Nvidia teases a 21 day countdown to unveil or release of the RTX 3000

[dburne]

5 minutes ago, chiliwili69 said:

 

The 47 seconds benchmark (which is temporary, SYN_Vander will provide final version) can be run in two modes: Stress CPU or stress GPU.

If you are going to test the 3090 vs 2080Ti, you should use the setting to stress GPU.

 

 

Ah perhaps I did not have that part set properly, let me try again cause based on GPU temp it was like it was hardly using it.

 

[chiliwili69]

6 minutes ago, dburne said:

Now the strange part on the IL-2 Benchmark mission

2080 Ti FTW3 - min 48  max 151  avg 85

3090 FTW 3 Ultra - min 46  max 153  avg 86

 

Yes, this is waht I said. You have probably tested the CPU test (in 1080p, low clouds, no AA, etc). Since your CPU is the same your results are indentical.

There is another post where I am collecting feedback to ellaborate the testing procedure which will created in a new thread with all instructions.

[dburne]

1 minute ago, chiliwili69 said:

 

Yes, this is waht I said. You have probably tested the CPU test (in 1080p, low clouds, no AA, etc). Since your CPU is the same your results are indentical.

There is another post where I am collecting feedback to ellaborate the testing procedure which will created in a new thread with all instructions.

 

I don't see a cpu versus gpu test selection?

[chiliwili69]

The mission benchmark when run in CPU stress mode was having my 1080Ti loaded at 30-40%.

So your 2080Ti was loaded probably at 25%, and your 3090 would be 15% perhaps. That´s why your temps and fans were low

1 minute ago, dburne said:

I don't see a cpu versus gpu test selection?

 

The misson is exactly the same, all depends the graphics setting you use.

 

For CPU stress (no GPU stress) you use: monitor 1080p, Ultra Preset, all maxed out with no mirrors, low Clouds, no AA, no HDR, No SSAO

 

For GPU stress (no CPU stress)  you use: monitor 4K, LOW Preset, all to minimun, with Extreme Clouds, MSAAx8, no HDR, no SSAO

 

see this post.

 

If you don´t have a 4K monitor you can try to use a 4K TV or I don´t know how to force IL-2 to run in a higher resolution than the one soported by the monitor.

[dburne]

I don't think you are understanding me, I used the exact same settings that I used when I tested with the 2080 Ti, the 3090 seemed

to go on vacation, no boost and temp of 32c during the run. The 2080 Ti did not do that.

[SYN_Vander]

58 minutes ago, chiliwili69 said:

 

Yes please, we wil see then when all building are loaded.

Congrats for your new beast!

 

The 47 seconds benchmark (which is temporary, SYN_Vander will provide final version) can be run in two modes: Stress CPU or stress GPU.

If you are going to test the 3090 vs 2080Ti, you should use the setting to stress GPU.

 

 

Here is v4. Keeping it simple...always works best

 

Benchmarkv4.zip

[dburne]

15 minutes ago, SYN_Vander said:

 

Here is v4. Keeping it simple...always works best

 

Benchmarkv4.zip 144.88 kB · 0 downloads

 

How many seconds is this one?

[SYN_Vander]

19 minutes ago, dburne said:

 

How many seconds is this one?

 

should be same 47 s

[sevenless]

1 hour ago, dburne said:

Now the strange part on the IL-2 Benchmark mission

2080 Ti FTW3 - min 48  max 151  avg 85

3090 FTW 3 Ultra - min 46  max 153  avg 86

 

I don´t think it is strange at all. If you did that test at 1080P as far as I understood things there is no benefit to be expected from the RTX 3090. That benefit will only become visible above 1440p and 2080p. Indeed your results tell me that all I need for my personal setting @1080p and perhaps in a few years @1440p a RTX 3070 card will be more than sufficient.

[dburne]

V4 a little lower for me.

Frames: 3786 - Time: 47000ms - Avg: 80.553 - Min: 44 - Max: 106

 

[sevenless]

On 11/8/2020 at 4:09 PM, chiliwili69 said:

.

Also, another thing. Could you make the mission a little bit longer. The variability effects will afect more in a short track.

Something around  80 seconds instead of 47.

 

Now that is interesting. This is my quick&dirty bench with my ROG STRIX 980 on a Asus Hero XI board with 9900k @ 5.0 Ghz and 32Gigs RAM. That tells me in this game @1080p CPU is king.

 

Edited November 9, 2020 by sevenless
added system specs

[Jaws2002]

2 hours ago, chiliwili69 said:

I don´t know how to force IL-2 to run in a higher resolution than the one soported by the monitor.

 

 If the monitor doesn't support 4k, you can create custom resolution in NVidia control panel. Once you create and save the 4k custom resolution, it will show up in game as an option.

[Alonzo]

2 minutes ago, Jaws2002 said:

 

 If the monitor doesn't support 4k, you can create custom resolution in NVidia control panel. Once you create and save the 4k custom resolution, it will show up in game as an option.

 

Yes exactly.

 

Right click desktop > NVidia Control Panel

Manage 3D Settings > Global Settings

Set DSR Factors as required.

Start IL2, select a higher than monitor resolution

 

[dburne]

Been working on my GPU overclock on my new 3090 FTW3 Ultra.
Looks like I am settling in at +169 on the GPU clock, looking good stressing it in Time Spy.
Looks like GPU temp topping out around 60c. Not bad at all.
Max Power draw total on my rig I am seeing is around 670w at peaks. EVGA T2 Titanium 1000w seems to be handling it like a pro.
Will work on the mem clock some tomorrow.

 

[Gambit21]

29 minutes ago, dburne said:

Max Power draw total on my rig I am seeing is around 670w at peaks. EVGA T2 Titanium 1000w seems to be handling it like a pro.
Will work on the mem clock some tomorrow.

 

 

I would hope so.

I went with a 750w

[TheSNAFU]

Congrats on the 3090 dburne.  Looking forward to a full assessment of it actually playing the game in VR. 

[Bernard_IV]

3080 + 4 core intel.  650w power supply crashes with a processor overclock.  With processor stock speeds I was able to get the 3080 to 2100mhz or so.  Power did seem unstable.  I'll see if I can get more speed with more power.  1000w power supply looks to be the way to go for this.

[Gambit21]

56 minutes ago, Bernard_IV said:

3080 + 4 core intel.  650w power supply crashes with a processor overclock.  With processor stock speeds I was able to get the 3080 to 2100mhz or so.  Power did seem unstable.  I'll see if I can get more speed with more power.  1000w power supply looks to be the way to go for this.

 

Yeah - that’s way too much.

Overkill by a big margin for that system would be 750w

[Bernard_IV]

26 minutes ago, Gambit21 said:

 

Yeah - that’s way too much.

Overkill by a big margin for that system would be 750w

You're misunderstanding my post.  650w isn't enough to overclock both my processor and my card.  750w might do it but at that point just buy a 1000w for future upgrades and to ensure clean delivery.  750w might be adequate but certainly isn't overkill.

[Alonzo]

19 hours ago, SYN_Vander said:

Here is v4. Keeping it simple...always works best

 

Benchmarkv4.zip 144.88 kB · 3 downloads

 

This, in VR, absolutely crushes my CPU. Like 12ms frame times on an 8086K @ 5.1ghz. Is that intentional?

[dburne]

1 hour ago, Bernard_IV said:

3080 + 4 core intel.  650w power supply crashes with a processor overclock.  With processor stock speeds I was able to get the 3080 to 2100mhz or so.  Power did seem unstable.  I'll see if I can get more speed with more power.  1000w power supply looks to be the way to go for this.

 

For a 3080 a  850 watt or so would be more than enough. Certainly 1000 would be plenty. 1000 is what I have been running for some time.

Get a good quality supply though. 

My total system with a 3090 max I have seen the system draw was 690 Watts, but these 3090's are power hungry.

Edited November 10, 2020 by dburne

[SYN_Vander]

51 minutes ago, Alonzo said:

 

This, in VR, absolutely crushes my CPU. Like 12ms frame times on an 8086K @ 5.1ghz. Is that intentional?

 

Must be the AI. Try recording it and then measure performance of the playback.

 

With my Rift-S I had the same experience though: The best way for me was to create a coop mission instead of a single mission and run it on my dedicated server (in same LAN). Then the framerate was much better.

Edited November 10, 2020 by SYN_Vander

[Gambit21]

13 hours ago, dburne said:

 

For a 3080 a  850 watt or so would be more than enough. Certainly 1000 would be plenty. 

 

Ya think?

[THERION]

14 hours ago, dburne said:

 

For a 3080 a  850 watt or so would be more than enough. Certainly 1000 would be plenty. 1000 is what I have been running for some time.

Get a good quality supply though. 

My total system with a 3090 max I have seen the system draw was 690 Watts, but these 3090's are power hungry.

 

1 hour ago, Gambit21 said:

 

Ya think?

 

It's some kind of hilarious and priceless to mention a 1000 Watts power supply, if you keep in mind that the EU commission decided in September 2017 to limit the

power for new vacuum cleaners at 900 Watts max. to reduce the domestic consumption of energy. So our mummies and wives have to battle the household with
900 Watts at max., whereas the sons and big boys enjoy their VR experience at 1000 Watts and maybe at 1200 Watts with the next generation of GPU/VR setups.

 

Just saying... 

[Alonzo]

5 hours ago, THERION said:

So our mummies and wives have to battle the household with 900 Watts at max., whereas the sons and big boys enjoy their VR experience at 1000 Watts and maybe at 1200 Watts with the next generation of GPU/VR setups.

 

Maybe you're trying to make a joke but I don't find it very funny. ?

[THERION]

18 minutes ago, Alonzo said:

 

Maybe you're trying to make a joke but I don't find it very funny. ?

 

- No, it's not a joke, there is a EU regulation for this.

- It seems to me, maybe, that you felt some kind of misogynist flavour in my last sentence. I'm not.

  I raised my two sons alone (sole custody) and did all by myself - no help/support wanted.

 

Anyway sorry for that.

 

Cheers

[41Sqn_Skipper]

Honestly just take 850w. Well above the recommended power. It's below vacuum cleaner limit and no one thinks it's excessive.

[chiliwili69]

On 11/9/2020 at 10:22 PM, SYN_Vander said:

Here is v4. Keeping it simple...always works best

Thanks. I will test it. I think that we could use this one to create the new thread.

On 11/9/2020 at 11:11 PM, dburne said:

V4 a little lower for me.

Frames: 3786 - Time: 47000ms - Avg: 80.553 - Min: 44 - Max: 106

Yes, this is because there is no high view at the begining which produce higher fps.

Now all view is from same camera to avoid transition spikes.

On 11/9/2020 at 11:46 PM, sevenless said:

That tells me in this game @1080p CPU is king

Exactly. I we want to know what CPU is best for VR as well.

On 11/10/2020 at 9:09 PM, Alonzo said:

This, in VR, absolutely crushes my CPU. Like 12ms frame times on an 8086K @ 5.1ghz. Is that intentional?

 

The purpose is to run it in two modes in monitor: Only stress CPU or only stress GPU.

Some should apply to VR, but need to test it as well.

 

[Bernard_IV]

Ok I got an EVGA 850GQ because it was at Best Buy and I don't feel like waiting.  I'll try to do some tests.

[Alonzo]

On 11/10/2020 at 1:59 PM, SYN_Vander said:

Must be the AI. Try recording it and then measure performance of the playback.

 

With my Rift-S I had the same experience though: The best way for me was to create a coop mission instead of a single mission and run it on my dedicated server (in same LAN). Then the framerate was much better.

 

I converted the mission to co-op and ran it on the Combat Box secondary server in my basement. There's definitely something odd still happening with CPU frame times in VR. When over the city, frame times are up above 16ms for the CPU (according to SteamVR Frame Timing window), but moving out of the city there's a sudden and very specific drop in the frame times to about 12ms, and then again to about 10ms. No AI running on my rig as far as I can tell.

 

All this has me a little disheartened. I think until we have an actual repeatable benchmark provided by the game, the best we can do is guesses ?

[chiliwili69]

24 minutes ago, Alonzo said:

When over the city, frame times are up above 16ms for the CPU (according to SteamVR Frame Timing window), but moving out of the city there's a sudden and very specific drop in the frame times to about 12ms, and then again to about 10ms

 

Why this is strange?

when you are in the city there are many building and more smoke, so this is CPU intensive, so your frametimes are high. When you go out of the center of the city there is less building and less smoke.

[dburne]

1 hour ago, Bernard_IV said:

Ok I got an EVGA 850GQ because it was at Best Buy and I don't feel like waiting.  I'll try to do some tests.

 

Not particular familiar with the GQ - EVGA power supplies though are typically top notch.

[SCG_Fenris_Wolf]

1 hour ago, Alonzo said:

 

I converted the mission to co-op and ran it on the Combat Box secondary server in my basement. There's definitely something odd still happening with CPU frame times in VR. When over the city, frame times are up above 16ms for the CPU (according to SteamVR Frame Timing window), but moving out of the city there's a sudden and very specific drop in the frame times to about 12ms, and then again to about 10ms. No AI running on my rig as far as I can tell.

 

All this has me a little disheartened. I think until we have an actual repeatable benchmark provided by the game, the best we can do is guesses ?

We all experience the same in this mission. It's a good stress test though. IL-2 does weird things from time to time, somewhere in the code they have a line that triggers AVX, and so on. We should find that line, should be possible by force-crashing the game right? I've read somewhere there'd be a way to gain insight. Would be great if they removed it.

 

Anyway: A good benchmark is not something that runs fluid, especially considering that VR has a cap on framerate logged. Or is there a way to capture average frametime over a set span? Maybe fpsVR has such a feature, haven't checked.

Edited November 11, 2020 by SCG_Fenris_Wolf

[chiliwili69]

21 minutes ago, SCG_Fenris_Wolf said:

Or is there a way to capture average frametime over a set span? Maybe fpsVR has such a feature, haven't checked.

 

With fpsVR you can have an histogram of the session. And I think they implemented a key to reset the history.

https://forum.il2sturmovik.com/topic/56485-benchmark-for-cpuram-performance-remagen-4002-to-4005/?do=findComment&comment=874714

 

Clearly fpsVR gives you frametimes which are not capped like with the FPS to 90.

[dburne]

32 minutes ago, SCG_Fenris_Wolf said:

We all experience the same in this mission. It's a good stress test though. IL-2 does weird things from time to time, somewhere in the code they have a line that triggers AVX, and so on. We should find that line, should be possible by force-crashing the game right? I've read somewhere there'd be a way to gain insight. Would be great if they removed it.

 

Anyway: A good benchmark is not something that runs fluid, especially considering that VR has a cap on framerate logged. Or is there a way to capture average frametime over a set span? Maybe fpsVR has such a feature, haven't checked.

 

That is interesting.

Would this have any ramification if one has no AVX offset?

[Jaws2002]

Nvidia RTX 3080Ti 20GB of GDDR6X due for release in January. 

Projected price $999.00

 

 

https://www.tomshardware.com/news/nvidia-geforce-rtx-3080-ti-rumored

 

 We'll see if that is going to be enough to keep the performance crown, since 6800xt /6900XT may be a great overcklockers. overclocked 6800xt cards are hitting 2500MHz boost clock and trading blows with 3090. 

 Strong competition, at last. ?

Edited November 12, 2020 by Jaws2002

[Alonzo]

22 hours ago, chiliwili69 said:

Why this is strange?

when you are in the city there are many building and more smoke, so this is CPU intensive, so your frametimes are high. When you go out of the center of the city there is less building and less smoke.

 

It's the magnitude of the difference that is surprising to me. The CPU frame time graph shows quite large "steps" -- a 16-18ms frame time is basically unplayable, the step down to ~12ms gives me 80 FPS if I am lucky, and the next step down to ~10ms is fine for playing on. A 60% (or more) penalty for flying over a city with smoke is really quite a big difference. I'd need to get my "normal" frame times down to the 6ms range to be able to handle combat over a city.

[dburne]

1 hour ago, Jaws2002 said:

Nvidia RTX 3080Ti 20GB of GDDR6X due for release in January. 

Projected price $999.00

 

 

 

Dayem...

 

[chiliwili69]

1 hour ago, Alonzo said:

It's the magnitude of the difference that is surprising to me

 

The city centre is really dense with many rail buildings with a lot of details which are captured by this low flight.

There are also 8+8+8+8 planes flying around, 8+8vehicles and a train.

It is not a typical scenario. 

Even Stalingrad city is much simpler since the buildings a really simple. The V5 of the benchmark is a bit less heavy.

[LukeFF]

1 hour ago, dburne said:

Dayem...

 

 

Yeah, I still am aiming for the RX 6800 XT. No way am I going to put down a grand for a card like that.