Masterclass I took advantage of Black Friday to go from 32 to a 64 GB kit to play Star Citizen with my 7800X3D and I am currently testing the manual timings to optimize performance. Kit of 6600CL32 that I run at 6200CL28 and I just work on the most stable FCLK ! Thank you, this motivates me to stay around 2200
Its fun to fiddle with this but in reality a 6000 CL32 kit at EXPO settings is within spitting distance of these hyper-tuned setups. I came to the conclusion a few years ago that the juice isn't worth the squeeze any more, especially if you're not running a 4090 which is the only thing with enough power to even show the differences. Lot of work here though, you got a sub from me.
Its pretty apparent that your FCLK beyond 2133mhz is not actually fully stable as most of your results regress the higher you go. The top priority as with all overclocks is 100% stability, so sticking with slightly lesser FCLK values will most likely save you lots of time, testing and potentially some performance regression. 6400/3200 is often very voltage hungry or not even possible for many 9800x3d samples so sticking with 6200/3100 with fclk of 2133/2166 should be preferred for people who just want stable yet still fast configuration of their cpus. And then crank down all the ram timings as far as they go and you should be good. 8000/2000/2000 can be great choice if you got the dud 9800x3d sample and need plenty of voltage to get even 6200. Because using 1:2 mode is way easier on memory controller so you can get away with significantly lower voltages yet still with good performance
I don't think it's performance regression. I think it just shows that some games favor the increased bandwidth of 2200fclk while other games favor the improved latency of synced 6400/2133fclk. Older games seem to favor latency while newer games seem to favor bandwidth.
@@jjlw2378 FCLK has ECC which *does* lead to performance regression. It also negatively scales with vSoc, which you need fairly high to stabilise 6400+ *1:1*.
@IIHydraII If 2200fclk was unstable, it would always lose to 2133fclk, but it actually wins several times. I believe that this testing shows that some games prefer bandwidth, and some games prefer latency. Feel free to do your own testing and see which you prefer. The "core tuning" or "latency killer" options available in 1.2.0.2b AGESA can also show this. Some games perform better with the "legacy" option while others perform better with Level 2/Auto. I just don't see anything in this test that leads me to believe that his 2200fclk is unstable.
What is a good FCLK stability test anyway? I know one ad hoc test is just to listen for pops and cracks in audio, but what about any of the stress test programs?
Everything is good, except for the fact, that you run 6400c28 (which is hardly achievable) and 8000c38, and not like 6400c32 vs 8000c38 or 6400c28 vs 8000c34
Wrong. It is easier to achieve 6400c28. I could not even achieve 8000c36, c38 was already a pain. Using a low end B650 board it was quite easy to archive 6400c28, if you have a high end bard and can achieve 8000c34, then you can probably do 6400c26 or even 6600mhz. 8000 on AMD is so pointless, especially when you are bandwith bound by the FCLK. If you can run 8000c34 what results are you getting?
@@gamedaytoday1 I been running c36 8000mhz 2000fclk. Hynix A die. completely rock solid stable, not reddit stable. I do however have to tinker with timings every single bios update on a x870e crosshair.
@@gamedaytoday1I’m at 8000c34 and very tight secondary timings (gdm disabled) and I’m happy with it considering that with the uclk only at 2000mhz, I could run with only 1-1.05v on soc and utilize only 7-8w on soc instead of 16-18w when at 3200/3300uclk and 1.25v on soc. I’m still playing with a 3090 (I’m waiting the new series), so I’m not cpu limited in any circumstances. However, testing with aida 64, I noted (ofc) an increase in memory bandwidth at 2200flck vs 2000, but with a 1ns loss in latency.
@gamedaytoday1 I don't know about 8000c38 being hard to achieve, I guess it depends on the cpu memory controller, mobo memory topology (I've heard that being lucky with your mobo is also a thing), your bios and the memory kit itself. My 8k expo just booted and it was stable. I'm only at the start but here are my timings at the moment /LsoNR4o.png, everything is loose now, even the voltage. It works fine with c38 on 1.4v-1.45v vdd, I might be too generous. I didn't try to run it in 1:1 mode yet. As I've mentioned, I'm only at the start. As far as I know, my x670e pg lightning is no different from b650pg lightning/pro rs/steel legend in terms of memory overclocking. And there are definitely b650 boards that could do better than mine. Also what results are you talking about? I don't play these games and I don't want to download this. And even if I would, there's no point comparing my results to yours, because we might be having different system configuration, different windows version (with different settings, debloat stuff etc), different kits (I have 2x24 m-die sr kit), small different settings here and there. Sorry if it sounded offensive, I had no such intention. If I'm not too lazy, I'll give you some results like in a week or so (maybe more). I need to get some stuff delivered to fully finish my build, then I need to spend a lot of time tuning the cpu and tuning/tightning the memory timings. I would need to find the best timings for both 1:1 and 2:1 and test the stability. I can also boot 8400c38, definetely not stable, but I might be able to achieve the stability by fiddling with timings and voltages. And I also need to know what to download so it would be both easy and indicatively at the same time. I'm not promising anything, maybe I will be lazy, I actually don't want to do this already, because I'll need to test too much. And I'm not trying to prove anything, just gathering some information. I only left a comment because you were the only one who mentioned syncing the uclk and fclk on 2:1 mode.
This video is interesting not only for gaming, but also for my threadripper setup. Especially since threadripper can overclock memory far more higher while keeping a stable system.
Just an FYI, if you're using MSI afterburner there's a bug with the power monitoring part of the OSD that impacts 1% lows on the 9800X3D by as much as 50%. For reference th-cam.com/video/bQH3DYNboM0/w-d-xo.html
I'd love to know the differences for productivity workloads. If im doing big computational workloads, would there be an improvement in work times if I go with the higher memory frequencies?
Even though this testing ends up being largely academic, it really scratches that itch in my brain wondering about probably-nothing performance differences. I'd be curious to see if the lowest possible latency helps the 0.1% lows when you're not fully CPU bound such as when they're at 1080/1440 and evenly utilized with the GPU and at 4K max reasonable settings. I understand how much time and effort goes into this testing, so I'm not requesting that you do it, but I'm definitely subscribing hoping for more (:
did you do stability stress test with that 6400: 1:1 flck 2200? do you run any issue bro? mine have hard time reaching fclk 2100, my pc wont even boot after memory training
2133 should be the right value as it’s a 1:3 ratio. So 6400/3 =2,133. Increase dram voltages to 1.4v and vsoc to 1.25-1.3. Make sure uclk:memclk should be 3200. Loosen your timings 32 98 98 102.
You really don't need to worry about mem oc on X3D chips, but if you do it just go with a decent config that is the easiest to get stable. 8000/2000 is the best and easiest config on my 9800X3D system with the games I play. Runs on 1.05VSOC and is completely rock solid stable. 6200/2067 and 6400/2133 was comparable in gaming and has worse latency performance. It's not worth dealing with a hassle if 8000 doesn't work on your system but it was incredibly easy and painless with my 9800X3D. 2:1 was not worth the headache on my 7950X3D system at all though.
Not one of those games require massive amount of RAM transfer, so only latency is the critical point here. You have to try X4 Foundations, Stellaris and Transport Fever 2 which are very CPU/RAM demanding compared to all the above games.
True, it is very option especially fir 2x32GB, but there is no way that CL38 is comparable, even now if you buy sticks they have like CL36 1.4V, and some like deltas 8200Mhz cl36 1.45V, with 1.5V cl34, and 1.62 around CL32. While 6400 CL28 is on the edge but some chips can do CL26, ultimatelly 8000 will be faster bcos if you compare 64xCL28 vs 80xCL34 you will get better avg. fps min fps. And explanatiin that 8000cl38 is barly stable is rudiculous bcos the 6400 CL28 works in 20-30% of X3D chips in 7800x3d and around 50-60% in 9800x3d, same 8000 works on 30-35% 7800x3d while on 9800x3d it is mostly stable on low SOC voltage in every chip. So basically thi clip only shows how much faster the 8000:2000 is. Bcos many people have CL34 and some got CL32.
@@jamesm568 You have to be able to achieve that, not everyone can, whats are your voltages and freq? Also, maybe 3% perf at most.. 6000 30 is perfectly fine for 9800x3d and all users should be capable of that
@@silverwerewolf975 The 9800X3D on a X870 motherboard can practically handle most RAM speeds you throw at it. Take advantage of it, but if you're conservative you can use 6000 MT/s RAM. I prefer to push every last drop of performance while staying in the 1:1 ratio.
It would be helpful if you'd let us know what voltages and scalar are you running on your CPU to be able to hold 2200fclk fully stable? Also what voltage to stabilize your RAM at 6400 C28?
yeah I benchmarked a 8000cl34 ( now cl32) and did a 6400cl28 fclk2200 tuning and was achieving around 7~10 more fps. for the price point I say its not worth it
This is the only smart comment I have read all day Smart Guy , I thought the same thing and tested it, very disappointing results because the motherboard could not do less than 22cl and could not boot bellow 26cl so the latency was in the 84ns. So yeah Did not work for me. More should try to see if they can make it work
Regarding Star Citizen Alpha 4.0, I get 120fps in 3K with my 7600XT 16GB vram. Gotta love framegen for this game 😅. Now if the game itself would be more stable...
With that crap b650, i can assure you if you open karhu ram test with 8000 ram you get error within 7 seconds or insta crash. Also, sub timings matter more than primary.
Given that AMD has the best performing chips in this situation (gaming) and the only our competitor Intel is now using basically the same technology called Universal Chiplet Interconnect Express ( or UCIe).... I'm not sure what there is to fix here... They'll fix things when they can to improve performance as the entire industry has been doing for 40 years it currently performs the very best of any CPU you can buy to ask them to fix it makes little sense
@@jaykoerner yes little buddy just because you are at the top doesnt mean you can still improve. the cpus where bound by this fabric bullshit forever. imagine the extra juice you can get if you can actually utilize 8000 mhz ram or even 10000 by that time. also just because i said "fix" doesnt mean that anything is broken lol chill
@@De2t3ny your wording was junk(you use the word bullshit implying you think it's bad), I don't agree with your actual conclusion since the entire reasoning for these memory discrepancies is the ability to give us higher core count chips at lower prices, increasing yields, allowing different lithography nodes to be used between core and io dies and probably a host of things I'm not qualified to understand, The fact that there is a minor 0 to 6% difference in performance depending on the mismatch between memory speed and The clock to the interconnect is honestly a feature, performance would be better if they could make it better, Roughly the same technology is used by amd Intel, apples m series, nivida server chips and I would assume others, The entire industry doesn't move in an direction for no reason
@IIHydraII It was in an article when the agesa first came out, hadn't kept up with news about it since and that article is now removed. Anywho, thanks again for the results, now I know that grabbing lower CL 6400 is definitely the way to go. Saves me a lot of money on my next build.
Can you provide links to your ZenTimings and HWInfo screenshots for your 6400 and 8000 settings?
Awesome. Thanks again for your advices! Great job as usual
Masterclass I took advantage of Black Friday to go from 32 to a 64 GB kit to play Star Citizen with my 7800X3D and I am currently testing the manual timings to optimize performance. Kit of 6600CL32 that I run at 6200CL28 and I just work on the most stable FCLK ! Thank you, this motivates me to stay around 2200
Its fun to fiddle with this but in reality a 6000 CL32 kit at EXPO settings is within spitting distance of these hyper-tuned setups. I came to the conclusion a few years ago that the juice isn't worth the squeeze any more, especially if you're not running a 4090 which is the only thing with enough power to even show the differences. Lot of work here though, you got a sub from me.
Its pretty apparent that your FCLK beyond 2133mhz is not actually fully stable as most of your results regress the higher you go. The top priority as with all overclocks is 100% stability, so sticking with slightly lesser FCLK values will most likely save you lots of time, testing and potentially some performance regression. 6400/3200 is often very voltage hungry or not even possible for many 9800x3d samples so sticking with 6200/3100 with fclk of 2133/2166 should be preferred for people who just want stable yet still fast configuration of their cpus. And then crank down all the ram timings as far as they go and you should be good. 8000/2000/2000 can be great choice if you got the dud 9800x3d sample and need plenty of voltage to get even 6200. Because using 1:2 mode is way easier on memory controller so you can get away with significantly lower voltages yet still with good performance
I don't think it's performance regression. I think it just shows that some games favor the increased bandwidth of 2200fclk while other games favor the improved latency of synced 6400/2133fclk. Older games seem to favor latency while newer games seem to favor bandwidth.
scrolled through video to find any mentions of used FCLK stability testing methodology, dont think there is any, so +1
@@jjlw2378 FCLK has ECC which *does* lead to performance regression. It also negatively scales with vSoc, which you need fairly high to stabilise 6400+ *1:1*.
@IIHydraII If 2200fclk was unstable, it would always lose to 2133fclk, but it actually wins several times. I believe that this testing shows that some games prefer bandwidth, and some games prefer latency. Feel free to do your own testing and see which you prefer. The "core tuning" or "latency killer" options available in 1.2.0.2b AGESA can also show this. Some games perform better with the "legacy" option while others perform better with Level 2/Auto. I just don't see anything in this test that leads me to believe that his 2200fclk is unstable.
What is a good FCLK stability test anyway? I know one ad hoc test is just to listen for pops and cracks in audio, but what about any of the stress test programs?
Everything is good, except for the fact, that you run 6400c28 (which is hardly achievable) and 8000c38, and not like 6400c32 vs 8000c38 or 6400c28 vs 8000c34
Wrong. It is easier to achieve 6400c28. I could not even achieve 8000c36, c38 was already a pain. Using a low end B650 board it was quite easy to archive 6400c28, if you have a high end bard and can achieve 8000c34, then you can probably do 6400c26 or even 6600mhz. 8000 on AMD is so pointless, especially when you are bandwith bound by the FCLK. If you can run 8000c34 what results are you getting?
I'm running 6400 cl28 @1.53v 2133 fclk on a 7800x3d and b650m aorus elite ax ice
Stock ram timings is 6000 cl36 ,but it's A Die
@@gamedaytoday1 I been running c36 8000mhz 2000fclk. Hynix A die. completely rock solid stable, not reddit stable. I do however have to tinker with timings every single bios update on a x870e crosshair.
@@gamedaytoday1I’m at 8000c34 and very tight secondary timings (gdm disabled) and I’m happy with it considering that with the uclk only at 2000mhz, I could run with only 1-1.05v on soc and utilize only 7-8w on soc instead of 16-18w when at 3200/3300uclk and 1.25v on soc. I’m still playing with a 3090 (I’m waiting the new series), so I’m not cpu limited in any circumstances. However, testing with aida 64, I noted (ofc) an increase in memory bandwidth at 2200flck vs 2000, but with a 1ns loss in latency.
@gamedaytoday1 I don't know about 8000c38 being hard to achieve, I guess it depends on the cpu memory controller, mobo memory topology (I've heard that being lucky with your mobo is also a thing), your bios and the memory kit itself.
My 8k expo just booted and it was stable. I'm only at the start but here are my timings at the moment /LsoNR4o.png, everything is loose now, even the voltage. It works fine with c38 on 1.4v-1.45v vdd, I might be too generous.
I didn't try to run it in 1:1 mode yet. As I've mentioned, I'm only at the start.
As far as I know, my x670e pg lightning is no different from b650pg lightning/pro rs/steel legend in terms of memory overclocking. And there are definitely b650 boards that could do better than mine.
Also what results are you talking about? I don't play these games and I don't want to download this. And even if I would, there's no point comparing my results to yours, because we might be having different system configuration, different windows version (with different settings, debloat stuff etc), different kits (I have 2x24 m-die sr kit), small different settings here and there. Sorry if it sounded offensive, I had no such intention.
If I'm not too lazy, I'll give you some results like in a week or so (maybe more). I need to get some stuff delivered to fully finish my build, then I need to spend a lot of time tuning the cpu and tuning/tightning the memory timings. I would need to find the best timings for both 1:1 and 2:1 and test the stability. I can also boot 8400c38, definetely not stable, but I might be able to achieve the stability by fiddling with timings and voltages.
And I also need to know what to download so it would be both easy and indicatively at the same time. I'm not promising anything, maybe I will be lazy, I actually don't want to do this already, because I'll need to test too much. And I'm not trying to prove anything, just gathering some information. I only left a comment because you were the only one who mentioned syncing the uclk and fclk on 2:1 mode.
In one of your videos, you said playing StarCraft II is kind of like a RAM benchmark. How much impact do you think CUDIMMs will have?
Thanks for doing tests that most people wouldnt want to. Ill just leave it stock and call it a day whenever I get my hands on that cpu
Finally listens to crimson about star citizen being a benchmark
This video is interesting not only for gaming, but also for my threadripper setup. Especially since threadripper can overclock memory far more higher while keeping a stable system.
Just an FYI, if you're using MSI afterburner there's a bug with the power monitoring part of the OSD that impacts 1% lows on the 9800X3D by as much as 50%. For reference th-cam.com/video/bQH3DYNboM0/w-d-xo.html
It's like these guys don't even pay attention lmao
I'd love to know the differences for productivity workloads. If im doing big computational workloads, would there be an improvement in work times if I go with the higher memory frequencies?
It doesnt do as much as none x3d chips sems like we need better interconnect afterall
Even though this testing ends up being largely academic, it really scratches that itch in my brain wondering about probably-nothing performance differences. I'd be curious to see if the lowest possible latency helps the 0.1% lows when you're not fully CPU bound such as when they're at 1080/1440 and evenly utilized with the GPU and at 4K max reasonable settings. I understand how much time and effort goes into this testing, so I'm not requesting that you do it, but I'm definitely subscribing hoping for more (:
wow thanks for the quick lesson. doing my first amd build, i hope my 5700x3d can do 1.9k fclk.
I have researched ram and cl28 is widely known to be a needless expensive ram for x3d chips.
I dont understand why is 2200 left on fclk? Still a bit lost on this video
did you do stability stress test with that 6400: 1:1 flck 2200? do you run any issue bro?
mine have hard time reaching fclk 2100, my pc wont even boot after memory training
2133 should be the right value as it’s a 1:3 ratio. So 6400/3 =2,133. Increase dram voltages to 1.4v and vsoc to 1.25-1.3. Make sure uclk:memclk should be 3200. Loosen your timings 32 98 98 102.
You really don't need to worry about mem oc on X3D chips, but if you do it just go with a decent config that is the easiest to get stable.
8000/2000 is the best and easiest config on my 9800X3D system with the games I play. Runs on 1.05VSOC and is completely rock solid stable. 6200/2067 and 6400/2133 was comparable in gaming and has worse latency performance. It's not worth dealing with a hassle if 8000 doesn't work on your system but it was incredibly easy and painless with my 9800X3D. 2:1 was not worth the headache on my 7950X3D system at all though.
Not one of those games require massive amount of RAM transfer, so only latency is the critical point here. You have to try X4 Foundations, Stellaris and Transport Fever 2 which are very CPU/RAM demanding compared to all the above games.
Keep it simple people, 6000CL30
True, it is very option especially fir 2x32GB, but there is no way that CL38 is comparable, even now if you buy sticks they have like CL36 1.4V, and some like deltas 8200Mhz cl36 1.45V, with 1.5V cl34, and 1.62 around CL32. While 6400 CL28 is on the edge but some chips can do CL26, ultimatelly 8000 will be faster bcos if you compare 64xCL28 vs 80xCL34 you will get better avg. fps min fps. And explanatiin that 8000cl38 is barly stable is rudiculous bcos the 6400 CL28 works in 20-30% of X3D chips in 7800x3d and around 50-60% in 9800x3d, same 8000 works on 30-35% 7800x3d while on 9800x3d it is mostly stable on low SOC voltage in every chip. So basically thi clip only shows how much faster the 8000:2000 is. Bcos many people have CL34 and some got CL32.
@@maveric19871 Intel chips run way faster with 8000mt/s though, even 8600mt/s still gains performance, even with the higher latency
I'm a flight simulator user so 6400MT/s at 28CL is the sweet spot. 6000 MT/s is a disgrace to the 9800X3D.
@@jamesm568 You have to be able to achieve that, not everyone can, whats are your voltages and freq?
Also, maybe 3% perf at most.. 6000 30 is perfectly fine for 9800x3d and all users should be capable of that
@@silverwerewolf975 The 9800X3D on a X870 motherboard can practically handle most RAM speeds you throw at it. Take advantage of it, but if you're conservative you can use 6000 MT/s RAM. I prefer to push every last drop of performance while staying in the 1:1 ratio.
Were there any changes to Vulkan multithreating in SC 4.0 patch?
What about if I'm running 6000mt/ s? What settings should I use then?
I run 6400mhz cl30 with the 9800x3d and it loves it
It would be helpful if you'd let us know what voltages and scalar are you running on your CPU to be able to hold 2200fclk fully stable? Also what voltage to stabilize your RAM at 6400 C28?
Its available in my discord for my supporters
Can you please turn down the music at least 30% next time? It is competing with the volume of your voice and at some points in your video, it wins.
Gotcha thank will balance it better next time
yeah I benchmarked a 8000cl34 ( now cl32) and did a 6400cl28 fclk2200 tuning and was achieving around 7~10 more fps. for the price point I say its not worth it
What soc for 6400 cl28?
1.3
@@gamedaytoday1 I asked him as he used 2200fclk
Recommended PC build for Star Citizen at ~1800 euro price range if you don't mind?
what you should do in the b slot run the bios in duel channel mode and put a 8000mt in b2 slot and 6400mt in b4 slot and see what it does.
also overclock the cpu
I'm really new to this memory thing, so what's the best memory to get for the 9800x3d? For the best performance.
Tuned is the inly way
So I guess my question would be if you can do 2200mhz would 6400 or 8200 be more advantageous?
very curious how the latency and performance might be if you downclock mclk on purpose to force 1:1:1 as fclk:uclk:mclk. ex) 2200:2200:4400
This is the only smart comment I have read all day Smart Guy , I thought the same thing and tested it, very disappointing results because the motherboard could not do less than 22cl and could not boot bellow 26cl so the latency was in the 84ns. So yeah Did not work for me. More should try to see if they can make it work
Regarding Star Citizen Alpha 4.0, I get 120fps in 3K with my 7600XT 16GB vram. Gotta love framegen for this game 😅. Now if the game itself would be more stable...
I think whit 8000 gear 2 probably the best is fclk 2000mhz!
Don't do this with your memory. Also don't worry about it because it's only 3% but still don't do it unless you want to because it's only 3%
Im just going to keep my 6000 c30 setup. Its stable at 1.35v and I wont crash when playing gray zone and lose my stuff.
What Is your vsoc? You should be very lucky to get a stable 9800x3d @ 3200 uclk and 2200 fclk
For 6400mhz its 1.3 for and 8000mhz its 1.1 since the UCLK is half the speed you don't need high SOC.
@gamedaytoday1 he meant how could you achieve 6400 with 2200 fclk
Like 6400 needs high soc which is bad for high fclk
should've benchmarked a game where memory latency really matters like factorio.
what is the background music? has a nice chill vibe
what about 6000 cl26
With that crap b650, i can assure you if you open karhu ram test with 8000 ram you get error within 7 seconds or insta crash. Also, sub timings matter more than primary.
so 8800 MT/s kit with CUDIMM it is.
fclk 2200 uclk 2200 mclk 4400 xD less stress on cpu so u get higher boosts
Ryzen cpus do not support CU-DIMMS. They will run at 3200mhz.
@@kevinerbs2778 nice troll xD
@@QuaK3RRR I'm not trolling at all. Those are both facts.
@@kevinerbs2778 Kingston says on their website it just needs a BIOS Update.
What is fclk?
@@Rashimotosan the infinity fabric
ty good job
it wasn't MT until a few years ago. In reality, it still IS a frequency. edit: show me micro transfers and I'll give them their MEGA transfers 😅
i just got a 6000MTs / CL30 kit for my new pc.. hope thats okay lol
It's fine and most likely stable.
@User-ys7cb cool beans c:
Same here, i hope its stable :( @@User-ys7cb
never use max flck to 2200mhz it is not 1:1:1 it is 1:1:2 u need all 3 to be close
i dont understand why you are testing 10 years old games with 2024 cpu, totally demented test, sorry
i haev 13900kkk with 6400 cl 32 workz noice
pls try warzone with these settings
Memory controller on AMD always been poor compared to its performance anything above 6000 pretty pointless even with high end motherboard
Star Citizen 3.34? Its false, there must be standing 3.24!
Nobody calls it a "Nine Thousand Eight Hundred X 3D". AI voiceovers...just stop already.
Trying running 4 sticks on AMD and get back to us. 😂
the only thing that this video shows is that ram doesnt matter at all on these cpus and amd needs to fix their infinity fabric bullshit
Given that AMD has the best performing chips in this situation (gaming) and the only our competitor Intel is now using basically the same technology called Universal Chiplet Interconnect Express ( or UCIe).... I'm not sure what there is to fix here... They'll fix things when they can to improve performance as the entire industry has been doing for 40 years it currently performs the very best of any CPU you can buy to ask them to fix it makes little sense
@@jaykoerner yes little buddy just because you are at the top doesnt mean you can still improve. the cpus where bound by this fabric bullshit forever. imagine the extra juice you can get if you can actually utilize 8000 mhz ram or even 10000 by that time. also just because i said "fix" doesnt mean that anything is broken lol chill
@@De2t3ny your wording was junk(you use the word bullshit implying you think it's bad), I don't agree with your actual conclusion since the entire reasoning for these memory discrepancies is the ability to give us higher core count chips at lower prices, increasing yields, allowing different lithography nodes to be used between core and io dies and probably a host of things I'm not qualified to understand, The fact that there is a minor 0 to 6% difference in performance depending on the mismatch between memory speed and The clock to the interconnect is honestly a feature, performance would be better if they could make it better,
Roughly the same technology is used by amd Intel, apples m series, nivida server chips and I would assume others, The entire industry doesn't move in an direction for no reason
@@jaykoerner "i dont agree with your conclusion here" then move on and stop typing? like i really dont care lol
@@De2t3ny you cared enough to reply ;)
AMD says it can run 8k memory 1:1, but apparently not. Thanks for showing this!
Where did AMD state that? Its still using the same Zen 4/5 memory controller.
? Where have they ever claimed that lol?
@IIHydraII It was in an article when the agesa first came out, hadn't kept up with news about it since and that article is now removed.
Anywho, thanks again for the results, now I know that grabbing lower CL 6400 is definitely the way to go. Saves me a lot of money on my next build.
@@Void_Tiger_ Fair enough, just keep in mind not many AM5 CPU's can do 6400 1:1. :)