CPU overclock - is it truely worth it for DCS?

[DocSigma]

I had my first gen. i7 920 oc'd to 4.0 for close to 10 yrs without one issue once i got it stable.

[Biggus]

Here is the thing,

I agree that OC done by an expert can benefit the system rather than shorten its life...but...most are just laymen who just try to bump up systems that shouldn't be OC or OC to the extent that it ruins them or lessen their performance.

Chips, motherboards and power supplies are optimized (kind of ;) ) to work best within certain parameters within a certain interaction.

When you try to take such a system beyond what it was built for you should just get hold of a different system that was built for that higher load/ capacity/ strain.

If you don't and unless you are an expert it shall be very similar to taking a car and modifying it for racing with a drilled engine.

It shall be able to perform better for a short while but since it was not built for that higher stress it just shall not hold it together for too long.

Are power cars fun?

Hell yes!

But most are short lived if pushed hard and DCS pushes real hard.

 

 

I don't agree with very much of this, either as an overclocker or car enthusiast. Nobody is suggesting that a person open up their bios and set massive vcore with absolutely zero knowledge or guidance or research, and the level of knowledge required to safely overclock a system is easily available today.

 

 

 

Why would anyone overclock their system to the point where in-usage performance is degraded? Components certainly have parameters that they need to operate within, but that's certainly accounted for by anyone with the most basic level of reading comprehension. Intel provide specific chipsets for motherboards with the design intent of overclocking. They provide CPUs that support overclocking. AMD support overclocking across three main families of motherboard chipsets today. The aftermarket supports a wide range of cooling options. The internet has a huge range of reading resources available for every modern consumer OC-capable CPU. If a person owns an overclockable system and has the time and motivation to learn, then I'd encourage them to do so. Historically, for myself and many others it has had nothing but benefits.

 

 

 

As for the car analogy, it's not really accurate to say that getting more power out of a motor will degrade it's longevity. It's sometimes true, but it regularly isn't.

[STRYC]

I'm running a 9900k on a EVGA Z390 Dark, overclocked on all cores to 5.1Ghz. Paired with watercooled, overclocked and power modded 2080 TIs. My Son has the same GPU paired with a 7700k stock clocks on a Z170. We get the exact same frame rate in DCS VR sitting on the deck of the super carrier.

[STRYC]

Sorry, but I don't get it.

You OC your system in order to get more juice out of your system because replacing is with a better system is expensive...but...in doing so you're shortening your system's life which will force you to spend the same money you couldn't or more if the damage is extensive.

And...all that for a few frames in a software that is as resources hungry as the half horse?

 

It's fun. I always power mod to heavily overclock the fastest GPU when It becomes available. I've never seen degraded hardware from overclocking, but I always slap waterblocks on my CPU and GPU. I have however kept my 9900k because I love my motherboard and the my CPU memory controller rocks! I'm running 16x2 32GB @4000mhz with 16cl timings and rock solid stable! That's insane! And fun!

[eatthis]

Sorry, but I don't get it.

You OC your system in order to get more juice out of your system because replacing is with a better system is expensive...but...in doing so you're shortening your system's life which will force you to spend the same money you couldn't or more if the damage is extensive.

And...all that for a few frames in a software that is as resources hungry as the half horse?

 

It's fun. I always power mod to heavily overclock the fastest GPU when It becomes available. I've never seen degraded hardware from overclocking, but I always slap waterblocks on my CPU and GPU. I have however kept my 9900k because I love my motherboard and the my CPU memory controller rocks! I'm running 16x2 32GB @4000mhz with 16cl timings and rock solid stable! That's insane! And fun!

 

what memory is that?

[Hoirtel]

I'm running a 9900k on a EVGA Z390 Dark, overclocked on all cores to 5.1Ghz. Paired with watercooled, overclocked and power modded 2080 TIs. My Son has the same GPU paired with a 7700k stock clocks on a Z170. We get the exact same frame rate in DCS VR sitting on the deck of the super carrier.

 

This means you are GPU bound (as you probably know) which is where you probably want to be and where my system is. This is kind of proof that new GPU upgrades will increase performance for those of us in this situation. I would think that if you both installed the same 3080 card you would start to get better fps on the PC with higher clocks as this will feed that card better!

[STRYC]

Originally Posted by eatthis

 

 

what memory is that?

 

 

GSkill TRIDENTZ F4-4000C19D-32GTZSW DDR4-4000 CL 19-19-19. I'm unable to clock it passed 4000 but I'm able to tighten up the timings @ Cl 16-16-16. Rock solid stable.

Edited October 1, 2020 by STRYC
Show Who I was Answering

[Thinder]

3D Mark Fire Strike Ultra 3840 × 2160 (4K UHD)

 

 

GPU OCed with MSI Afterburner. CPU OCed with Ryzen Master. Both moderate, no special tweaking.

 

Differences are a lot greater at lower resolutions and less demanding tests but i try to figure what to expect when I will play DCS with the HP Reverb G2, so I focus on high resolutions while preparing for a much better cooling for my case (not delivered yet) which should help if I want to push the system further.

 

 

Graphics score

6 995 vs 6 870

Graphics test 1

38.76 FPS vs 37.58 FPS

Graphics test 2

25.03 FPS vs 24.79 FPS

Physics score

19 041 vs 18 804

Physics test

60.45 FPS vs 59.70 FPS

 

The Physics test runs 32 parallel simulations of soft and rigid body physics on the CPU.

With MSAA turned to 2, Oculus Tray Tool set for 2 as well and new profile for both GPU and CPU (Still nothing special, no manual tweaking) I got this:

 

Graphics score

7 226

Graphics test 1

40.26 FPS

Graphics test 2

25.76 FPS

Physics score

18 677

Physics test

59.29 FPS

 

If the gain isn't enormous I think it provides with a solid performance for VR, resolution for this test are higher than what is required for the G2, also, the more I push the GPU, the more the CPU seems to have to work in the Physic tests.

 

Edited October 4, 2020 by Thinder

[eatthis]

Originally Posted by eatthis

 

 

what memory is that?

 

 

GSkill TRIDENTZ F4-4000C19D-32GTZSW DDR4-4000 CL 19-19-19. I'm unable to clock it passed 4000 but I'm able to tighten up the timings @ Cl 16-16-16. Rock solid stable.

 

is speed or timing more important for dcs vr?

[Motomouse]

When I am proficient with all DCS airplanes, perhaps I will consider overclocking. Until then I enjoy flying.

[hansangb]

is speed or timing more important for dcs vr?

 

 

I can't find it anymore, but there was a comparison of faster memory in VR. And there was material difference in FPS.

[Thinder]

Progresses made with O.C, Benchmark is 3D Mark Fire Strike Ultra.

 

MSAA from 2 to 1 to x 0 at 3840 x 2160.

 

Lowest score was obtained in stock settings, mid scores profile 1, highest score profile 3.

https://forums.eagle.ru/showthread.php?p=4512075#post4512075

 

Graphics score

7 360 vs 6 995 vs 6 870

Graphics test 1

41.22 FPS vs 38.76 FPS vs 37.58 FPS

Graphics test 2

26.15 FPS vs 25.03 FPS vs 24.79 FPS

Physics score

20 381 vs 19 041 vs 18 804

Physics test

64.70 FPS vs 60.45 FPS vs 59.70 FPS

 

 

Pick GPU temperature: 76°

 

Pick CPU temperature: 67.25°

 

 

 

3D Mark Fire Strike Ultra

 

Max AF anisotropy = 16

 

MSAA x 2

 

Tessellation detail level 10 )Maxi_

 

Max tessellation factor = 24

 

Shadowmap size 2048

 

Volumetric illumination quality 10 )Maxi_

 

Volumetric illumination sample count = 1.5

 

Particle illumination quality = 10 (Maxi)

 

Ambient occlusion quality = 10 (Maxi)

 

Depth of field quality = 10 (Maxi)

 

Color saturation = 100

 

 

 

So if you ask me, with proper cooling, going from 37.58 FPS with MSAA set to zero to 41.22 FPS with MSAA set to 2 at 3840 x 2160 with temperature not going over 76/67.25°, yes O.C is worth it.

 

In DCS V.S using an Oculus CV1, I noticed the overall increase in sharpness and the smoothness of the game.

 

 

......

Edited October 4, 2020 by Thinder

[STRYC]

is speed or timing more important for dcs vr?

Typically speed trumps timings. However, you will get diminished returns as you increase speed and therefore will want to tighten up your timings at some point. I was previously running Gskill 4600mhz 16GB (2x8gb). I decided to increase my memory to 32gb, but at the time, the fastest ram I could find was Gskill 4000mhz (2X16gb). MY motherboard is a EVGA Z390 Dark it only has two dimm slots, and was specifically built that way for overclocking. I ran benchmarks to see if I would get better performance in DCS VR with the increase in ram 16gb vs 32gb. Sadly, I did not get any performance increase or decrease in DCS VR in any configuration I tested. The 16gb at 4600mhz with the same timings as the 4000mhz 32gb, no change. Reducing the 4600mhz 16gb down to 4000mhz vs 32gb same timings, no change.

 

Even though my system was using almost all 16gb of ram with the 16gb set and up to 25gb with the 32gb set, in DCS VR, there was zero performance change from one configuration to the next. I gave my 4600mhz ram to my Son to add to his system, which put him from 16gb to 32gb (4x8gb) on a Z170 with 7700k and a watercooled 2080ti. He saw no performance change in DCS VR.

[BitMaster]

It depends what trumps what. A 3200CL14 is as fast or faster in delivering the first few bits and bytes as as 3600CL16. The 3600CL16 will win in brute throughput if large amounts of data are called from memory, but if it is only few MB's then the 3200CL14 might be the better deal.

 

There are equations on goggle that let you calculate actual real world latency and throughput based on Latency and Speed.

 

The DDR4 sweetspot is at 3600/CL16. Latency, Throughput and Price come to a good relation.

Also, not to forget, the faster and less latency they have, the lesser are your chances they actually work as intended in your motherboard and with your IMC. So if you overdo the modules, like 4800MHz/CL16 for 999€ a 32GB kit, chances are they wont work in 99% of the boards. ThatÄs why I would not exceed 3600-3800 MHz for said reasons.

[eatthis]

It depends what trumps what. A 3200CL14 is as fast or faster in delivering the first few bits and bytes as as 3600CL16. The 3600CL16 will win in brute throughput if large amounts of data are called from memory, but if it is only few MB's then the 3200CL14 might be the better deal.

 

There are equations on goggle that let you calculate actual real world latency and throughput based on Latency and Speed.

 

The DDR4 sweetspot is at 3600/CL16. Latency, Throughput and Price come to a good relation.

Also, not to forget, the faster and less latency they have, the lesser are your chances they actually work as intended in your motherboard and with your IMC. So if you overdo the modules, like 4800MHz/CL16 for 999€ a 32GB kit, chances are they wont work in 99% of the boards. ThatÄs why I would not exceed 3600-3800 MHz for said reasons.

 

why is that? iv bought 4 x8gb 4400 cl 19 kits for my new build, you reckon i should aim get it running 3800 at cl16?. ps il probably end up with a decent x570 mobo IF the new amd cpus beat intels for dcs

[Thinder]

The DDR4 sweetspot is at 3600/CL16. Latency, Throughput and Price come to a good relation.

 

What really matter is the ability for your RAM to pass data to the CPU/GPU link, so it's about bottlenecks, ultimately the difference between 3200 and 3600 MHz is not the limiting factor, especially with good quality RAM one can overclock without problem.

 

 

On a budget, I recommend this:

 

 

I tested mine at 3200 and 3600 MHz and find similar results, about 5 frames/sec difference.

 

 

 

 

 

https://uk.crucial.com/memory/ddr4/bl2k16g32c16u4b

 

 

https://www.techpowerup.com/review/crucial-ballistix-gaming-memory-ddr4-3200-mhz-cl16/

 

 

PS do NOT forget to take cooling into account, you need proper cooling to make sure you can push your components safely.

https://forums.eagle.ru/showthread.php?p=4511015#post4511015

 

 

 

......

Edited October 8, 2020 by Thinder

[BitMaster]

why is that? iv bought 4 x8gb 4400 cl 19 kits for my new build, you reckon i should aim get it running 3800 at cl16?. ps il probably end up with a decent x570 mobo IF the new amd cpus beat intels for dcs

 

You have to run the equation to see if 4400MHz with a CL19 penalty outruns a quick 3600 CL16 kit.

 

Not to forget, subtimings come into play if you need to move more than small bits and bytes.

 

So a CL16-16-16-36 ( 1 or 2 T ) is better than a 3600 CL16-18-18-36 ( or often higher than 36 ).

 

You gotta check google for RAM MHZ vs. latency calculator, there are a few versions out there with slightly different approaches mathmetically, but they all should give you an idea if the 4400 was a bad buy or if you can surpass a slower kit but with lower latency. I would think they are really close, CL19 at 4400MHz might well beat the CL16 3600er...but do the math yourself, then compare the prices of both kits and ask yourself if it was worth it.

 

I have spent far too much money on RAM that in the end was wasted money. Happy to have a true 16-16-16-36 3600 kit...and not so happy that my damn board refuses to load XMP with more than 2 modules installed. So down the line, your board plays a big factor too if it all will go as planned. 4400MHz is still considered way above what runs without issues on all boards, not to speak of more than 2 modules, that is a totally different issue itself ( T-Topology vs. Daisy Chain topology..pros and cons for each topology )

 

a quicj 3200MHz CL12 aint bad either ;) OK,, I stop here :smilewink:

 

..and yes...a tripple fan GPU offloading all it's heat into the case is a bad co-player for fast and stable RAM or overclocking. There are good YT vids about that issue, it's worth thinking about that issue, best WC your GPU and CPU to avoid ANY heat inside your case, best with an external cooler. I have that setup and with 2 modules installed only I can OC my RAM at least to 4133 with a decent latency and easy approach.

Blow hot air over them and you have a totally different thing going on, no BS.

Edited October 8, 2020 by BitMaster

[eatthis]

You have to run the equation to see if 4400MHz with a CL19 penalty outruns a quick 3600 CL16 kit.

 

Not to forget, subtimings come into play if you need to move more than small bits and bytes.

 

So a CL16-16-16-36 ( 1 or 2 T ) is better than a 3600 CL16-18-18-36 ( or often higher than 36 ).

 

You gotta check google for RAM MHZ vs. latency calculator, there are a few versions out there with slightly different approaches mathmetically, but they all should give you an idea if the 4400 was a bad buy or if you can surpass a slower kit but with lower latency. I would think they are really close, CL19 at 4400MHz might well beat the CL16 3600er...but do the math yourself, then compare the prices of both kits and ask yourself if it was worth it.

 

I have spent far too much money on RAM that in the end was wasted money. Happy to have a true 16-16-16-36 3600 kit...and not so happy that my damn board refuses to load XMP with more than 2 modules installed. So down the line, your board plays a big factor too if it all will go as planned. 4400MHz is still considered way above what runs without issues on all boards, not to speak of more than 2 modules, that is a totally different issue itself ( T-Topology vs. Daisy Chain topology..pros and cons for each topology )

 

a quicj 3200MHz CL12 aint bad either ;) OK,, I stop here :smilewink:

 

..and yes...a tripple fan GPU offloading all it's heat into the case is a bad co-player for fast and stable RAM or overclocking. There are good YT vids about that issue, it's worth thinking about that issue, best WC your GPU and CPU to avoid ANY heat inside your case, best with an external cooler. I have that setup and with 2 modules installed only I can OC my RAM at least to 4133 with a decent latency and easy approach.

Blow hot air over them and you have a totally different thing going on, no BS.

 

i dontr know what im doing with teh calcs. my current and next pc will have cpu and gpu watercooled so alot of the heat is going out of the case

theres a block you can get to watercool the ram. are they worth it?

[hansangb]

i dontr know what im doing with teh calcs. my current and next pc will have cpu and gpu watercooled so alot of the heat is going out of the case

theres a block you can get to watercool the ram. are they worth it?

 

 

I had a fan over them because my Corsair modules came with it. But I'm not sure if water block makes sense for memory.

[Thinder]

 

..and yes...a tripple fan GPU offloading all it's heat into the case is a bad co-player for fast and stable RAM or overclocking. There are good YT vids about that issue, it's worth thinking about that issue, best WC your GPU and CPU to avoid ANY heat inside your case, best with an external cooler. I have that setup and with 2 modules installed only I can OC my RAM at least to 4133 with a decent latency and easy approach.

Blow hot air over them and you have a totally different thing going on, no BS.

 

 

You need to blow the hot air out of the case and feed your case fans with as cool and clean airflow as possible.

 

If your RAM is directly between your CPU fan and front case fan, it doesn't always help to add a fan over them because the airflow from CPU is hotter and two fans in close proximity doesn't necessarily work well together.

 

You're far better off increasing the airflow from front fan to back fan and I did find a solution to this.

 

https://forums.eagle.ru/showpost.php?p=4511015&postcount=1

 

 

I have tested my RAM at 3200 MHz and 3600MHz without any significant increase of temperature and at full load (maximum GPU core and VRAM settings) the whole system remains cool.

 

I am planning to add a second Noctua NF-F12 120mm at the front to increase the airflow further before sticking to a 3600MHz setting for my RAM, just to stay in the safe side, but it appears that the RAM is using its sinks to best effect.

 

All the RAM needs is to be able to work within a cooler environment, sinks can then keep working to the best of their capabilities, the cooler the airflow around them, the better.

 

 

......

Edited October 9, 2020 by Thinder

[Flextremes]

Interesting thread, I also use an older 6700k @ OC to 4.5 Ghz.

 

Will upgrade in the next few months to either 5950x or 10900k. I wonder what ZEN 3 OC potential is.

 

But this is not main reason why I reply to this thread.

 

I just want to point out for 6700k, 7700k and 8600k (and older) owners that it is really worth to DISABLE meltdown and spectre vulnerability fixes. This will improve your IPC by a whopping 10 to 15%. Use inspectre for this, disable both and reboot.

 

I am not too worried about security, since there are hardly any active malware/virus out there that use this vulnerability. If you are, just temporarily disable for a DCS session. Only downside is you need to reboot for enabling/disabling.

 

My cinebench score improved by 15% > I am confident DCS will benefit also.

[Thinder]

Interesting thread, I also use an older 6700k @ OC to 4.5 Ghz.

 

Will upgrade in the next few months to either 5950x or 10900k. I wonder what ZEN 3 OC potential is.

 

But this is not main reason why I reply to this thread.

 

I just want to point out for 6700k, 7700k and 8600k (and older) owners that it is really worth to DISABLE meltdown and spectre vulnerability fixes. This will improve your IPC by a whopping 10 to 15%. Use inspectre for this, disable both and reboot.

 

I am not too worried about security, since there are hardly any active malware/virus out there that use this vulnerability. If you are, just temporarily disable for a DCS session. Only downside is you need to reboot for enabling/disabling.

 

My cinebench score improved by 15% > I am confident DCS will benefit also.

 

 

Thanks for the tip.

 

I know I am using a newer AMD but apparently it does help with O.S stability as well, so it's all gain anyway.

 

 

.......

[BitMaster]

Had to try this out.

 

Defaulted Bios to stock settings, NO OC applied.

 

3385 points with both enabled

3438 points with both disabled

 

The Microcode Update remained in action, I don't think we can turn that off easily.

 

BTW, my "old" Z370 chipset got another ME ( Management Engine ) Firmware upgrade just a few weeks ago, which is unusual as all other drivers etc. on Asus page are "2019".

 

No way I am buying Intel again in the forseeable future, neither will I recommend one to customers. Let's see if they take the crown November 5th, I wish they do, honestly, deserves both right, for the downfall of Intel and for the success for AMD.

Edited October 10, 2020 by BitMaster

[Brainless]

The improvement in the older ones (like Sandy Bridge) is even bigger.

https://forums.eagle.ru/showthread.php?t=260180

[STRYC]

It depends what trumps what....

 

 

Although paraphrased, that's what I was saying. And you're correct about 3600mhz being the sweet spot. At least for DCS.

 

 

 

Ram timings are built in latency. Overclocking ram, which for ME, includes tightening up timings, will give a performance increase up to a point depending on the game or application. Again "up to a point". That said, If you grab let's say Gskill ram 3600mhz at stock XMP vs 4000 stock XMP, which will have higher latency; the 4000mhz is a faster module. Gskill is not going to put out higher frequency ram with a timing handicap making it slower than any lower frequency modules in their lineup.

 

 

 

As for DCS, I mentioned in my previous post what I experienced with various ram configurations, i.e., overclocking; testing various speeds and speed vs timings.

 

 

 

"I decided to increase my memory from 16GB to 32GB, but at the time, the fastest ram I could find was Gskill 4000mhz (2X16gb). I ran benchmarks to see if I would get better performance in DCS VR with the increase in ram 16gb vs 32gb. Sadly, I did not get any performance increase or decrease in DCS VR in any configuration I tested. The 16gb at 4600mhz with the same timings as the 4000mhz 32gb, no change. Reducing the 4600mhz 16gb down to 4000mhz vs 32gb same timings, no change."

Edited October 12, 2020 by STRYC