7800X3D, 7900X3D, 7950X3D..

[trevoC]

8 hours ago, Mr_sukebe said:

TrevoC>

Thanks for that.  Sorry, I didn't mean the fastest clocked, rather the faster processing.  I'm assuming that you're saying the CCD with the extra Cache can run DCS faster.

Does make me wonder whether DCS with MT would be faster with 2 CCDs (16 cores) all with the same lower cache (7950x), or with a single CCD (8 cores) with the additional cache (7950x3d).  Is that a question that's already been answered?

oh ok... "faster processing" is dependent on whether or not the software can take advantage of the vCache which not all software can so it can be confusing.

right now (I've tested all scenarios... both CCDs independently, both with parking enabled and both with parking disabled... all cores available).

At the end of the day, just leave the PC alone to do what it does. That is the fastest scenario which is nice because it requires no messing about.

DCS with MT is not faster when disabling parking and allowing all 16 cores to be available according to all my tests on an x3D and all other things being equal would not be faster on a non x3D chip like the 7950x as I ran the test on only CCD2 (without vCache) which would emulate a 7950x (half the cores) but DCS doesn't use the extra cores anyway as it is not a true multi-threaded application.

[Greekbull]

10 hours ago, Mr_sukebe said:

TrevoC>

Thanks for that.  Sorry, I didn't mean the fastest clocked, rather the faster processing.  I'm assuming that you're saying the CCD with the extra Cache can run DCS faster.

Does make me wonder whether DCS with MT would be faster with 2 CCDs (16 cores) all with the same lower cache (7950x), or with a single CCD (8 cores) with the additional cache (7950x3d).  Is that a question that's already been answered?

It is absolutely faster on the CCD with the V Cache...DCS screams with the V Cache CCD.

[Mr_sukebe]

So both of you believe that DCS at present runs faster on 8 (higher performance) cores, than 16.

Definitely worth knowing

[Mr_sukebe]

Does make me wonder about the viability of getting windows to:

- run the DCS client on the primary 8 cores

- run the DCS server once it’s MT (it’s currently single threaded) on secondary cores

- run peripheral tools (eg Discord, voice attack, Simshaker) on secondary cores

[skypickle]

@trevoC

@some1

have you seen any comparisons of the 5800x3d vs 7950x3d for DCS?

[Fakum]

I am a little confused in what I have read so far, perhaps more time is needed for a better conclusion due to the Multi Threading? I am getting parts together for a new build and I am wondering if I should actually consider going AMD instead of Intel i9-13900K? I am in need of really trying to knock down the cost anyway as I am just over budget with my current parts list.

[brain_delay]

Anyone not considering an X3D processor is throwing away performance. I haven't seen someone with a 7K00X3D with a 7900XTX to compare to my 5800x3d but in general the 7950x3d(7800x3d) is 10-25% faster depending. 

 

Spoiler

[hhhttt]

8 hours ago, Fakum said:

I am a little confused in what I have read so far, perhaps more time is needed for a better conclusion due to the Multi Threading? I am getting parts together for a new build and I am wondering if I should actually consider going AMD instead of Intel i9-13900K? I am in need of really trying to knock down the cost anyway as I am just over budget with my current parts list.

Based on my testing, in cpu limited mission, 7950x3d on 3d vcache core 20% faster than on non x3d core (aka 7950x)

it just harder to trigger cpu limited situation now due to multi threading

in VR my cpu time reduce from 8ms to 6ms (Apache Mariana )  but that is not needed because gpu frametime is 10ms in my setting.

But thinking in the next 5 years. Planes would became more complex but you can always drop graphic setting but you can’t make cpu faster  

 

 

[Fakum]

Thanks for the response, Im not really sure what you are comparing here "7950x3d on 3d vcache core 20% faster than on non x3d core (aka 7950x)". I know nothing right now about AMD processors, but will apply myself if I discover that AMD is the best route to take on a new build for DCS. Thats my 1st step, is it more sensible performance wise to switch?

[Th3ChosenOn3]

18 hours ago, brain_delay said:

I haven't seen someone with a 7K00X3D with a 7900XTX to compare to my 5800x3d

  Reveal hidden contents

My results earlier in the thread are with a 7950X3D and 7900 XTX, if you want to do a comparison.

[Hoirtel]

51 minutes ago, Fakum said:

Thanks for the response, Im not really sure what you are comparing here "7950x3d on 3d vcache core 20% faster than on non x3d core (aka 7950x)". I know nothing right now about AMD processors, but will apply myself if I discover that AMD is the best route to take on a new build for DCS. Thats my 1st step, is it more sensible performance wise to switch?

As I am in the same situation I think it's sounding like you can go either way. MT may have made things more difficult to know but I think the result is the same. Seems the 12th and 13th gen intels are having some trouble ensuring it's only p cores than run MT DCS and 79XXX3D are having trouble ensuring it's only the vache cores than run DCS. I am now hedging my bets that the 7800X3D will be the one to get. Good all round and low power too. However I don't think the win will be massive. I have waited so.long for X3D, part of me wishes I had just gone 13th gen. As I now own an X670 board I just gotta be quick on April 6th.... May even end up with a 7700X at this rate. 

Edited March 23, 2023 by Hoirtel

[Fakum]

Whats happening April 6th?

 

My current parts list is:

-Intel Core i9-13900K
-ASUS ROG Maximus Z790 Hero
-MSI Gaming GeForce RTX 4090 SUPRIM Liquid X 24G
-G.Skill Trident Z5 RGB Series (Intel XMP) 32GB (2 x 16GB) 288-Pin SDRAM DDR5 6000 CL36-36-36-96 1.35V
-SAMSUNG 990 PRO SSD 2TB PCIe 4.0 M.2 Internal Solid State Drive
-MSI MPG A1000G PCIE 5 & ATX 3.0 Gaming Power Supply
-Tower: Lian Li PC-O11DW 011

[BitMaster]

vor 11 Minuten schrieb Fakum:

Whats happening April 6th?

 

 

AMD will release it's 7800X3D chip afaik.

[Hoirtel]

13 minutes ago, Fakum said:

Whats happening April 6th?

 

My current parts list is:

-Intel Core i9-13900K
-ASUS ROG Maximus Z790 Hero
-MSI Gaming GeForce RTX 4090 SUPRIM Liquid X 24G
-G.Skill Trident Z5 RGB Series (Intel XMP) 32GB (2 x 16GB) 288-Pin SDRAM DDR5 6000 CL36-36-36-96 1.35V
-SAMSUNG 990 PRO SSD 2TB PCIe 4.0 M.2 Internal Solid State Drive
-MSI MPG A1000G PCIE 5 & ATX 3.0 Gaming Power Supply
-Tower: Lian Li PC-O11DW 011

Sorry, bad assumption that everyone would know 7800X3D release. It's etched in my mind... 

[BitMaster]

That's why we are a bunch of guys and not only one guy that has to know it all. The strength is in the "together"

[Fakum]

LOL,, I can imagine the anxiety by most. I only considered AMD 2 days ago when someone suggested it to me. Outside of that, I have very little knowledge of AMD CPUs, But i am trying to get an understanding. Thanks

[Fakum]

Man, I already dont get their numbering scheme! 7800X3D being released soon? Yet they have already surpassed that number. I dont get the hype. 

 

 

[Baldrick33]

27 minutes ago, Fakum said:

Man, I already dont get their numbering scheme! 7800X3D being released soon? Yet they have already surpassed that number. I dont get the hype.

 

Someone will explain it better but there is a belief the 7800X3D maybe the best option for VR given how the cores access the additional cache or something like that. Failing that it may well provide similar performance for VR gaming as its big brothers for a lower cost.

[Hoirtel]

6 minutes ago, Baldrick33 said:

Someone will explain it better but there is a belief the 7800X3D maybe the best option for VR given how the cores access the additional cache or something like that. Failing that it may well provide similar performance for VR gaming as its big brothers for a lower cost.

Yeah sort of although generally speaking the single CCD ryzens have the advantage of lower latency but also have less cores. Single CCD means the scheduler is not really important either. 

[brain_delay]

12 hours ago, Th3ChosenOn3 said:

My results earlier in the thread are with a 7950X3D and 7900 XTX, if you want to do a comparison.

I followed your 1080p/High Preset/Multithread/Mirrors/Tacview test. At 161.5 avg that's a 32% increase for the 7950X3D wew. I'll assume the 1%/.1% lows being better are due to me running the latest patch, or scheduling problems. Did you do this test using just the vcache core? Sorry if I missed it this thread has been going a while. 

 

null

Edited March 24, 2023 by brain_delay

[Th3ChosenOn3]

17 hours ago, brain_delay said:

I followed your 1080p/High Preset/Multithread/Mirrors/Tacview test. At 161.5 avg that's a 32% increase for the 7950X3D wew. I'll assume the 1%/.1% lows being better are due to me running the latest patch, or scheduling problems. Did you do this test using just the vcache core? Sorry if I missed it this thread has been going a while. 

 

No worries, I do all of the tests with all the cores enabled, and let game bar handle the scheduling. It would be a bit of a waste to turn off CCD1 on the 7950X3D. You're not showing the 1% and 0.1% lows in the statistics, those are the 1 and 0.1 percentiles in your screenshot.

Here is a picture of all of the stats on latest MT patch using 1080p High Preset with Tacview and Mirrors enabled:

 

Same thing, but with CCD1 disabled (7800X3D Simulation):

CapFrameX-DCS-7800X3D-Simulation.json CapFrameX-DCS-7950X3D.json

Edited March 24, 2023 by Th3ChosenOn3

[Anubis_94]

20 hours ago, Th3ChosenOn3 said:

No worries, I do all of the tests with all the cores enabled, and let game bar handle the scheduling. It would be a bit of a waste to turn off CCD1 on the 7950X3D. You're not showing the 1% and 0.1% lows in the statistics, those are the 1 and 0.1 percentiles in your screenshot.

Here is a picture of all of the stats on latest MT patch using 1080p High Preset with Tacview and Mirrors enabled:

 

Same thing, but with CCD1 disabled (7800X3D Simulation):

CapFrameX-DCS-7800X3D-Simulation.json 3.25 MB · 0 downloads CapFrameX-DCS-7950X3D.json 3.38 MB · 0 downloads

 

I really like your feedbacks .

Is it possible to complete these two benchmarks with a run on the CCD1 only (CCD0 disabled) ?

In order to have the perf on natural 7950, "vcache" 7950 and "freq" 7950.

Edited March 25, 2023 by Anubis_94

[trevoC]

15 hours ago, Anubis_94 said:

I really like your feedbacks .

Is it possible to complete these two benchmarks with a run on the CCD1 only (CCD0 disabled) ?

In order to have the perf on natural 7950, "vcache" 7950 and "freq" 7950.

 

Its faster by a smidge on vCache. Ran above test without gains. Thought maybe the higher frequency cores might do better than the vCache for DCS. I was wrong. In raw performance the vCache was at least 2.5% faster or more depending on the scenario (never less than 2.5% more) which is actually more than 2.5% if you consider the vCache cores are slower to start so they more than make up for their slower clock speeds.

[Th3ChosenOn3]

20 hours ago, Anubis_94 said:

I really like your feedbacks .

Is it possible to complete these two benchmarks with a run on the CCD1 only (CCD0 disabled) ?

In order to have the perf on natural 7950, "vcache" 7950 and "freq" 7950.

 

Thank you. Here is the same benchmark with CCD1 only (V-Cache CCD disabled):

 

The fully enabled 7950X3D with it's V-Cache cores has a 45% higher average and 15% better lows (1%, 0.1%). Here's a chart with all three of them for comparison:

CapFrameX-DCS-7950X3D-CCD0-Disabled.json

Edited March 26, 2023 by Th3ChosenOn3

[Anubis_94]

Thank you