PC upgrade, worth it?

[swissa340 1]

Hi everyone

I‘m considering upgrading my CPU because I‘m not happy with my performance anymore. I have an i7 4790k OC@4.7Ghz and an ASUS GTX1080ti. P3D is on a separate SSD. My settings are rather moderate. Autogen all sliders on medium to high settings. I’ve done a lot of research concerning performance in P3D and I think that my CPU just can’t handle things anymore. Core 0 is always at 100% while the other three range between 20-50% usage. Do you think it’s worth upgrading? I thought about an i7 9700k or maybe even an i9 9900k. Any comments about this?

 

[Bob Scott 7,721]

Well, the 4790K is still no slouch.  I still use my 4790K machine (also with a 1080Ti) for FPS and driving games. 

That said, P3D really runs well on 6 or more cores, enough so that I think it would make an upgrade from your Devil's Canyon CPU worthwhile. 

 

[Noel 4,501]

2 hours ago, swissa340 said:

Hi everyone

I‘m considering upgrading my CPU because I‘m not happy with my performance anymore. I have an i7 4790k OC@4.7Ghz and an ASUS GTX1080ti. P3D is on a separate SSD. My settings are rather moderate. Autogen all sliders on medium to high settings. I’ve done a lot of research concerning performance in P3D and I think that my CPU just can’t handle things anymore. Core 0 is always at 100% while the other three range between 20-50% usage. Do you think it’s worth upgrading? I thought about an i7 9700k or maybe even an i9 9900k. Any comments about this?

 

Your 4790k is ~4.5% faster in single-threaded performance than my now 6.5 y/o 3930K when running at the same clock speed.  I can only do 4.44Ghz w/ HT enabled.

A 9900K will be about 19% faster than your 4790k at the same clock speed.

FWIW, my Core 1 is generally always below 100% because I always adjust settings to insure that happens, which is required for completely stutter-free fluid video performance.  It appears we both have similar settings, except I have GPU related sliders down very low.  Quite frankly in a side-by-side comparison you will hardly note a difference as the bang for buck on the GPU is very poor IMO.   Run a GPU utilization monitor to see how often as you approach max GPU utilization as frame rate and freedom from stuttering suffers immensely.

To me 19% starts being meaningful, and as you will likely get to 5Ghz you can expect a bit more than 19%, maybe 20%.  But then again, P3D is replete w/ slider settings that can take anything down.  So in the end, it's always going to be a case of staying w/in the processing budget you have to work with.   Since GPU related settings offer so little for such a high cost I would consider upgrading your CPU to 9900K, but skip the GPU upgrade for now.   Maybe when RTX 3080Ti arrives reconsider, though once again GPU related improvements relative to the cost of the GPU are pretty paltry.  For my 3930K HT enabled offers the superior performance.  I'm still using the original GTX Titan, now also over 6y/o.    I'm building a new machine next fall if my current old box makes it that far.

Edited December 28, 2019 by Noel

[ryanbatc 12,856]

Are you at 4k?

I've got an identical rig....  But at 1080p - it's been really enjoyable and a nice upgrade from my 2500k.

Usually I'm in the 25 fps area around Orbx socal and fsdt klax.  With payware and asp4 clouds/ENVTEX.

Edited December 29, 2019 by ryanbatcund

[swissa340 1]

21 hours ago, Noel said:

Your 4790k is ~4.5% faster in single-threaded performance than my now 6.5 y/o 3930K when running at the same clock speed.  I can only do 4.44Ghz w/ HT enabled.

A 9900K will be about 19% faster than your 4790k at the same clock speed.

FWIW, my Core 1 is generally always below 100% because I always adjust settings to insure that happens, which is required for completely stutter-free fluid video performance.  It appears we both have similar settings, except I have GPU related sliders down very low.  Quite frankly in a side-by-side comparison you will hardly note a difference as the bang for buck on the GPU is very poor IMO.   Run a GPU utilization monitor to see how often as you approach max GPU utilization as frame rate and freedom from stuttering suffers immensely.

To me 19% starts being meaningful, and as you will likely get to 5Ghz you can expect a bit more than 19%, maybe 20%.  But then again, P3D is replete w/ slider settings that can take anything down.  So in the end, it's always going to be a case of staying w/in the processing budget you have to work with.   Since GPU related settings offer so little for such a high cost I would consider upgrading your CPU to 9900K, but skip the GPU upgrade for now.   Maybe when RTX 3080Ti arrives reconsider, though once again GPU related improvements relative to the cost of the GPU are pretty paltry.  For my 3930K HT enabled offers the superior performance.  I'm still using the original GTX Titan, now also over 6y/o.    I'm building a new machine next fall if my current old box makes it that far.

I don‘t want to upgrade my GPU only my CPU. Right now, my 1080ti is at about 40% during a flight which is pretty low. I‘ll probably go for the i7 9700k and the ASUS ROG Strix Z390-F Gaming Mobo with new DDR4 RAM. 

 

4 hours ago, ryanbatcund said:

Are you at 4k?

I've got an identical rig....  But at 1080p - it's been really enjoyable and a nice upgrade from my 2500k.

Usually I'm in the 25 fps area around Orbx socal and fsdt klax.  With payware and asp4 clouds/ENVTEX.

No, I’m running 2k at the moment bc my monitor only has this resolution. Yes, normally I’m in that area as well. During cruise I even get ~40FPS but as soon as I approach any airport I have slow loading Autogen/Scenery/Night Lighting and between 15-20FPS. 

[Quasimodo 505]

I had a Haswell system for 6 years. i7 4770K at 4.8ghz, 780ti, 2400mhz CL9-11-11. At the time I built if it, it was second to none. I could even get it stable at 4.9ghz but kept it at 4.8 because the extra 100mhz didn't make any difference and require quite the Vcore boast. In the spring I got an itch to build a new system because I wasn't getting the performance I wanted in VR. At that point I had already been through a 980ti and onto a 1080ti. Everything else in the system was the same. I recall not being overwhelmed by the improvement with upgrades but there was improvement to measure for sure. When I started sniffing around for parts (used and/or Open box because I'll build one generation behind from here on out) I got a good deal on a used Asus Maximus X Apex and an open box 8086K. Fortunately, I had a lot of good stuff to recycle. Case, PSU, THermalRight Silver Arrow, 1080ti but I still ended up braking the bank with some new stuff: 3 x 1TB Samsung 970 Evo Plus drives and 16gb of G.Skill DDR4 4400Mhz CL18-19-19

After all that I was lucky again with the silicon Lottery ( on my next build the odds are that I am getting a dud CPU). I am getting 5.3ghz on all six cores stable. I've done 5.4 too but again the Vcore performance trade off isn't there. All in all once I measured the performance difference between the new system and old system it wasn't as much as I had expected it to be. It was around a 20%+ improvement if I recall correct. I was hoping for about 30% improvement. Looked better in game play though. I was seeing 30+fps where I was preciously seeing 24fps

Still I found that the impact of bad weather with my favoured setting (which are high😉) took its toll. The system bottleneck went from the CPU to the GPU. So a few weeks ago I went a hunting for a used 2080ti and found one. Again buying used is not a risk free endeavour. Not everyone is honest! As it turned out the 2080ti I got was faulty but the seller was honest and took care of the RMA for me and I got back a brand new 2080ti.

Here is where the build has come into its own. The 2080ti contrary to many posts I had read in the past made a huge difference. Not in scenarios where the CPU was the bottle neck i.e. fair weather, mid day scenarios but in overcast conditions and/or at dawn or dusk where the 1080ti was the bottle neck, the 2080ti made a huge difference. In the same overcast test scenario with the exact same setting where I was seeing ~ 20fps with the 1080ti when taxing from gate to runway I was seeing ~30+ fps with the 2080ti. So in the ballpark of a 50% performance increase in GPU intensive scenarios.

So in conclusion I spent CAD$3000 on a system built from a combination of recycled, used and new part. I have found a buyer for my 1080ti who is paying me CAD $700 tomorrow. So that gets me down to CAD$2300. I still have the i7 4770K, the Asus Z87 Sabertooth and the 16gb of G.Skill DDR3 2400mhz. At the time I had decommissioned that system I had put them up for sale but I was in the process of preparing for a big cross country move and now 6 months later I still have them. Anyway I do need another PC in my house so there you have it.  Had everything been new I'd guess CAD$4500+. Was it worth it? Yes it was. I'll guess that had I ran that same overcast test scenario on my Haswell system with the 1080ti it would have been the CPU that bottle necked. I'll guess ~18 FPS taxing to rwy. So if that guess is close, since I got the itch I've gained about 60% in FPS in scenarios that hit both the CPU and the GPU Hard.

With Microsoft Flight Simulator coming and Prepar3D V5 and XP11.50 I'd say I'm good for a few years. My only concerns is 6 cores and 16gb of ram as apposed to 8 cores and 32gb of ram.

That i7 4790k system if you built it 5 years ago owes you nothing. Scratch that itch!

Edited December 29, 2019 by Avidean

[Noel 4,501]

36 minutes ago, Avidean said:

Still I found that the impact of bad weather with my favoured setting (which are high😉) took its toll. The system bottleneck went from the CPU to the GPU.

That, of course, is 100% a function of what your GPU driver settings are at, and of course some of those nifty GPU-related slider settings many to most of which can't hardly be perceived by the user when enabled or not.  I just flew out of KDEN HD, in blizzard pure white-out conditions, in the PMDG 777, using a GPU that is 50% slower than the 1080Ti.  Solid smooth stutter-free taxi at 30 frames.   After TO frames dipped down to 25 or so for a few minutes until we got near the Front Range west of Denver OTW to Kona HI.   The frame rate dump was secondary to my now 6.5y/o CPU at 4.42Ghz and had zero to do with the GPU which was chugging along at about 70% utilization.   On a 3440x1440 curved screen using DSR 4x via nV driver setting, aniso at 16x, FXAA off, MSAA 4 Samples, and have freedom from jaggies and a crystal clear image. 

I bring this up to emphasize the fact it is not difficult to obtain an excellent, stutter-free, ultra high quality simulation with less than ideal hardware.   If you were having ANY difficulty obtaining fabulous performance w/ the 1080Ti in ANY kind of severe weather I strongly doubt you were using settings that deliver on visuals w/o the type of impact you were seeing.  Put another way, I'll bet $$$ you could have easily set up that 1080Ti, with its full 50% greater thru put than my old GTX Titan 6Gb, such that it could easily have handled any kind of weather thrown at it.  My old Titan certainly does.  I'm using ActiveSky Next w/ 5 layers, high density out to 100m.

[pgde 571]

On 12/28/2019 at 9:54 AM, w6kd said:

Well, the 4790K is still no slouch.  I still use my 4790K machine (also with a 1080Ti) for FPS and driving games. 

That said, P3D really runs well on 6 or more cores, enough so that I think it would make an upgrade from your Devil's Canyon CPU worthwhile. 

 

Just FYI, the 4790K is a Haswell not Devil's Canyon. 

[Bob Scott 7,721]

1 hour ago, pgde said:

Just FYI, the 4790K is a Haswell not Devil's Canyon. 

Well, It's actually sort of both...but the 4790K most certainly *is* a Devil's Canyon CPU, along with the i5-4690K.  The Devil's Canyon processors used binned Haswell CPU dies with additional capacitors on the package for better rectification and power delivery, and were also fitted with a much-improved thermal interface.  That CPU really rocked...my 6-year old 4790K machine is still holding its own quite well as a gaming platform.

https://ark.intel.com/content/www/us/en/ark/products/80807/intel-core-i7-4790k-processor-8m-cache-up-to-4-40-ghz.html

https://www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k

[Quasimodo 505]

6 hours ago, Noel said:

That, of course, is 100% a function of what your GPU driver settings are at, and of course some of those nifty GPU-related slider settings many to most of which can't hardly be perceived by the user when enabled or not.

Couldn't agree more but whether something can be perceived or not varies from one person to another. In my case as a VR simmer the system I have now with the setting I use give me a pretty consistent and smooth 30fps in any soup and without really compromising on the setting. It is however in the eye of the beholder. And with the same setting on my previous system with the 1080ti in VR it was a slide show.

Edited December 30, 2019 by Avidean

[swissa340 1]

First of all thanks for your opinions. I think it’s quite helpful when someone tells his experience from upgrading hardware. 

10 hours ago, Avidean said:

Couldn't agree more but whether something can be perceived or not varies from one person to another. In my case as a VR simmer the system I have now with the setting I use give me a pretty consistent and smooth 30fps in any soup and without really compromising on the setting. It is however in the eye of the beholder. And with the same setting on my previous system with the 1080ti in VR it was a slide show.

But let’s talk about Scenery an Autogen for a moment because I think this is the most strenuous task for the CPU. What about your settings there. My aim is to have high to very high settings. Will a new CPU help me with that. My FPS are between 20-25 when I’m at an airport with the FSLabs and full weather etc. 

[Noel 4,501]

Swissa, 

You can expect somewhere on the order of a 20-21% increase in frame rate impacted by scenery sliders, all else being kept the same, and provided, and this is critical, your GPU is not approaching 100% utilization, if you moved up to 9700K at 5Ghz or so, +/- a Ghz or two.   I've accepted the fact my now 6.5 y/o system can't handle too much metropolitan complexity so I'm careful to set up scenarios to live w/in my processing budget.   I have Scenery Complexity at Extreme, Autogen at Dense, and Vegetation at High, I think they are called.  In yesterday's flight of PMDG 777 out of KDEN HD, in FTX Central Rockies scenery, I was taxiiing to rwy 35 with a liquid smooth frame rate of 30.  Shortly after take off and for a few minutes until I got out of Denver area and up over Front Range OTW to Kona HI I saw a frame rate of around 25, w/ resultant loss of liquid smooth video performance.  Were I to be using a 9700K or 9900K at 5Ghz +/- this would have given me roughly a 30% increase in CPU power over my 3930K@4.4Ghz, therefore I would have seen 25fps * 1.3 = 32.5fps or so, in which case I would have maintained perfect stutter free performance at 30fps.  GPU utilization in blizzard conditions was only 70% or so.   For that entire 7h flight the ONLY time I saw anything short of perfect fluid video was shortly after TO from Denver for just a few minutes.  Not a giant price to pay for being able to continue using a 6.5 y/o system.   I'm careful to fly in and out of specific areas w/ a plane that matches the overhead of the scenery I will be in.

The secret to perfect video performance for me is to adhere to the following:

1.  Frame rate UNLOCKED in-sim

2. VSYNC to 30mHz vertical refresh set up via nVidia driver support--my Dell 3440x1440 screen will run at 30mHz refresh

3.  Monitor GPU utilization to insure I don't approach 100% utilization (so often I have terrain/cloud shadows dialed back or even off, situation dependent).  I do use bloom and cloud reflections and those my old GPU can easily handle.  

4.  HT enabled w/ the correct affinity mask absolutely runs best compared to HT off

If you are getting down to 24fps in the worst conditions, I think you will see 24 * 1.21 = ~29fps, and if you were setup as I am, would not have smooth video performance, because maintaining 30 is required w/ VSYNC to 30mHz refresh. It's very hard to know for certain because as you know everything you're running will come into play, so it's possible you may need to dial back autogen slightly as required in the worst of scenarios.   This is the reason I don't get too concerned w/ getting to the bleeding edge of performance because it's still going to be possible to overwhelm your system and you will ultimately have to live w/in your processing budget despite the bleeding edge hardware.

Edited December 30, 2019 by Noel

[Quasimodo 505]

I have seen indications in my latest optimization testing that 8 cores might be better than 6. I've seen situation where Core 0 and the GPU are both roughly in the high 70's low 80's but core's 2,3,4 & 5 are all intermittently at 100% during climb out over cityscape. I'd expect that has something to do with scenery loading. First time I've ever seen anything other the the GPU or Core0 being the bottle neck!

[Noel 4,501]

On my old 3930K there is a very clear improvement in using 12 LPs versus 6 cores provided I use the right AM.   I'm going to a 9900K for just that reason, just in case, and also in case MSFS 2 can use more LPs.

[JcSk8 0]

Already being at 4.7 Ghz I think the difference would be minimum.

I wouldn´t do it .Wait for Vulkan, see what happens ,and than maybe. Paired with your 1080ti your setup is as good as possible for now.