Noel

Tell me this isn't so! 3930K v 7700K single core performance

Recommended Posts

I just looked at this Passmark chart:  https://www.cpubenchmark.net/singleThread.html

By my calculations my 3930K at 4.42Ghz versus 7700K at 4.5Ghz gets a whopping increase of 8% in that single-core benchmark.  And the 3930K has been on a Noctua air cooler.  This is very discouraging as my 3930K box just died, and I'm thinking it's the ASUS P979X WS mainboard w/ its beep code saying, 'Hardware Component Failure'.   The 3930K could still be working!  This leaves me wondering if I shouldn't pick up a new ASUS P979X WS which is available on eBay right now and maybe this means I can resus the entire system w/ minimal work, no software reloading, etc.  Very attractive.  I could continue to use my 32Gb of DDR3 at 2400mHz and my CPU, versus upgrading to a new mobo/ram/cpu.  The problem of course is that I really don't know which 'Hardware Component Failed', but I'm assuming since there were beep codes at all (one continuous f/b 4 short beeps) it must be the mainboard.

I was all set to upgrade to 7900X or some comparable 6+ core solution, but this single core benchmark is exceedingly underwhelming, especially considering I will need to change cooling solutions to water which I've never done before and no nothing about. 

Am I reading all of this correctly?

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

18 minutes ago, Noel said:

I just looked at this Passmark chart:  https://www.cpubenchmark.net/singleThread.html

By my calculations my 3930K at 4.42Ghz versus 7700K at 4.5Ghz gets a whopping increase of 8% in that single-core benchmark.  And the 3930K has been on a Noctua air cooler.  This is very discouraging as my 3930K box just died, and I'm thinking it's the ASUS P979X WS mainboard w/ its beep code saying, 'Hardware Component Failure'.   The 3930K could still be working!  This leaves me wondering if I shouldn't pick up a new ASUS P979X WS which is available on eBay right now and maybe this means I can resus the entire system w/ minimal work, no software reloading, etc.  Very attractive.  I could continue to use my 32Gb of DDR3 at 2400mHz and my CPU, versus upgrading to a new mobo/ram/cpu.  The problem of course is that I really don't know which 'Hardware Component Failed', but I'm assuming since there were beep codes at all (one continuous f/b 4 short beeps) it must be the mainboard.

I was all set to upgrade to 7900X or some comparable 6+ core solution, but this single core benchmark is exceedingly underwhelming, especially considering I will need to change cooling solutions to water which I've never done before and no nothing about. 

Am I reading all of this correctly?

I guess not. 7700K is newer tech and faster than 6700K. And I am dead certain the 6700K is faster than a 3rd gen ivy bridger

Share this post


Link to post
Share on other sites
2 hours ago, Piotr007 said:

I guess not. 7700K is newer tech and faster than 6700K. And I am dead certain the 6700K is faster than a 3rd gen ivy bridger

"Faster" applies when it's 0.5% faster, right?  Sure.  Actually 3930K is 2nd generation I-7.   Of course that's only Passmark.  If someone has objective evidence that this 8% is woefully under representing single core performance AS IT APPLIES to P3D, please direct me to a source.    Everyone can agree CPU performance as it related to P3D is LARGELY, not completely, single-core performance dependent.  So that matters a whole lot!  It appears i9-7900X and I7-7700 are neck and neck for single core.  I do want to have at least 6 cores as I am a staunch believer in it having enjoyed 3930K for these 4.3y and performance has been real good.

Share this post


Link to post
Share on other sites
9 minutes ago, Noel said:

"Faster" applies when it's 0.5% faster, right?  Sure.  Actually 3930K is 2nd generation I-7.   Of course that's only Passmark.  If someone has objective evidence that this 8% is woefully under representing single core performance AS IT APPLIES to P3D, please direct me to a source.    Everyone can agree CPU performance as it related to P3D is LARGELY, not completely, single-core performance dependent.  So that matters a whole lot!  It appears i9-7900X and I7-7700 are neck and neck for single core.  I do want to have at least 6 cores as I am a staunch believer in it having enjoyed 3930K for these 4.3y and performance has been real good.

I am interested in a hexa-core as well. Do you think P3D V4 will benefit more from 6 cores than from 4? I am using i7 6700K now at 4.5 ghz. I guess this is easily achieveable with a 6-core, to get 4,5ghz speed?

The downside is, more cores is more heat generated at OC so less headspace to overclock. That makes a 6-core a useless upgrade if it is true...

 

I should also mention that with ORBX, FSDT KJFK, Ultimate Live traffic, ASP4, I get 100% CPU usage on all 4 cores at moments. Causes many lag graces when flying near New York.

Share this post


Link to post
Share on other sites

My sense all along has been, Gee, I have an almost 5 y/o 2nd generation i7 based system and I think from what I read here I was getting very decent performance.  I know my GPU is a major bottleneck to being able to do more GPU intensive slider settings (Lighting group predominantly) but I have to think my CPU is still right in there w/ current offerings--only 8% worse if this one benchmark applies to P3D.  It's unfortunate that single-core performance has improve so little in nearly 5 years.   This being said, I do believe P3D benefits enough from multicore to have at least 6 cores if not more, provided you don't lose much single-core performance to get there.  

With SteveW's and GSalden's help I found the right affinity mask for my hexacore running HT enabled at 4.42Ghz, and performance is really excellent.  I do respect the performance ceiling by avoiding mixing too many complex attributes in a given flight plan.   So as a result I have liquid smooth stutter-free performance almost always.  I will benefit from a 1080Ti so eventually will go there.

  • Upvote 1

Share this post


Link to post
Share on other sites

I don't overclock and I am doing 7700 on air (no K). :smile:

blaustern

 

 

Share this post


Link to post
Share on other sites

You're comparing an older extreme chip to a newer mainstream chip with less cores and gimped memory bandwidth, hard to see how it all pans out but I've studied it hard and don't see that much more benefit unless you put that 7700k on water and get it to 5Ghz. P3D and FSX - clock speed is king.

There is now the new X299 platform coming from Intel too that looks like a better replacement for an X79 system than a mainstream Z270 chipset

Share this post


Link to post
Share on other sites
1 hour ago, Slayer said:

You're comparing an older extreme chip to a newer mainstream chip with less cores and gimped memory bandwidth, hard to see how it all pans out but I've studied it hard and don't see that much more benefit unless you put that 7700k on water and get it to 5Ghz. P3D and FSX - clock speed is king.

Yes, I threw in i7-7700K only because it had almost identical single core perf as does i9-7900X and I will be looking at picking up either 7820X or 7900X.   If 7900X can sustain at least one core at 4.6Ghz or more I would set all other cores to something real low, like 3.6Ghz which I would imagine can easily manage peripheral processes and terrain texture loading.  Perhaps cooling needs can then easily be managed with a simpler AIO water cooler.

My system will benefit greatly by 1080Ti or Titan Xp as well.  I think it will be best to get one of these installed on day 1 because from what I see in benchmarks what I will see will largely be underwhelming without adding the new GPU, and I hate to spend this kind of $$ for underwhelming!

Share this post


Link to post
Share on other sites

I believe I read the higher end chips can OC 2 cores now too, it will be interesting to see the impact on sims.

Share this post


Link to post
Share on other sites
1 hour ago, lownslo said:

I'm not sure how they have set up this comparison.   If you look at their 'Peak Overclocked Bench' the 7700 is 29% faster than the 3930K.  But the average overclock submission for the 7700 was 5.15Ghz, whereas for the 3930K it was only 3.6Ghz.  My 3930K runs at 4.42Ghz, so the +29% increase in mixed single-core performance goes down to a grand total of +5.4% if you use those numbers.   As I say, I can't quite tell what they are comparing.  The first metric, 'Effective Speed' apparently was based on the stock speeds of the 7700 at 4.2Ghz, and the stock speed of the 3930K of 3.2Ghz (!!), so the +41% goes out the window big time.

I'm afraid progress CPU single-core performance has sadly come to near halt over the past 5 years which is why simulator developers should think out of the box enough to figure out how to better access all of this poorly tapped multicore potential.   

  • Upvote 1

Share this post


Link to post
Share on other sites

Given raw math performance the i7 39s are good, I'll upgrade my 3960x @4.1 when it goes pop.

Intel Core i7-3960X @ 3.30GHz = 12686, 12686 / 3.3 * 4.2 = 16145

Intel Core i7-3930K @ 3.20GHz = 12027, 12027 / 3.2 * 4.2 = 15785

Intel Core i7-7700K @ 4.20GHz = 12178, 12178 / 4.2 * 5.15 = 14932

Can the performance be extrapolated by way of frequency in this way? Yes, more or less, but the overall performance gains for recent hardware are through motherboard components and improvements to chipsets.

Share this post


Link to post
Share on other sites
32 minutes ago, SteveW said:

...but the overall performance gains for recent hardware are through motherboard components and improvements to chipsets.

Plus the improved IPC of the newer CPU's can certainly be advantageous.  Only comparing clock per clock is simply not enough for a valid performance discussion.

Regards,

Greg

Share this post


Link to post
Share on other sites

I wonder how much sim add-on developers work w/ LM to work out a way for those add-ons, or components of them, to be able to sync w/ P3D and run outside of P3D.  I understand Majestic Software's Dash 8 Q400 has some significant part of it running outside of P3D and that this helps account for its truly stellar performance.  Why doesn't PMDG do this, that is, if they don't already?   GSX seems to have the potential for a negative impact slight as it may be within P3D, so can it be made to run outside P3D's main thread?  Maybe it does already but it doesn't really behave like it at first glance.  So yes, since the constraint is the slow rate of improvement in single-core performance it seems obvious the place to spend effort is in multicore to the max.   

Share this post


Link to post
Share on other sites
8 hours ago, lownslo said:

Plus the improved IPC of the newer CPU's can certainly be advantageous.  Only comparing clock per clock is simply not enough for a valid performance discussion.

Regards,

Greg

Absolutely, for example why a 3960x outperforms a 3930 at the same frequency.

Share this post


Link to post
Share on other sites
On 28/06/2017 at 8:33 AM, Noel said:

I'm afraid progress CPU single-core performance has sadly come to near halt over the past 5 years

I think that statement is essentially the gist of the story, although IPC and memory improvements play an important role as well. 

 

Personally, I'm sitting on a good 4770k (4.6Ghz)  with 2400Mhz RAM and I haven't seen anything recently that convinced me it's worth my time and money to upgrade. Hopefully the next upgrade cycle will change that. 

  • Upvote 1

Share this post


Link to post
Share on other sites

I go to and use the Guru3d forums where the smart kids only overclock for bragging rights and numbers. There are some real gurus there that know their stuff so when I asked about trading my 5930k in for a 7700 they universally replied that I was crazy and that it would be a down grade. 

To go to all the expense and work to get new mb and ram and cpu for a couple fps is nuts so I will wait until something that is really much faster to hit the sales floor. 

Share this post


Link to post
Share on other sites

Has anyone done a 'core utilization' type of analysis of P3D V4 to confirm that  single threaded performance is still a key CPU requirement for best sim experience ? I would appreciate a link to any such analysis. 

Share this post


Link to post
Share on other sites
1 hour ago, warbirds said:

I go to and use the Guru3d forums where the smart kids only overclock for bragging rights and numbers. There are some real gurus there that know their stuff so when I asked about trading my 5930k in for a 7700 they universally replied that I was crazy and that it would be a down grade. 

To go to all the expense and work to get new mb and ram and cpu for a couple fps is nuts so I will wait until something that is really much faster to hit the sales floor. 

Well, that is where I am at right now, poised to spend serious $$ to do an 'upgrade' that barely is.  The GPU bump will be nice, but CPU/Mobo/Ram--nada.  I sent my dead ASUS motherboard for my 3930K in last week and should hear from them within a week as to whether something can be fixed on it, else off we go to do a new build.  The preposterous part is cooling needs have actually gone up!

I'm hoping Threadripper comes before I've lost my patience for not having a simming PC--I miss it!  I've been sim flying for 20+ years almost daily!

Share this post


Link to post
Share on other sites
4 hours ago, odourboy said:

Has anyone done a 'core utilization' type of analysis of P3D V4 to confirm that  single threaded performance is still a key CPU requirement for best sim experience ? I would appreciate a link to any such analysis. 

All you have got to do, is run the Windows Task Manager..

Yes, P3DV4 still is heavily single core dependent (although more is being offloaded than previously).

Share this post


Link to post
Share on other sites

Thanks Bert. I've monitored core and GPU with V4 but a) I see some behaviors I don't understand (not being privy to the guts of P3D) and b) my old 3570K may not be really indicative of how a more modern HT multi core processor runs the sim.

Share this post


Link to post
Share on other sites
21 hours ago, odourboy said:

Has anyone done a 'core utilization' type of analysis of P3D V4 to confirm that  single threaded performance is still a key CPU requirement for best sim experience ? I would appreciate a link to any such analysis. 

Yes, and it is, but that's far from the whole story. The sim splits up its processes depending on how many logical processors (LPs) it sees.

It is important to understand how these processes split out across the CPU since some processes can happily share the LPs of the same core, and yet others are not so happy sharing a core as they want full bandwidth of that core.

So we can look at how the sim utilises cores for loading scenery, and how that can affect the rendering stage of the sim when the CPU is HT enabled - without care and the appropriate Affinity Mask we can instead lose performance.

Have a look at the following diagram for a four core HT enabled running P3D v4. HT disabled is basically ignored as in all systems it won't be so good on performance and anyway no Affinity mask is required for an ideal setup. Instead this diagram of Task Manager graphs concerns only four core HT enabled, which will work better with the correct affinity mask, and as always attention must be paid to where the addon exe apps run for the final outcome:

 

P3Dv4_four_core_HT.jpg

 

  • Upvote 1

Share this post


Link to post
Share on other sites

1. In above diagram could you clarify the result with AM=85 as I think it would give two loading cores similar to 245 and make rendering free from loading?

2. If I change p3d to high priority in task manager which should not hold up the process for other things as addons and some OS work, but with HT on and sharing one LP with p3d and other LP of the same core with addon or other app having a cpu demand, will it choke the p3d process despite the high set priority as from what I read windows sees them as separate physical cores and that should instead assign the addon process on the same LP as p3d thread to let windows priority work normally?

Share this post


Link to post
Share on other sites
On ‎08‎/‎07‎/‎2017 at 3:22 AM, him225 said:

1. In above diagram could you clarify the result with AM=85 as I think it would give two loading cores similar to 245 and make rendering free from loading?

2. If I change p3d to high priority in task manager which should not hold up the process for other things as addons and some OS work, but with HT on and sharing one LP with p3d and other LP of the same core with addon or other app having a cpu demand, will it choke the p3d process despite the high set priority as from what I read windows sees them as separate physical cores and that should instead assign the addon process on the same LP as p3d thread to let windows priority work normally?

 

In the graphs the situation is loading scenery and objects. The red dots show where an LP is used when loading. In the second graph 85 with four LPs allocated, shows that LP0 top left is rendering and also charged with loading. The graph 245 with six LPs shows no loading on the first two LPs. However 85 has three cores loading but 245 only has two cores loading (4 LPs) which intensify work on those two cores and care should be taken when pushing an overclock, check temperatures during scenario loading.

If you change P3D to high priority it can't run any faster. Instead other processes are not provided the same bandwidth and since many of these are providing resources to the sim, irrespective of addons, this can hold up the sim.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now