Anyone using Ryzen 7 for P3D v4?

[Sethos 4,304]

Ryzens are still behind high-end Intel when it comes to the one thing that matters for the flight sims; single-thread performance. Something like an 1800X still has the same performance as an average overclocked 5960x. 

[himmelhorse 1,276]

Hello guys,

I will concede that AMD is a wee bit slower with single thread performance, but, does anyone have any further info on the prospects of P3D utilising multi threads in a better way. Also, does XPlane11 perform better using multi thread .

AMD have had an almost soundless release of the Vega64 and is this any better than a GTX1080ti with its 16gb HBM VRAM and a Zotac GTX 1080ti OC.

I was seriously thinking of going completely AMD too but blew up a SSD and R9 290 GPU from constant power fails. I replaced it with a Zotac GTX1080ti amp extreme (yet to be fitted) Whils I have absolutely no regrets with my purchase, I am wondering whether or not to eventually upgrade to an Intel system or AMD when the time and money comes to replace my aging system.

If anybody has the Ryzen/Vega64 system I would really like to get some further information.

Thanks Guys n Girls

Regards to all

Tony Chilcott

 

[swiesma 625]

From all information I have I am really impressed with the Ryzen CPUs, not so much with AMDs new GPUs.

If I had to buy a new PC now I would go for a Threadripper and a 1080ti. If money is a non-issue I would go for the i9 7900 and one of the newer X299 mainboards with VRM coolers like the Asus Rampage VI Apex.

[331BK 96]

11 hours ago, swiesma said:

From all information I have I am really impressed with the Ryzen CPUs, not so much with AMDs new GPUs.

If I had to buy a new PC now I would go for a Threadripper and a 1080ti. If money is a non-issue I would go for the i9 7900 and one of the newer X299 mainboards with VRM coolers like the Asus Rampage VI Apex.

the big question is if a 4.2 ghz threadripper can catch up with a 4.8 ghz 7820x.

if raw ghz still rules in p3d v4 i think not.

but 32 threads could be usefull if you run a lot of other tools like as16, .......

 

[Sethos 4,304]

1 hour ago, 331BK said:

the big question is if a 4.2 ghz threadripper can catch up with a 4.8 ghz 7820x.

if raw ghz still rules in p3d v4 i think not.

but 32 threads could be usefull if you run a lot of other tools like as16, .......

 

Wouldn't count on it. The Threadripper, despite the higher clock, is slower than the 1800X single-core. Too much core latency to contend with. 

In single-threaded performance, the Ryzen falls pretty far behind Intel; 

http://pclab.pl/art75073-22.html

[wiler 394]

I am doing tests w p3dv4 w tfdi 717 on my ryzen 1700x 3.8ghz oc and 1080ti. Runs really well getting 50to 60fps w ftx global installed. Havnt  gone into complex paywares yet. I do see one core very active 80-100 % and a 2 or 3 more being lightly used (30-40%) the rest are pretty inactive. In xp11 all 16 threads are active with 2 cores a Bit higher then the rest. P3d still highly dependant on single thread but its gotten much better than previous versions. Will try out some more complex airports and the ngx and see how it runs. I mainly use xp11 but want to provide feedback to the p3d users. 

[331BK 96]

if we only would know in which direction LM will go regarding multi threading.

when they give more Support to more cores i buy a Threadripper otherwide a 7820x would also be more than enough.

 

[wiler 394]

11 hours ago, Sethos1988 said:

Wouldn't count on it. The Threadripper, despite the higher clock, is slower than the 1800X single-core. Too much core latency to contend with. 

In single-threaded performance, the Ryzen falls pretty far behind Intel; 

http://pclab.pl/art75073-22.html

I dont see a huge difference TBH between my 6600k @ 4.5ghz and ryzen 1700x @ 3.9 3200 ram in p3dv4 w tfdi 717.  P3dv4 is using more cores then previous versions according to my process lasso activity. 

[Sethos 4,304]

10 minutes ago, wiler said:

I dont see a huge difference TBH between my 6600k @ 4.5ghz and ryzen 1700x @ 3.9 3200 ram in p3dv4 w tfdi 717.  P3dv4 is using more cores then previous versions according to my process lasso activity. 

At peak overclocks, the 1700X is actually 22% slower than the 6600K in single-threaded performance; http://cpu.userbenchmark.com/Compare/Intel-Core-i5-6600K-vs-AMD-Ryzen-7-1700X/3503vs3915

And while P3Dv4 does use more cores, it's not seeing any real performance gains from it. Uses those extra threads to queue up terrain / autogen rendering, however the framerate and smoothness of the simulator still hinges on the single-core performance. 

Overall, if you goal is improving your performance in P3D or flight simulators in general, Ryzen is a really poor choice.

[wiler 394]

Just now, Sethos1988 said:

At peak overclocks, the 1700X is actually 22% slower than the 6600K in single-threaded performance; http://cpu.userbenchmark.com/Compare/Intel-Core-i5-6600K-vs-AMD-Ryzen-7-1700X/3503vs3915

And while P3Dv4 does use more cores, it's not seeing any real performance gains from it. Uses those extra threads to queue up terrain / autogen rendering, however the framerate and smoothness of the simulator still hinges on the single-core performance. 

Overall, if you goal is improving your performance in P3D or flight simulators in general, Ryzen is a really poor choice.

With all due respect, im the one with the ryzen coming from the 6600K and reporting that I am seeing equal performance overall. P3d has never ran as smooth for me until now. I dont know what speed ram this CPU comparison is using or what "peak" clocks they are using but  I would be willing to bet its at default Ram speed for ryzen( 2133 ram speed) most likely. The difference between using 2133 ram and 3200 on ryzen is HUGE in xp11 it was at least at 20-25% gain for me, on the intel machine it didnt make a huge difference to run at 3200 ram. In P3d having ASN open as well as some other apps doesnt even make a dent in this 8 core processor like it did on my 4 core 6600k, so for me I am winning. My goal was to move to a platform that runs well today (on par or very close to intel)and is ready to be fully utilized (xp is going Vulkan in the next 9/12 months). Again, if you dont want to beleive what I am reporting, then dont. I am just trying to provide real world perspective to the question the OP had. Thanks. 

[Sethos 4,304]

Just now, wiler said:

With all due respect, im the one with the ryzen coming from the 6600K and reporting that I am seeing equal performance overall. P3d has never ran as smooth for me until now. I dont know what speed ram this CPU comparison is using or what "peak" clocks they are using but  I would be willing to bet its at default Ram speed for ryzen( 2133 ram speed) most likely. The difference between using 2133 ram and 3200 on ryzen is HUGE in xp11 it was at least at 20-25% gain for me, on the intel machine it didnt make a huge difference to run at 3200 ram. In P3d having ASN open as well as some other apps doesnt even make a dent in this 8 core processor like it did on my 4 core 6600k, so for me I am winning. My goal was to move to a platform that runs well today (on par or very close to intel)and is ready to be fully utilized (xp is going Vulkan in the next 9/12 months). Again, if you dont want to beleive what I am reporting, then dont. I am just trying to provide real world perspective to the question the OP had. Thanks. 

That's all fine but it's an anecdotal reference at best, with zero factual evidence, no numbers. Your simulator not running smooth could be a hundred other factors, especially when 'upgrading' to a factually slower single-threaded performing processor, it lends more credibility to the fact that it could be something else. And no, these speeds are based on an average of user reported results and then a median is created. So I very much doubt it's all based on 2133MHz Ram. Even having ASP or ASN open uses a minuscule amount of rendering time, so little you could move it off to one of your last threads on a 4C / 8T cpu and have no performance impact. You don't need 8 cores for that. I've used affinity masking to remove half of my 8 cores from P3D and there's absolutely zero performance difference.

It's fine you had a great experience but all the Ryzen reports so far are anecdotal with zero merit in facts. No metric, no benchmarks back up anything you or any other Ryzen user says when it comes to single-threaded performance. They excel at highly multi-threaded applications which the ESP-Platform certainly is not. No Ryzen users are doing any real benchmarks, so it's hard to find anything credible. 

Then you also have to contend with the very human nature that no one is going to admit their brand new hardware might not have been the best choice, so you'll obviously have people saying it runs oh so much smoother, literally XX% better -- Despite not having done any actual comparative tests, it's just percentages they grab from thin air based on loose testing. It's just not credible in any shape or form. 

And no, X-Plane 11 is not going fully Vulkan within a year. And even with a full Vulkan implementation, it doesn't magically means it'll utilise 8+ cores and see a huge performance increase. That's a completely misunderstood concept of Vulkan and low-level APIs. 

[wiler 394]

Just now, Sethos1988 said:

That's all fine but it's an anecdotal reference at best, with zero factual evidence, no numbers. Your simulator not running smooth could be a hundred other factors, especially when 'upgrading' to a factually slower single-threaded performing processor, it lends more credibility to the fact that it could be something else. And no, these speeds are based on an average of user reported results and then a median is created. So I very much doubt it's all based on 2133MHz Ram. Even having ASP or ASN open uses a minuscule amount of rendering time, so little you could move it off to one of your last threads on a 4C / 8T cpu and have no performance impact. You don't need 8 cores for that. I've used affinity masking to remove half of my 8 cores from P3D and there's absolutely zero performance difference.

It's fine you had a great experience but all the Ryzen reports so far are anecdotal with zero merit in facts. No metric, no benchmarks back up anything you or any other Ryzen user says when it comes to single-threaded performance. They excel at highly multi-threaded applications which the ESP-Platform certainly is not. No Ryzen users are doing any real benchmarks, so it's hard to find anything credible. 

Then you also have to contend with the very human nature that no one is going to admit their brand new hardware might not have been the best choice, so you'll obviously have people saying it runs oh so much smoother, literally XX% better -- Despite not having done any actual comparative tests, it's just percentages they grab from thin air based on loose testing. It's just not credible in any shape or form. 

And no, X-Plane 11 is not going fully Vulkan within a year. And even with a full Vulkan implementation, it doesn't magically means it'll utilise 8+ cores and see a huge performance increase. That's a completely misunderstood concept of Vulkan and low-level APIs. 

You speak about having no evidence or credibility,  but where are the exact RAM or CPU speeds used  in these benches ? yea thats right you said you doubt its at 2133, but do you know for sure or not , you dont!  Most people cant post above 2667 with 3200 ram right now and even thats a miracle with a lot of ram. Also when were these benches done... prior to or after all the latest bios updates based on microcode improvements by AMD which make huge FPS gains in gaming. These are very important factors to consider to really understand if these benches are accurate or not. Please get your facts straight before you attack me and say my contributions have no merit. 

[Sethos 4,304]

21 minutes ago, wiler said:

You speak about having no evidence or credibility,  but where are the exact RAM or CPU speeds used  in these benches ? yea thats right you said you doubt its at 2133, but do you know for sure or not , you dont!  Most people cant post above 2667 with 3200 ram right now and even thats a miracle with a lot of ram. Also when were these benches done... prior to or after all the latest bios updates based on microcode improvements by AMD which make huge FPS gains in gaming. These are very important factors to consider to really understand if these benches are accurate or not. Please get your facts straight before you attack me and say my contributions have no merit. 

You are now contending, that a well-known database of accumulated user benchmark scores, meant to create baseline median scores to give you an overview of the performance of this hardware isn't credible. Yet your numbers, that seem grabbed from thin air at best, are more credible? Hmm. The database, as I told you in the previous post, is based on user submitted scores and are a median calculation. Then you have the peak scores, the scores based off the fastests results, meaning results with 2133MHz RAM definitely wouldn't make the cut. So you can be sure the peak scores are based off the fastest user submitted hardware, i.e fast ram, fastest BIOS revisions etc. So it's a lot more credible than the results you are making up on the spot as an indicator of your off-the-cuff experience. AGESA 1006, the one that fixed a lot of the prior issues, was also released over a month ago, so results have definitely stabilised in favour of Ryzen's speed potential.

Also funny that your entire argument not only hinges on excuses for situations that makes Ryzen as slow as possible but also the notion that everybody buys the absolute slowest RAM, just to make Ryzen look bad. 

Maybe you need to "get your facts" straight here and stop making so many excuses, thinking the rest of the hardware world that is incapable of providing actual benchmarks, based on every end of the hardware speed spectrum. Yet your completely unconfirmed remarks, that no benchmark can verify or back up, are the gold standard in how well Ryzen is performing. That's the beauty of hardware, you can measure it and get an actual number. No need to rely on people's baseless anecdotes and subjective opinions. 

[wiler 394]

3 minutes ago, Sethos1988 said:

You are now contending, that a well-known database of accumulated user benchmark scores, meant to create baseline median scores to give you an overview of the performance of this hardware isn't credible. Yet your numbers, that seem grabbed from thin air at best, are more credible? Hmm. The database, as I told you in the previous thread, is based on user submitted scores and are a median calculation. Then you have the peak scores, the scores based off the fastests results, meaning results with 2133MHz RAM definitely wouldn't make the cut. So you can be sure the peak scores are based off the fastest user submitted hardware, i.e fast ram, fastest BIOS revisions etc. So it's a lot more credible than the results you are making up on the spot as an indicator of your off-the-cuff experience. AGESA 1006, the one that fixed a lot of the prior issues, was also released over a month ago, so results have definitely stabilised in favour of Ryzen's speed potential.

Maybe you need to "get your facts" straight here and stop making so many excuses, thinking the rest of the hardware world that is capable of providing credible, factual metrics of performance aren't able to providing a picture of a piece of hardware's performance, yet your random anecdotes that NO benchmark is able to back up, are the gold standard. That's the beauty of hardware, you can measure it and get an actual number. No need to rely on people's baseless anecdotes and subjective opinions. 

Just trying to help the other users and the OP who are interested in switching over to Ryzen make an informed choice. By all means dont believe what I write, but your bench isn't representative of real world results , which I have found out myself based on having both these chips and running the same scenerios before and after in XP11 and p3dv4.  I can tell you the results I have found are not made up sir. Its offending that I am told that I pull results out of thin air because they dont align with what you want.   Lets agree to disagree and move on, im spending my time and efforts  to help push this community further  and not argue with people who have already decided what the results are without even having the hardware in their possesion......I do wish ROB A would test a ryzen system, because he has the respect and credibility in this community and if he did find similar results then maybe people would believe. 

[Sethos 4,304]

Just now, wiler said:

Just trying to help the other users and the OP who are interested in switching over to Ryzen make an informed choice. By all means dont believe what I write, but your bench isn't representative of real world results , which I have found out myself based on having both these chips and running the same scenerios before and after in XP11 and p3dv4.  I can tell you the results I have found are not made up sir. Its offending that I am told that I pull results out of thin air because they dont align with what you want.   Lets agree to disagree and move on, im spending my time and efforts  to help push this community further  and not argue with people who have already decided what the results are without even having the hardware in their possesion......I do wish ROB A would test a ryzen system, because he has the respect and credibility in this community and if he did find similar results then maybe people would believe. 

But it is a fair representation of real-world results, especially when compared to specific cases where that exact power is used. I advise you to take a look at the benchmark I posted earlier, the one featuring ArmA III; http://pclab.pl/art75073-22.html. The difference between the fastest, overclocked Ryzen results with 3200MHz RAM and the fastest Intel CPU is around 25% performance -- Which is a lot. That aligns very well with the results on userbenchmarks, because those are very effective and credible guidelines / rules of thumb for the overall performance of a CPU. The ArmA III results could easily be applied to our flight simulators as they, just like ArmA, are now based on engines that rely heavily on single-core performance.

Again, that's the beauty of hardware, it can be measured and weighed without even having it in your possession because it's based off cold, hard facts. Not anecdotes, not a feeling in your stomach or some subjective reasoning. Just beautiful numbers. I'm just providing a counter-argument against Ryzen, because it is based on every available piece of metric and fact, a poor choice for a flight simmer -- At least if you can afford an equivalent Intel counterpart. 

But I think the point has been made, then OP or whoever pops by can decide what they'd want to base their potential purchase decision off of.