Best CPU for Prepar3dv4.2

[supersym 52]

On 5/28/2018 at 6:36 PM, sultanofswing said:

I went from a 7820x@ 4.6ghz to a 8700k@ 5ghz and saw a slight improvement with the 8700k.

only slight ?

could you tell me how much fps you have more, or if you only get better smoothness ? 

[AnkH 2,262]

36 minutes ago, supersym said:

only slight ?

could you tell me how much fps you have more, or if you only get better smoothness ? 

Do the maths. From 4.6 to 5.0GHz is an increase of 8.6%. Means, even if 100% of this increased clock speed is translated into performance gain (rather unlikely), this would mean that you get instead of 20FPS a whopping 21.76FPS. Even if your cutoff is 30FPS, what you get is 32.5FPS. As he said, "only slight". Considering that in the end, such a increase in clock speed results in reality in about 5% performance increase, you are down to 1 or 2 FPS more, with the fluctuations in the sim basically nothing...

[supersym 52]

Doing the math isn't the best way to evaluate 😉 

most of things can influence especialy the work between the cores, the bandwith within memory, the cache...

Those 2 processors have different structures 

[Nemo 1,305]

My 7820x runs at 4.8GHz (all cores) and I cannot imagine to get a better performance with a 8700K @ 5.0 GHz..To many parameters go in, especially MB design and memory connection (4 channels) and not to forget two additional cores.

[TechguyMaxC 516]

On 5/29/2018 at 5:12 PM, American 833 Heavy said:

Surprised to see NOT ONE recommendation for any of the new Ryzen chips which have comparable and in many cases better performance for price than quite of a few of the intels.

I've used Intel chips my entire life. This past summer decided to give AMD's new Threadripper 1950x top of the line 16 core CPU a go. Upgrading from a 3770K. I got it in August as my 3770K was on the way out and didn't want to wait for Q4 '17 i7 Extreme's to be released.

Let me tell you the ESP platform has never run better. I run it with a Zenith Extreme Mobo, 3466mhz CL15 DDR4 32GB G Skill Ram, 2 GTX1080TIs in SLI and have my OS and P3Dv4 on 2 separate Samnsung 960 EVO 2TB m.2 SSDs. I'm able to pretty much lock 30FPS even in the most intensive areas with the entire Orbx Global and FTX suite, Open LC, FS Global 2018 FTX with lots of addon scenery.

It's also phenomenal in multi tasking (best I have ever used), Audio Mixing/Mastering/Editing and Encoding/Decoding as well as Video Encoding/Decoding. Not to mention being able to record P3D with OBS with virtually zero FPS loss.

Really an awesome alternative to some of the top tier Intels and I'd do it all over again if I had the choice. No regrets heading over to AMD for the 1st time.

Ryzen is a good product line.  AMD did a lot of work to become competitive with Intel again, and it's paying off.

That being said, Intel still holds the lead in clock speed, and instructions per clock, the two single largest contributors to performance in any ESP-based sim, at least.  

Can Ryzen run e.g. P3D well?  Sure.  Can it do it better than say, an 8700k?  Not without exotic (sub-zero) cooling, which could also be applied to an Intel chip raising its performance even further (thus negating this argument).  

Basically, unless AMD's manufacturing partner, Global Foundries, can improve their manufacturing processes sufficiently to allow AMD's chips to hit the same or higher clock speeds than Intel, workloads that are largely single-threaded (i.e. flight sims) will always run faster on Intel.  Maybe that difference is only 10-20%, but the price difference between say a 2700x system and an 8700k system is minimal, maybe $50 which means one doesn't have to pay a whole lot more for this additional performance in flight sim.

As for Ryzen's validity in other, more multi-threaded workloads it is undeniably an excellent value.  I, like many others here, have a separate/dedicated PC for flight sim though and do not use it for much else.  I have a 10-core 7900x system that I use for my video rendering needs and that does quite nicely.  For awhile I considered a 16-core Threadripper but when I was ready to build this system X399 motherboards were simply unavailable at my local Micro Center, which swayed my decision towards X299.  There's always next time though.  By the time 32-core Threadripper chips are out that may be enough to sway me.

[westman 257]

2 hours ago, Nemo said:

My 7820x runs at 4.8GHz (all cores) and I cannot imagine to get a better performance with a 8700K @ 5.0 GHz..To many parameters go in, especially MB design and memory connection (4 channels) and not to forget two additional cores.

I can, prefer the 5.4ghz 8700K before my 7920X and 7980XE both at 5ghz

[TechguyMaxC 516]

37 minutes ago, westman said:

I can, prefer the 5.4ghz 8700K before my 7920X and 7980XE both at 5ghz

As usual, Hasse gets it right.

 

Allow me to ellaborate:

The 7820x and other Skylake-X (SKL-X) chips are very different from regular Skylake/Kaby Lake/Coffee Lake (*Lake) chips.  SKL-X has a new cache architecture (compared to the mainstream *Lake chips) which is both large and "wide" in that it can move a lot of data within a given time period (this is known as bandwidth).  Initial access to data within these caches is slower than on mainstream chip architectures however, a byproduct of widening the lanes to the caches as well as the implementation of a ring-bus which allows any processor core to access the contents of any slice of cache.  Additionally, the cache clock domain (Uncore in Intel terminology) runs at a lower frequency on SKL-X vs. *Lake, further increasing latency.  Further discussion on this topic is well beyond the scope of this post.  All that being said, SKL-X is optimized for workloads that are:

1) well-threaded

2) constrained by bandwidth to:

a) cache

b) system memory

 

The *Lake families of CPUs are designed to accommodate workloads that are:

1) few-threaded

2) constrained by *latency* to:

a) cache

b) system memory

 

To give you an automotive analogy, think of these chips as vehicles.  Take a 7820x and an 8700k.  The 7820x is like an SUV, even a really fast one like the Jeep Trackhawk (has the Hellcat engine).  It's going to be able to carry a lot of stuff and it can get there pretty quickly.  The 8700k is like a Dodge Demon.  It's got the same Hellcat engine but in a lighter body and with more aggressive gear ratios.  It can't carry much stuff but man can it get you there in a hurry.

To illustrate the point back in CPU land, the benchmarking program Cinebench is well-known for its ability to represent the performance of various CPUs in both single and multi-threaded workloads. 

My 7900x has 10 cores/20 threads and runs at a stock clock speed of up to 4.5GHz on 2 cores and 4.0GHz on all 10 cores.  It achieves a multi-threaded score of 2197 in this benchmark. 

My 7700k on the other hand is a 4 core/8 thread chip and it runs at 5.2GHz on all 4 cores.  It achieves a multi-threaded score of 1133, only about half that of the 7900x. 

Here's the kicker though:  

The single-threaded score on the 7900x is 183

The single-threaded score on the 7700k is 226

 

With 700-1200MHz more clock speed on tap and lower latency to cache and system memory, the 7700k is able to blow past the 7900x in lightly-threaded workloads.

Edited June 2, 2018 by TechguyMaxC

[ozboater 6]

OK, most of you guys are experienced Intel users, but I'll give you my AMD Ryzen 7 1700 with GTX 1080 experience at 4K resolution. 16 Gb G.Skill 3200 14-14-14-34 memory and boot disk is Samsung M.2 960 Evo 512 Gb.

P3Dv4 at 4K resolution most sliders maxed. ORBX - all of it. Usually I fly with FPS set to 30, and I get an incredibly smooth performance everywhere.

This morning I thought I'd see what setting FPS to unlimited would be like.

Flying from Just Flight Tromso in Carenado PA 31T Cheyenne   got 90 to 120 FPS at the airport and 120 to 180 FPS flying away over ORBX Norway. I fly mostly in Scandinavia and Australia.

No overclocking, except game mode selected in Bios. P3Dv4 is on its own Samsung 850 SSD (should be on my 1 Tb M.2 Intel 600p, but Win 10 1803 meant I had to swap it out for the SATA Samsung).

Get This. On the free stock cooler. You Intel guys pay real solid bucks for a cooler to get 5 GHz, on top of an expensive chip and motherboard combo. To me Intel 5 GHz seems like an expensive (fast) platform.

At FlyTampa Copenhagen and Vidan Real Denmark scenerywith 50% airline and GA traffic and high ground vehicles I got 40 to 50 FPS. Over 60 FPS flying away from the airport.

Checking CPU load, 15 of the 16 threads were working. I was also simultaneously downloading a 3 Gb SACD classical album.

My purpose for posting - Ryzen AM4 makes a good platform for flight simming. No need to be afraid of it.

Intel is tops for 1080p gaming, no dispute, but if you are using higher resolution, the choice between Intel and AMD CPUs for flight simming is not much different than "I like the colour of my car more than yours".

Biggest advantage for AMD is that you can get this performance at a much lower platform cost with future upgradability without throwing away half the computer. Then you might spend the savings on buying a better graphics card in the first place. Thats is undeniably the easiest way to get improved simulator performance.

Better still, the new Ryzen 2 cpus are faster than my first gen chip at no extra cost.

Anyway, after this I set the max framerate back to 30 FPS synced and enjoyed the beautiful smooth flight most relaxed. Everywhere.

[GSalden 4,656]

3 hours ago, TechguyMaxC said:

As usual, Hasse gets it right.

 

Allow me to ellaborate:

The 7820x and other Skylake-X (SKL-X) chips are very different from regular Skylake/Kaby Lake/Coffee Lake (*Lake) chips.  SKL-X has a new cache architecture (compared to the mainstream *Lake chips) which is both large and "wide" in that it can move a lot of data within a given time period (this is known as bandwidth).  Initial access to data within these caches is slower than on mainstream chip architectures however, a byproduct of widening the lanes to the caches as well as the implementation of a ring-bus which allows any processor core to access the contents of any slice of cache.  Additionally, the cache clock domain (Uncore in Intel terminology) runs at a lower frequency on SKL-X vs. *Lake, further increasing latency.  Further discussion on this topic is well beyond the scope of this post.  All that being said, SKL-X is optimized for workloads that are:

1) well-threaded

2) constrained by bandwidth to:

a) cache

b) system memory

 

The *Lake families of CPUs are designed to accommodate workloads that are:

1) few-threaded

2) constrained by *latency* to:

a) cache

b) system memory

 

To give you an automotive analogy, think of these chips as vehicles.  Take a 7820x and an 8700k.  The 7820x is like an SUV, even a really fast one like the Jeep Trackhawk (has the Hellcat engine).  It's going to be able to carry a lot of stuff and it can get there pretty quickly.  The 8700k is like a Dodge Demon.  It's got the same Hellcat engine but in a lighter body and with more aggressive gear ratios.  It can't carry much stuff but man can it get you there in a hurry.

To illustrate the point back in CPU land, the benchmarking program Cinebench is well-known for its ability to represent the performance of various CPUs in both single and multi-threaded workloads. 

My 7900x has 10 cores/20 threads and runs at a stock clock speed of up to 4.5GHz on 2 cores and 4.0GHz on all 10 cores.  It achieves a multi-threaded score of 2197 in this benchmark. 

My 7700k on the other hand is a 4 core/8 thread chip and it runs at 5.2GHz on all 4 cores.  It achieves a multi-threaded score of 1133, only about half that of the 7900x. 

Here's the kicker though:  

The single-threaded score on the 7900x is 183

The single-threaded score on the 7700k is 226

 

With 700-1200MHz more clock speed on tap and lower latency to cache and system memory, the 7700k is able to blow past the 7900x in lightly-threaded workloads.

You are forgetting one thing : there is more to it than just clock speed and you already pointed to it : cpu cache. Cache is memory inside the CPU and is way faster than the main memory we put on the mobo. More cache means less usage of the slower main memory.

A 7700k has 8mb cache , while a 7900x has 13.75mb cache. If you are doing single-threaded benchmark tests, surely the higher clock speed of 5.2 Ghz from the 7700k will outperform the 4.5 Ghz of the 7900x.

Hasse has a 7940x running at 5 Ghz on all cores. What if he would compare that to a 7700k running at 5.2 Ghz ?

Regarding P3D : beside the main thread clock speed, which is very important for the framerate, we also have a scenery to fly over. Scenery is calculated by the other cores.

If you have 1 HD screen then a 7700k or 8700k will do very nice, but what when using a dual or triple view setup with highly detailed visuals ?  Then 4 cores for everything are very little. The 8700k will do much better and the 7900x will do even better.  

Like it has been said many times ; it all comes down to what setup someone has , what scenery and how detailed it has to be. And then choose the cpu that fits you ( at this moment.... ).

 

 

 

  

Edited June 2, 2018 by GSalden

[American 833 Heavy 132]

On 6/2/2018 at 9:02 AM, TechguyMaxC said:

Ryzen is a good product line.  AMD did a lot of work to become competitive with Intel again, and it's paying off.

That being said, Intel still holds the lead in clock speed, and instructions per clock, the two single largest contributors to performance in any ESP-based sim, at least.  

Can Ryzen run e.g. P3D well?  Sure.  Can it do it better than say, an 8700k?  Not without exotic (sub-zero) cooling, which could also be applied to an Intel chip raising its performance even further (thus negating this argument).  

Basically, unless AMD's manufacturing partner, Global Foundries, can improve their manufacturing processes sufficiently to allow AMD's chips to hit the same or higher clock speeds than Intel, workloads that are largely single-threaded (i.e. flight sims) will always run faster on Intel.  Maybe that difference is only 10-20%, but the price difference between say a 2700x system and an 8700k system is minimal, maybe $50 which means one doesn't have to pay a whole lot more for this additional performance in flight sim.

As for Ryzen's validity in other, more multi-threaded workloads it is undeniably an excellent value.  I, like many others here, have a separate/dedicated PC for flight sim though and do not use it for much else.  I have a 10-core 7900x system that I use for my video rendering needs and that does quite nicely.  For awhile I considered a 16-core Threadripper but when I was ready to build this system X399 motherboards were simply unavailable at my local Micro Center, which swayed my decision towards X299.  There's always next time though.  By the time 32-core Threadripper chips are out that may be enough to sway me.

Let me remind you I am NOT an AMD shill and have used Intel my entire life. My previous 3770K based rig was terrible with FSX/P3D platform. I Got the 3770K right when it came out years ago and finally upgraded to the 1950x in the Summer of 2017. I usually upgrade PCs every 3-4 years.

Now, let's not get carried away.... 10-20%? With a Threadripper or even mid-high tier Ryzens? I doubt that very much. Remember the GPU plays a big role here too.

P3D takes advantage of multiple cores for background processes and scenery loading etc. Unlike the FS9 days it's not just only about single threaded performance. Granted ESP is till single thread/clock speed biased... but the Threadrippers and other Ryzens do an outstanding job. And when paired with a 1070/1080 type card. They among the top of the top.

These new AMDs overclock nicely with maintenance free closed loop CPU coolers like the Corsairs. Even at the stock 3.4GHZ on the 1950x which it will turbo-boost to 4.0, the 2 metrics you noted are a virtual wash on the top line AMDs vs Intels with the Intels a slight edge.

Multi threaded performance the Threadripper line beats anything Intel has to offer except their $2000 18 Core Extreme which has a lower base speed than the 1950x BTW. And even then the Multithreading peformance is marginally better than the Threadripper at a $1000+ price premium. I'd take the 1950x and a 1080TI with that extra 1000.

The X399 was the hardest to get chipset mobo 2nd half of last year for a reason. The Threadripper system was simply a better value in the higher end performance segment and does better in overall computing. I do sound design so I need fast encoding/decoding and the 1950x delivers it in spades. And as I said it runs P3D/ESP the best I have ever had by far.

I think had you held out for that mobo a little longer you would have been extremely happy with the 1950x. Easily overclockable to 4GHZ stable on a good cooling setup.

On the otherhand though I am glad I did not end up waiting for last winters Q3/Q4 7th Gen Extremes to be released as they've already been passed over in favor of the 8th gen architecture and I wouldn't have ever had the privileged of seeing what AMD has been working on all these years.

We have finally, for the 1st time in many, many years. Maybe since the early to mid 2000s virtual 1 to 1 parity with the AMD/Intels. And it's great for us consumers.

It's a shame people seem not be open minded and trying something different. This thread should be renamed "Best Intel CPU for running P3D" loi.

All jokes aside. All of us have benefited from this new found parity in the CPU world. Hopefully in a year or 2 if AMD can close the gap to Nvidia as they have done with Intel it would be a dream come true. Competition drives these companies to put out the absolute best they can and for the consumer it's a win-win.

Edited June 4, 2018 by American 833 Heavy

[Guest]

So long as my now 4 year old 4770K keeps going at 4.8ghz I will not upgrade my CPU until I can get a CPU with a 6ghz main thread. Even that  25% increase wouldn't help much. All it would do is make it easier to maintain 30fps with maybe slightly higher eye candy setting and extra cores I guess would help to keep textures sharp. Anything less would not be worth the $$$ IMO

[jabloomf1230 2,445]

1 hour ago, Avidean said:

I will not upgrade my CPU until I can get a CPU with a 6ghz main thread

I wonder if that will ever happen on the desktop, unless the technology changes radically. I don't know whether Intel either has run up against a cost/technology wall in trying to improve single thread performance or they just don't care. For servers, more cores makes more sense. And for gaming, the CPU is no longer king as most games rely on the GPU for the heavy lifting. Here's the latest Passmark single thread scores:

https://www.cpubenchmark.net/singleThread.html

[newtie 533]

4 hours ago, Avidean said:

So long as my now 4 year old 4770K keeps going at 4.8ghz I will not upgrade my CPU until I can get a CPU with a 6ghz main thread. Even that  25% increase wouldn't help much. All it would do is make it easier to maintain 30fps with maybe slightly higher eye candy setting and extra cores I guess would help to keep textures sharp. Anything less would not be worth the $$$ IMO

You and me brother.

I'm at 4.4 with EvoPro x 2 (250 and 500mb) w/32mb GSkill.

On three Dell 27 UltraSharps @ 2550x1440

Runs perfect.

Cheers,

Mark

[Greggy_D 935]

I just ordered a 2600x.  Just to try something different.  (Coming from a 6600k)

[DylanM 1,069]

This may be the CPU you're looking for... 😉

https://www.neowin.net/news/intel-teases-mammoth-single-socket-28-core-processor-for-pcs