This will be my soap box moment.... so please bear with me....

Can we use GHz (CPU speed) as the only indicator for performance?

The answer is No.

GHz has become less of a yardstick to measure performance over the years.

The graph below shows single core performance between a i7-9700k 3.6 GHz and an i7-4790 4 GHz. Here, the i7-9700k is 400 MHz slower than the i7-4790k at stock speed. Notice that the i7-9700K still outperforms the i7-4790k in single core benchmarks at stock speed.



One explanation is that the CPU architecture improved over the years, thus making them more efficient in processing instructions (e.g. IPC).


But with multi-core, the performance gap becomes almost night and day.


Will overclocking help?

The answer is it depends. If you're only speaking about single thread applications, then overclocking an older processor (e.g. 4th generation) might help to reach the performance of a non-overclocked newer processor (e.g. 9th generation). Again, it depends on how far the generation gap between processors is.

This graph is an example of stock and overclocked CPUs comparing FPS.



(Cost to performance ratio is not discussed)

@Orbyx,

I appreciate these technical soap-box moments and I thank you for the post - while I'm no "pjuter" specialist - more of a tinkerer - I'm pleased to see this post since it reminds me of the good old days of RISC vs. CISC discussions that were prevalent on the internet of the 1990s.

Specifically, I'm thinking of the "PowerPC" chips manufactured by IBM in the 1990s and used in many of the beige Macintosh computers of the time, of the Gil Amelio era (before Jobs returned to the scene) - RISC as in reduced/improved instruction set, compared with the CISC (complex instruction set) of the whole family of Intel Pentium processors (x86) of those days - Pentium Pro, MMX, Pentium II, III, etc., etc.

There was a heated debate back then going on specifically with regard, let's say, to the PowerMac 9500 model (with a 604 PowerPC chip running at 150MHz) - and how it was possibly a faster chip than something like the Pentium Pro running at 166MHz, the latter being a typical chip on which to advertise the fps goodness of "Duke Nukem 3D." Extreme comparisons went so far as to suggest that even the mid-range PowerMac 7600, running at 120MHz, had the "equivalent" speed of the 166MHz Pentium Pro, etc.

Detractors of course said that such things are not true - that 166MHz is still faster than 120 MHz, no matter the type of processing instruction set - but no firm conclusions were drawn in the 1990s, and then the whole debate disappeared once Intel (CISC) chips went into Macs too - around 2005.

As indicated in one of the other technical threads here, WOFF runs just fine on my 6-core and quad-core processors - whether or not this is the result of an extra core or two, or speeds above 3 GHz - I have no way of knowing - but I look forward to reading your posts gents.

Happy flying and tinkering in WOFF,
Von S

One interesting thing I noticed in the graph above is that a stock i9-9900k (3.6 GHz) generates more FPS than an overclocked i7-9700k (5.1 GHz). The difference wouldn't be noticeable in game, but it's interesting to note that overclocking is NOT a magic bullet.

Those of us who actually build systems for others - often, for competive gamers and streamers - don't ever rely strictly on clock speeds anyway.

I use a very detailed mix of benchmarks and documented real-world performance to help my customers decide what it's best to spend their hard-earned money on.

And oddly enough, it is NEVER the newest, retail thing. Plus, only two of the machines I've built in the past 3-4 years had more than 4 cores...and yet, they're all still out there, kicking a$$ and running everything they were built for just fine, with only 4 cores (and their owners saved hundreds by not throwing money at the latest retail thing).

The trend toward more cores is absolutely one technical thing and one marketing thing, that's it. Technically, you can't dissipate heat fast enough as a package size decreases, so the only way to gain any performance is more cores, spread out over a die still smaller than the heat spreader, and lower the frequency. This has been known/done for decades now.

Marketing wise...well, 8 cores *MUST* be better than 6...right??? (Hint: No, not always...in fact, not for most common use to this day). You have to consider how much software people run that can actually use the extra cores to the extent it justifies the extra cost - which is something your comparisons continue to ignore.

Maybe you can afford to ignore the cost factor...oh, wait: You already said you can't afford one of these 9900s.

So, maybe some people - a minority, by miles - can afford to ignore the cost vs performance factor, but I don't know any who can. And I am absolutely sure that the majority on this very forum cannot afford to ignore cost vs. performance, either.

Bottom line: As long they ignore the cost as a factor, your comparisons are simply invalid.

Actually, you don't need to own technology to discuss it. That's what reviewers have been doing for centuries. I posted a graph showing real world performance difference between processors running a game.

Interestingly, this website compares dual-core, quad-core, and octa-core processors between various game titles:. Are Quad-Cores Dead in 2019? ... and his conclusion "a cheap quad-core will get you by.". ..."Bottom line, quad-cores are perfectly fine as entry-level parts" .... and I'll add to that, don't expect the performance of newer processors.

I decided not to discuss money because that wasn't the point of this thread. I was showing the performance difference between different processors without taking into account money. I don't see that as invalidating the results of the tests. I work with high end processors all the time. Just like the article I linked doesn't invalidate the result of that reviewer's tests unless he verified that he owned each processor or take into account their cost differences.

Edit: saving money is always a good idea, but it's not always black or white like how you portray it to be. People splurge on their hobbies for various reasons without considering money as a factor.

.



Orby, I won't argue that point at all. However, I will say that in all the years I have been playing CFS3/OFF/WOFF I have never run across anyone who had that attitude when it came time to build a new system to run this series of sims. It's always been, to varying degrees, trying to build the best they can on a limited budget. But oh to have a blank check, eh?

.

Well Lou,

Varying degrees...

When I purchased and personally assembled my system back in 2016, I spent $3452 to future proof it.
I do believe it is money WELL spent.
Four years later it is still going strong.
I bought what thought I needed to achive my goals, how much it cost was not a decision factor.
Was having a system that could run every game over 120fps part of that need - no.
Was having a system so I could enjoy modern games and not have to worrry about upgrading in a couple year part of that need - yes.
Could I have spent more sure - did I need to - I do not think so.

Of course this does not include the thousands of dollars and hundreds of hours building my pit.

But his is my hobby and it is where I choose to spend my dollars and my time.

My friend liked my system so much, he had me build an identical one for him.

To be clear: I completely, absolutely agree that you don't need to own (anything) to discuss it. That said, you began this series of comparisons over in another thread, where you were strongly trying to suggest the 9900F was, without question, the best choice in a CPU (and I'm pretty sure you did say "best" more than once - if not, I apologize).

What I keep saying is "best" means different things depending on the circumstances. Underwater, the "best" thing you can probably have is a way to breathe. On a roller coaster, best is probably a seatbelt.

You can't shouldn't claim a CPU is 'best' as advice on a public forum without including cost as a factor. Or, at least, you owe it to the reader to specify "If money is no object" or "If cost is not your primary concern". What I've consistently said is that there are better options than the newest retail offers, dollar for dollar. And I'll stand by that advice, and re-affirm it applies to a vastly greater number of people than "money is no object" comparisons.

The 9900s are great, no doubt. Best? For what? Under what circumstances? At what cost?

And I respect why you excluded cost. I am still up in the air about whether single-core or multi-core performance is better for this sim. OBD and most players here have long insisted it's single-core speed, but some more recent suggestions are that multi-core is better. I've relied on the 'official' recommendation thus far, but that's not cast in stone for me.

Finally, I strongly disagree with the language that (all) "Quad cores are "entry-level parts". I've got a 4790k, a 7700k, a 6700k and even a 7600k right here in the room with me ATM that would all beg to differ - and they'd #%&*$# well be right. These are NOT 'entry-level' anything. This is (probably) some snotty reviewer who looks at it like money is no object and everyone can afford to completely rebuild a computer every year.

Bit more disclosure about my Mac Pro 2013 rig., to "come clean" as it were - was stimulated by this thread that Orbyx posted to find out what exactly is inside. As indicated in my previous posts in other threads, WOFFpe runs just fine on it, especially with the GPU Tuner Patch installed, ave. fps hovering between 80 and 120 or so - depending on airfield utilities congestion, type of weather, etc. External views while flying often jump into the 150s fps too, etc.

Specs, taken from internet:

Processor Speed: 3.5 GHz
Processor Type: 6 Core Xeon E5-1650v2
Details: The 64-bit 22-nm "Xeon E5" processor, or more specifically the "Ivy Bridge-EP" based E5-1650v2, has six independent "cores" each with a dedicated 256k level 2 cache and 12 MB of level 3 "Smart Cache." It also supports "Hyper-Threading" -- which "allows two threads to run simultaneously on each core" (so OS X recognizes twelve total cores on this model, six real and six virtual) -- and "Turbo Boost" -- which "automatically boosts the processor speed based on workload" up to 3.9 GHz for this model.

Ram info.
4 sticks of 8GB 1866MHz DDR3 PC3-14900 ECC (total of 32GB Ram - can upgrade to 128GB Ram at slightly lower 1333MHz DDR3 PC3-10600 ECC standard but I see no point to that)

Vid. card.
Two AMD FirePro 700 cards, with 6GB of GDDR5 Vram (2048 stream processors / 384 bit wide memory bus / 264 GB/s memory bandwidth / 3.5 teraflops performance)

Storage:
1TB PCIe-based SSD

Other setup is fairly standard, with the old Logitek Attack 3 still functioning well (Extreme 3D Pro bought recently but still in storage ) - and two Anker SoundCore Mini speakers (the smaller, first ver. of the speakers). Monitor is an ASUS 27-inch one at 1920 x 1080 resolution, by the way, 5ms latency. Nothing fancy about the external gear at all but it gets the job done. The 2012 quad-core Mac Mini has a 24 inch ASUS, also at 1920 x 1080 resolution, attached to it, same latency of 5ms, by the way. I've noticed that anything under about 5 or 6 ms latency is just fine - had an ACER monitor, 27 inches, very briefly - at 7ms latency - and did notice mini-stuttering on that one, even in FE2 - but I think that was more a result of the "VA" (vertical alighment) technology of the panel so I switched back to the good-old IPS technology of the ASUS monitors - much better. (VA is on the other hand the best for photo viewing/editing, and perhaps also for best reproduction of blacks in movies - but I was frustrated by its performance in my flight sims.)

Moral of story: a good mid-range monitor with carefully combined mid-range components, at a minimum, should be just fine for WOFF - and you will be able to run most settings at max. (I've used the phrase "mid range" since the 2013 Mac Pro's specs., by today's standards of a top-of-the-line computer, are probably mid-range too.)

Happy flying all,
Von S

I have to agree with both points and Stache,s too regarding his hobby of Pit and PC building..
I know you enjoy your wonderful collection of books Lou,is that your main hobby mate? Do you sometimes purchase a title you want/like without thinking of the cost?

I build a system to my needs but where my football (soccer) is concerned then money,s no cost,I wish it was different but it's in mine and my teenage lads blood. I've been all over the world to see them even pre season games. It's a good job were not that much of a successful team like Liverpool, Man Utd or I'd have been bankrupt many years ago

No one's saying you can't spend more if that's what you want and that's what you can afford. By all means.

I'm saying you don't have to do that, in order to get performance that's plenty capable, and you can save 30-50% off retail cost, while still getting 80-90% the same performance. Its called "value", or "bang for the buck". These top end 9series CPUs are still around $500. I can use a CPU that costs half that, and there is no way a 9900k is going to perform twice as well. So it's not just about who can afford what, it's about the actual performance for the cost. You're not getting twice the goods, but you're paying twice the cost.

200% the cost for 20% more performance? Doesn't seem smart to me.

In the case of the $300 9700F from the other thread, 14% more performance for 67% more in cost? Doesn't strike me as terribly wise either.

I'm saying the less expensive way applies to far more people than the more expensive way does. Guaranteed.

And once again, I'm saying it doesn't do OBD any favors when people come here asking for advice on running this sim, and someone's trying to convince them a $500 CPU in a $3500 computer is the way to go.

You mentioned a 4790k, a 7700k, a 6700k and even a 7600k. Are you speaking about $70 savings compared to the 8th or 9th gen processors?

The cheapest CPU out of that bunch is the i7-6700K (~$300) which was released in 2015 compared to an i7-9700k which was released in 2018 (~$370). Even the motherboard prices are about the same for both processors.

I don't understand that reasoning. And that's why it doesn't make sense to save $70 for an older processor that clearly is inferior in every way to the newer models. Anyways, this thread is about performance, not price.

I agree with this reasoning. The monitor and PC components must be similar in order to get the best performance. It doesn't make sense to buy a high end monitor and the PC components cannot utilize it's full potential. Having said that, I still believe the better your hardware, the better performance you'll get in any application and not just WOFF.

I'm talking about used hardware, I guess that was implicit when we're talking about 4th-gen CPUs. No real difference in new/used, if you do business with the right person. I give warranty and support on everything. And perfectly capable of the prices/performance that I've mentioned.

No one, not even the OP, said this conversation was limited to strictly new hardware. You can't expect economy and the best overall value without making choices. But again, if you must have new, good for you - bring your wallet. Meanwhile, I'll be partying with guys who got much better GPUs with the money they saved on my advice, than if they'd insisted on new, top-end hardware.

Very honest and dependable review from Gamers Nexus on the best and worst CPUs of 2019

I admit the video quality is very poor and doesn't represent what I actually witnessed. I suspect the lag is due to my monitor's high latency which shows up worse on the recording. I couldn't get the sound to record either. Seems I need a lesson on how to upload good youtube videos.

Now to the hardware monitor.... there were a few dips in FPS and only one was actually noticeable for a split second inside the game. It's when the FPS dropped from 57 to 49. The whole experience was very fluid. The graph does show a few drops of 3 to 4 FPS during gameplay, but I didn't notice those.

I made sure to shut down all applications before starting WOFF. Nothing else was running. In WOFF, I set All workshop settings were maxed out and I had supersampling at x8 and frames limited to 57.

Takeaways from the graph:

-WOFF menu and screen navigation did not use all 8 cores, however, ALL 8 cores were being used while CFS3 was running and flying the mission.

-While CFS3 was running, only one CPU hit 100%, one hit 77%, and the rest were at around 50% or so.

-Memory usage was at 3.97 mb while WOFF was active and 5.47 mb while CFS3 was running.

If I had locked one core to CFS3, would that have made a difference? in the number of Frame drops? maybe less fluctuation in FPS?.... I don't believe it would have made any noticeable difference. My guess is that older hardware might benefit from locking CFS3, TrackIR each to one core.

.

Interesting numbers and info, Orby. I am wondering though what was going on with your recording, it is quite choppy and with a fair amount of noticeable hesitations. Does Afterburner eat up that much in resources to have caused those issues? My old system produced a considerably smoother recording, even with TrackIR running, using OBS Studio to record. Still not as smooth as what I am actually privy to when running PE without video capture, but not bad either.

.

.

My OBS recording was choppy also when doing this.
Note - I'd only toyed with OBS before.
I did get it much smoother - at the expense of larger files.
I found a video talking about such - made changes based on that and got much smoother results.
Will post settings later in a new thread.

You have to remember that kknowsitall talks about USED PC parts because he sells used PC parts, so naturally he thinks those are better.

.

PR, I'm not exactly sure what the issue is there. I've built numerous systems myself with gently used components and saved myself hundreds of dollars while getting excellent gaming results. I've no problem with them at all, as long as you're smart and do your research before buying them.

.