Jump to content

Kaby Lake / Z270 RAM sweetspot


Recommended Posts

now look RAM-OC 3866

 

The CPU was @ 5G for the XMP 3600 ( L1 - L3 is faster then ) whereas the 3866 was done with the CPU in default mode @ 4.2/4.5

Unbenannt1.JPG.2dd50491193366f23054994b6dd91c74.JPG


Edited by BitMaster

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

Properly done now with 3866@5G

 

some fluctuations with L2 +/- a few....

Unbenannt2.JPG.308b804c2e1a28511e990be97ac6cc7f.JPG


Edited by BitMaster

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

Properly done now with 3866@5G

 

some fluctuations with L2 +/- a few....

 

I like this 7700K build i'm thinking of

video

 

***edit...my 100MB are full...no more pictures....:(

 

Well that should be extended :thumbup:

Control is an illusion which usually shatters at the least expected moment.

Gazelle Mini-gun version is endorphins with rotors. See above.

 

Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.

Link to comment
Share on other sites

Look at the 3rd picture, the system was nuts !

 

Either Aida64 failed, the Bios or something else but with 5G/3866 it was just as fast as 3600 setting STRANGE INDEED....LoL

 

I didnt check the result, just took a snapshot and posted it. It's NOT what I wanted to show but things happen.

 

I have no idea why it was 4GB/sec slower just because the CPU was faster !???

 

Anyway, the 3866 is an unstable setting that I do NOT use, just did it to show the scaling in bandwidth. Might well be that something clocked down for whatever reason.

 

I am priming it now to make sure it got no errors at my usual settings, you never know !

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

Alright after some bios updates and tweaking here is the bottle neck for me.

 

look at the GPU usage.

 

Normandy_Gazelle.PNG

 

reasonable FPS occasional drop frames overall Normandy and NTTR look great single player.

 

Seems 2.1 is a major upgrade and Normandy map is awesome, I actually get Dory syndrome what with all the detail.

 

Dory syndrome, you know coming in over the tree tops and well seeing something and going and having a look at it, or following a tree line and well seeing something and having a look at it,...mm

 

Just keep flying just keep flying Flying flying,.. :P

 

recursivewhat.png

 

 

<edit>

So recommend 16GB mem 32GB if you can, but graphics the above is with a 980Ti on a 5930@4.2Ghz so looks like a 1080Ti or two,.. :)


Edited by FragBum
<edit>

Control is an illusion which usually shatters at the least expected moment.

Gazelle Mini-gun version is endorphins with rotors. See above.

 

Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.

Link to comment
Share on other sites

When the GPU usage drops it's not the GPU that is too slow, it is your system not feeding your GPU well enough imho.

 

Any GPU should run at 99% unless you cap the fps or have a CPU too slow to feed it if it is the NASA 3080GTX-Double-Ti-Turbo HBM4 version you wanna feed ;)

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

Since the timespan was longer than 1 second...maybe your system did SOMETHING else while you gamed and thus the fps went down ??

 

Show the WHOLE graph with all cores to have a chance to see what happened.

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

Nah I think it's GPU,.

 

System

X99 i7 5930 @ 4.2Ghz

32GB DDr4 @2667Ghz

SSD Intel 750 400GB NVME PCIe

 

Slightly clocked GPU/RAM better, but seriously everything else lazing ATM. :thumbup:

 

Thing is wait or SLI 1080Ti??

Control is an illusion which usually shatters at the least expected moment.

Gazelle Mini-gun version is endorphins with rotors. See above.

 

Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.

Link to comment
Share on other sites

If the GPU was too small you'd never have good fps at all, think logic !

 

If it breaks in randomly it is likely not the GPU, but I dont own Normandie and wont until stable so I cant tell you from 1st hand experience with that map.

 

What output do you use again ?? VR or 4k, 1440p, FullHD ???

 

I would wait until the map is stable and decide from there on. Throwing bucks at a beta stage product gives no garanty that your expenses will result in better performance.

 

just my 2 cents...but hey, if money is not a question, get the 1080Ti ;) ..just maybe it wont change much

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

Okay good call I just had this conversation with my partner and I'm going to wait it out for better VR/GPU :thumbup:

Control is an illusion which usually shatters at the least expected moment.

Gazelle Mini-gun version is endorphins with rotors. See above.

 

Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.

Link to comment
Share on other sites

Output == Rift :thumbup:

Control is an illusion which usually shatters at the least expected moment.

Gazelle Mini-gun version is endorphins with rotors. See above.

 

Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.

Link to comment
Share on other sites

Okay good call I just had this conversation with my partner and I'm going to wait it out for better VR/GPU :thumbup:

 

I know this burning feeling..haha

 

Wait & see is always a good advice but hardly ever followed by youngsters with hot blood and the desire to aquire a new gadget.

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

  • 2 weeks later...

OK so these seem to be my options for dual channel 32GB 3200MHz kits:

 

CAS16

Corsair LPX 16-18-18-36 £275

G.Skill Ripjaws V 16-18-18-38 £250

 

CAS15

G.Skill Ripjaws V 15-15-15-35 £280

G.Skill Trident Z 15-15-15-35 £285

 

I'm tempted to just go for the CAS16 G.Skill at £250 since I don't think the CAS will make any difference in a flightsim(???): 32GB CAS14 kits are over my budget.

Regards, Django.

| BMS | DCS OB | A-10C II | AV-8B | F-16C | F/A-18C | FC3 | Persian Gulf | Supercarrier | Tacview | XP11 | FF A320 | FF 757 |

| I7-9700K + NH-D15 | RTX3080Ti 12GB | DDR4-3200 16GB | Aorus Z390 Ultra | 2X Evo 860 1TB | 850W | Torrent Case |

| Warthog HOTAS + CH Pedals | 32" TV 1080p 60Hz | TrackIR5 |

Link to comment
Share on other sites

The ED devs could/should answer how memory is accessed by DCS

 

 

This articles on page 2 explains latency pretty well in easy terms:

 

http://www.guru3d.com/articles-pages/amd-ryzen-7-agesa-1006-performance-update-review,2.html

 

* you still might be able to lower latncy by UPPING the volts a little...may be worth a shot.

 

Make sure they are listed in your QVL from mobo !


Edited by BitMaster

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

anything above 2133 respectively 2400 for newer CPU`s is considered overclocking !

 

Those infamous X299 boards support up to 4266MHz RAM with XMP profile, which means, if those modules are in the QVL they should work out of the box.

 

2133 vs 4266 = D O U B L E the bandwidth, instead of pushing ca. 35GB/sec those push 70GB/sec.

 

As I wrote in that mail to you, some apps do actually love fast RAM, like rendering, video editing ( mining and priming as well ) etc etc. Word wont type faster ! YT vids will be the same, but your render job may speed up by 1/3 or such with such differences. Couple that with low latency and you have a nice rig ;)

 

Anything past 2933 gets expensive, the lower the latency the higher the price, mine are like 400€ now, I got them for 340€ end of January....and those are "only" 3600. Those 4266 will be real expensive.

 

The other trade you have to accept is module size. You will not find a 16GB module with 4266, expect the largest one to be 4GB, lucky if they make them 8GB now, havent checked lately. So max you could do was 16GB for 2-channel or 32GB for quad pumped X99/299.

 

Certified for use just means, WE DONT GARANTY THAT IT WILL WORK...but it should as long as the modules are in QVL and you do NOT OC the CPU to the highest limit it can go as the Integrated Memory Controller will tilt away at some point if you do both to an extreme, may or may not work. 5.2/4166 would be monster setup indeed !


Edited by BitMaster

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

The ED devs could/should answer how memory is accessed by DCS

 

Yeah I wish developers of all sims would give more quantitative feedback on hardware requirements. OK some other considerations:

Bandwidth GB/s = MHz x 8 x 2

Speed ns =CAS x 1000 x 2 / MHz

 

3000MHz 48GB/s 10ns

£240 G.Skill Trident Z 15-15-15-35

£260 Corsair LPX 15-15-15-36

£280 Corsair Dominator 15-15-15-36

 

3200MHz 51.2GB/s 10ns

£250 G.Skill Ripjaws V 16-18-18-38

£275 Corsair LPX 16-18-18-36

 

3200MHz 51.2GB/s 9.4ns

£280 G.Skill Ripjaws V 15-15-15-35

£285 G.Skill Trident Z 15-15-15-35

 

Hmm... perhaps go for either 3000MHz or 3200MHz CAS15... Dominator 3000MHz vs Trident Z 3200MHz???


Edited by Django

Regards, Django.

| BMS | DCS OB | A-10C II | AV-8B | F-16C | F/A-18C | FC3 | Persian Gulf | Supercarrier | Tacview | XP11 | FF A320 | FF 757 |

| I7-9700K + NH-D15 | RTX3080Ti 12GB | DDR4-3200 16GB | Aorus Z390 Ultra | 2X Evo 860 1TB | 850W | Torrent Case |

| Warthog HOTAS + CH Pedals | 32" TV 1080p 60Hz | TrackIR5 |

Link to comment
Share on other sites

Honestly you want as much cache as possible. Nothing beats the transaction lookaside buffer.

Basically just nothing.

 

RAM overclocking is great for synthetic loads, for FPS, its measurable, but in the end not really something which is going to give one a FPS boost thats worth the effort and or monies.

I did go with my CPU from 3.6 to 4.2 and with ram from 2133 to 2933. The difference is neglectable. That 1.5 even benefits from that stuff is mostly down due to its terrible terrible outdated engine.

 

On 2.1 for example I get 70% util of 12 threads.

 

Also, that a XMP profile is there, doesnt mean it works. I am running 64GB of 3200, but its maximum runs stable at 2933.

 

I do overclock for grid computing and stuff like that. For gaming, well GPU>everything. And even 3.6ghz with 2133 is perfectly capable of feeding the GPU enough for full util in 2.1.

 

EDIT: Fast ram is kinda important for AM4 due to infinity fabric being clocked at the same rate as the RAM.

And since data transfer between CCX's is extremely expensive time wise, higher ram clock leads to quite substantially better performance in terms of context switches.


Edited by microvax

[sIGPIC][/sIGPIC]

 

*unexpected flight behaviour* Oh shiii*** ! What ? Why ? What is happening ?

Link to comment
Share on other sites

 

RAM overclocking is great for synthetic loads, for FPS, its measurable, but in the end not really something which is going to give one a FPS boost thats worth the effort and or monies.

I did go with my CPU from 3.6 to 4.2 and with ram from 2133 to 2933. The difference is neglectable. That 1.5 even benefits from that stuff is mostly down due to its terrible terrible outdated engine.

 

But it makes the numbers look better, :P

 

On 2.1 for example I get 70% util of 12 threads.

 

I have just started using MSI Afterburner and noticed that my i7 5930 @ 4.2Ghz overall CPU utilisation is hovering around that level. Where as Task manager only indicates 16 to 18% CPU utilisation for "same" conditions.??

 

Although I have noticed more core utilisation in 2.1.x, it could also just be the measuring tools too..:music_whistling:

Control is an illusion which usually shatters at the least expected moment.

Gazelle Mini-gun version is endorphins with rotors. See above.

 

Currently rolling with a Asus Z390 Prime, 9600K, 32GB RAM, SSD, 2080Ti and Windows 10Pro, Rift CV1. bu0836x and Scratch Built Pedals, Collective and Cyclic.

Link to comment
Share on other sites

Honestly you want as much cache as possible. Nothing beats the transaction lookaside buffer.

Basically just nothing.

 

RAM overclocking is great for synthetic loads, for FPS, its measurable, but in the end not really something which is going to give one a FPS boost thats worth the effort and or monies.

I did go with my CPU from 3.6 to 4.2 and with ram from 2133 to 2933. The difference is neglectable. That 1.5 even benefits from that stuff is mostly down due to its terrible terrible outdated engine.

 

On 2.1 for example I get 70% util of 12 threads.

 

Also, that a XMP profile is there, doesnt mean it works. I am running 64GB of 3200, but its maximum runs stable at 2933.

 

I do overclock for grid computing and stuff like that. For gaming, well GPU>everything. And even 3.6ghz with 2133 is perfectly capable of feeding the GPU enough for full util in 2.1.

 

EDIT: Fast ram is kinda important for AM4 due to infinity fabric being clocked at the same rate as the RAM.

And since data transfer between CCX's is extremely expensive time wise, higher ram clock leads to quite substantially better performance in terms of context switches.

 

 

When I load DCS with default speed and hop into the cockpit I always have MSI AFterburner on and watch my performance data. The default is 4.5G/3600 and my 980GTX 1350/7000.

If I then Alt-Tab into Asus AI-Suite and OC the rig to CPU/RAM_5G/3600 and GPU-1500/8000 I can see a ~10% increase in fps, in every module. I can reproduce this anytime.

4.5G and default GTX is already fast enough to enjoy, no question, I can just clearly state and reproduce the gain in fps when I add another 500MHz on the CPU and 150MHzGPU + 1000MHz on the VRAM.

 

I assume that 3.6 to 4.2 etc is just not fast enough to overcome the load from scripts and stuff and thus the CPU/system cannot utilize the extra MHz for more fps. YOu might have to be at a level where the base-work DCS has to do is already powered adequately and THEN add more MHz on top that will eventually result in higher fps.

 

That's how I explain this behaviour of my machine and statements from others. Might be totally wrong tho :doh:

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

When I load DCS with default speed and hop into the cockpit I always have MSI AFterburner on and watch my performance data. The default is 4.5G/3600 and my 980GTX 1350/7000.

If I then Alt-Tab into Asus AI-Suite and OC the rig to CPU/RAM_5G/3600 and GPU-1500/8000 I can see a ~10% increase in fps, in every module. I can reproduce this anytime.

4.5G and default GTX is already fast enough to enjoy, no question, I can just clearly state and reproduce the gain in fps when I add another 500MHz on the CPU and 150MHzGPU + 1000MHz on the VRAM.

 

I assume that 3.6 to 4.2 etc is just not fast enough to overcome the load from scripts and stuff and thus the CPU/system cannot utilize the extra MHz for more fps. YOu might have to be at a level where the base-work DCS has to do is already powered adequately and THEN add more MHz on top that will eventually result in higher fps.

 

That's how I explain this behaviour of my machine and statements from others. Might be totally wrong tho :doh:

 

As said 1.5 is plain terrible in terms of efficiently using nowadays hardware.

2.1 is a lot better.

 

But for how FPS increase works I would very much disagree with your interpretation as that makes no sense from a programming and execution standpoint at all.

I would suspect the +10% come very much from the roughly +150mhz GPU since GPU clock and FPS scale a lot more linearly since technically there is rarely a reason to be constantly CPU limited. CPU overclock simply resolves the spikes in processing time faster and thus the GPU can render the next frame sooner in these spiky cases.

 

 

on the topic of being cpu bottlenecked... basically with a 60 euro cpu not a thing until post 1070.

Its the same as with pci-e x8 vs x16 for GPUs. One is going to loose like 1-3% for 50% of the lanes.

 

But overall stuff is quite dependant on implementation, but again well programmed games are mainly GPU limited nowadays. One obviously can always get some %s with lower latencies. But basically even with 2133 GPU is still more reasonable to invest in then faster RAM. Unless one has a special use case where faster RAM makes more sense.

 

Its always fun to try to push down render times as far as possible, but for someone just asking what to buy to get out of the box best performance without great tweaking, GPU is the main thing.

 

As said 1.5 atm kinda exception because very old engine, but thats going to change pretty soon with 2.1 where the whole thing is so GPU limited and overall as it seems well optimised.

[sIGPIC][/sIGPIC]

 

*unexpected flight behaviour* Oh shiii*** ! What ? Why ? What is happening ?

Link to comment
Share on other sites

No, no and 3rd time NO.

 

I loaded up DCS 1.5, KA-50 Free Flight just now to test what you say and it is REPRODUCABLY the other way round.

 

I start with 5G and 1500/8000 86-88fps -> set CPU to default and get 75fps with GPU still in full OC --->then set GPU to default and only loose 2-4 fps--->then set CPU back to 5G and again 83-86fps--> add thr GPU OC and I gain another 2-4 fps.

 

I have done this for 2 min back and forth up & down the ladder. If you want I can record it to proof it.

 

The very limiting factor for at least 1.5 IS THE CPU.

 

There is a roughly 10fps difference on my end when I switch from 4.5 to 5G and about 2-3, maybe 4 fps when I oc the GPU from 1350/7000 to 1500/8000. Each scaling is INDEPENDANT from the other, which tells me I am GPU bounjd now as I can unleash more frames if I oc my GPU, regardless if I oc from 4.5 or 5G.

 

As I said, I can reproduce this ANYTIME to show you the effect. Words and statements mean nothing to me unless you can reproduce it and proof it with evidence. I have BOTH at hand.


Edited by BitMaster

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

I have made the video and in about 90min it's uploaded. I will post the link. My YT name is Bitmaster69.

 

 

 

*edit*

 

In addition to show the scaling beween 4.5 and 5G I also did a flight with as low as 4.2 but that setting is not covered in the video.

 

The result was as expected, it scales up & down the ladder, 4.2 has the least fps, 4.5 is better and 5G is max.

 

I will do a 5.2G run after uploading to see if 5.2 gains even more fps...its just too risky to BSOD while uploading ;)


Edited by BitMaster

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

You didnt listen to what I said or did not understand it or I did word it ambigious.

 

1.5 is CPU limited, its CPU limited hard. GPU is basically bored most of the time.

As said 1.5 is the exception. Because its terribly outdated in terms of engine, thus it profits from high core clock and memory frequency.

 

2.1 should not scale that way. And since even todays newsletter basically puts 1.5 at its place as an obsolete piece of tech which is going to go away soon I do not think its reasonable to build a system with specs that are nice on paper but soon are going to be unnecessary or not even really benefitial anymore.

 

For 1.5 its all very much correct, but for any modern engine, including 2.1 it should simply not be that much of a factor anymore and it isn't.

Thats why I did link the video, I do not think designing a system for 1.5 is reasonable anymore at this point in time.

 

What really doesnt make any sense is "I assume that 3.6 to 4.2 etc is just not fast enough to overcome the load from scripts and stuff and thus the CPU/system cannot utilize the extra MHz for more fps. YOu might have to be at a level where the base-work DCS has to do is already powered adequately and THEN add more MHz on top that will eventually result in higher fps.".

 

That could be the case if a frame would take longer to render then 7th of a second, which is obviously not the case. Since a frame increase by your model would only be able if simulation + rendering can be done in the cycles added by overclocking. Since obviously more then 6FPS are beeing achieved that doesnt make any sense. As I said, its measurable, just not really perceivable without having a FPS counter on.

[sIGPIC][/sIGPIC]

 

*unexpected flight behaviour* Oh shiii*** ! What ? Why ? What is happening ?

Link to comment
Share on other sites

I see your point, I saw it before as well.

 

Performance cant harm is my POV and I will try 2.1 NTTR as well as the beta ( once available for downlaod, I still cant seem to hook up to it with the updater, maybe later today )

 

FYI..my GPU ( ok, a 980 is not a 1080Ti ) is usually at 99% in 1.5. Since they moved to DX11 with 1.5 I can see a HUGE increase in GPU usage and a hefty UP in fps.

 

Dont want you as an enemy, I am just eager to point the right thing out for others with less capabilities to sort things out, your input is very welcome and sobering as well.

 

Bit

 

edit+ addendum:

The Ka-50 is maybe the worst to show scaling when you call out for frametimes. When I am in BF, flying the Mi-8 through vallies, I often see 140-180 fps with frametimes as low as 5-7ms.

In FC3 or Su25T it is also mostly way above 100fps, depending on scenario it also goes up to 180fps. Would that matter then according to your math, calculation time vs frametime ?

 

At least for helicopters I am looking for the maximum fps I can gain. they just fly so much smoother with higher fps. Ok...140 or 180 fps is arguing at a very high level, I totally agree, but since there are monitors out for years now that do 166Hz or 180Hz on QHD-2k it does matter.


Edited by BitMaster

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...