The path forward to a CPU hyper-jump is perhaps yet to come, in 2018

[SpiritFlyer 987]

Fellow system deciders,

This recent information was scattered elsewhere on different threads, but perhaps it can be best useful served more in one place.

Like some I have an older system but the latest software and FS products and looking forward to a great hardware jump so I can upgrade. However, I am (personally) underwhelmed with the 7700K and 8700K marginal P3Dv4 performance improvements being reported, being a few percentage points faster at best (if any), even with many of them testing with reduced P3D settings. I had hoped and expected much better from the first hopeful rumours.

Unless there is a huge surprise the new CPU performance numbers required for me to change from my rock steady 2700K and GTX1080 combo have not yet reached the tipping point. Although I am sure most (and possibly all) new equipment might exceed the performance shown here, it remains doubtful that it will be by sufficient margins to warrant the expense and trouble to do so...yet. 

Part of the reasons are explained and documented in the below link by Hardware Canucks testing the 2700K vs the 8700K in gaming scenarios:

http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/76333-i7-2600k-vs-i7-8700k-upgrading-worthwhile.html

Please note that the test 2600k used above was not clocked as high as many Sandy Bridges safely run at, including my 2700k (see below).

Please don't get me wrong. It is time for a trade, so I look forward to being lured into a new build when the next true flight simulation performance increasing CPUs arrive,

but until then, here are my specs, settings, and numbers:

Processor: 2700K @ 4.938Ghz (as shown)
Motherboard: Asus P67 Deluxe (Revised and replaced by Asus)
Video Card: GTX1080 Overclocked with MSI Afterburner runs between 90% and 100% load at 120% speed (as shown)
RAM 16G at 2196 (as shown)
Primary Drive: 240G SSD
Flight Sim Drive: 480G SSD
Monitor: 4K 55inch curved screen Samsung HDTV Smart TV
Flight Simulator: P3Dv4.1 (over 340G installed)
Settings: (as shown)
Results: Entirely satisfactory with smooth framerates between 22 and 30

Please see the P3Dv4.1 settings and screenshots pictured below:

https://www.flickr.com/gp/156377156@N06/S536g0 

New Information:

Nvidia claims the new 388.00 driver (and forward) significantly improves SLI performance.

So perhaps a 2nd card for 2x1080's in SLI would change things even more since my single card runs wide open?

In any case, am looking forward to exciting new horizons to be reached by those now testing the latest, better and best! 

Thank you for your patience.

Kind regards,

Stephen

[martin-w 3,235]

Quote

Part of the reasons are explained and documented in the below link by Hardware Canucks testing the 2700K vs the 8700K in gaming scenarios:

 

The sim isn't an ordinary "gaming scenario" though Stephen. It's still very dependant on CPU horsepower. You would need to see a direct comparisons "in the sim" between the two CPU's. Personally, given the significant increase in IPC between the two CPU's, I would be surprised if there wasn't a worthwhile improvement in the sim.

Conventional games are VERY GPU biased and one can quite easily get away with a less than top notch CPU. So I'm not surprised at all that the conventional games tested at Hardwarecanuks showed little improvement. Precisely what I would have expected. 

 

Quote

Results: Entirely satisfactory with smooth framerates between 22 and 30

 

Very subjective, but given my sensitivity to low frame rate, I would be disappointed with that. Single core performance comparing 2700K and 8700K is 39%. Does that mean you will see up to 42 frames per second? Too many variables to say, but in my opinion, I would say feasible. 

 

Quote

 

Nvidia claims the new 388.00 driver (and forward) significantly improves SLI performance.

So perhaps a 2nd card for 2x1080's in SLI would change things even more since my single card runs wide open?

 

 

I'm no expert in SLI, but conventional wisdom has it that gains from SLI in the sim are small and only at high res. SLI can also be troublesome. As I say, I'm not an expert in SLI in the sim though, and it would be wrong of me to be definitive with no experience. 

 

[Anxu00 176]

I too will be looking at HW change next year.  I have the same CPU as Stephen but the GPU is older.  However, my monitor is 2560x1600 not as high as a 4K, so my GPU is fine at the moment.  For people like me, 2018 looks like a sweet spot.

[aentwis 61]

Vu, I am running the 2700K at 4.8GHz as well, though I know have it paired with a GTX1070Ti and 16Gb of RAM (at 1866 timings)

I have been looking at upgrading for about a year now. One way forward is to go for a selected CPU taht can run at a guaranteed clock of say 5 or 5.1GHz... that way, the performancce gain is quite noticeable my guess... particularly when paired with more recent GPU technology like the 1080or 1070Tis...

As an aside though, what voltage are you running your 2700K to get 4.8GHz?

A

[lownslo 592]

58 minutes ago, martin-w said:

The sim isn't an ordinary "gaming scenario" though Stephen. It's still very dependant on CPU horsepower. You would need to see a direct comparisons "in the sim" between the two CPU's. Personally, given the significant increase in IPC between the two CPU's, I would be surprised if there wasn't a worthwhile improvement in the sim.

Conventional games are VERY GPU biased and one can quite easily get away with a less than top notch CPU. So I'm not surprised at all that the conventional games tested at Hardwarecanuks showed little improvement. Precisely what I would have expected.

+1  The review HardwareCanucks offerd was quite nice... for gamers.  But it didn't tell us simmers all that we need to know before spending our money on an upgrade... specifically, how the systems compare.  The newer systems will have faster memory, faster buses, faster... everything.  All these attributes combined plus the all-important enhanced IPC of the 8700K over an older CPU will offer strong dividends.

The last thing we should do in deciding to commit to a new build is compare CPU speeds.  They are just one (simple) part of the equation.

Greg

[martin-w 3,235]

3 hours ago, lownslo said:

 

The last thing we should do in deciding to commit to a new build is compare CPU speeds.  They are just one (simple) part of the equation.

Greg

 

Yep. 

[martin-w 3,235]

3 hours ago, aentwis said:

One way forward is to go for a selected CPU taht can run at a guaranteed clock of say 5 or 5.1GHz... that way, the performancce gain is quite noticeable my guess...

 

 

Bear in mind though, that the difference in performance in terms of frame rate between, lets say, 4.8 and 5 GHz is small. It's pretty much linear as long as the system is reasonably balanced. 

So yes, achieving the much sort after 5 GHz is nice, but it's the CPU's IPC that's the big deal. 

[Anxu00 176]

4 hours ago, aentwis said:

As an aside though, what voltage are you running your 2700K to get 4.8GHz?

1.32 v I think.  Very stable for almost 6 years now.  If you want exact number, I will have to boot to BIOS to know it (or may be looking into CPU-Z) as it has been a while. From which GPU did you go from to get to 1070Ti?  I am interested also to know that experience too.

[SpiritFlyer 987]

6 hours ago, martin-w said:

 

The sim isn't an ordinary "gaming scenario" though Stephen. It's still very dependant on CPU horsepower. You would need to see a direct comparisons "in the sim" between the two CPU's. Personally, given the significant increase in IPC between the two CPU's, I would be surprised if there wasn't a worthwhile improvement in the sim.

Conventional games are VERY GPU biased and one can quite easily get away with a less than top notch CPU. So I'm not surprised at all that the conventional games tested at Hardwarecanuks showed little improvement. Precisely what I would have expected. 

REFUTED BY HARDWARE CANUCKS VIDEO

Very subjective, but given my sensitivity to low frame rate, I would be disappointed with that. Single core performance comparing 2700K and 8700K is 39%. Does that mean you will see up to 42 frames per second? Too many variables to say, but in my opinion, I would say feasible. 

NOT ACCORDING TO HARDWARE CANUCKS VIDEO

 

I'm no expert in SLI, but conventional wisdom has it that gains from SLI in the sim are small and only at high res. SLI can also be troublesome. As I say, I'm not an expert in SLI in the sim though, and it would be wrong of me to be definitive with no experience.

Yes it would 

 

I disagree with your general discounts of the Canucks hardware testing being relevant to those using fast Sandy Bridge CPUs within Flight Simulation.

These variables you mention are well accounted for and measured within the testing. Even P3Dv4.1 is still very much like the games that are tested and I bet the findings are very similar as well. The CPUs have great single core performance!

As far as my subjectivity interfering within my "subjective" assessment, I guess your subjective denials are just as valid as my subjective positives. The difference is I see exactly what I am saying and provided documentation, and you are only reading my words.

I am a bit disappointed actually that the objective measurements are under such a cloud of dismissal. The truth is the latest chips are not rocket-ships, but they win in comparisons, but only by a nose within similar PC games and flight simulation examples.

That will change when better hardware emerges, and at that point so will my subjective opinions and objective purchasing decision when that time comes, probably in 2018.

Anyway, what does it matter? By the time 2019 comes around nearly everyone still breathing, flying and writing on AVSIM will likely have new rigs, including me, Lord willing!  

Kind regards, 

Stephen

 

[G-RFRY 1,710]

I think the GPU is going to play a bigger role in future upgrades with the 64bit platform, and it is possible next years GPU could big as big leap as this year with new dia size and DDR6 rumoured. 

[sebastorama 2]

Something is seriously wrong inside P3D. I know it's  not a video game, but man, the 1-thread CPU limitation is ridiculous. The focus should be Spawn the aircraft sim in one (even the aircraft can be divided in multiple workloads), the render workers in other, weather in other, AI ... etc I would bet that the SDK in P3D (and FSX) is so Monolithic, that a change like that would be messing with a monstrous dependency spaghetti. The only hope for the future (I really doubt that we will see single-thread huuuuge increments on the CPU-side, multithreaded is a whole other story, Threadripper is a reality) is a complete rewrite, a new competition, X-Plane getting some serious stuff like Active Sky / PMDG... Dreams dreams... 

Also, a more comprehensive way to separate addons would be great, every addon require no anti-virus, run as admin, some specific setup, whitelist on firewalls, and so on and so on... When I get my sim together I get that impression that everything is glued together, so fragile, hard to know what and when anything is going to break =/ .... BUT! ... I dunno exactly what is, but when you're completely immerse in a new situation, see something beautiful out of the window, dominate that plane you've beeing studying for days... That's priceless, and no other kind of "game" can do that :P 

[GHarrall 494]

All I can say is my jump from 2600k at 4.6 ghz to a stock speed 6700k was very noticeable, even more so when the overclock to 4.5 ghz went on. 

Obviously the switch to 3200 MHz DDR4 helped as well, but it’s clear that the IPC on the 6700k was far superior to the 2600k.

 

[martin-w 3,235]

15 hours ago, SpiritFlyer said:

 

Kind regards, 

Stephen

 

 

REFUTED BY HARDWARE CANUCKS VIDEO

 

I've just read the entire article.  Where were my comments refuted by Hardware Canucks? I see exactly what I said, conventional games, which are graphics card biased performing admirably with both CPU's. However, when we look at more CPU biased rendering tasks the 8700K is massively better. And worth mentioning that some of those rendering applications utilise both CPU and GPU, just as P3D does.

Yes, I understand that P3D utilises the GPU more than the sim used to, but CPU is still the primary bias isn't it? Or am I wrong?  Has there been a huge reversal in the bias in P3D and it's now 90% GPU? I've not heard that, but then I have been out of simming for a protracted period. 

Yes it would 

Regarding SLI...  You asked the question and then immediately rejected my opinion as wrong. I'm curious as to why you asked the question in the first place, if you were convinced you already knew the answer? 

Even P3Dv4.1 is still very much like the games that are tested and I bet the findings are very similar as well. 

Really, do you have any evidence of that?  The games tested are VERY GPU biased, as evidenced by the test results. Way more GPU biased that P3D that has been tweaked to utilise somewhat more of the GPU I would have thought. 

I am a bit disappointed actually that the objective measurements are under such a cloud of dismissal. The truth is the latest chips are not rocket-ships, but they win in comparisons, but only by a nose within similar PC games and flight simulation examples.

"Flight sim examples"... as I said, can you show us evidence of that?

I would have to say that to claim that games, well known to be very, very GPU biased, are equivalent to P3D is a stretch to be honest. Utilise the GPU more than it used to yes, but have Lockheed Martin really "completely reversed" the  sims CPU bias so it is now very heavily GPU biased? That would surprise me greatly. 

I'm willing to be convinced if you have evidence.

 

 

Quote

 

 

[martin-w 3,235]

8 hours ago, GHarrall said:

All I can say is my jump from 2600k at 4.6 ghz to a stock speed 6700k was very noticeable, even more so when the overclock to 4.5 ghz went on. 

Obviously the switch to 3200 MHz DDR4 helped as well, but it’s clear that the IPC on the 6700k was far superior to the 2600k.

 

 

Yep, to be honest I noticed a slight improvement when I jumped from a 3770K to a 6700K, a mere one architecture jump, rather than Stephens four architectures jump. Only slight, but it was there.

[martin-w 3,235]

Ignore.

 

Quote