• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Poll: Ryzen 7950X3D, 7900X3D, 7800X3D

Will you be purchasing the 7800X3D on the 6th?


  • Total voters
    191
  • Poll closed .
Just bought the 7950x3d. What’s the best way to run this for gaming? I’ve seen people say that Process Lasso is the best way, but surely that’s not what AMD recommend. Would installing all the drivers as per AMD’s recommendations work just as well? I have watched some YouTube videos but it seems to be different information every time, and the channels are not so great.

Is there a setup guide on here? I’m reading through this thread but it just seems to be arguing about things.

This is the best one I’ve seen in the time I’ve looked.
Don't use Process Lasso, in my testing it lowers performance over what is achievable using the game bar/game mode method as recommended by AMD.

I've tested Warzone 2 and Modern Warfare 2 (amongst other games) using game bar, which means the game runs on the cache CCD, along with Windows stuff. I then installed Process Lasso and put everything bar the game file on the second CCD, kept the game on the cache CCD (as COD/Warzone loves cache) and saw worse average FPS and worse 1% lows than I did with game bar. Noticeably worse.

One scenario in which process lasso might be useful is for games that prefer frequency but for some reason are using the cache CCD. In my experience there are very few games out there that meet this criteria, I've yet to find any in fact but there are probably some obscure games where this might be the case. In this scenario you could use process lasso to make it use the frequency CCD. Of course, you can just enable that CPPC option in the BIOS and set it to frequency and this will have the same effect as using process lasso, so process lasso is just another service running using resources. However, it does stop the need to enter the BIOS and change this option which makes it worth consideration in this scenario.

Another scenario would be for someone that is live streaming and wants the game on the cache CCD and they want their streaming software and perhaps some other services related to that on the frequency CCD. However, this will still likely perform worse overall.

If someone is saying they use process lasso to get better performance with a dual CCD X3D, it usually means they haven't got things setup correctly. The only setup you need to do is:
  • Latest BIOS
  • Make sure CPPC in BIOS is set to Auto (it is by default)
  • Make sure UEFI/Secure Boot are enabled and CSM is disabled
  • Use official latest Windows Installation Media
  • Use the default Windows Balanced Power Profile
  • Use the latest Chipset drivers from the AMD website
  • Make sure game bar and game mode are enabled (they are by default)
  • Lastly and most importantly, update game bar via the Windows Store until no more updates are available
Once that is done, you are good to go. Check for game bar updates periodically as new games come out that you want to play. If an obscure game is not detected as a game and is using the frequency CCD, you can make it use the cache CCD to improve performance by bringing up gamebar with Win Key + G, and check the option that says this is a game. That's literally all there is to it and in my testing 99% of games require no intervention at all.

Anyone who tries to proclaim this is too much needs to get in the waste bin, and stay there.
 
Last edited:
Don't use Process Lasso, in my testing it lowers performance over what is achievable using the game bar/game mode method as recommended by AMD.

I've tested Warzone 2 and Modern Warfare 2 (amongst other games) using game bar, which means the game runs on the cache CCD, along with Windows stuff. I then installed Process Lasso and put everything bar the game file on the second CCD, kept the game on the cache CCD (as COD/Warzone loves cache) and saw worse average FPS and worse 1% lows than I did with game bar. Noticeably worse.

One scenario in which process lasso might be useful is for games that prefer frequency but for some reason are using the cache CCD. In my experience there are very few games out there that meet this criteria, I've yet to find any in fact but there are probably some obscure games where this might be the case. In this scenario you could use process lasso to make it use the frequency CCD. Of course, you can just enable that CPPC option in the BIOS and set it to frequency and this will have the same effect as using process lasso, so process lasso is just another service running using resources. However, it does stop the need to enter the BIOS and change this option which makes it worth consideration in this scenario.

Another scenario would be for someone that is live streaming and wants the game on the cache CCD and they want their streaming software and perhaps some other services related to that on the frequency CCD. However, this will still likely perform worse overall.

If someone is saying they use process lasso to get better performance with a dual CCD X3D, it usually means they haven't got things setup correctly. The only setup you need to do is:
  • Latest BIOS
  • Make sure CPPC in BIOS is set to Auto (it is by default)
  • Make sure UEFI/Secure Boot are enabled and CSM is disabled
  • Use official latest Windows Installation Media
  • Use the default Windows Balanced Power Profile
  • Use the latest Chipset drivers from the AMD website
  • Make sure game bar and game mode are enabled (they are by default)
  • Lastly and most importantly, update game bar via the Windows Store until no more updates are available
Once that is done, you are good to go. Check for game bar updates periodically as new games come out that you want to play. If an obscure game is not detected as a game and is using the frequency CCD, you can make it use the cache CCD to improve performance by bringing up gamebar with Win Key + G, and check the option that says this is a game. That's literally all there is to it and in my testing 99% of games require no intervention at all.

Anyone who tries to proclaim this is too much needs to get in the waste bin, and stay there.
Thank you for the detailed response.

This is what I was hoping for, because it is what AMD intended. I don’t like using Process Lasso.

My primary game is BF2042, so hoping I get good results.

I will be sure to use everything you recommend.
 
Last edited:
Thank you for the detailed response.

This is what I was hoping for, because it is what AMD intended. I don’t like using Process Lasso.

My primary game is BF2042, so hoping I get good results.

I will be sure to use everything you recommend.
BF2042 is one of the few games out there where the CPU utilisation is high enough (at least when using 1080P low settings with a 7900 XTX) that one or two cores on the second CCD are used. See my video recording below.

The second video shows Ryzen Master Tool running in the background to accurately tell you which cores are truly awake and active and which are sleeping. If you studied my (MSI AB) overlay on the video above, you'd think all cores were awake and active but that is not the case.

How it works is that primarily all workloads are pinned to one CCD, however if CPU utilisation is high enough then one or more cores from the second CCD may become active and wake up from sleep.

The big misconception is that this should never happen, but that is not the case.

Cyberpunk 2077 is another game that does this but to a lesser degree at 1080P Ultra with RT enabled with the 4090. If you force the game to use only the cache CCD in BIOS or via process lasso, performance goes down vs the game bar configuration which primarily uses CCD0 but sometimes wakes up one or two of the fastest cores on CCD1.
 
Last edited:
BF2042 is one of the few games out there where the CPU utilisation is high enough (at least when using 1080P low settings with a 7900 XTX) that one or two cores on the second CCD are used. See my video recording below.

The second video shows Ryzen Master Tool running in the background to accurately tell you which cores are truly awake and active and which are sleeping. If you studied my (MSI AB) overlay on the video above, you'd think all cores were awake and active but that is not the case.

How it works is that primarily all workloads are pinned to one CCD, however if CPU utilisation is high enough then one or more cores from the second CCD may become active and wake up from sleep.

The big misconception is that this should never happen, but that is not the case.

Cyberpunk 2077 is another game that does this but to a lesser degree at 1080P Ultra with RT enabled with the 4090. If you force the game to use only the cache CCD in BIOS or via process lasso, performance goes down vs the game bar configuration which primarily uses CCD0 but sometimes wakes up one or two of the fastest cores on CCD1.
That is great to see. I honestly was considering going with the 7800x3d but I have a feeling in some games the 7950x3d will be better, maybe Cyberpunk and Spider-Man are better on this? Maybe others also down the line.

Do you have some advice regarding the SOC voltage I should aim for? I have a Hero board which is ASUS, and I don’t want to change boards just because of the recent things online.

That result in BF2042 is awesome.
 
Last edited:
That is great to see. I honestly was considering going with the 7800x3d but I have a feeling in some games the 7950x3d will be better, maybe Cyberpunk and Spider-Man are better on this? Maybe others also down the line.

Do you have some advice regarding the SOC voltage I should aim for? I have a Hero board which is ASUS, and I don’t want to change boards just because of the recent things online.

That result in BF2042 is awesome.

7950X3D & x670e Hero is an excellent combo. CPU is fantastic all around and monstrously efficient. I have a 13900k in another rig, and the 7950x3D just sips power in comparison.

I'm using a 6000Mhz C30 kit, EXPO, SOC is 1.25v, not felt the need to manually tune it, it's under the 1.3v limit so all is good. It's a very stable motherboard.
 
That is great to see. I honestly was considering going with the 7800x3d but I have a feeling in some games the 7950x3d will be better, maybe Cyberpunk and Spider-Man are better on this? Maybe others also down the line.

Do you have some advice regarding the SOC voltage I should aim for? I have a Hero board which is ASUS, and I don’t want to change boards just because of the recent things online.

That result in BF2042 is awesome.
It depends what memory frequency and FCLK you will be running. If you are on the latest BIOS, the board should limit the SOC to 1.3v as the maximum possible voltage. In my experience this will easily be enough for 6000/6200Mhz speed.

Many overclockers have been running SOC voltage at 1.4v and higher without issue since Ryzen 7000 series launched. What recent events have shown us is that much SOC voltage is not needed at all. I was running 1.3v - 1.35v previously on my 7950X3D without issue. But since this all came to light, I'm now running the SOC voltage at 1.162v, fully stable. Quite a drop.

What I would recommend is update your BIOS to the latest and use auto values to start with, which will probably be 1.25v - 1.3v depending on memory speed used. Get things dialled in, memory timings, frequency etc, then you can experiment with lowering the SOC voltage once you've confirmed everything's stable. However, if you are just planning to apply EXPO and change nothing else, then you can get straight into seeing how much lower you can reduce SOC voltage before things get unstable. 1.2v sounds plausible for most samples, potentially even lower.

One thing to bare in mind is that FCLK going higher than 2000Mhz (which is a sweet spot) may require a higher SOC voltage to gain stability. Depends on the silicon lottery. My 7950X3D can do 6200MhzC26, with FCLK at 2167Mhz fully stable at 1.162v SOC which is great. But some samples may need higher so you just have to test it and see what yours is capable of.

As Dave above, I'm also using a Hero and I'm still using the older BIOS versions too as it works fine for me so no need to change BIOS as I use manual voltages anyway for the SOC, etc.
 
Last edited:
7950X3D & x670e Hero is an excellent combo. CPU is fantastic all around and monstrously efficient. I have a 13900k in another rig, and the 7950x3D just sips power in comparison.

I'm using a 6000Mhz C30 kit, EXPO, SOC is 1.25v, not felt the need to manually tune it, it's under the 1.3v limit so all is good. It's a very stable motherboard.
That’s good news because I bought the Corsair C30 kit from Overclockers. I may tune the timings, but I am not going too in depth because I just want to game and not take a long time tuning.

Have you tried the Curve Optimiser and PBO boost?
 
It depends what memory frequency and FCLK you will be running. If you are on the latest BIOS, the board should limit the SOC to 1.3v as the maximum possible voltage. In my experience this will easily be enough for 6000/6200Mhz speed.

Many overclockers have been running SOC voltage at 1.4v and higher without issue since Ryzen 7000 series launched. What recent events have shown us is that much SOC voltage is not needed at all. I was running 1.3v - 1.35v previously on my 7950X3D without issue. But since this all came to light, I'm now running the SOC voltage at 1.162v, fully stable. Quite a drop.

What I would recommend is update your BIOS to the latest and use auto values to start with, which will probably be 1.25v - 1.3v depending on memory speed used. Get things dialled in, memory timings, frequency etc, then you can experiment with lowering the SOC voltage once you've confirmed everything's stable. However, if you are just planning to apply EXPO and change nothing else, then you can get straight into seeing how much lower you can reduce SOC voltage before things get unstable. 1.2v sounds plausible for most samples, potentially even lower.

One thing to bare in mind is that FCLK going higher than 2000Mhz (which is a sweet spot) may require a higher SOC voltage to gain stability. Depends on the silicon lottery. My 7950X3D can do 6200MhzC26, with FCLK at 2167Mhz fully stable at 1.162v SOC which is great. But some samples may need higher so you just have to test it and see what yours is capable of.

As Dave above, I'm also using a Hero and I'm still using the older BIOS versions too as it works fine for me so no need to change BIOS as I use manual voltages anyway for the SOC, etc.
Thank you for the advice. It looks like you have a great chip.

Would you advise the Sha3 for the ultimate CO stress test?
 
Thank you for the advice. It looks like you have a great chip.

Would you advise the Sha3 for the ultimate CO stress test?
That's good for testing stability for curve optimiser.

I personally use Karhu for testing memory timings and FCLK stability 8-12 hours minimum. I use y cruncher 1-7-0 for testing curve optimiser, and sometimes do a few runs of sha3 too.
 
That's good for testing stability for curve optimiser.

I personally use Karhu for testing memory timings and FCLK stability 8-12 hours minimum. I use y cruncher 1-7-0 for testing curve optimiser, and sometimes do a few runs of sha3 too.
Is Sha3 just a normal version or do you have to download a specific or previous version?

Are there any other things you would recommend enabling or disabling when testing curve optimiser?
 
If you want something that you can just leave stressing single cores I can recommend CoreCycler, you can set it to just stress test a single core with either Prime95 or ycruncher. It does also support Aida64 but I haven't really tried it, I just manually binded affinities for SHA3 when I was testing. Stressing single cores at a time can sometimes be useful for finding issues because they're more likely to boost higher if only one core is being stressed. That said all core stress tests are also still useful for stability testing, the best thing you can do is just do a wide range of tests for as much time as you can really.

Some specific settings for CoreCycler I've found good for it on my 7950x3d are FFT Size set to 4-5 and AVX2 instruction set on Prime95. As for Ycruncher, Kizuna is the one for Zen 4 iirc, though some people have mentioned Kagari is decent too so I tried both.

In terms of how to set it up for gaming I've mostly used the driver really. There's a few games which benefit from turning off the second CCD and also SMT such as the new Star Wars game, but that's more because the game is super unoptimised and isn't using cores well at all. I have tried the process lasso set up but I also haven't really noticed any real gains from it to be honest. I'd say it's likely only a benefit if you plan to do something quite CPU intensive in the background so you'd want to tie that to the frequency side and then have the game locked to the first ccd. That's a very specific usecase which I'd imagine not many people would come under though.

Even for games that might benefit from the frequency core instead you don't necessarily have to use process lasso too. When I was looking into the driver I noticed there is a section of the registry where you can add entries to basically set games to a CCD.

Code:
Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\amd3dvcache\Preferences\App

By default there was an entry for League of Legends, which I guess works better on the frequency core, as it's set to use that. I can't say I've tested eitherway as I don't play LoL, but interesting never the less. Personally I've added a few emulator exes in there to use the frequency ccd instead as there is some of those that don't benefit as much from the cache. Some emulators do like the cache though, so it's very trial and error.

I'd say my my main issue with the AMD driver method is that a lot of games aren't detected by the Game Bar as being games even with the latest version of game bar. So you have to go into the Win+G screen, tick the game box and then restart it. Hopefully Windows will improve on detection as time goes on. Anything on Game Pass seems to be fine as I've never had to do it for any of those, but I've had to go through that process for a lot of Steam games. Which is a bit of a minor inconvenience.
 
If you want something that you can just leave stressing single cores I can recommend CoreCycler, you can set it to just stress test a single core with either Prime95 or ycruncher. It does also support Aida64 but I haven't really tried it, I just manually binded affinities for SHA3 when I was testing. Stressing single cores at a time can sometimes be useful for finding issues because they're more likely to boost higher if only one core is being stressed. That said all core stress tests are also still useful for stability testing, the best thing you can do is just do a wide range of tests for as much time as you can really.

Some specific settings for CoreCycler I've found good for it on my 7950x3d are FFT Size set to 4-5 and AVX2 instruction set on Prime95. As for Ycruncher, Kizuna is the one for Zen 4 iirc, though some people have mentioned Kagari is decent too so I tried both.

In terms of how to set it up for gaming I've mostly used the driver really. There's a few games which benefit from turning off the second CCD and also SMT such as the new Star Wars game, but that's more because the game is super unoptimised and isn't using cores well at all. I have tried the process lasso set up but I also haven't really noticed any real gains from it to be honest. I'd say it's likely only a benefit if you plan to do something quite CPU intensive in the background so you'd want to tie that to the frequency side and then have the game locked to the first ccd. That's a very specific usecase which I'd imagine not many people would come under though.

Even for games that might benefit from the frequency core instead you don't necessarily have to use process lasso too. When I was looking into the driver I noticed there is a section of the registry where you can add entries to basically set games to a CCD.

Code:
Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\amd3dvcache\Preferences\App

By default there was an entry for League of Legends, which I guess works better on the frequency core, as it's set to use that. I can't say I've tested eitherway as I don't play LoL, but interesting never the less. Personally I've added a few emulator exes in there to use the frequency ccd instead as there is some of those that don't benefit as much from the cache. Some emulators do like the cache though, so it's very trial and error.

I'd say my my main issue with the AMD driver method is that a lot of games aren't detected by the Game Bar as being games even with the latest version of game bar. So you have to go into the Win+G screen, tick the game box and then restart it. Hopefully Windows will improve on detection as time goes on. Anything on Game Pass seems to be fine as I've never had to do it for any of those, but I've had to go through that process for a lot of Steam games. Which is a bit of a minor inconvenience.
Thank you for the info.

It’s not too much to effort for me to check a box if it is not in fact detected. Something to keep an eye on tho.

There is so many recommendations for Process Lasso being a must and people are put off by that. Such a shame that misinformation can be spread by a few people that don’t understand.

I’m sure you guys are using exotic cooling, those days are over for me and only using an AIO for this 7950x3d. Hope temps are fine.
 
Thank you for the info.

It’s not too much to effort for me to check a box if it is not in fact detected. Something to keep an eye on tho.

There is so many recommendations for Process Lasso being a must and people are put off by that. Such a shame that misinformation can be spread by a few people that don’t understand.

I’m sure you guys are using exotic cooling, those days are over for me and only using an AIO for this 7950x3d. Hope temps are fine.
7950X3D runs cool for the most part, about 5-10c cooler than my 7950X. Should be fine with any decent AIO cooler tbh. I use a H150i Elite 360.
 
Thank you for the info.

It’s not too much to effort for me to check a box if it is not in fact detected. Something to keep an eye on tho.

There is so many recommendations for Process Lasso being a must and people are put off by that. Such a shame that misinformation can be spread by a few people that don’t understand.

I’m sure you guys are using exotic cooling, those days are over for me and only using an AIO for this 7950x3d. Hope temps are fine.
Nah, I'm just using my NH-D15 from my previous CPU. It does the trick. I've only seen it go high in very intensive all core workloads that last a while. For example a 10min run of Cinebench r23 will get it to go to the cap of 89C. In fact I bet if I had a case with more airflow it wouldn't even hit that but there's limited options that allow for a 5.25 drive bay still. I have to say I'm not worried about it running that hot either way. Running to the limit when under full load has always been by design for Zen 4. My Cinebench scores aren't really any lower than any of the reviews. In terms of games the temps get no where near that too, generally the temps are low 70s during gaming. Max like 80C in a game that actually uses more cores.
 
That’s good news because I bought the Corsair C30 kit from Overclockers. I may tune the timings, but I am not going too in depth because I just want to game and not take a long time tuning.

Have you tried the Curve Optimiser and PBO boost?

To be honest, no. I've just set EXPO (to load Corsair 6000Mhz C30profile) and have been gaming away. The amount of time investment to squeeze out single digit % increases is not worth it for me, I'm busy with work so don't have as much free time to tinker as before.

The 7950X3D runs very well out of the box (with the setup @LtMatt advised, software wise). Couple it with a good GPU (I'm using a 4090) and you can't go wrong, you'll just get rock solid stability, 60-80W power draw (package power) in games, while having the same (or greater) fps than a 13900KS.
 
Last edited:
Thank you for the info.

It’s not too much to effort for me to check a box if it is not in fact detected. Something to keep an eye on tho.

There is so many recommendations for Process Lasso being a must and people are put off by that. Such a shame that misinformation can be spread by a few people that don’t understand.

I’m sure you guys are using exotic cooling, those days are over for me and only using an AIO for this 7950x3d. Hope temps are fine.

7950X3D runs so cool that cooler choice is (almost) irrelevant. I'm also a NH-D15S on mine and it's complete overkill.

It's not like a 13900K/KS than runs at 100C on an AIO, these chips are in a different league!
 
Thanks so much for all awesome info. I know where to come if I have any issues for sure.

I am very much looking forward to this build now. I much expected people to say about the ASUS problems and to not use this board. Great news.

Thank you guys.
 
Turns out the Gskill c30 6000 are a die so I will go with them I think - F5-6000J3040G32GX2-TZ5NR. Although the F5-6400J3239G32GX2-TZ5RK are also A die and the same price. I know I wont be able to overclock much past 6000 on my 7950x3d but which ram do you think will be the better starting point or nothing in it? I know the 6400 ram was a lot more money when launched (+£110) so perhaps they are the better binned dies?
mine F5-6000J3038F16GX2-TZ5N is M die, probably the new batch is A die, If I was in your position I'd get the 6400 version just to future proof a little because it most probably will overclock better.
Sorry for late reply I don't know why I didn't get a email notification.
 
Last edited:
Finally got my 7800x3D up and running last night, loaded optimised settings and then did a run of timespy just to compare to my i7 9800 and it scored 29k compared to 21k with the i7 so I'm more than happy with that.
So far this morning I've just enabled expo and medially set the voltage to 1.2v and need to start fine tuning.
 
Back
Top Bottom