Gtx 1080 Slow!!!

[Enduro14]

Well all three versions of dcs it runs awesome in 4K big step up from the 980 in all 3 locked out at 60 until the engine limits with processor kicks in

[adrianstealth]

every DCS version same

 

bad utilisation, ground textures or what ever it is I'm guessing its something pointless need to be switchable off

 

a decent consistent fps is really needed for VR ( & vr is becoming more and more common place )

 

constant dips to 45fps when everything set on low no AA with a very high end system

( gtx 1080 etc ) is a no no

[BitMaster]

 

every DCS version same

 

.....

( gtx 1080 etc ) is a no no

 

 

Dont think so, in 2.x I have considerable more CPU usage due to DX11 handing the work out to multiple cores, thus freeing some of that ONE tortured core away that now has more juice for DCS which results again in higher fps...until again either CPU or GPU is flooded. In 2.x my GPU is the limit, not my CPU. GPU often at 99% usage.

 

When 2.x gets final I may get a new GPU, till then the 980 is plenty fast for 1440p

[Jesper]

Happinness changed fast.

 

Yeah I hope they fix this. I don't understand. Looking straight, 90fps, looking left 45fps :( :(

http://sv.tinypic.com/r/6fmede/9

http://sv.tinypic.com/r/4lfvat/9

 

Why can't I post a pic here? It's just a X instead of a pic?

Edited July 9, 2016 by Jesper

[Hadwell]

you need to put a direct link in to the .jpg.

[Jesper]

Well I did, it worked before but now I only got a box with a X.

Hmm well thanks :)

[sunski34]

GTX1080 MSI Gaming X & DCS V1.5.4 vs V2.0.3

 

Hi,

I had 970GTX before !

I have a 2600K too, and did some tests, with CPU OC (4.5GHz) and without OC (3.8GHz auto)...

 

I confirm that on V1.5.4 (CPU OC or not) fps are sometimes bad (setup trees at 10000) for such a card. No real change between 970GTX and 1080GTX at low altitude, over trees. But if you move your head (track ir) to look in another direction, fps can go up quickly with the GTX1080 (not with the GTX970) even if the scene have a lot of trees too !! (tests done with Mig21b or M2000C, better with F5 but not a lot)

but no real change between 3.8GHz (no OC) and 4.5Ghz (OC)

 

But on V2.0.3, it's always perfect !! The 1080GTX give 2x fps than the 970GTX. Over Vegas, 60 fps is a min (CPU not OC, GPU not OC). All to max in setup.

So I think that V2.0.4 is well optimized for such a GPU.

 

I hope that 2.5 Caucasus map will have same performance as Nevada Map !

 

Wait and see. :cry:

[sunski34]

Rear mirror OFF on 1.5.4

 

Much better fps with setup rear mirror OFF on 1.5.4.

 

No change (ON/OFF) in 2.0.3

 

:thumbup:

Edited August 24, 2016 by sunski34

[dburne]

Hi,

I had 970GTX before !

I have a 2600K too, and did some tests, with CPU OC (4.5GHz) and without OC (3.8GHz auto)...

 

I confirm that on V1.5.4 (CPU OC or not) fps are sometimes bad (setup trees at 10000) for such a card. No real change between 970GTX and 1080GTX at low altitude, over trees. But if you move your head (track ir) to look in another direction, fps can go up quickly with the GTX1080 (not with the GTX970) even if the scene have a lot of trees too !! (tests done with Mig21b or M2000C, better with F5 but not a lot)

but no real change between 3.8GHz (no OC) and 4.5Ghz (OC)

 

But on V2.0.3, it's always perfect !! The 1080GTX give 2x fps than the 970GTX. Over Vegas, 60 fps is a min (CPU not OC, GPU not OC). All to max in setup.

So I think that V2.0.4 is well optimized for such a GPU.

 

I hope that 2.5 Caucasus map will have same performance as Nevada Map !

 

Wait and see. :cry:

 

That has pretty much been identical to my experience as well.

My performance in DCSW2 went up drastically.

1.5 not so much.

[Feuerfalke]

The video posted above doesn't really tell you that much. That's the skylake-chips making a runout against each other and one AMD. If you got the "legendary" 2600K though, you'll most likely running the SandyBrigde, which has a much lower memory throughput than those CPUs. (Effectively almost half that of a modern CPU).

 

Since the GTX 1080 scales pretty dramatic with CPU and memory adressing, the 2600K is most surely the bottleneck. That has not even to do with maximum usage, but with physical limitations of the chip.

 

 

But it strikes me, that you are limited to 50 fps and mentioned that the 970 was in the same area. Are you sure that you have vsync disabled?

50 fps sounds a lot like vsync enabled on a flat-screen-tv via standard-HDMI cable.

[sunski34]

Perhaps but not sure... that the CPU is the problem (or Sandy Bridge).

 

Yes Vsync is OFF. And in 1.5.4, turning off rear mirror option solve my pb even if the GTX1080 isn't in full use !

 

The problem doesn't exist in V2 which seens to use much GPU capabilities. So I will see with 2.5 !

Edited August 24, 2016 by sunski34

[TAW_Impalor]

Same experience. Returning my Strix 1080 after two weeks. On the low settings it needs to run DCS at 90 FPS my old MSI 760OC also runs at 90 FPS! Will wait couple of years to upgrade my whole rig to next gen CPU/RAM, or until DCS optimizes for VR... (((

[BitMaster]

The video posted above doesn't really tell you that much. That's the skylake-chips making a runout against each other and one AMD. If you got the "legendary" 2600K though, you'll most likely running the SandyBrigde, which has a much lower memory throughput than those CPUs. (Effectively almost half that of a modern CPU).

 

Since the GTX 1080 scales pretty dramatic with CPU and memory adressing, the 2600K is most surely the bottleneck. That has not even to do with maximum usage, but with physical limitations of the chip.

 

 

But it strikes me, that you are limited to 50 fps and mentioned that the 970 was in the same area. Are you sure that you have vsync disabled?

50 fps sounds a lot like vsync enabled on a flat-screen-tv via standard-HDMI cable.

 

Run a benchmark against your RAM and post the results. I honestly doubt what you say as a DUAL CHANNEL DDR3-XXX has the same bandwidth on 2nd gen as on 3rd or 4th generations i7's.

 

In fact, when I tested my RAM at 2133 it was almost on par with DDR4-2400 due to its higher latency on DDR4. YOu need really high clocked DDR4 to make the change worth.

For same or almost same speed DDR3 is usually crisper ;)

 

Single Core Performance of a 2600k is still on par or just a tad below with ANY CPU released by

Intel since. Charts and screenshots that proof this are plenty on this forum, incl.my own.

[Hadwell]

not really... the 2600K is like... 30% slower than a 6700K at the same MHz, that's not enough of an increase to justify an upgrade in my opinion...

[BitMaster]

not really... the 2600K is like... 30% slower than a 6700K at the same MHz, that's not enough of an increase to justify an upgrade in my opinion...

 

yep....and usually you will never ever get a 6700k clocked that high and thus the SandyBridge can almost close the gap by just turning like nuts :)

 

Also the RAM....DDR4 is same or even slower on same MHz...you need the FAST DDR4 modules to make it worth, like 3000+ MHz. Various tests show this very clear.

With the same speed DDR3 is faster.

[BitMaster]

not really... the 2600K is like... 30% slower than a 6700K at the same MHz, that's not enough of an increase to justify an upgrade in my opinion...

 

just you wont get a 6700k clocked that high, so bottom line, the 6700k wont pull away if you can OC the Sandy well.

[Hadwell]

just you wont get a 6700k clocked that high, so bottom line, the 6700k wont pull away if you can OC the Sandy well.

 

I'll rephraise, the 2600k at 4.6 GHz is the same speed as a 6700k at 3.2 GHz. 2600k at 5ghz is the same speed as a 6700k at 3.6 ghz...

 

Anyway the 2600k compared to the 6700k is today like what the amd FX chips were when the 2600k came out IPC wize

Edited August 25, 2016 by Hadwell

[BitMaster]

Are there any single thread benchmarks available for the 6700k ??

 

Cause the 2600k @ 5GHz is faster than a 4970k @ 4.5 GHz, tested and published here in forum.

 

So I really really doubt that the 6700k is that much faster, wouldnt that be the 30+% we waited for.

 

How high can you o/c a 6700k before the lesser quality heat-transfer layer goes up in smoke ?

 

Anyone having a 6700k clocked sky-high ??

 

 

edit::

 

4790k faster single threaded than 6700k : https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7-6700K+%40+4.00GHz

 

Haha... ooops... maybe not such a good CPU if compared !? the 6700k is sweating to keep up with the 4790k...and at 4GHz it is inferior to the 2600k.

 

You have to o/c that machine just as I have to mine to gain a leap ahead, if at all. The 6700k is a lesser CPU internally, the heatplatter is not of the same quality as the x99-2011v3 or Sandies.... not much headroom for cooling, unless you unpack the core and go HARD CORE cooling on a naked die.

Edited August 29, 2016 by BitMaster

[msalama]

To whom it may concern -

 

http://www.cpubenchmark.net/common_cpus.html

 

The latest PassMark comparison table for your perusal. Enjoy :)

[Penshoon]

To whom it may concern -

 

http://www.cpubenchmark.net/common_cpus.html

 

The latest PassMark comparison table for your perusal. Enjoy :)

 

The single thread performance table should be more useful if main application is DCS?

http://www.cpubenchmark.net/singleThread.html

[aairon]

The single thread performance table should be more useful if main application is DCS?

http://www.cpubenchmark.net/singleThread.html

 

You're right the Passmark CPU mark tests many other things that don't reflect the way it's used here. The single core mark seems to be the most accurate way to tell how the CPU will actually perform in this sim world.

 

According to the info in that first chart PassMark CPU MARK.

 

Single threaded performance on the top three listed there goes as follows.

i7 4930k 1997

i7 5820k 2008

i7 3930k 1934

i7 4790k 2527

So we see that best overall cpu mark has little to do with single threaded performance.

Edited August 29, 2016 by aairon

[pacotito]

You're right the Passmark CPU mark tests many other things that don't reflect the way it's used here. The single core mark seems to be the most accurate way to tell how the CPU will actually perform in this sim world.

 

According to the info in that first chart PassMark CPU MARK.

 

Single threaded performance on the top three listed there goes as follows.

i7 4930k 1997

i7 5820k 2008

i7 3930k 1934

i7 4790k 2527

So we see that best overall cpu mark has little to do with single threaded performance.

 

It also depends how well it ocs. As you can see here my 5820k@4.5 just beats bitmasters Sandy at 5.0ghz. Judging by the numbers he provided we could assume a Sandy at 4.5 would give somewhere around 2200-2400

https://forums.eagle.ru/showthread.php?p=2863782#post2863782

So once you consider the randomness of overclockablility and the down turn it's taken recently the new processors aren't really much that much faster in terms of single thread. Granted most Sandy's won't hit 5 0

Edited August 30, 2016 by pacotito

[BitMaster]

The only CPU I would now buy was a 4790K from a vendor that may charge a bit more but sells them as "4.5" or "4.8" etc. versions. Would need to see how much they ask more but a non-oc'ed 4790k would make plain no sense.

 

The 7700k looks promising, with 4C/8HT 4.2/4.5 GHz it's already pretty nice clocked. Once available we will see how far you can push the silicon and if we can crack the 3k Passmark hurdle ;)

 

 

Gawwwd, those days with 100°F+ are hopefully over....even WC comes to a halt if you cant AC your rooms.

[Penshoon]

The anticipation of waiting for a new chip to arrive and check how fast it runs is part of the experiance of upgrading a pc for me!

Sure buying from a binning company might make more sense as you will pay for what you get but my inner gambling addict always persuades me to go for a plain boxed chip hoping I will win the silicon lottery :P

[Goldsmack]

Wow....why has ED not looked at this? performance was fine before? I was able to get solid 60fps with everything on High and view distance on ultra. Even people with an I7 and GTX 1080 hare struggling to hit 60fps...and this is not just a couple people with this problem there are countless post about it lately. And I am getting sick of this whole "your 300$ cpu is not good for dcs get this $500 cpu and it should work better..." or "put everything on low"..you/I sill get the exact same frame dip no mater what settings.