Hey everyone. Title says it all. Currently have the following system:

4770k @ 4.4 GHz (can’t go higher OC than this. Believe me, I’ve tried. Lost silicon lottery lol)

Asus Maximus Hero VI

EVGA 1080 Ti SC2

Corsair Vengeance Pro 32 GB DDR3 1866 MHz 9-10-9-27

Noctua NH-U14s

TX850W

Finally got my hands on a 10900k, but having second thoughts about building a rig around the Z490 platform, especially with the lack of PCIe 4.0 support, thereby jeopardizing the potential longevity of the system. With Zen 3 due later this year and the purported 10-15% increase in IPC and 5-10% clock speed increases, AMD will likely match or slightly exceed the single-threaded performance of the 10900k and have PCIe 4.0 support along with (perhaps) official support for faster DDR4 (maybe 3300-3600 MHz). 

Wait for 4700x/4800x/4900x or go with the 10900k? I’ll be getting (hopefully) the RTX 3090. Primary use of my new rig will be FS2020 in VR with HP Reverb G2. Also XP11 in VR. Thanks for any replies!

-Elliot

Edited September 5, 20205 yr by Eman

There will always be something "better" around the corner.  And if you wait, when the time comes, there will then be something even better still around the next corner.  etcetera, etcetera, etcetera...

FWIW, the 10900K really is an exceptional CPU.  Unless you're planning something like a serious triple-SLI setup, PCIe 4.0 is not really a factor, as no single GPU is coming anywhere close to saturating 16 lanes on PCIe 3. 

Thanks Bob. Been reading your posts for a long time (been around here since 2004), and I appreciate your insights. Glad you're happy with your 10900k build. Yeah, my new 10900k is staring at me right now in its pretty packaging, just waiting to be installed in a new mobo. Looks like I'm gonna be an Intel guy for the foreseeable future 😎 Thanks again, man.

-Elliot

IRT PCI 4.0 you might be interested in this: https://www.tomshardware.com/news/phisons-new-e18-ssd-controller-rivals-samsung-980-pro-with-7-gbps-of-throughput

On 9/6/2020 at 8:11 AM, pgde said:

IRT PCI 4.0 you might be interested in this: https://www.tomshardware.com/news/phisons-new-e18-ssd-controller-rivals-samsung-980-pro-with-7-gbps-of-throughput

SSD throughput really isn't a factor in sim performance though...there's really no discernable difference in sim performance right now between an nVME and a SATA III SSD, despite the nVME device being more than an order of magnitude faster than the SATA device.  We're limited by how fast the CPU and GPU can fuse the layers and compose the frames.  Lookahead buffering makes access time for scenery data a noncritical factor for real-time sim performance.  So if you do something that involves frequent transfers of really *massive* files, it might prove worth having.  For simulation...not so much.  The only reason I spend the extra $$ for nVME drives right now is the elegance of their physical simplicity--no mounts, no cables, just a small board not much bigger than a stick of chewing gum mounted under a plate on the mobo.

 

1 hour ago, w6kd said:

We're limited by how fast the CPU and GPU can fuse the layers and compose the frames.

Bob, On your system is this a limitation of the CPU, GPU, motherboard, or a combination of all of the above? I'm just curious where the bottleneck is on the latest Gen systems.

Ted

55 minutes ago, Ted Striker said:

Bob, On your system is this a limitation of the CPU, GPU, motherboard, or a combination of all of the above? I'm just curious where the bottleneck is on the latest Gen systems.

During startup, I think it's mostly a limitation in the software itself.  I don't see a lot of parallelism going on when the sim first loads--it's typically just one or two processors showing a load until late in the startup when the scenery loads.  While running, the lookahead buffering is bringing the next scenery tiles into RAM far in advance of actually needing them, making the speed at which that happens (storage throughput) noncritical, at least so long as it happens fast enough that the CPU isn't starved of data.  The data is generally in memory when it's needed, even on systems with HDD storage.  The question of where the bottleneck(s) occur is a complex one that depends on the various hardware components and software elements and settings at that moment of time.

I can force GPU-limited operation if I slam the AA up too high e.g. trying to run 8xSSAA at 4K with 5+ layers of heavy weather and at night with dynamic lighting.  When I see GPU utlization near 100% and stutters appearing, that's tell-tale GPU-limited operation.  And almost always the difference between GPU hitting the wall or not is internal to the GPU, specifically AA settings.  When I back down the AA, the GPU load drops and things smooth out, and tells me that I am not limited by PCIe bus bandwidth--the GPU is getting the data from the CPU in time, but it can't post-process the frames internally fast enough to keep up.  Similarly, if I jam the LOD range way up with high autogen density and 100% AI traffic in the LA basin or NYC metro area, I'll see one or more CPU cores at or near 100% when the stutterfest begins--there I know I'm seeing CPU-limited operation.  It just depends on the circumstances and the settings...my goal is to have a good operating combo of high-performance hardware and software settings that allow me to power through the peak loads without either/both of the CPU or GPU hitting the wall, causing frame rate drops and long frames.  I'm having pretty good success doing just that.

 

Hi Bob.  A couple questions for you...

1 - Are you overclocking to 5.2 GHz on all cores, or is that just a single core boost?

2 - I assume you're running 4K at 30Hz for continuity of frames.  For MSFS in particular, how close would you say you're getting to being able to run consistently at 60Hz on that system?  I'm curious if the upcoming 3080s or <gulp> 3090s might be the hardware to finally break the 4K/ steady 60 fps barrier.

I'm really being tempted by the 10900K; but I also know that Zen 3 may be on the same street... not just around the corner.

Edited September 8, 20205 yr by MadDog

21 hours ago, w6kd said:

goal is to have a good operating combo of high-performance hardware and software settings that allow me to power through the peak loads without either/both of the CPU or GPU hitting the wall

Thanks for that reply Bob. It sounds like things have not changed much with the new hardware as I do the same with my old system, albeit at much lower settings overall than you can probably operate with.

I apologize for the late response. I got tied up yesterday getting the house and yard ready for the onslaught of sudden winter.

Ted

1 hour ago, MadDog said:

Hi Bob.  A couple questions for you...

1 - Are you overclocking to 5.2 GHz on all cores, or is that just a single core boost?

2 - I assume you're running 4K at 30Hz for continuity of frames.  For MSFS in particular, how close would you say you're getting to being able to run consistently at 60Hz on that system?  I'm curious if the upcoming 3080s or <gulp> 3090s might be the hardware to finally break the 4K/ steady 60 fps barrier.

1.  5.2 on all 10 cores, HT off
2.  I don't think I'm even remotely close to being able to consistently hold 60 fps...I'm happy now that I can punch my way through ORBX SoCal with very few long frames or frame rate drops while VSync locked to my 30Hz hardware refresh rate.

1 hour ago, Ted Striker said:

Thanks for that reply Bob. It sounds like things have not changed much with the new hardware as I do the same with my old system, albeit at much lower settings overall than you can probably operate with.

I apologize for the late response. I got tied up yesterday getting the house and yard ready for the onslaught of sudden winter.

Yeah, we're supposed to get as much as a foot of heavy wet snow tonight.  But then again, the NWS has predicted 12 of the last 3 snowstorms, so...

 

3 hours ago, w6kd said:

1.  5.2 on all 10 cores, HT off

May I ask why you turned off HT? I thought for MSFS 2020 its best to leave it on?

5 hours ago, captain420 said:

May I ask why you turned off HT? I thought for MSFS 2020 its best to leave it on?

There's a point of diminishing returns...with HT off and 10 cores running at 5.2, the per-core loading is low enough across all the cores that adding additional headroom with HT doesn't produce a noticeable improvement.  On a quad or hex core CPU, especially at stock speeds, per-core loading is high and running MSFS with HT on seems to be the way to go. 

Also, after a brief hiatus to experiment with the new MSFS, I am back to running P3D as my primary simulator platform, so my system optimizations remain focused on P3D.

FYI, I finally have all the parts to put my 10900K system together next weekend. Sure I could have waited to see what is around the corner but then I would always be waiting because something new always comes out. 

Build, baby, build!!  😁