CPU overclock - is it truely worth it for DCS?

[Tordox]

Hi everyone.  I was wondering if any of you did the test in DCS of how the FPS varies with the oc cpu and with the stock cpu.  I am a vr user  and how it has been said several times (each FPS matters).  so depending on whether it is worth it or not (in DCS) I Will do oc or not.  

Thanks in advance

Edited September 21, 2021 by Tordox

[Lurker]

9 hours ago, Tordox said:

Hi everyone.  I was wondering if any of you did the test in DCS of how the FPS varies with the oc cpu and with the stock cpu.  I am a vr user  and how it has been said several times (each FPS matters).  so depending on whether it is worth it or not (in DCS) I Will do oc or not.  

Thanks in advance

 

 

I haven't done detailed testing, but as a VR user myself it definitely helped with performance. My overclock is modest, and I use a pretty decent cooling solution (BeQuiet 240mm Watercooling). I can probably push it higher but I don't want to go overboard with voltage. I'm using a i5-9600k and have overclocked it using Intel's XTU to 4.8Ghz stable, with a cache overclock to 4.5 Ghz. 

Edited September 22, 2021 by Lurker

[Hoirtel]

DCS is single core priority so an overclock that improves single core performance will help. Maybe not by massive amounts but all helps. Worth thinking that if you have hyper threading, disabling this might help you achieve a higher and cooler all (physical) core over clock. Yes you have reduced the performance of your cpu in all multi threaded tasks. But you should be able to get more in single core applications, like dcs. 

Edited September 22, 2021 by Hoirtel

[Mr_sukebe]

Take a scan in task manager and split the usage by core.  For me, none of my CPU cores are maxed out, it’s my GPU that’s the bottleneck.  So overclocking my CPU  wouldn’t help me at all.

Edited September 22, 2021 by Mr_sukebe

[Lurker]

11 minutes ago, Mr_sukebe said:

Take a scan in task manager and split the usage by core.  For me, none of my CPU cores are maxed out, it’s my GPU that’s the bottleneck.  So overclocking my CPU  wouldn’t help me at all.

 

 

What exactly does CPU core usage tell you about CPU performance? Let me answer that. Just one thing. That your CPU is in use. 

Edited September 22, 2021 by Lurker

[sirrah]

With chip prices to an all time high, I haven't really dared to try overclocking my 8700k (because, well, I'm that guy that will probably fry his CPU)..

 

But seeing this thread popup again planted the seed again to go and try it...

 

A question before I start this enterprise :

 

How can I best create some sort of a benchmark mission to test if the overclocking has any effect?

 

I mean, I can imagine that when I just place 30 moving Tomcats within my field of view, that will just cripple my GTX1080TI, resulting in maximum GPU load and maybe only limited strain on my CPU?

Should I just place lot's of moving units and have them fire at each other outside my field of view? (which I'd expect to cause little strain on my GPU and more on my CPU)

 

Any advice here?

[Lurker]

1 hour ago, sirrah said:

A question before I start this enterprise :

 

How can I best create some sort of a benchmark mission to test if the overclocking has any effect?

 

 

 

Just download the XTU utility from Intel and take it from there. It has convenient benchmarks which will score your overclocks. If at any time your PC becomes unresponsive, you can reboot your PC and it will default to the last stable overclock. 

Edited September 22, 2021 by Lurker

[sirrah]

30 minutes ago, Lurker said:

 

Just download the XTU utility from Intel and take it from there. It has convenient benchmarks which will score your overclocks. If at any time your PC becomes unresponsive, you can reboot your PC and it will default to the last stable overclock. 

 

But will that test my VR performance?

 

I use my pc dedicated for DCS, so wouldn't it be better to test in DCS maybe?

[Lurker]

14 hours ago, sirrah said:

But will that test my VR performance?

 

I use my pc dedicated for DCS, so wouldn't it be better to test in DCS maybe?

 

Well yeah I guess you could do that. However it would be very clunky difficult and slow. Since there are so many variables involved, I never test within DCS World itself. If I can get a stable overclock, with software designed to do that, I do that and test that. Then only once I do so, do I enter DCS World and play a scenario that I use to "test" my changes so to speak. I put "test" in parenthesis because I never test with FPS counters or frame-time counters or anything else within DCS World either. I usually just play the game and see how responsive and smooth it is in comparison to the last time I played it. I choose one single player scenario with a relatively large unit count and drop some weapons on them. Usually with demanding maps\modules. Lately Syria+Hind. 

Edited September 23, 2021 by Lurker

[sirrah]

5 minutes ago, Lurker said:

 

Well yeah I guess you could do that. However it would be very clunky difficult and slow. Since there are so many variables involved, I never test within DCS World itself. If I can get a stable overclock, with software designed to do that, I do that and test that. Then only once I do so, do I enter DCS World and play a scenario that I use to "test" my changes so to speak. I put "test" in parenthesis because I never test with FPS counters or frame-time counters or anything else within DCS World either. I usually just play the game and see how responsive and smooth it is in comparison to the last time I played it. I choose one single player scenario with a relatively large unit count and drop some weapons on them. Usually with demanding maps\modules. Lately Syria+Hind. 

 

Ok, thanks @Lurker

[BitMaster]

sirrah,

 

I oc my 8700k ever since I own it, it was ment to be overclocked because it is one of the "older" CPU's where you could still gain some serious MHz with modest effort.

 

The 2 things you should care about is Volts and Heat, directly connected to each other. But hey, even if you can cool that CPU with a super cooler it is no good idea to pump more than 1.35v through it under load and run it like that for months and years to come.

 

I would start upping the MHZ first until it fails, that produces less heat. Increase Volts only if need in 0.01-2v increments.

 

Use Adaptive Voltage for the CPU and set it to 1.30, set your CPU to 4.7GHz all cores and test it, if all green, do 4.8, then 4.9..until it fails, if it fails increase voltage by 0.01 or 0.02v and try again but always check that temps dont mainifest in the high 80s. 80-85°C max is still ok under full loadbut I wouldnt run a setting that pushed it into the high 80s into the 90s °C.

 

Things to may tweak are VCCIO and Systam Agent Voltage, both may get auto set to values way above 1.25v to 1.35v, that puts a lot of stress on the IMC and you may set it to 1.15 both manually if you start to get RAM errors or postings issues.

 

 

Other than that, most do 4.8G with a good air cooler and really good ones do 5G on air but that is borderline if the 8700k is not delidded ( mine is ). With a good AIO or DIY loop most do 5G or even more if delidded and lucky.

 

 

One more thing, by ALL MEANS, I would NOT boot my Win10 do test this until you have somewhat stable OC. Overclocked-Reboots often cause file corruption, I have ruined many many Win10 installs this way. Best if you makle a bootable USB stick with a Linux and boot that. If that fails or crashes nothing happens to your install. If you Win10 errors, do  sfc and dism

commands to check file integrity ( google )

 

 

 

Edited September 23, 2021 by BitMaster

[Lurker]

@BitMaster do you overclock in BIOS or do you use the XTU utility from Intel? 

[BitMaster]

Never really used XTU tbh, I use Bios and also Asus Ai-Suite sometimes. For a 24/7/365 oc you definitely want to do it in Bios.

 

There are so many test suites and benchmarks around, pick a few to find out obvious weaknesses in your OC, the real test is many many days w/o reboot and lots of different tasks inbetween, from DCS/Gaming, browsing, backup over many hundreds of Gigabyte... if it does all that incl. not failing the tests it might be stable  LoL

 

Tbh, Win10 itself is BSODing enough, sometimes you wont be able to tell easily why it happend.  I see a lot of 10 installs and honestly, Microsoft ....ahhh I stop here..LoL. You get the idea.

[BitMaster]

What I forgot to say about the Volts...

 

Asus, Asrock, Gigabyte, MSI and all others use different approaches to Load Line Calibration. If you are not familiar to that, google it up real quick. The essence is what your dialed in Volts

under load say and not the value that you dialed in. That may, and likely are, 2 different pairs of shoes. LLC is the corrective to that !   In order the read the real & present value under load I personally use HWinfo in Sensors Mode Only setting. You can read the correct value there when you stress it. The more it drops from the set value under load the less stable it may get, use a higher LLC setting to correct the drop under load ( some vendors use 1-7 low to high and some do it vice versa iirc. ). Buffer the drop so when you dial in 1.33 it wont read any lower than 1.32 under prime small FFTs. With Z370 and Asus there is a Bios setting to pre-calibrate the LLC and Adaptive Volts offset ( and others may too ) which has Intel default, Best, Medium and WCS ( Worst Case Scenario). To be as linear as possible use BEST, as it will add no extra Voltage to whatever you dial in anywhere. WCS i.e. adds some extra Volts and Heat is likely the reason why Auto-Overclock seldomly brings excellent results vs. DIY with BS&T. With HWinfo you can easily read what each setting adds vs. Best ( or equivalent term in your Bios if present ) with a fixed LLC. I use Best Case to no have Asus add their juice to my soup when dialing in Volts.

 

It makles a considerable difference in the amount of heat it produces and the LLC level you have applied under full load. You may fail 4.8, 4.9 or even 5.0G just because LLC adds way too much volts and heats up the CPU so much that it tilts. With the correct Adpative Volts max 1.35v, correct pre-calibration setting if available and the right or near right LLC setting volts don't go much higher and not much lower than what you dialed in under load. Adaptive Volts and leaving all Speed Step settings enabled enables windows to lower the frequenc & volts in energy save modes, if you want you can switch to high performance mode, the CPU will kick into 5G and 1.35v in idle, when you load it fully it should not drop more than 0.01-2 volts and not get too hot. That setting is the sweetspot for my 8700k. Depending on the silicon yours may need more or less volts to run stable. Delidding certainly helped to reduce the heat on mone by the high 1-digit values across all cores with a beefy DIY Loop. Not delidded and with air the limit under full load may well be under 5G but tbh for DCS you could try 100-200MHZ higher than what WS-stable is. I can play DCS at 5.1-5.2G for at least 2h ( i hardly ever use it longer ) but it is way off from stable under full load because to the heat due to the extra 0.05 Volts it needs to push the die to 5.2G. I havent run the chip above 5G for a long time but it runs 24/7/365 at 5G ever since and only reboots if needed. With VMware I often Kill it, trash it & really really need every little bit of power it's got and it really delivered under full tilt loads. When it crashed and BSOD'ed it did under no load, YT & idle etc. I will upgrade pretty soon and 128GB seems more likely than 64GB and more cores, as much as trhe budget allows, due to VMware, not DCS. Since DDR5 has less than stellar DDR5 latency leaks I think it's still I good time to go team red and get a Ryzen-9. Dial in PBO2 and call it a day. Overclocking is fun but at the endof the day I paid way way too much ti get that 8700k to where it is now ( either that or lesser silicon ). Delidding kit, Liquid Metal thermal paste, WC I think AMDS's appoach with PBO2 is better. Give it a good cooler and lean back for best single core performance, what DCS needs.

 

 

 

Edited September 26, 2021 by BitMaster

[Mouse_99]

As a retired IT tech, I do not believe in overclocking for the most part, it's like running your car at redline all the time, it does work, but the results take a toll and eventually your engine breaks because of the extra stress you put on it. If you need more, then get a bigger, faster CPU.

[Lurker]

11 hours ago, Mouse_99 said:

As a retired IT tech, I do not believe in overclocking for the most part, it's like running your car at redline all the time, it does work, but the results take a toll and eventually your engine breaks because of the extra stress you put on it. If you need more, then get a bigger, faster CPU.

 

That's interesting. I have never had a CPU die on me. Ever. Some of them I've used for many years. And I've overclocked all of them. Granted I've never taken the overclock too high, and I've always used aftermarket cooling solutions but the overclock was always higher than the rated max turbo boost speed. 

[Hoirtel]

2 hours ago, Lurker said:

 

That's interesting. I have never had a CPU die on me. Ever. Some of them I've used for many years. And I've overclocked all of them. Granted I've never taken the overclock too high, and I've always used aftermarket cooling solutions but the overclock was always higher than the rated max turbo boost speed. 

Most dodgy overclocks result in CPU degradation. You have to do something fairly stupid to get an immediate death. This is the real danger with overclocking as the degradation can be slow enough that you don't know until it sets in. And then its too late, you will have to dial it back and possibly at that point stock speeds may be difficult.

 

Anyone interested in overclocking, I would watch buildzoids YT vids "ramblings" on CPU degradation, load line calibration and voltage regulation. All very interesting. Mostly the two on load line calibration are especially worth it. Fairly long, but all very good info. Either that or just leave well alone and plan an upgrade.

[Lurker]

25 minutes ago, Hoirtel said:

Most dodgy overclocks result in CPU degradation.

 

How is this manifested? I have never had this happen to me either. I have had degradation in performance due to thermal paste losing some of it's thermal conductivity (easily fixed by applying fresh paste every 12 months or so, but this is also subjective and not really grounded in physics) and I have had dust clogging up my PC case and\or fans, again very easily fixed. Once these two fixes are applied performance is back to previous levels. How exactly does a CPU degrade? Because in 20+ years that I've built my own PCs this has never, ever happened to me, or rather it was always something else that caused this. (New software that requires a beefier CPU, or Windows updates, or dust\dirt or user error etc.)

 

To me, this sounds like an internet "myth" that is repeated online ad nauseum, by people who do not understand the basics of semi-conductor technologies. Even if is this is something that is actually a thing, most gamers will upgrade their CPUs long before it ever becomes an issue. 

Edited September 27, 2021 by Lurker

[Hoirtel]

17 minutes ago, Lurker said:

 

How is this manifested? I have never had this happen to me either. I have had degradation in performance due to thermal paste losing some of it's thermal conductivity (easily fixed by applying fresh paste every 12 months or so, but this is also subjective and not really grounded in physics) and I have had dust clogging up my PC case and\or fans, again very easily fixed. Once these two fixes are applied performance is back to previous levels. How exactly does a CPU degrade? Because in 20+ years that I've built my own PCs this has never, ever happened to me, or rather it was always something else that caused this. (New software that requires a beefier CPU, or Windows updates, or dust\dirt or user error etc.)

 

To me, this sounds like an internet "myth" that is repeated online ad nauseum, by people who do not understand the basics of semi-conductor technologies. Even if is this is something that is actually a thing, most gamers will upgrade their CPUs long before it ever becomes an issue. 

 

Slightly over high voltages, too much heat, too much current. All gradual. You have clearly avoided the worst of this. From my understanding all CPU/GPU chips degrade, but usually very very slowly - longer than their usable lifespan. Overclock will likely speed this up, again not usually a problem but it can be if you don't take care. Lots of variables in place too. Vid is on chip degradation as it covers GPUs too.

 

 

Edited September 27, 2021 by Hoirtel

[Lurker]

Yes all of those could become problems if you practice bad hardware "hygiene" especially with sub-par cooling, but the result of any of those three issues becoming critical is chip failure. Not chip degradation! A chip either works, or it doesn't. Your chip might downclock due to inbuilt safety measures, but those are usually disabled in overclocked chips, so that in the event of an actual failure, that chip....well fails. 

 

You know what would convince me? An actual benchmark of "old" vs "new" chips. Where you can consistently show a difference. I've only seen this one video, where the differences are within the margin of error. (Sometimes in favor of the "used" cpu)

 

 

P.S.

 

The longest I've had a chip was an i5-3570K, that thing worked like a dream overclocked to oblivion for a full 5 years. It's still alive in a my friend's hand-me-down PC, which he uses only occasionally I grant you but still.... (I actually gave him the chip, and the mainboard, and the ram when I upgraded years ago). 

Edited September 27, 2021 by Lurker

[Hoirtel]

1 hour ago, Lurker said:

A chip either works, or it doesn't.

This is kind of contradictory to your "bad hygiene" line so not sure what this bad hygiene refers to. A poor overclock? We agree then? Its worth noting that the variables of this are not just the voltages, temps or power draw. Also the time under load makes a difference. If you use your computer for a few hours a week, this is less degradation than 24/7 and a "poor overclock/bad hygiene" will take much longer to have an effect. Just to be clear I'm not saying you have a bad overclock!

 

I'm definitely no expert. I have no qualifications in this area but my research about overclocking and chip lifespan points to degradation being pretty normal in CPUs and GPUs. I've come across it many times in various postings and videos - Including the one I linked (one of the best actually) although I appreciate it is 1 hour - but I did watch it all. As I said I'm not a professional and happy to take your knowledge as different. Your experiences of overclocking are encouraging to me as I am possibly a little over worried about degradation so I take care and it sounds like you do too.

 

One of the general lines of information that I seem to come across again and again is that an incorrect "stable" voltage might not be stable after a year or two or three, and then needs more voltage to stay at the same frequency or the frequency reducing. This is degradation although also a slight simplification. This is supported as one of the three routes of degradation discussed in the video above. I cannot verify the content of buildzoid's videos but I have watched a lot of them and find them really insightful in his "rambling format" and he seems to have the verification of other widely regarded PC media channels such as gamers nexus.

Edited September 27, 2021 by Hoirtel

[Lurker]

Yeah it looks like we were maybe talking around each other, while actually agreeing. My point was that while degradation might occur within chips, performance degradation is not a thing per se.

 

In any case, if you take care of your whole system (mostly with regards to proper cooling), use unlocked chips (mostly intel, no idea if AMD does those) and don't take your overclocks too far, there is very, very little danger of a CPU failure occurring. I've never had it happen to me, and I've been doing this since when you had to use a pencil on your CPU to make it overclockable. 

Edited September 28, 2021 by Lurker

[Hoirtel]

51 minutes ago, Lurker said:

My point was that while degradation might occur within chips, performance degradation is not a thing per se.

Yeah ok, we can argee to agree? or maybe disagree... If a chip's internal structure is degraded enough it will not perform the same way. This can take a long time or a short time, more liely a very long time. Many many variables. The idea that chips are 100% fine working as new or they completely fail isn't correct.

 

You are of course correct about proper cooling, but why is cooling important? why do CPU's thermal throttle? Its not to prevent complete outright failure but its to avoid the heat degrading the chip which will then reduce its performance. Just google "CPU heat degradation". This is also the first segment of buildzoids video. I think you only need to watch 20mins for this to be reasonably well explained. 

 

"don't take your overclocks too far, there is very, very little danger of a CPU failure occurring" - I agree but as overclocks go further the go from safe, to degrading to, then outright failure if you go high enough. The latter is unlikely and requires some fairly silly inputs. The real danger is the degradation zone.

 

Maybe we are discussing the same point, anyhow this is a bit of a thread highjack. Apologies. In short to the OP, it can be worth OC'ing really depends. Be careful.

[ShaunX]

On 9/22/2021 at 3:40 PM, sirrah said:

How can I best create some sort of a benchmark mission to test if the overclocking has any effect?

 

There is actaully a very accurate method of benchmarking in DCS for VR. I got this method from Speed-Of-Heat's's excellent 3090/G2 thread and use it all the time. It does require FPSVR.

 

1. Fly a simple mission, quite low over some mixed terrain for perhaps a minute or so. Save the Track file from the mission. That's your baseline file.

2. With your Headset on it's stand or hook (Not on your head). Launch FPSVR and DCS in VR mode. You MUST run this process from your PC screen so that your headset view doesn't change. Obviously press the VR recentre button so you have a centered view on your screen.

3. Make sure you turn motion smoothing/reprojection OFF (otherwise you'll just see mostly 45FPS in the results)

4. Launch the baseline trackfile in Replay.

5. As the replay starts click CTRL+F11. This initiates FPSVR capturing FPS and Framerates every fraction of a second.

6. When the track finishes, press CTRL+F11 again. FPSVR will write the results to a csv file. (The output folder is defined in FPSVR settings). For a minute long track, the file will have between 7000-10000 rows, a pretty decent sample.

7. Open the csv file in Excel. You'll need to trim off rows at the start and end of the file where the rate was settling and after the mission ended. It'll be obvious how many to remove by their values.

8. Use edit/replace to globally remove all the " characters from the file (otherwise calculations won't work)

9. Use =Average to take average of the FPS, GPU framerate and CPU framerate columns.

 

And there's your benchmark results.

I usually run the track file 3 times and take the average from the 3 runs.

 

Now make your changes and rerun the process.

 

I find this gives really accurate deltas between any changes I make, mods I add, or between DCS OB updates.

 

It's a little cumbersome but once you've run a few times it's easy to repeat.

 

If you're looking for an extreme stress test, obviously create a more complex mission/track file (or just fly Marianas )

Edited September 28, 2021 by ShaunX