I’ve been running a 9950x for a week now and it’s absolutely crushing my compilation workloads in Scala….great chip for professional software developers
Would like to see you do the curve shaper on the 9600x and also see how high the infinity fabric can go for both, as well as how high you can get the uclk on the 9950x and what the performance differences are between stock and the max speeds you can achieve while being stable.
The reason I didn't do the curve shaper on the 9600x, is because it's 5.4 max boost and when adding the +200 max freq and all core negative optimizer it hit 5.6ghz all core easily And there were no stability issues. So it's simpler just to set the all core and let it run. You arent going to get more then the +200 fmax. And if its stable then its stable. On the 9950x stability was immediately a problem when trying to get the +200. So the curve shaper helped a ton. When I dip into it a little but more I will put out a video on the 9950x with how I tuned the settings.
@@JohnnyRage303 So i'm not really sure what you mean by you won't get more than +200 mhz, you can manually set the cpu clock frequency to 5.7 or 5.8ghz right?? so will the curve shaper not let you do that or am i missing something?? also I understand you said in the video as well as here you set the all core undervolt and the +200 and it ran it with ease and no stability issues, if you used the curve shaper is there a efficiency bonus there or maybe a frequency bonus??
@kodemdotexe4925 I am currently not using a all core locked frequency. You can set the all core multiplier higher. That’s much more of a traditional overclock and you would set a constant voltage. The settings I am referring to is the pbo fmax offset. It is limited to+200. The advantage is you can usually get a higher frequency and a variable voltage. But the system is ultimately using inputs such as temp and load to determine the frequency. Curve optimizer and shaper just instruct pbo to run a higher frequency at a given voltage. If you look up most overclockers that go for a manually set multiplier end up performing better than stock but less than pbo tuned settings.
@@JohnnyRage303 got it so even with the curve shaper you can't get pass that pbo limit of +200mhz and I've seen all core OC do better in multi threaded workloads but do worst in single threaded workloads if the all core clock is lower than pbo boost clock so maybe if you did a all core oc of 5.7 or 5.8 on the 9600x you would see some additional performance
Low key wanting to upgrade to DDR5, but the latency on my DDR4 kit is just too good. I'm willing to sacrifice some bandwidth if it means i can access bytes in significantly fewer cycles.
To me it's just about the cpu gen, if you're am5 then It's obviously gotta be ddr5 and i like trying out the new speeds comming out. If your am4 and you're rocking ddr4 and the cpu meets your needs then heck ya just keep using that low latency. I wouldn't change the whole system just for different ram though.
I am still doing testing, I've been out of town a bit lately though. My initial runs just playing with FLCK and Cinebench was the increased FCLK did help performance more than marrying them up. I am going to be working on getting a video out with some performance info once I get my settings finalized. Also to complicate this more AMD keeps releasing more BIOS updates to improve performance and there's the windows update that will improve performance as well. which is a good thing but as they roll out I will have to see the overall impacts and continue to tweak settings.
Once I get the oc stable I'm gonna test the 8000 vs the 6000 I normally run on zen 4. 6400 seems to be a good speed on single ccd zen 5. That's what I am running on my 9600x.
im new to underatanding thinga like memory latency, ipc, core to core latency etc. im currently on a 5800x3d paired with 3600 cl14 gskill b die, aida64 shows 63ns latency which im pretty sure isnt great. My question is am5 better as it relates to system latency
Ddr5 inherently has a higher Cas latency, but higher frequency, so the real answer is it depends. The real trick is just finding the setup that works best with you current cpu or planned cpu. As I get more time I am going to test the ddr5-6000 cl30 vs the ddr5-8000 cl36 on the 9950x to see which one is better in gaming. For am4 it's just about matching your Infiniti fabric to ram speed and memory controller because you need to be 1:1:1. Then the lowest cl you can afford. Which it seems like you have a good kit.
The 9600x is in a b650e riptide from asrock and the 9950x is in a msi mag tomahawk x670e. Honestly I didn't even mess with the llc on these 2 builds yet. On the 9950x i might try to tune the shaper more to try and get a little more out of it and tightening the llc up might help the stability. I'm still playing around with the 9950x a bit.
@@JohnnyRage303 Tech Yes City found fairly large improvement from using Windows 10. Win 11 harms all chips performance, but especially the 9000 series. ruclips.net/video/mVpv-EpEoGM/видео.html
@@JohnnyRage303the total length of the traces matters as well... If you would watch buildzoid talking about it why you would understand... Please start watching buildzoid videos and learn don't spread missinformation
Thanks for taking time to make this video while being busy with work
I’ve been running a 9950x for a week now and it’s absolutely crushing my compilation workloads in Scala….great chip for professional software developers
It is a beast for sure, thanks for watching!
awrsome video man realpy enjoy your commentary, subscribed.
Thanks!
Would like to see you do the curve shaper on the 9600x and also see how high the infinity fabric can go for both, as well as how high you can get the uclk on the 9950x and what the performance differences are between stock and the max speeds you can achieve while being stable.
The reason I didn't do the curve shaper on the 9600x, is because it's 5.4 max boost and when adding the +200 max freq and all core negative optimizer it hit 5.6ghz all core easily And there were no stability issues. So it's simpler just to set the all core and let it run. You arent going to get more then the +200 fmax. And if its stable then its stable. On the 9950x stability was immediately a problem when trying to get the +200. So the curve shaper helped a ton. When I dip into it a little but more I will put out a video on the 9950x with how I tuned the settings.
@@JohnnyRage303 So i'm not really sure what you mean by you won't get more than +200 mhz, you can manually set the cpu clock frequency to 5.7 or 5.8ghz right?? so will the curve shaper not let you do that or am i missing something?? also I understand you said in the video as well as here you set the all core undervolt and the +200 and it ran it with ease and no stability issues, if you used the curve shaper is there a efficiency bonus there or maybe a frequency bonus??
@kodemdotexe4925 I am currently not using a all core locked frequency. You can set the all core multiplier higher. That’s much more of a traditional overclock and you would set a constant voltage. The settings I am referring to is the pbo fmax offset. It is limited to+200. The advantage is you can usually get a higher frequency and a variable voltage. But the system is ultimately using inputs such as temp and load to determine the frequency. Curve optimizer and shaper just instruct pbo to run a higher frequency at a given voltage. If you look up most overclockers that go for a manually set multiplier end up performing better than stock but less than pbo tuned settings.
@@JohnnyRage303 got it so even with the curve shaper you can't get pass that pbo limit of +200mhz and I've seen all core OC do better in multi threaded workloads but do worst in single threaded workloads if the all core clock is lower than pbo boost clock so maybe if you did a all core oc of 5.7 or 5.8 on the 9600x you would see some additional performance
Low key wanting to upgrade to DDR5, but the latency on my DDR4 kit is just too good. I'm willing to sacrifice some bandwidth if it means i can access bytes in significantly fewer cycles.
To me it's just about the cpu gen, if you're am5 then It's obviously gotta be ddr5 and i like trying out the new speeds comming out. If your am4 and you're rocking ddr4 and the cpu meets your needs then heck ya just keep using that low latency. I wouldn't change the whole system just for different ram though.
Helo hiw is boot for x870e hero ..is have a problem?
could u test on the single ccd and dual ccd cpu: infinity 2000 ulck 2000 mclk 4000 ddr5 8000mts ???? so fclk = uclk ?
I am still doing testing, I've been out of town a bit lately though. My initial runs just playing with FLCK and Cinebench was the increased FCLK did help performance more than marrying them up. I am going to be working on getting a video out with some performance info once I get my settings finalized. Also to complicate this more AMD keeps releasing more BIOS updates to improve performance and there's the windows update that will improve performance as well. which is a good thing but as they roll out I will have to see the overall impacts and continue to tweak settings.
Is there any benefit to DDR5 8000? These max out at 6400.
Once I get the oc stable I'm gonna test the 8000 vs the 6000 I normally run on zen 4. 6400 seems to be a good speed on single ccd zen 5. That's what I am running on my 9600x.
im new to underatanding thinga like memory latency, ipc, core to core latency etc. im currently on a 5800x3d paired with 3600 cl14 gskill b die, aida64 shows 63ns latency which im pretty sure isnt great. My question is am5 better as it relates to system latency
Ddr5 inherently has a higher Cas latency, but higher frequency, so the real answer is it depends. The real trick is just finding the setup that works best with you current cpu or planned cpu. As I get more time I am going to test the ddr5-6000 cl30 vs the ddr5-8000 cl36 on the 9950x to see which one is better in gaming. For am4 it's just about matching your Infiniti fabric to ram speed and memory controller because you need to be 1:1:1. Then the lowest cl you can afford. Which it seems like you have a good kit.
a 7800x3d with an b650 hdv gets 8000mt ddr5 easily, best price/performance out there
could you make a video about it, and show your testing??
What board are you using and what is you llc set at?
The 9600x is in a b650e riptide from asrock and the 9950x is in a msi mag tomahawk x670e. Honestly I didn't even mess with the llc on these 2 builds yet. On the 9950x i might try to tune the shaper more to try and get a little more out of it and tightening the llc up might help the stability. I'm still playing around with the 9950x a bit.
@@JohnnyRage303 Tech Yes City found fairly large improvement from using Windows 10. Win 11 harms all chips performance, but especially the 9000 series. ruclips.net/video/mVpv-EpEoGM/видео.html
What motherboards do you use?
With the 9950x I have the msi mag tomahawk x670e, with the 9600x I have the asrock b650e riptide.
@@JohnnyRage303 Impressive. You got 8000 working on 4 dimm board.
It's not about how many are available it's about the amount used. I'm using the teamgroup xtreem ddr5-8000 kit 2x24gb. So double dimm still.
@@JohnnyRage303the total length of the traces matters as well... If you would watch buildzoid talking about it why you would understand... Please start watching buildzoid videos and learn don't spread missinformation
Are you using Win 10 or 11?
Windows 11
@JohnnyRage303
That explains it. I'm getting an incredible uplift from my 7950x for dev work on linux. Only the gamers are crying like babies 😅😅😅😅😅