I've read through this thread and didn't see any technical discussion about the source of the results.  Does anyone have any thoughts about the difference between the 7900X and the 9900K?  Frequency scaling (5.0 vs 5.3 GHz) would suggest about a 6% improvement at best.  

22 minutes ago, silentsage said:

I've read through this thread and didn't see any technical discussion about the source of the results.  Does anyone have any thoughts about the difference between the 7900X and the 9900K?  Frequency scaling (5.0 vs 5.3 GHz) would suggest about a 6% improvement at best.  

Hi Silent -

I think Rob's post is more about P3D 4.3 vs 4.4 than the 7900x vs the 9900K.

2 hours ago, silentsage said:

Frequency scaling (5.0 vs 5.3 GHz) would suggest about a 6% improvement at best.

It's a bit more complicated than that ... I couldn't get my 7900X to run on 5Ghz  HT OFF without generating a lot of heat, it was stable, but I would frequently see 105C peaks, as a result I ran 4 cores at 5 Ghz and staggered the rest around 4.8 to 4.9 and one at 4.7.  In addition I could NOT bump up the cache frequency at all in the 7900X without suffering instability ... and I think the default cache frequency was pretty low in the 7900X (was it 2700/3200 or something like that) ... even with AVX offset of 4 or 5.  Cache frequency helps "significantly" in P3D.  The i9 9900K also has "some" hardware level protection for Meltdown and Spectre so no performance hit from the patched 7900X.

Cheers, Rob.

Rob - 

That makes sense.  Do you know what the cache frequency was for the 9900K?

By the way, thanks for all your benchmarking and performance optimization efforts.  I always look for your postings.  

The default cache frequency for 9900K is 4.3Ghz, I run it at 4.8Ghz with CPU at 5.3Ghz all cores. AVX 5.1Ghz

1:1 would be ideal, but I’ve never seen a CPU that can do that unless you reduce CPU frequency.

Cheers Rob

Edited December 22, 20187 yr by Guest

6 hours ago, Rob Ainscough said:

I couldn't get my 7900X to run on 5Ghz  HT OFF without generating a lot of heat, it was stable

Mine isn't at 5.0, even just on one core (with the rest on a mix of 4.8 and 4.6 as you suggested to me once). All 10 cores at 4.8 is okay -- been using it like that for over a year.

6 hours ago, Rob Ainscough said:

Cache frequency helps "significantly" in P3D. 

Looking forward to that! (My 9900K processor arrived today BTW).

Pete

 

Edited December 22, 20187 yr by Pete Dowson

1 hour ago, Pete Dowson said:

Looking forward to that!

So did it arrive in a blue plastic cheesy hexagon in a black bag ... or did you get a small nameless white box with some foam in it?  Did you pay 10,000 pounds sterling for "the one" that can do 5.5 Ghz ? 😉  Tuning will take the most time once setup ... but I have a pretty good base to work up from.

Cheers, Rob.

7 hours ago, Pete Dowson said:

Mine isn't at 5.0, even just on one core (with the rest on a mix of 4.8 and 4.6 as you suggested to me once). All 10 cores at 4.8 is okay -- been using it like that for over a year.

Looking forward to that! (My 9900K processor arrived today BTW).

Pete

 

Still using my 7940x here : 8x 4.8 Ghz for P3D and 6x 4.0 Ghz for the addon apps.

Temps stay below 65 degrees Celcius with my EKWB Xtreme set..

Edited December 23, 20187 yr by GSalden

Something is not right here.

Given P3Dv4x only uses 5 or 6 threads, there is no way more cores, or v4.4 has resulted in anything like a 60% boost.

Smoothness..... maybe, if core load is better distributed.

Everyone else reports little change .... if anything, a decrease, due to some hidden improvements in quality.

For cripes sake, don't go out spending stoopid money on 50 core cpus. Only the clockspeed is relevant beyond 6 physical cores.

"A fool and his money, are.........." ?

You finish the sentence..... and wait for at least 6 Ghz processors before upgrading. Even that will only increase FPS by 25%.

Cripes, switching off (the pug-ugly) autogen, gives me a 50 % boost over FTX region. And that's free !

 

 

Edited December 23, 20187 yr by Gabe777

10 hours ago, Rob Ainscough said:

So did it arrive in a blue plastic cheesy hexagon in a black bag

Yes. It's actually a regular dodecahedron (12 sided, each side being a pentagon. Daft!

11 hours ago, Rob Ainscough said:

or did you get a small nameless white box with some foam in it?

That'd be the OEM one. Not much cheaper. And the retail one was available on Amazon Prime with free next day delivery.

11 hours ago, Rob Ainscough said:

Did you pay 10,000 pounds sterling for "the one" that can do 5.5 Ghz ? 😉  Tuning will take the most time once setup ... but I have a pretty good base to work up from.

Well, for nearly twice the price I could have got a "pre-binned" version from Overclockers, rated at 5.1 I think. But you said the pre-binning isn't worth paying for.

1 hour ago, Gabe777 said:

if core load is better distributed.

P3D 4.4 definitely seems to use more cores quite well -- two or three others (after the high core 0 usage).

2 hours ago, Gabe777 said:

For cripes sake, don't go out spending stoopid money on 50 core cpus. Only the clockspeed is relevant beyond 6 physical cores.

The 10 cores on my 7900X are used quite well. I'm not only running P3D on that PC, but also Prosim, ActiveSky+ASCA, UTLive, SkyForce3D, and a couple of drivers for my cockpit hardware. I've arranged them on the cores not used much by P3D. They don't noticeably affect P3D performance or smoothness so I'm reasonably pleased with the 7900X. I just want more, so I can increase some of the settings, and maybe even use the Orbx country sceneries I've bought!

Pete

 

Extra cores are certainly better for 3rd party apps in the background.

I am surprised that 4.4 utilises more cores thsn 4.3. I am sure they would have mentioned this in their release blurb.... surely ? I have seen large increases in CPU useage with lots of cores, but it is only "apparent", due to polling intervals used by the FPS monitoring software. This was demonstrated when the 8700 came out, numerous times, and no actual increase in performance was noticeable. Moreover, it could be argued thst retricting P3D to 5 or 6 threads, actually benefits it, a little like putting blinkers on horses !

Anyway, I simply do not believe the OP's purported 60 percent increase in pure FPS, in a like-for-like scenario..... something else is clearly going on there.

 

Edited December 23, 20187 yr by Gabe777

Well ...I did notify the OP that this thread was not for the average joe blow.  some thoughts:

1- I'm still not clear what the subject matter is

2- I made a similar a claim when I went from an I5 4670k running 4.7 / nb 4.5 on 2400 cl 10 mems to the 8600k at 5.0 / nb 4600 on 4133 cl17 mems ......and pretty much got the "you're on crack cocaine" reaction

3) I can show a 10% fps boost just by switching mem speed from 3200 cl 14 to 4133 cl 17

4) looking at the 7900x config the OP was reportedly running makes me think my old Haswell was faster

5) Westman reported similar results to the claims made by the OP in this thread

6) Westman has reported that  "he prefers" the 7900x over his highly tuned and clocked 8700K ...though his posts over the past year seem to flip-flop between the 7900 and the 8700

 

 

 ....I'll be back with more thoughts <g>