I made a pilot to fly a regular campaign mission in order to take more hardware tests.



Campaign Date: 7 March 1918 8:48am Flying for Britain in RFC-46
My ride: Camel
Wingmen: 5
Patrol Enemy Front Lines
Air start over target
Active war
Total Flight Time: 10 minutes

Heavy air activity
Highest workshop settings

Average GPU temp 39 C

Average CPU temp: 47 C

GPU usage Average: 29%
CPU usage Average: 20%

Framerate locked to 59.

Frame drops: I don't understand why there are so many minor frame drops. The mission was not stressing my system at all!

.

Orby, I've a hunch those FPS dips go right back to that old CFS3 pipeline. You can only push so much through it so fast, and sometimes it still won't be enough to keep up with all that beautiful action and eye candy going on in WOFF.

.

I'm 100% with you on this Lou. Just really a shame this happens on systems that are easily capable otherwise.

Orbyxp, your conclusion regarding frame drops, GPU/CPU loading is consistent with very similar testing I've done. Even using a fairly mid/low end GPU, I've seen 50% or less actual load. I don't know why the frame drops occur either (I really wish I did), but this clearly illustrates that even conservative systems are easily capable, but something is going on that causes FPS drops even though the system isn't really loaded much at all. Thank you for showing this with the clarity you have.

Edit: I wanted to say, looking closer at the graphs, I'm not sure how "minor" those frame rate drops really are (assuming you're referring to what I think you are; the negative spikes that are visible on the graph). For example, although it's not happening all the time, there appear to be a couple pretty severe dips. They don't look too bad, stretched out over time like they are. But the scale, if I'm looking at it correctly, is actually 200 FPS, and your frame rate is locked at 59. However, zooming in on those spikes shows they're >10% at times, and the couple worst ones look to be around 30% (drop in actual frame rate). This is going to be a very noticeable drop, and nothing I'd call minor TBH. The horizontal grid is what, 5s/div? Looks like some instances where there are drops >10%, several times a minute.

Also, I'm sure you know any measurement tool has resolution or sample rate (how often it's actually recording a number, and how much time that recording takes, itself). So I'd even go so far as to say that, when these drops are happening, that by the time it's being sampled/recorded, it could possibly have been on the 'upswing' from a much lower number - and I'd be willing to bet that, on screen, the perceived pause is longer (and maybe 'deeper', losing more FPS) than even what the chart shows at times.

Thank you for this post/thread Orbyx, since otherwise I never would have found out about "Afterburner MSI." I did use some level-2 or level-3 cache overclocking tweaks back in the good old days of Mac OS 9 and earlier - but never bothered to read up on this for Win - and also because there aren't any good software overclocking programs, that I know of, and that are designed to work on Macs that feature integrated Intel vid. cards. Now that I'm running WOFF on the Mac Pro - I have a short technical question for kksnowbear, our resident specialist in computer components, installations, etc.

Is it safe to (slightly) overclock the main clock and memory on a workstation vid. card - specifically on the FirePro D700 in my case (the consumer equivalent seems to be the FirePro W9000, by the way). I won't be tampering with voltages, and I am only overclocking within the "factory overclocking band" in Afterburner - nothing fancy.

Default settings are:

GPU clock: 850 MHz
Memory clock: 1370 MHz

Tweaks that I've managed to set, successfully - but have not tried them out yet in WOFF and FE2:

GPU clock: 920 MHz
Memory clock: 1445 MHz

(Further info.: I will not tamper with overclocking the CPU since it boosts up to 3.9 GHz anyway when necessary, 6-core Intel Xeon - and I don't want to damage it.)

Hopefully my "overclocked" specs. for the two AMD FirePro D700 cards are not excessive, but I can always drop them to something like 900 MHz for the GPU, and 1420 or so MHz for the memory clock.

(The factory limits, by the way, within the factory overclocking band, are 950 MHz for the GPU clock, and 1475 MHz for the memory clock.)

Thanks in advance for any comments/info. that you gents can provide. Always much appreciated.

Von S

P.S. Fuller specs for the vid. card (https://www.techpowerup.com/gpu-specs/firepro-d700.c2555).

Well, thanks for letting me answer...and (for a change) I'll try to offer a short answer (but we all know that, in my usual idiom, a long answer is coming right behind it )

So, here we go: Is it safe to (sightly) overclock the main clock and memory on a workstation vid. card ?

In a word, "Yes".

And cue the long answer...

It's arguably 'safe' to overclock most anything. The differences come from how much of an increase in speed, for how long, with what enhancements in cooling if any...blah, blah, blah.

But, here's the thing (at least in my mind): If your overclock is so mild as to not ever require any voltage bumps, etc. the I'd submit that, reasonably, any performance gained is going to be perhaps negligible. Let me clarify this ("Atta boy Luther..." ):

If I take a CPU and get a (very aggresive) 30% overclock out of it, it will pass the various benchmarks at roughly 30% higher (depending on various junk) because that's what those benchmarks do: Measure that performance specifically. However, the overall increase in the computer's performance, as a function of noticeable difference to a human user...well, that might run around 5%. The reason is that the CPU maybe have bumped up 30% but it's contribution to the computer's overall performance isn't that much to begin with. Is it more than the SSD? Yup. More than the GPU? Usually not. So you see, if a CPU is 25% (total) of a computer's gaming performance ability, and you increase that by 25% with an overclock, then overall the gaming performance only gained 6.25%. And that's going to be difficult to notice - it doesn't even necessarily translate directly to 6.25% increase in teh almighty FPS (which is misleading of itself, IMHO).

And there's no way you're getting 30% from a CPU without increasing voltage.

That's with CPUs. TBH I have limited experience with overclocking GPUs, and even less with Macs I have done it with GPUs - using the same Afterburner tool orbyxp has (great tool BTW). But, I still feel fairly certain you're not going to see anywhere near 30%, or even half that, in a GPU overclock - especially without a voltage increase. Voltage bumps are all but hand-in-hand for overclocks, so without voltage increasing, I can't imagine getting very far. I've only been able to get a few percent out of it the few times I tried, and I was actually increasing the voltage slightly. (And stability at load is a whole 'nother discussion).

Now as with any component, there are notable exceptions. And most GPUs these days (if not all) have a complicated on-board management system to allow 'automatic' overclocking within the constraints of thermal management. The card's firmware will temporarily bump speeds up with loads if it can within thermals, and bump then down as temps begin to increase (all the while, controlling fan speed as well). But even with that, you do still have some ability to overclock GPUs a bit further. SInce GPUs typically are a bigger factor in overall gaming performance than CPUs, then a smaller GPU bump can arguably make a bigger difference than a bigger CPU bump. But really, none of this would be likely without voltage increases.

For me, though, I'll tell you: The performance you can get from a good CPU overclock is (arguably) worth what has to be considered (temps, increased cooling costs, stability and the testing needed...). But I hardly ever try to overclock memory or GPUs anymore because a. I don't seem to have time for what I could get out of it, and b. Given a., it doesn't seem 'worth it' to me. Out of 350+ records in my 3DMark spreadsheet o' results, I think maybe 2 or 3 have overclocked GPUs (and I'm pretty sure one was aggressive enough I wound up bumping it back due to rare stability problems).

If you really really want to get every last drop of got-go out of your setup, then you'd have to do the CPU, memory, GPU, maybe LN2 or even oil submersion cooling, etc...but you're certainly not going to get all that without bumping up a voltage

I hope it helps.


PS: Looking at your figures, you've gotten roughly 8% overclock on the GPUs, if I follow. What do the thermals look like? Benchmarks? And (perhaps most of all) any stress testing to confirm stability? Sometimes remarkable things are possible, but I usually test any overclock for at least 8 hours and sometimes 24 or more.
EDIT: Fat-fingered calculator and did your GPU overclock % wrong. It's closer to 8% than the 10 I originally wrote. Which, incidentally, reinforces the questions re: what does it do actual performance (synthetics are best here, because even if they're "not real world" they're going to be uniform in terms of what's tested and how, as well as the empirical result - as opposed to often subjective human observations)

Much obliged gents and thank you KK for the detailed and thorough response. Will look into ave. fps and gpu temperatures later today to see if this minor overclock is worth it - in stock mode, temps. average around 80 degrees Celsius if running at max settings in a flight sim, for about an hour or two.

Von S

No problem I hope it helps. If it will work in your environment you should seriously look at a gaming graphics benchmarking application. 3DMark is sort of a standard and there's a free version.

Without such a program, there's little way to do accurate, apples v apples comparison testing. The old "I get x FPS by changing this setting" is grossly subjective and almost impossible to use for any meaningful comparison because there are too many variables that aren't held constant, and our eyes just are not fast enough or sensitive enough to accurately measure anything (much less document it). This is why I say use synthetic benchmarks. You can always do FPS in-game testing as subjective augmentation to proper quantative testing.

I always use to measure WoFF onscreen, while using Afterburner, only I never did use FPS nor one graph for the complete CPU.
Switch to frame times and let all your CPU cores log the load. Be aware when you find a peak from around 90% on one core this core could be overflowing. There is also a logging between time if I still can remember.
Watch what happen ingame if something does peaking, like CPU peak is there suddenly an massive AI activity. I can remember something that when I was flying low at lots of ground activities it was always fluctuating a lot.

If not did this, turn on the RAM’s XMP in your bios, free extra performance, and see what happen.

read this about frame times: https://techreport.com/review/21516/inside-the-second-a-new-look-at-game-benchmarking/

BTW, What is your system hardware?

I agree that the frame drops (typically visible as micro stutters) is something in the underlying CFS3 engine. I've chased after eliminating these for years, and no amount of hardware upgrades and tweaking of settings has ever made them go away entirely.

They are most evident when flying low to the ground looking perpendicular to the direction of flight. This has the image on the screen needing to blit big chunks of data laterally with large relative offsets. I suspect that it's related to a memory buffer clearing routine in the game engine that is not optimized to take advantage of today's hardware capabilities. (e.g. much of my GPU/CPU memory sits there unused when flying)

In some of my test scenarios I've been able to get the hiccup to occur like clockwork at precise time intervals, as the buffer fills up to a limit and resets. Changing the level of detail, altitude, or flight speed changes the frequency of the interrupts, but maintains their consistent pacing.

This is such an incredibly precise description it's almost scary. I have to be honest, even though I have done lots of work with software and developers, I've no professional experience with graphics/game engine software - so the part about blitting etc is kinda outta my league. But, I can absolutely see what you're trying to describe, and this is (by miles) the best assessment of the situation I've ever read. In fact, I recall the first time you posted (something similar), I read it thinking "Hallelujah, I am not losing my mind, and it's not just me or my hardware..."

A couple of specifics I'd like to comment on:

One, your remark regarding hardware upgrade and tweaking of settings: I think this remains entirely true. I don't know what your current system is like, but the issues I can replicate (see next point) haven't shown any change regardless of whatever system I've tried on, all the way up to and including i7-8086k / 1080Ti (which are both very capable devices, even by today's standards). Sure, the system and the sim run better overall compared to more modest setups, no doubt. But it appears to have zero impact on the actual frame drops we're discussing.

Two, concerning the test scenarios: I've done exactly what you describe, with scenarios that precisely show the drops to occur at the same points. I've described many times how it's necessary to have such a test scenario that can do this, predictably and reliably, in order to do testing with any real, meaningful conclusions. (Among other things, If there's no "before", how can you test an "after"?)

I am honestly just extremely grateful that your experience and comments seem to be exactly what I've tried to communicate. Whether it's the CFS3 engine (I'm sure we all believe that it is) has never really been my concern, though I can say I saw this behavior way back in CFS3. I just wish it weren't happening now.

FWIW I do sincerely wish that OBD's obvious talent, passion, and skill weren't affected by this, and I do still support their efforts in the most meaningful way I can afford. I think that such support can exist, while still acknowledging some stuff we'd all like to see gone.

I genuinely appreciate your input MM. For reference, would you be willing to share your system details (and/or that of the units you may have tested on), please?

EDIT: Something else I should've mentioned...some time ago, I decided to do a test to rule out storage access time as a cause, having already upgraded to various storage setups many times faster than standard hard drives. I actually set up a RAM drive on my (32G) machine, and it benchmarked exactly as expected: Not HDD speeds, SSD speeds, or even NVMe speeds - it ran at actual system RAM speed; easily 9-22 times faster than a Samsung 950 Pro NVMe SSD depending on the test. I set up the RAM drive to create a volume big enough to hold WOFF, and ran the entire game from that RAM drive (which is also functionally equivalent to any 'preloading' of sounds etc, because this way, the entire thing is loaded in RAM.) But I still saw the exact same drops, at the exact same places.

KK and MM

Thanks for all the effort and data you have brought to this thread. Most of it is over my head but the link MM provided is very detailed and easily understood. I too am beginning to believe that the CFS3 engine is the culprit. That said, some of us are achieving levels of performance that are acceptable to us as a result of various tuning techniques and better hardware choices.

I still get some mixro stutters but I live with and accept it in favour of all the pleasures of the sim.

I too wish that OBD could get license access to the engine source code in hopes that the code could be improved but that is not likely to happen. I do fail to see why MS holds the license so close when the CFS3 is no longer marketed.

Just my two cents worth.

Thanks folks!

Best Regards

Robert,

I believe the point you're making is true in that there are also some issues that tuning etc *can* improve. But I think it's important to distinguish between the "lesser" problems that might dealt with by tuning, TrackIR tweaking, better hardware etc... and those that seem persistent regardless. I've long since been able to "tune" graphics very well; I don't have (nor would I tolerate) stuttering from overclocking or whatever else. It's the stuff that isn't changed by any of the things we can do that has always been a concern for me.

Acknowledging the difference between these two types of issues saves people beating their heads against the wall to no avail, or even far worse, heaving money better spent elsewhere toward a problem it isn't going to solve.

Moreover, I believe it actually in *any* game developer's best interest to advocate lower system requirements, and thus increase the potential market, by keeping the 'price to play' down. I'm all but convinced even very modest systems will get as much improvement from this sim as you're going to get, and that anything beyond that is cost that's progressively disproportionate to the gain. Just as an example, I feel you can have a <$750 machine that will do just about as well in this sim as a machine costing three times as much can do in this sim. A third the cost, for the vast majority of the performance, and neither will be completely free of the remaining issues anyway**

I hope you can see what I'm trying to say.

**EDIT To be absolutely clear, this has nothing to do with new or used hardware, or the source one chooses for it. It's strictly an observation of cost v performance.

KK

I understand you point and I have no argument with your views. I merely wished to state that I found the link by MM very informative and that in my particular case the performance of WOFF acceptable to me, that is not to say I don't wish it was better because I do. I receive considerable pleasure from the sim and I am willing to live with the small stutters I receive when flying it. That of course is my particular stance and others may find it unacceptable. To each his own.

I also was just acknowledging that I also believe the CFS3 engine is the primary culprit and that I am somewhat puzzled that OBD is not able to acquire licensing for the source. If it is no longer being developed I don't know why MS cannot offer it as open license for other developers to build on. Just my thoughts!

Best Regards

Robert, good, no argument here, and certainly no offense intended. Just civil discussion, far as I'm concerned.

I think everyone shares your curiosity regarding licensing, engines, etc. - but it's regretfully not a subject OBD has had much to say about thus far. I wouldn't even know if there's been any contact with MS (strictly as an assumption, I doubt it). I do feel, though, that if anyone approaches MS formally about using any part of their product, I wouldn't be at all surprised that they suddenly insist on getting paid - even if not being actively developed, it's still their intellectual property. If no one says anything, they can choose to look the other way if they want. But if it were to come up, they might have grounds to make a fuss. A lot of this does depend on 'source code' and whether any of the modding could've been possible without 'reverse engineering', that I do know from related experiences with software licensing in other industries.

Again, as far as CFS3 and MS, this is all speculation; I do not know any of this to be fact.

One thing - I'm not sure I see the MM link you've referred to. Did I miss something?

KK, Sorry, My bad, That link was in Dutch post above. You probably saw that.

Very interesting article indeed, thank you. It discusses in detail something I've mentioned often (above, in this same thread, in fact): The fallacy of (what I call) the "Almighty FPS". This article shows that FPS alone is misleading and modern measurements don't usually rely strictly on an FPS number.

I do have to say though, that changing one's memory to XMP isn't likely to make a tremendous difference in the 'final' on-screen experience. Unless it's already *very* poorly setup (as in, worse than even BIOS defaults), XMP can potentially increase RAM speed a fair amount - but the effect on the system overall might do well to change by 2% in a standardized test. And, as discussed elsewhere, just changing this setting by itself can introduce instability. I've never set XMP alone and had it pass MEMTEST loops without failing, unless I manually bumped the RAM voltage. I am sure this is because the memory manufacturers don't make the motherboard, and vice-versa. It's also less likely to be stable if you populate all the DIMM slots, due to what's called "fan out" in electronics.

XMP is supposed to be one-click, but it doesn't always work that way. Of course, YMMV and it's going to vary with board mfr and memory mfr, too.

Currently I'm running a water cooled Ryzen 7 3700X with a GTX1070 Super and the program is loaded on an NVME SSD.
This set-up lets me run a 3440 x 1440 Gsync monitor at up to 100MHz with 4X SSGA which would not have been anywhere near possible with any of the previous gaming systems I've put together over the years.
The image quality is outstanding, and the game plays with great responsiveness, but those little frame pacing hitches persist.


I've chased after the HDD versus SSD versus NVME versus RAM Drive idea, and like you, found it was not a relevant factor.

I've isolated / optimized the mouse, joystick, and TrackIR activity to eliminate the usb interrupts as a source.

I played with single core versus multi threading processors to no effect.

Running at half refresh rate so both the CPU and GPU have plenty of excess capacity does not make them go away.

I've gone the other way and tried loading the system down with as much to do as possible (8x SGAA, High Terrain and Object Density, etc.). The system never shows a hint of slowing up from its work, and it still happens,

I've also explored all of the various FPS limiters like RTSS, or those available through Nvidia Profile Inspector, as well as the one built into the game engine itself, and their routines for regulating the pace of displaying each frame doesn't resolve this.

I've even tried a frame pacing routing written for Reshade that was specifically intended to make everything buttery smooth like a game console, and it failed to clear it up. (though I'll admit that the FPS timing was absolutely rock steady between the little hitches)


I'm satisfied that I getting all I'm going to out of this old bird. After all, I'm an engineer, so my diligent efforts at trying to determine the root cause of this behavior has been just my cup of tea.

Please, don't forget, this is typically going to be most noticeable when you're just tooling along looking around at the scenery.

When you're doing a white knuckled check six in the middle of a dogfight, you're hardly ever going to notice a little hitch in the display.


Get a strong enough system so it looks good and you're not getting too bogged down, and then just go have some fun!

Very interesting discussion. I hope my enjoining does not throw it off the rails. To that end, let me start by offering KKSnowbear a public apology for my recent disgraceful behavior. No matter how much passions are stirred, there's no excuse for that.. I am sorry KK.

Major McGee, I concur 100%. You have used the term 'hiccups' to describe what you see. I've used that term myself in the past and I'm very relieved that it appears to have been accepted in this thread as a legitimate term for describing a key issue with the CFS3 engine. After all my tuning and experimenting I do still have those as well. Typically there's one hiccup I can set my watch by at ~ 93 seconds into the mission (it varies a little depending upon the campaign region) followed by one or two less dramatic ones, until around 200 seconds in, from which point I will seldom see any more of them through to the conclusion of the mission. Sounds very similar to what MM is describing.

Now here's the part where I hope the discussion can stay civil: if those occasional hiccups were grouped together to become stutters every time I was in low combat,, i.e. the issue that KK's original QC showed so well, I wouldn't be able to play this game. But they don't, for me at least, no matter what terrain I'm over, or the weather, or anything. Sure, sometimes maybe not 100% smooth, but still very playable. As the Major says, "When you're doing a white knuckled check six in the middle of a dogfight, you're hardly ever going to notice a little hitch in the display."

Whatever terms we use for less than 100% smooth performance, It looks like pretty much everyone agrees that it's the engine that's the root cause. I would say that if you can put up with those hiccups and you are prepared to put the time and effort into it to make everything else very smooth and playable at high detail settings, this sim has an abundance of enjoyment to offer.

Thanks to Lou and Stache I was able to configure OBS to get a pretty decent 2K recording that's very close to what I actually experience on my monitor. No question there's a bit of a performance hit with OBS but it's ok for demo purposes. I now have a video that I'll upload later that shows those early hiccups very clearly.

Again, hope this stays civil.

Thanks all.

I7 9700
2080 ti

By the way, all this reference to not needing a modern CPU does omit some differences. A modern CPU can do some things that simply can't be done with older CPUs:
1. Fly at 60 FPS (90% of the time) in 1918 on heavy air activity. So, no need to fiddle with air activity once 1918 starts.
2. Fly at 60 FPS with x12 time compression at ultra heavy in 1917. Gets to target quicker.
3. Any Frame drops last a second or less.
4. Precipitation in any year with any air activity has zero effect on FPS. Stutters occur at the same times with or without precipitation
5. M.2 (ssd or nvme, not cpu mostly) loads mission faster. E.g Ultra Heavy in 1918 missions spawns your aircraft on the field in under 1 minute.
6. What about when you want to play another modern game, then the differences become even more pronounced.
So, if none of these make any difference, then you're good.

On the subject of omitting differences, there are some pretty significant ones you continue to omit, for example:

1,2: I haven't specifically tested time compression, but I also don't know there's any proof that it's mandatory to 'fiddle with air activity' just because you don't have a 9-series CPU. You have proof of a requirement to fiddle with air activity settings for everyone who doesn't have a 9-series CPU?

3. This is completely misleading. I've rarely (if ever) seen a frame drop last more than 1 second, regardless of the CPU (at least going back to some 2nd and 3rd gen i7s, which I have tested not that long ago). Even these CPUs still show the same drops everyone has seen, yourself included, essentially regardless of CPU. The drops last .25-.5 seconds, maybe sometimes slightly more. Your statement "Any frame drops last a second or less" applies to any CPU I've seen/tested since 2nd gen (like Lou has) - not just 9-series CPUs.

4. "Stutters occur at the same times with or without precipitation" I haven't specifically tested precipitation. Again, I'd welcome proof of your specifics. If you're saying precipitation doesn't worsen the specific frame drops we're discussing in this thread, I think it's been accepted that drops still happen with a 9-series CPU, just like other CPUs. Not sure what impact precipitation has on this; but if they all show it....well, they all show it. Maybe it gets worse still on older CPUs, but I've never seen proof of this - the stutter seems to be what it is, regardless.

5. Again, this is misleading. I was using verious methods of storage beyond a simple SSD 10 years ago, and have been ever since. I always make suggestions that system builds for gaming include 'performance' storage, which can vary depending on platform.

First of all, M.2 is nothing more than a form factor. It is not inherently faster than even plain SATA drives; in fact, some M.2 slots actually still operate at SATA 6Gs speeds (no faster than a 2.5" SATA SSD). Long before NVMe capabilities on boards (starting ~Z97), I was using/building RAID arrays with multiple SSDs on real, hardware RAID controllers (with very little CPU load by design), and getting 1200+ read speeds when SSDs were still second-gen (SATA 300G/s) And many gamers were still using standard hard drives, many (including some here today) claiming it "only improves load times". This is obviously incorrect, and since that time, almost everyone now says to use SSDs (funny, since many at the time insisted it didn't matter).

Also, most any board since ~Z68 will have PCIe 2.0 x4 slots, which means cards that can deliver 1500+Mb/s read rates, or >3x faster than SATA 6G/s. My original NVMe drive - a very fast Samsung 950 Pro - gave read speeds at ~2000, so I can do 75% of that speed for no more in cost, on a board that dates back to 2nd-gen processors. Done it, in fact, many times. No doubt NVMe is fast; my newest Samsung delivers up to about 3200 Mb/s, but it (and the hardware required to support that speed) costs a lot more than a 2nd-gen Z86. Like I've said many times, a high percentage of the performance at a much lower cost. Whenever I build a system for gaming, storage performance gets plenty of attention, and capable hardware recommendations. Still, it's up to the person paying for it. In spite of the benefit to gaming, fewer people decide to include these units than really should, and for one reason: Cost. Many have opted to run games from even standard hard disks, to save money. While I can (and do) recommend otherwise, what this illustrates is that for the majority, you cannot exclude cost from comparisons.

6. The CPUs I've mentioned here, and recommended to actual clients, are more than capable of modern games. For example, before my son's recent upgrade to an overclocked 8086k/RTX2070 Super, he was using an overclocked 4790k/GTX1070 and was keeping very good frame rates on high settings, and while streaming, no problem, with very demanding recent titles. Certainly more than this sim requires, per the evidence we've all seen that CPU load isn't typically high at all in WOFF. (And that's only a 4790k; I've also mentioned others into the 6 and 7 series as well; which can be had for much less than newest retail costs).

So, even if the things above make a difference, you can still save money and get performance that's reasonably comparable, and a better value for the cost...unless, as I've said before, money is no object.

Once again, all this might be accurate if you're ignoring cost. But cost is a factor for almost anyone. All these discussions about how the latest CPUs are 'best' are really without merit, unless someone can actually afford to spend $500+ on just a motherboard and CPU (no memory, no drives, no PSU, no GPU...which collectively can push the cost upwards toward $2000). But the performance still isn't likely to be in proportion: You're not getting twice the performance for twice the money. You might get 15-30% more for 200% the cost.

I had a 6700k/1080 Ti before my current setup and played WOFF UE with benchmarks to death, so the comparison I'm giving is from my own experience. If my own experience is not enough, then I can't give any more proof than that.

Steam has games like the latest Tomb Raider with it's own benchmark which measures CPU/GPU performance. The latest Hitman also has a benchmark which measures FPS. There are many more out there. So, yeah, I've seen 200% improvement on modern games with max settings when I went with the 9 series/RTX Ti. Again, my own experience from owning a 6700k/1080 Ti.

EDIT: Also, I'm referring to M.2 PCI Express SSDs (20Mb/s) which the standard M.2 SATA SSDs (6Mb/s) can't compete no mater what brand or speed.

EDIT2: Ask anyone on this forum with an older CPU, 6th gen and below to verify if they can run 1918 with Ultra air activity at 60 FPS. I couldn't do it with my 6700k. So, if I started playing on Ultra air activity in 1916 and wanted 60 FPS, I'd have to lower it a notch in 1917, then two notches in 1918. That's my proof.

EDIT3: To cut a long winded story, and from reading the comments in this thread, I think this discussion is going nowhere. I made this thread in an attempt to figure out where the stutters were coming from, but I can't. I've decided not to pursue trying to figure out why there are stutters and just focus my time on modding whatever I can.

I asked if you had proof that fiddling with air activity is required for someone who doesn't have a 9-series CPU. Did you ask everyone who doesn't have a 9-series CPU about this? Also, there are many CPUs that are better than a 6700k without being a 9-series.

Just because the SATA M.2 drives are slow as Christmas, doesn't mean by any stretch that there aren't any other options that are faster. I named at least two. (By the way, BIOS have been modified to run and even boot NVMe storage on many older boards as well, going back to Z86...and if you have a PCIe 3.0 x4 slot, that means high speed NVMe without an M.2 slot at all.

In any case, you said "If none of this matters", and I've now shown that even older hardware can absolutely compete with CPU usage, storage speeds, and other factors, cost for cost.

You're omitting things that don't fit your perspective, like the fact that PCIe drives that are faster than SATA SSDs have been around since way before NVMe. Just because not everyone is familiar with the details of this stuff doesn't mean there aren't other alternatives. Like I said, I can and have put many PCIe SSDs in much older machines that don't load the CPU by design, and still get much better read speeds compared to a SATA SSD and without CPU overhead. (I didn't even mention RAM drives that will absolutely put any NVMe storage to shame, with 15x the read performance).

But, as I've pointed out many times, cost is always a factor, and buying the most expensive isn't always the 'best'.

No apology is necessary; I do sincerely appreciate your gesture. I try not to take this stuff personally, and I really wish no one did.

I strongly agree with MMs first post (this thread) regarding his own assessment of the stutters.

I don't necessarily agree with the concept that you're not going to notice it if you're in the middle of things. First, (I think) he's not saying it doesn't happen, it's merely his own assessment of whether it's noticeable. And that's going to vary by not only the individual, but the situation. Also, he doesn't say you won't notice; rather that you're hardly ever going to notice.

What's the difference, you might ask. Fair question.

Here's an example: >LINK<

At :31, a stutter occurs, causing the player to be unable to control or react to what's happening in the game. This is a fairly big pause, and more important, when the game resumes normal behavior, the player is getting pounded by gunfire that he could've otherwise potentially evaded.

I don't think this can be fairly described as something you're 'hardly ever going to notice'. There's no horizontal view movement, no head-turning, and this isn't during startup. It's just happening, at random. It adversely affects the outcome, regardless of player skill or action, and in fact the player has no control over it. Even if you'd panned your view around 10 seconds before, this can and does still happen.

This perfectly illustrates the problem with these drops: It causes you to 'lose track' in your input responses such that they are no longer appropriately timed/sequenced around the actual situation. Pilots refer to this as "situational awareness"; it is an absolutely crucial part of combat flying, and when this happens it's taken completely out of your control.

Now, maybe this is someone else's idea of something you'd hardly notice, but if we assume survival is everything in this sim (as it was in reality), then I have a hard time seeing how this could be overlooked.

As for remaining civil, I know of no reason it shouldn't remain civil. I discuss, at length, but within forum rules; I try to present evidence and cite fact, I sometimes give opinions (noted as such)...and we can all do this, indefinitely if desired...but none of that is uncivil, and it doesn't ever have to get that way as far as I'm concerned. I think vigorous debate can and should be expected, but no one has to get personally offended.

I don't claim to be perfect, and if I actually break forum rules then I'm sure it will be called out. But I don't believe it's necessary for anything to be uncivil.

I am absolutely certain your efforts regarding the stutters are appreciated. As I mentioned earlier, I am grateful for your post showing the graphs, and said so back on page 1. I'm truly sorry you can't figure out where the stutters were coming from, because I'd truly like that to happen. Believe me, if it was apparent the solution was that everyone should buy 9700/2080Ti systems, I'd certainly have no trouble making that recommendation. As it stands, I'm dealing with guys who want systems for less than a half (or even a quarter) of the costs.

KK, that video clip is 5 years old. It shows a perfect example of what I refer to as a hiccup, albeit a very extreme one. While the CFS3 game code has remained static, a lot else has changed in that time: hardware performance has improved. The WOFF product has improved. Peoples' efforts to subdue the annoying hiccups / stutters / etc., have improved. Ever since WOFF Platinum was released I haven't experienced a single instance of anything like as severe during a mission as can be seen in that video, no matter what the weather, the date, or the plane. I've put a lot of stick time in with campaign missions. People have posted dozens of videos in the current campaign thread. I've watched most of them and I have not seen a single example of anything even close to that moment in the video you linked to. This was true for me on my GTX 1070, and unsurprisingly, it is equally true with my GTX 2080.

I've seen your offers to help newcomers on this forum with hardware setups. Can you give an rough example of a budget setup that you would typically recommend? Along with this, I'm curious as to how you set their expectations WRT playing WOFF PE on a budget setup?

Thanks

H

Very true. That is the key. I don't believe you're likely to get the same fluidity from a 9th generation CPU, PCIe SSD, RTX GPU and 240hz monitor as you would with an overclocked 6th generation CPU, SATA SSD or HDD, GTX GPU and 60hz monitor. .

Monitor hz is often overlooked as a major factor in game performance. Search YouTube or Google on 60 vs 144hz or 240hz. It doesn't only apply to FPS games. It even makes your OS smoother!

HH

You and I already discussed the matter of how often the drop occurs during a flight. As I've explained, the problem with looking at it that way is that it can happen any time; there is nothing saying it can't happen more than once in rapid succession, and you essentially have zero control over when it does happen (or how long it lasts, or when it happens again). Looking around, even all the time (to 'load textures') doesn't solve this problem, because airplanes are pretty much required to be moving forward all the time (else no lift). Hence, you encounter new textures. You can look around 360 degrees, and still have this happen a second later.

The issue with all the other tests you mentioned that don't show drops, is - and this was actually pointed out in your test thread - that none of the videos show any drops to begin with; therefore no one can say whether anything was there to 'solve' in the first place. Everyone agrees (I believe) that drops are caused (or at least worsened) by 'texture loading' (just as a generic expression). So, it absolutely follows that whether/how severely any single flight encounters the drops will depend on factors like where/when you're flying, and what else is being rendered/displayed. That means unless you duplicate exactly what's in that first video, you aren't necessarily going to see the same drop. And, if the videos don't show stuttering to begin with, then there's really no proof it would've been present in that specific scenario (It could've just been the other factors, which we all understand will alter the effect; how severe, etc).

But let's set that specific video aside for a moment. Here's a newer one (about a month old): >LINK< . Look around 1:08. I would infer this is the most recent WOFF version, since it was posted by one of the OBD team. And I seem to recall Pol saying he had a fairly powerful updated machine, although this doesn't specifically say it was created by him or on his machine. Yes I know it's WIP etc, but that has no bearing at all in the context of frame dropping. Granted, the pause isn't as long as the one I linked before, but this is more likely because of the exact environment, textures, etc than differences in the system would reflect, IMO. In any case, we can't say it's not there, and we can't say that any difference was strictly due to the machine.

I've tested with a range of CPUs over the years, some many times stronger than others, and I've seen little to no change in these particular frame drops. And, in spite of what some might claim, even the latest 9900K setup isn't going to perform at a level overall that is 200% more powerful than some of the 6th, 7th and 8th-gen setups I've used (there is plenty of evidence of this over at the 3dMark website, BTW).

To me, it's exceptionally unlikely that I could run 15 different CPUs, some more than twice as strong as others, and see zero difference in the drops...and then, just by going to a setup that's (at best) maybe 30% faster than ones I've already tested, the drops suddenly disappear. Certainly, the range of processor performance overall would be (at least close to) linear with respect to drops...IOW, I'd expect a CPU 2x more powerful than a 'baseline' to be less likely to show drops, and one 4x more powerful to show even less.

But, your theory about the latest CPUs implies that the rate of improvement stayed exactly the same (near zero) through 6 generations of CPUs, then suddenly jumped through the roof with the latest generation. That's not consistent with the scale increase in their relative performance, and it doesn't make a lot of sense to me, for that reason.

To be clear: I'm not saying I don't believe the newer hardware won't improve the game at all - I certainly believe it will. And, I could see (a number of reasons) someone might believe the newest components would overcome these issues. However, what seems to be missing is an accounting for the "scale factor" I described above, particularly when people other than myself have already stated they have the same drops at times, in spite of some pretty high-end hardware.

If you please, the question about an entry level budget system is a great question, but I'll need to get back to it. I am actually looking forward to it, because it is a very relevant question (perhaps *the* question IMHO) just a little pressed for time ATM.

HH, cont.

As regards the budget setup and expectations - again, let me say this is a great question.

In the recent past, I have spec'd out a couple builds/upgrades that wound up using 3rd-gen CPUs, GTX780s, boot SSD/storage HDD, and 8G DDR3 RAM. I can do/have done systems like this for $400-500+, depending on exact CPU/GPU, exact memory, exact chassis, size/type storage, custom storage, and cooling setup. I have first-hand test results that show roughly 70FPS average, and that the CPU/GPU are not loaded much beyond about 50% with all settings maxed out (which kind of echos my own feeling that more isn't really helping with drops much, overall, and certainly isn't necessary). These results have been corroborated across multiple systems, and by others using systems I worked with them on, in addition to my own observations. This seems to make that level of system fairly reliable as a good budget performer; subjectively, people seem to be pleased with what they get.

Part of the reason I like this question so much is that it goes to show exactly how conservative a system can run the sim with decent performance (and, at least anecdotally, without any tweaking or mods being necessary). Interestingly, this is also close to what OBD says on their website for a 'recommended' system. And I firmly believe it's in everyone's interest that such minimal requirements can be effective. Really a bad idea, IMHO, to give the impression you need thousands of dollars of hardware to get decent performance out of any PC game.

As far as expectations go...

Well, if I'm being honest, I am essentially compelled to specifically mention frame rate drops. I will usually say something like this is a system closely matched to the required/recommended specs; testing shows it will yield in the range of X-x FPS. I am usually completely forthcoming about hiccups, frame drops or whatever we're allowed to call them.

Depending on the level of interest, I usually try to provide as much detail - including test figures - as I can. Some don't care about all that, and I try to gauge this during the 'pre-sale' discussions. For example, if a guy's coming from a complete slide-show, he'll be tickled silly with 70FPS, and the occasional stutter isn't likely to bother him. But I have also built very fast machines for gaming, and if there are going to be issues, I'd be wise to bring all this up in advance. Especially as the price goes up. And this is why I'm so interested in it, because people aren't just paying me for hardware. I have to advise them on how to get the best for their money, and sometimes that means saying "You're not getting past that problem, regardless of how much it costs, because..." Or, "If you want 144FPS consistently, you're going to have to spend more than $400..."

I do have to be somewhat concerned with performance-related issues, because I offer money-back guarantees, which could kill me if someone's not happy. I make nothing to begin with, and that's the truth. If I incur much loss, I'd be paying people to take hardware. And - as with ALL games and software - I generally specify that, while part of what I'm selling is support, and I will try to help, I also cannot be responsible for specific performance issues in any software (because I don't create the software). In some cases, I've spent way more time than it was worth trying to set up/optimize a build for a certain game(s) (not referring to WOFF here...ask me about Rainbow6 sometimes...or PUBG...or RUST...)

Either way, I do have to know - I have to be able to recognize, that is, what is a problem in the hardware and what's not. Usually by the time someone takes delivery, that setup has been so thoroughly vetted, cleaned, inspected and tested it's not even funny (which is the biggest reason I make nothing). I have seen various little anomalies in various games; they often look similar but aren't always the same. Texture loading isn't unheard of, but not so bad that very fast storage won't all but eliminate it, and the behavior is different besides that. If a RAM drive won't even change it, then something's wrong beyond what I'd call 'typical' texture loading.

I'm not sure this answers your question but I've tried to be thorough. Discussion/questions welcome.

Incidentally, I have to say everything else aside I genuinely appreciate this question and the chance to answer it.

EDIT: BTW I certainly didn't intend, in my exuberance with the question, to say that this stuff only matters because or when I sell something. The reason I thought it such a great question was because it addresses the subject of expectation, and expectation vs cost - which is a big factor as far as I'm concerned, regardless of whether you buy new or used, build yourself, off the shelf, online or local...

I offer this video with an afterburner overlay for consideration.
July 1918 - DVII's vs Bentley Camels

Stache, super video. You clearly have your system dialed in extremely well.

KK, thank you for the detailed answers. Your knowledge of PC hardware is clearly very comprehensive. I can only speak to my own experience. A combination of experimentation with settings, combined with key upgrades to my monitor, my GPU and my CPU and later versions of WOFF have made me very happy playing this game, so in my case I can say I'm glad I've spent the time and money on it.

Yes there's a very minor hiccup / stutter or whatever in that Polovski clip, but it's an external view with WIP code. In any case, I am not saying these problems are gone. I am saying that IMO better hardware diminishes them for me, and others seemingly.

I think you'd be providing your potential customers a more complete service by testing your builds with the latest version of WOFF, which I don't believe you are currently doing? I found that PE generally performed better than UE did right out of the box, but that's just a feeling. It is the release that most new players will want to purchase and play, I would imagine.

Well, after all that, this doesn't seem to be such a big deal, does it? We both agree there are niggling performance aspects still hanging around in the old game engine. We're both at least somewhat in agreement that better hardware does make it run 'better', albeit not offering a complete cure for stutters etc., aren't we? Problems only seem to arise when newcomers to the sim come onto the forum and complain about stutters. My hope is that in future we can all lean a little more toward optimism rather than pessimism in those circumstances. Stache's video, and many others, show how well a good machine can run WOFF PE and there's an opportunity for you right there to advise and sell more systems to help others get similar results.

EDIT: I just saw your recent response to a new forum visitor here: https://SimHQ.com/forum/ubbthreads.php/topics/4518451/new-pc That's very balanced IMO, thank you.

Yes indeed.

Thanks Harry,

This is not to say I do not get an occasional hiccup, I do like that term better as IMO, it indicates an event of a very short duration.
As infrequently as a hiccup may occur they certainly do not affect my enjoyment of WOFF.
There are somethings that have to be lived with. (Like that streching of side monitors - yeah I see it now)

Perhps OBD will be able to address this in the future, but depending on the resources, I might prefer they spend their time improving other aspects.

I do not believe I have done anything magical in my settings or hardware and I am running a much higher resolutions that most users here on what is now four year old technology.

Yes you're getting great results with your machine spec, I agree.

Quick follow-up to a couple of questions I had regarding vid. card overclocking, on page one of this thread. Info. may be helpful if anyone is out there running WOFF on one or two FirePro D700 cards.

For those running a 2013 Mac Pro as I am, with the two FirePro D700 cards that are de-tuned for workstations (consumer ver. of the card is the FirePro W9000) - you will find that the core clock and memory nos. are 850/1370 MHz. Feel free to download the "MSI Afterburner" program and jig up the nos. to 940/1380 MHz for an extra 8 to 10 fps or so in WOFFpe - also gives four or five extra fps, or so, in other sims like FE2/SF2/WOTR/ROF/IL2_1946/BoX - and an added sense of "smoothness" when flying, which I find more important than ave. fps.

The good thing about this tweak is that it is still within the factory overclock specs. for the two vid. cards and doesn't require tampering with voltage and other things that may wreck the cards. (Don't go beyond 1380 MHz for the memory overclocking since these cards are (strange?) and don't tolerate big tweaks to the memory frequency - I would get occasional freezes at 1385 MHz on the memory, for example.)

Also, set the "Macs Fan Control" program (https://crystalidea.com/macs-fan-control) in your BootCamp install on the Mac to point to "sensor-based value/gpu diode 2/ temp. range of 60 to 85 degrees Celsius" as range to activate extra fan speed and max. temp. allowed. You'll find that all components in the Mac Pro hardly go over 80 degrees that way, even after an hr. or two of intensive flight simming.

(Thank you Orbyx for beginning this thread since otherwise I would not have found out about "Afterburner.")

Von S

P.S. There is a "bug" in ROF and BoX that disables "Crossfire" mode on dual AMD cards unless "v-sync" is set to "on" in the in-sim menu. Happy flying all.