 and with these you can go to 3800 but like I have said you need to increase timings in particular you need to insanely increase second and third therefore there is no advantage, in fact is the opposite, latency kills frequency.

[St. Jimmy 272]

Would something like these 19-25-25-45 4000mhz modules be in the right direction?: 

http://www.gskill.com/en/press/view/g-skill-conquers-4000mhz-speed-barrier-and-releases-new-trident-z-and-ripjaws-v-series-ddr4-memory

Wow second and third timings are 25. Nice speed and insaine high timings. It's not step in the right direction until the timings are also kept low.

[R3vo 2654]

There has been alot of discussion about timings here in this thread, so I just want to remind you that you can't compare times between ram modules. To do a proper comperison you can use the following equation:

 

time = 1000 / clock speed [mhz] * 2 * timing

This will give you the time in nanoseconds.

 

Example:

 

Module 1:

Clock speed: 4000mhz

Timing (cl): 19

 

time = 1000 / 4000 * 2 * 19  = 9.5ns

 

Module 2:

Clock speed: 3600mhz

Timing(cl):17

 

time = 1000 / 3600 * 2 * 17 = 9,444ns

 

Even though the timing on the second module is lower by 2 it's barely faster than the first module.

[heavygunner 179]

Wow second and third timings are 25. Nice speed and insaine high timings. It's not step in the right direction until the timings are also kept low.

Yeah, I expected this.

I assume that the timings will improve a lot over time, so maybe with Skylake-E, Pascal GPU and some nice DDR4 we will get a stable 60 FPS. :D

 

 

[jumpinghubert 49]

Ofcourse it's not true, since Bratwurste is known for spreading misinformation.

I can confirm this :P

 

@fabio chavez

forteh asked you some postings back if you have v-sync enabled in your benchmarksessions. Its important to know this... ;)

[I give up 152]

Wow second and third timings are 25. Nice speed and insaine high timings. It's not step in the right direction until the timings are also kept low.

And not only those timings but also tRFC, for a game with Arma 3 architecture this one is crucial, needs to be set with insane values.

While with DDR3 around 2000 MHZ, tRFC is set around 70/100, with DDR4 3000 needs to be set somewhere close to 250/300.

If you don't know what is tRFC google it, but basically determines the timing of refresh cycles (lower=better performance).

Also Gskill announced DDR4 at 4131/4226 (no one knows when will be available..).

http://www.legitreviews.com/g-skill-prepares-tridentz-ddr4-4266mhz-and-ddr4-4133mhz-memory-kits_171043

Now look at tRFC, 450/660. This is just insane and also explains why they did not provide benchmarks for these,, surely with these timings the performance cant be good.

[fabio_chavez 103]

I can confirm this :P

 

@fabio chavez

forteh asked you some postings back if you have v-sync enabled in your benchmarksessions. Its important to know this... ;)

 

vsync was off during benchmarking!