7800X3D, 7900X3D, 7950X3D..

[trevoC]

4 hours ago, Th3ChosenOn3 said:

You claimed v-cache doesn't give anyone any gains, and now you're saying you don't know if v-cache has a performance improvement or not. It might be beneficial to figure that out first before making such claims, as what you're doing doesn't help anyone.

 

You shouldn't be testing for a CPU bottleneck in a very GPU bound situation like you're doing on Marianas. I'm also hoping you're using an application to capture data to analyze instead of just looking at a hardware monitor and Nvidia's performance overlay.

 

Its called an honest conversation, and I'm sharing my data as I go. I said "looks like", not it doesn't or I conclude. Its why I'm going back to basics.

I do understand that you are looking to nitpick my comments and I find it very strange that you are saying anything about "claims" after yours which you then admitted was false.

I stand by everything I say, and If you don't like the process of learning what we have or the information I'm sharing then you can move on.

Also... please don't put words in my mouth

"you claimed v-cache doesn't give anyone any gains," < I never said anything close to that.

 

[trevoC]

4 hours ago, Th3ChosenOn3 said:

You shouldn't be testing for a CPU bottleneck in a very GPU bound situation like you're doing on Marianas. I'm also hoping you're using an application to capture data to analyze instead of just looking at a hardware monitor and Nvidia's performance overlay.

 

I'm not testing a CPU bottleneck on the Marianas, who said I was. Its why I'm running two (actually three) track files. I think the biggest thing that isn't helping is criticizing things that didn't happen. Everything you have commented on is something I didn't say or is answered in the post. I didn't make claims that vCache gives you 75% gains that's for sure.

[BitMaster]

This is getting toxic and btw I never claimed 70ish gains, all I said is comparing things is not always easy.

Anyway, too much aggression for my taste

[trevoC]

Maybe I misunderstood, but it seems like that is exactly what was said. All I said was I find that hard to believe.  Which it is.

 

 

On 3/13/2023 at 7:07 AM, Th3ChosenOn3 said:

I've done a run on the Normal CCD of the 7950X3D, and got an average of 120, after I fixed my game bar issue and DCS was running on the 3D V-Cache CCD, my average jumped to 170 pre framecap update, and over 200 after the frame cap update.

You should see substantial gains if you're on the 3D V-cache CCD, unless you're being bottlenecked somewhere else.

 

[trevoC]

1 hour ago, BitMaster said:

This is getting toxic and btw I never claimed 70ish gains, all I said is comparing things is not always easy.

Anyway, too much aggression for my taste

I think this is a misunderstanding of a misunderstanding.... I never said you said that, but this is going nowhere which is why I said lets move on.

[trevoC]

that's too bad, because there was some nice contributions in this thread that has now been hijacked by people trying to have a gotcha moment and no real contributions. Meanwhile, the people actually spending hours trying to help others have good setups can't have an honest conversation.

Too many threads like this end, and then everyone goes their separate way and the rest of the testing never gets shared. I sure as hell don't need to spend my time sharing my outcomes or memory timings or anything else. A day ago I was sharing findings, and now I have to reply to comments like "you claimed" why are you doing that.

Good luck with your setup guys, see you in the skies.

 

[Azhek Ahriman]

Hello everyone. I need your help with core parking using Ryzen 9 7950X3D with dcs. Basically the damn thing won't park non V-Cache cores with DCS. It worked for some time but right now that damn game bar just refuse to remember dcs as a game. I can tick the "Remember this is a game" option as much as I want, but next time I open the game bar this option is unselected.  This doesn't happen with any other game - I don't even need to open the game bar. But with DCS as it is a niche title it just won't work, I have no idea what happened since it did work earlier. 

I have used all the tricks AMD has provided but it just won't work.

Please help

[chev255]

8 hours ago, trevoC said:

that's too bad, because there was some nice contributions in this thread that has now been hijacked by people trying to have a gotcha moment and no real contributions. Meanwhile, the people actually spending hours trying to help others have good setups can't have an honest conversation.

Too many threads like this end, and then everyone goes their separate way and the rest of the testing never gets shared. I sure as hell don't need to spend my time sharing my outcomes or memory timings or anything else. A day ago I was sharing findings, and now I have to reply to comments like "you claimed" why are you doing that.

Good luck with your setup guys, see you in the skies.

 

trevoC, I would like to encourage you to continue sharing as there are many of us who are looking for info on how 7950x3d performs. Unfortunately we sometimes become a little too passionate 

Myself I have a 9700K and a 2080Ti from 2018 and I am trying to figure out whether to go 13900K or 7950X. Since MT just halved my CPU frametimes I just pulled the trigger on a 4090 that I can use on my current build until I figure the CPU out. Last time I bought an AMD was 2005 and I wouldn't mind doing it again but all the discussions about core parking kind of worry me.

Cheers

[Kreutzberg]

18 hours ago, Azhek Ahriman said:

Hello everyone. I need your help with core parking using Ryzen 9 7950X3D with dcs. Basically the damn thing won't park non V-Cache cores with DCS. It worked for some time but right now that damn game bar just refuse to remember dcs as a game. I can tick the "Remember this is a game" option as much as I want, but next time I open the game bar this option is unselected.  This doesn't happen with any other game - I don't even need to open the game bar. But with DCS as it is a niche title it just won't work, I have no idea what happened since it did work earlier. 

I have used all the tricks AMD has provided but it just won't work.

Please help

I had the same issue. I set my bios to prefer cache, which gets the right cores working although the other ones are still not parked.

In my ASUS motherboard, the setting is under Advanced\AMD CBS\SMU Common options

[okletsgo]

For any of you optimising new systems and new windows installs, this might be of interest. Up to 15% performance gain in MSFS. Who knows how DCS reacts, worth a shot. 

https://www.tomshardware.com/news/windows-vbs-harms-performance-rtx-4090

I would try it, but I am dealing with a small leak issue on my 7950x3D/4090 custom loop rig (no damage, thankfully).

 

Edited March 15, 2023 by okletsgo

[trevoC]

On 3/14/2023 at 7:03 PM, chev255 said:

trevoC, I would like to encourage you to continue sharing as there are many of us who are looking for info on how 7950x3d performs. Unfortunately we sometimes become a little too passionate 

Myself I have a 9700K and a 2080Ti from 2018 and I am trying to figure out whether to go 13900K or 7950X. Since MT just halved my CPU frametimes I just pulled the trigger on a 4090 that I can use on my current build until I figure the CPU out. Last time I bought an AMD was 2005 and I wouldn't mind doing it again but all the discussions about core parking kind of worry me.

Cheers

Thanks chev, I'll share my latest insights... another dozen or so tests.

These tests I wanted to see how much of a difference vCache made in my setup and if parking the cores at all made a difference or just prevent all cores from parking ever.

For the vCache difference, I used DCS ST as there is some weird behaviour with parking cores on DCS MT that I haven't resolved yet but some of the tests are starting to paint a broad pictures at the moment.

First, the difference between DCS ST running on CCD0 or 1 is negligable (less than 1.5% in favor of vCache). That being said, it always seemed to be in favor of vCache.

Also remember that the vCache cores run at lower speeds, so considering they always performed a bit better (margin of error a bit), I conclude that they are actually doing something. If they were not doing anything I would expect the same reduction in performance than the chips are reduced in speed which isn't the case. vCache although not some magic buillet in DCS unfortunately (as they are in some games) doesn't hinder the performance at all by running on the slower chips. vCache seems to make up for the difference and then some... a little some, but some. I consistantly got 1-1.5% more performance on the vCache cores.

I then wanted to test in MT DCS if not parking any cores at all would be better than allowing cores to be parked. This would allow the entire chip to be used by DCS as it saw fit.

What I found confirmed my above results also when compared to a prior set of data that was just MT runs.

When the core parking was completely disabled via registry, it would just act like a 12 or 16 core chip (depending on your model). Of course the first 1/2 run slower so this would absolutely come into play.

What i found was DCS got assigned to the faster cores by default which gave me about 1% less performance (also sort of confirming the above results).

I didn't find it running any faster on plazma torture map than MT with parked cores enabled (even though they park very weirdly). I'm guessing even though DCS can use multiple cores, its not using more than 6 cores at the moment enough to disable core parking and have more cores available. This was my guess in an earlier thread months ago as I doubted that DCS would be a true multi-threaded application.

I've concluded from these and prior rounds of tests the following so far. Obviously this can be different depending on your situation.

-MT DCS is worth running as I saw gains of 4-5% in small scenarios like instant action and 10-40% (around 25% avg.) on plazma torture map. This is what ED said it would do so Its a result that matches the expectation. I have no further desire to test this when receiving the expected result.

-vCache is doing very little in DCS. It does do something, so if you have an X3D chip I wouldn't go to lengths to disable parking cores, but from my tests I wouldn't be running out to get one either. I'm betting there would be very small differences between the 7900x and 7900x3D in DCS. Obviously if you are playing other games this might be a factor worth considering as the x3D chips are wildly different across games and much more successful with others.

-Disabling the parking of your cores does not yield positive results, so I wouldn't bother.

-Core parking works perfectly on ST DCS but seems to work a little strange in MT. That being said, the numbers seem to state that the MT version of DCS is being run on the vCache cores even though windows is reporting otherwise sometimes. Even though the MT core parking looks a hot mess, the numbers confirm that they are consistant with being run on vCache properly and Multi-threading itself is a benefit.

In the end, I'm running the MT version of DCS and ingnoring the mess that is being reported by windows, logging tools, DCS etc... It is garnering better results.

-------

I still want to test a number of items including GPU scheduling settings and other windows optimization settings but I wanted to do this with a baseline that would be more consistant with what I'd be using going forward. That is now DCS MT so stay tuned and I'll post my results.

I did record all this, but its hours and hours of boring footage that I'm not interested in editing. If you are interested in seeing some of it I could probably just upload it raw but you wouldn't have access to the logs in the background that I used, just the onscreen data.

--------

Hope this all helps someone, feel free to ask me any questions. If i know I'll answer.

[Hoirtel]

6 hours ago, trevoC said:

Thanks chev, I'll share my latest insights... another dozen or so tests.

These tests I wanted to see how much of a difference vCache made in my setup and if parking the cores at all made a difference or just prevent all cores from parking ever.

For the vCache difference, I used DCS ST as there is some weird behaviour with parking cores on DCS MT that I haven't resolved yet but some of the tests are starting to paint a broad pictures at the moment.

First, the difference between DCS ST running on CCD0 or 1 is negligable (less than 1.5% in favor of vCache). That being said, it always seemed to be in favor of vCache.

Also remember that the vCache cores run at lower speeds, so considering they always performed a bit better (margin of error a bit), I conclude that they are actually doing something. If they were not doing anything I would expect the same reduction in performance than the chips are reduced in speed which isn't the case. vCache although not some magic buillet in DCS unfortunately (as they are in some games) doesn't hinder the performance at all by running on the slower chips. vCache seems to make up for the difference and then some... a little some, but some. I consistantly got 1-1.5% more performance on the vCache cores.

I then wanted to test in MT DCS if not parking any cores at all would be better than allowing cores to be parked. This would allow the entire chip to be used by DCS as it saw fit.

What I found confirmed my above results also when compared to a prior set of data that was just MT runs.

When the core parking was completely disabled via registry, it would just act like a 12 or 16 core chip (depending on your model). Of course the first 1/2 run slower so this would absolutely come into play.

What i found was DCS got assigned to the faster cores by default which gave me about 1% less performance (also sort of confirming the above results).

I didn't find it running any faster on plazma torture map than MT with parked cores enabled (even though they park very weirdly). I'm guessing even though DCS can use multiple cores, its not using more than 6 cores at the moment enough to disable core parking and have more cores available. This was my guess in an earlier thread months ago as I doubted that DCS would be a true multi-threaded application.

I've concluded from these and prior rounds of tests the following so far. Obviously this can be different depending on your situation.

-MT DCS is worth running as I saw gains of 4-5% in small scenarios like instant action and 10-40% (around 25% avg.) on plazma torture map. This is what ED said it would do so Its a result that matches the expectation. I have no further desire to test this when receiving the expected result.

-vCache is doing very little in DCS. It does do something, so if you have an X3D chip I wouldn't go to lengths to disable parking cores, but from my tests I wouldn't be running out to get one either. I'm betting there would be very small differences between the 7900x and 7900x3D in DCS. Obviously if you are playing other games this might be a factor worth considering as the x3D chips are wildly different across games and much more successful with others.

-Disabling the parking of your cores does not yield positive results, so I wouldn't bother.

-Core parking works perfectly on ST DCS but seems to work a little strange in MT. That being said, the numbers seem to state that the MT version of DCS is being run on the vCache cores even though windows is reporting otherwise sometimes. Even though the MT core parking looks a hot mess, the numbers confirm that they are consistant with being run on vCache properly and Multi-threading itself is a benefit.

In the end, I'm running the MT version of DCS and ingnoring the mess that is being reported by windows, logging tools, DCS etc... It is garnering better results.

-------

I still want to test a number of items including GPU scheduling settings and other windows optimization settings but I wanted to do this with a baseline that would be more consistant with what I'd be using going forward. That is now DCS MT so stay tuned and I'll post my results.

I did record all this, but its hours and hours of boring footage that I'm not interested in editing. If you are interested in seeing some of it I could probably just upload it raw but you wouldn't have access to the logs in the background that I used, just the onscreen data.

--------

Hope this all helps someone, feel free to ask me any questions. If i know I'll answer.

Thanks for sharing again. Glad you are still here. It's really making me question if I should pull the 13900k rip cord on my wait. I could just buy a board and chip and get going with my DDR4 ram. 

Do you have any information on frametimes? Or if you use cap frame X do you have the graphs from the runs. I wonder if there is more consistency that could be seen. Also I have seen on 5800x3d tests that FPS can be the same but CPU frametimes are lower still (than non 3xd parts). Unfortunately capframe X doesn't show separate frametimes but overall would be interesting.

[Kreutzberg]

 

17 hours ago, trevoC said:

Thanks chev, I'll share my latest insights... another dozen or so tests.

These tests I wanted to see how much of a difference vCache made in my setup and if parking the cores at all made a difference or just prevent all cores from parking ever.

For the vCache difference, I used DCS ST as there is some weird behaviour with parking cores on DCS MT that I haven't resolved yet but some of the tests are starting to paint a broad pictures at the moment.

First, the difference between DCS ST running on CCD0 or 1 is negligable (less than 1.5% in favor of vCache). That being said, it always seemed to be in favor of vCache.

Also remember that the vCache cores run at lower speeds, so considering they always performed a bit better (margin of error a bit), I conclude that they are actually doing something. If they were not doing anything I would expect the same reduction in performance than the chips are reduced in speed which isn't the case. vCache although not some magic buillet in DCS unfortunately (as they are in some games) doesn't hinder the performance at all by running on the slower chips. vCache seems to make up for the difference and then some... a little some, but some. I consistantly got 1-1.5% more performance on the vCache cores.

I then wanted to test in MT DCS if not parking any cores at all would be better than allowing cores to be parked. This would allow the entire chip to be used by DCS as it saw fit.

What I found confirmed my above results also when compared to a prior set of data that was just MT runs.

When the core parking was completely disabled via registry, it would just act like a 12 or 16 core chip (depending on your model). Of course the first 1/2 run slower so this would absolutely come into play.

What i found was DCS got assigned to the faster cores by default which gave me about 1% less performance (also sort of confirming the above results).

I didn't find it running any faster on plazma torture map than MT with parked cores enabled (even though they park very weirdly). I'm guessing even though DCS can use multiple cores, its not using more than 6 cores at the moment enough to disable core parking and have more cores available. This was my guess in an earlier thread months ago as I doubted that DCS would be a true multi-threaded application.

I've concluded from these and prior rounds of tests the following so far. Obviously this can be different depending on your situation.

-MT DCS is worth running as I saw gains of 4-5% in small scenarios like instant action and 10-40% (around 25% avg.) on plazma torture map. This is what ED said it would do so Its a result that matches the expectation. I have no further desire to test this when receiving the expected result.

-vCache is doing very little in DCS. It does do something, so if you have an X3D chip I wouldn't go to lengths to disable parking cores, but from my tests I wouldn't be running out to get one either. I'm betting there would be very small differences between the 7900x and 7900x3D in DCS. Obviously if you are playing other games this might be a factor worth considering as the x3D chips are wildly different across games and much more successful with others.

-Disabling the parking of your cores does not yield positive results, so I wouldn't bother.

-Core parking works perfectly on ST DCS but seems to work a little strange in MT. That being said, the numbers seem to state that the MT version of DCS is being run on the vCache cores even though windows is reporting otherwise sometimes. Even though the MT core parking looks a hot mess, the numbers confirm that they are consistant with being run on vCache properly and Multi-threading itself is a benefit.

In the end, I'm running the MT version of DCS and ingnoring the mess that is being reported by windows, logging tools, DCS etc... It is garnering better results.

-------

I still want to test a number of items including GPU scheduling settings and other windows optimization settings but I wanted to do this with a baseline that would be more consistant with what I'd be using going forward. That is now DCS MT so stay tuned and I'll post my results.

I did record all this, but its hours and hours of boring footage that I'm not interested in editing. If you are interested in seeing some of it I could probably just upload it raw but you wouldn't have access to the logs in the background that I used, just the onscreen data.

--------

Hope this all helps someone, feel free to ask me any questions. If i know I'll answer.

 

10 hours ago, Hoirtel said:

Thanks for sharing again. Glad you are still here. It's really making me question if I should pull the 13900k rip cord on my wait. I could just buy a board and chip and get going with my DDR4 ram. 

Do you have any information on frametimes? Or if you use cap frame X do you have the graphs from the runs. I wonder if there is more consistency that could be seen. Also I have seen on 5800x3d tests that FPS can be the same but CPU frametimes are lower still (than non 3xd parts). Unfortunately capframe X doesn't show separate frametimes but overall would be interesting.

My testing with a similar system (7950X3D and 4090) shows different results, with some improvements to frame times using the vCached CCD. 

I tested in Operation Cerberus North Mission 13A to make my results comparable to mjfrisby's data on page 6 of this thread: 

In that CPU bound scenario in VR (that's how I can measure frame times), I'm pretty consistently seeing about 3.5 ms better frame times with CCD0 only versus CCD1 only with Tacview on or off. That translates to something somewhat over 10 fps in the range of framerates we're looking at. However, DCS is currently pushing these types of performance numbers to 45 fps regardless. That might be causing some confusion if people are not seeing better frame rates. 

Even if locked at 45 fps or some other arbitrary number, better frame times give a buffer for more traffic or whatever. In any case, I do think the vCache is helping meaningfully in DCS.

[trevoC]

1 hour ago, Kreutzberg said:

 

 

My testing with a similar system (7950X3D and 4090) shows different results, with some improvements to frame times using the vCached CCD. 

I tested in Operation Cerberus North Mission 13A to make my results comparable to mjfrisby's data on page 6 of this thread: 

In that CPU bound scenario in VR (that's how I can measure frame times), I'm pretty consistently seeing about 3.5 ms better frame times with CCD0 only versus CCD1 only with Tacview on or off. That translates to something somewhat over 10 fps in the range of framerates we're looking at. However, DCS is currently pushing these types of performance numbers to 45 fps regardless. That might be causing some confusion if people are not seeing better frame rates. 

Even if locked at 45 fps or some other arbitrary number, better frame times give a buffer for more traffic or whatever. In any case, I do think the vCache is helping meaningfully in DCS.

I'm getting extremely low frame times across the board. Didn't install tacview this time around (I think from your testing from another thread).

Interesting you mention locked at 45.

I am experiencing a strange issue where it seems to lock to 60 fps or 90 fps (75 also I think). Either way, it will act normal (by normal I mean floating fps number up and down constantly in the 70's range, then all of a sudden it will lock at 60 fps and you can see that neither GPU or CPU are being fully utilized so there is room to jump back up to 70-80's but it won't.....

It does the same at 90 fps for some reason, just not as often. Then for no apparent reason it unlocks and the fps floats around again while the GPU runs full tilt.

[trevoC]

I'll have to check the logs on frame times but I've never seen them this low. In DCS when it locks to 60 fps (not GPU or CPU bound) the frame times are a line across the bottom like they aren't registering. when they do register, the graph is the bottom 1/6 of the chart. I haven't run a single test where the frame times were of any significance whatsoever. Major improvement here. I'll have to look back to see how much less or greater they were between vCache core and non, which I'll do if I can easily find the video and logs. (they really started piling up)

[okletsgo]

Hi all - sorry for the basic question - how do you guys know that xbox game bar is recognising DCS? 

I am seeing DCS loading my cores 8 to 15 rather than 0 to 7. If I set the affinity in the task manager, I can force DCS onto cores 0-7. Is something wrong on my end, or is this normal behavior?

I have a 7950x3D/4090/64GB on Win 11 with the latest BIOS, latest xbox game bar version and the chipset & CPU drivers from the AMD website. 

EDIT: I'm using the MT version of DCS OB.

Edited March 19, 2023 by okletsgo

[hhhttt]

Share some of my finding with DCS MT, 7950x3d,4090

I raise the max frame limit to 300 fps, select 1080p resolution (to remove gpu bottle neck), 2d, mostly high setting, 2xMSAA

Measure F-16 Ramp start, Caucasus

Using process lasso, force DCS MT on either cache or high freq cores.

1. cache cores, average fps 266

2. high freq cores, average fps 220

So x3d cache does help maybe 20% when CPU bound.

 

Secondly, core parking does work with DCS MT

If I disable process lasso, most of the process are in cache cores, however, DCS seems to span too many thread, and some of them are wake up parked high freq cores, and was put there.

However, average fps are still around 266, 1% low are also similar. at 126 fps, it is possible that those thread that on high freq cores are not in heavy communication with main thread, so moving them in to high freq core won't incur cross ccd penalty that much. 

 

Conclusion:

1. Not worth to use process lasso or tinker around, since majority of the usage are still on cache cores, just play the game normally.

2. x3d cache does help when GPU is not bottle neck.

 

[trevoC]

1 hour ago, hhhttt said:

Share some of my finding with DCS MT, 7950x3d,4090

I raise the max frame limit to 300 fps, select 1080p resolution (to remove gpu bottle neck), 2d, mostly high setting, 2xMSAA

Measure F-16 Ramp start, Caucasus

Using process lasso, force DCS MT on either cache or high freq cores.

1. cache cores, average fps 266

2. high freq cores, average fps 220

So x3d cache does help maybe 20% when CPU bound.

 

Secondly, core parking does work with DCS MT

If I disable process lasso, most of the process are in cache cores, however, DCS seems to span too many thread, and some of them are wake up parked high freq cores, and was put there.

However, average fps are still around 266, 1% low are also similar. at 126 fps, it is possible that those thread that on high freq cores are not in heavy communication with main thread, so moving them in to high freq core won't incur cross ccd penalty that much. 

 

Conclusion:

1. Not worth to use process lasso or tinker around, since majority of the usage are still on cache cores, just play the game normally.

2. x3d cache does help when GPU is not bottle neck.

 

I agree (re workarounds) as I've mentioned earlier.

Even though MT looks like a mess (if you are watching cores park and un-park) it seems to be running on the cache cores (at least the main thread does) as my frame rates indicate the same performance as ST which parks nicely.

My conclusion was the same... don't set affinity, don't disable CCD's, don't disable parking in registry, just let it be. It looks a mess but seems to be working. My cores jump around like.. some smart and funny analogy about something that jumps around a lot.

[Kreutzberg]

6 hours ago, okletsgo said:

Hi all - sorry for the basic question - how do you guys know that xbox game bar is recognising DCS? 

I am seeing DCS loading my cores 8 to 15 rather than 0 to 7. If I set the affinity in the task manager, I can force DCS onto cores 0-7. Is something wrong on my end, or is this normal behavior?

I have a 7950x3D/4090/64GB on Win 11 with the latest BIOS, latest xbox game bar version and the chipset & CPU drivers from the AMD website. 

EDIT: I'm using the MT version of DCS OB.

 

The only totally foolproof method is to pull up Resource Monitor and see if the other cores are listed as "parked." Even if it does that, though, just about everyone is reporting that in MT, some of the threads are on the wrong CCD.

Game bar itself can be pulled up with Win+G, click on settings, and see if it's recognizing DCS as a game. I've had a lot of trouble with it remembering it, so I have my bios set to prefer cache, which works perfectly in ST. I also think that ST feels subjectively smoother in the current build.

[Mr_sukebe]

Apologies chaps for jumping in.

Is the current consensus that with the 7950x3d, that using only the faster CCD (i.e. 8 cores) is beneficial to gaming performance, and specifically DCS?

[okletsgo]

9 hours ago, Kreutzberg said:

Game bar itself can be pulled up with Win+G, click on settings, and see if it's recognizing DCS as a game.

Thanks @Kreutzberg  , but how do you establish that game bar has recognised DCS as a game? For me, when I hit Win+G in just in Windows, I don't see much (any?) difference.

 

[skypickle]

has anyone seen any tests of the 5800x3d vs the 7950x3d in DCS?

 

This shows the 5800x3d beating the 7950x3d in far cry

but this shows the 7950x3d beating the 5800x3d in most games at 4k with a 3090 (which I thought would have been gpu limited at 4K)

 

so how can I trust anything I see on the net about this?

 

Edited March 20, 2023 by skypickle

[some1]

Both videos are most likely fakes. No hardware shown, only a bunch of numbers overlaid on a game footage. Take any video like that with a grain of salt.

[trevoC]

5 hours ago, Mr_sukebe said:

Apologies chaps for jumping in.

Is the current consensus that with the 7950x3d, that using only the faster CCD (i.e. 8 cores) is beneficial to gaming performance, and specifically DCS?

No, the faster CCD is worse than the vCache CCD in all my tests by a few percentage points (around 2% slower if I remember correctly). Honestly though, its very insignificant.

That being said, the moral is just let the pc do its thing.

2 minutes ago, some1 said:

Both videos are most likely fakes. No hardware shown, only a bunch of numbers overlaid on a game footage. Take any video like that with a grain of salt.

These are all fake videos for views. Many of them were posted before even review units were released.

[Mr_sukebe]

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?