As someone who is currently struggling to run a stable 96GB DDR5-6400 CL32 (F5-6400J3239F48GX2-RS5K) on an AsRock X870E Nova with a 9800X3D, I want to thank you a lot for this video. As always you are saving my ass while I'm trying to do something with the RAM!
@@Chicken-o5e Yes, but IMHO go with a 6000MT kit as Builzoid said on this Patreon. There is almost no performance uplift vs 6400MT and it will be way easier to run and tune.
4:21 maybe AMD could expose how much Infinity Fabric error correction is happening (preferably eroor counters: on the fly correctable errors if those could happen and uncorrectable errors that require retransmission) so overclocking FCLK could be easier.
@@ezioauditore1588 no, it shouldn't. Error reporting could be implemented by couple (for IOD-> CCD transfers - physycally located at CCD, for CCD-> IOD transfers - physycally located at IOD) of registers (initialized with zeroes at boot) that gets incremented each time infinity fabric error is found. On each error - data is retransmitted and there is definitely time to increase a counter by one.
Yeah I'm surprised we did not get a few OC flagship 2-dimm boards this cycle, given the RAM density increases. Nitropath is a nice innovation, though maybe short lived between temp exclusivity and if things transition to CAMM2.
@@Nunkuruji Nitropath is a good thing, yeah, but I doubt it could ever beat a board with just 2 normal DIMM slots. Sure there is a benefit for Workstation Boards. But I fail to undestand why they go with 4 slots for the majority of the desktop lineup, given that the people building systems with it generally (should) have the required knowledge to know what is the best for a given use-case.
They weren't able to squeeze that much silicon into the RAM chips yet, memory just isn't shrinking as much with new silicon nodes as it used to. They'll make it happen eventually but for now we have to be satisfied with the 50% increase instead of the typical 100% increase.
@@catonturtle3202 tyvm random internet stranger! 💖Mine was actually one step too high apparantly, (spotify was cracking a bit, never noticed) I reduced it by one and my audio seems fine now.
Perfect... man, this is exactly my setup! Only difference is that I have Dominator Titanium @ 6400. Same motherboard, 96GB, 9950x... And yes, it's a workstation. I don't even have a graphics card. I'm going to copy all your settings and see what happens. Am I expecting any better CPU performance with better memory timings? Also, as you have a 6000 kit going to 6400, whereas I have the higher speed out the box, can you recommend any alternative changes? I believe I've got really good quality silicon because in many other videos, people are overclocking and tuning in order to hit 44,000 in Cinebench R23. Whereas, out the box, I set the maximum temperature to 88 C (I was really shocked when it ramped up to 95C in 5 seconds - and I have a really high performance AIO water cooler) My scores are 45,000 - with no tuning whatsoever... I'm really just seeing what else I can squeeze out of it and also because my work benefits from memory speeds (audio processing in real time) edit: Copied all settings and although I can boot into Windows, my wallpaper was flickering. I still completed an R23 multicore test just fine though with a new score of 46,000! I've reset everything back though as I don't want to cause any damage and getting an extra 500 points isn't the gains I was hoping for. Otherwise, the system runs fantastically as-is. I was worried I'd damaged something as I have read that flickering windows can be permanent if you've done something too extreme. But resetting everything has solved the issue.
can please, at least see some performance results on videos like this? it would be really helpful to know what type of performance could be expected from this setup. thanks buildzoiid!~
Similarly - for core oc's, SFT v3 (v4 is still beta and doesnt seem to hit the cores hard enough as of this posting) will get you failures faster than splitting it up. Also, SFT >>> P95.
Y cruncher is giving me a coefficient is to late error 20 mins in to a test, Karhu also gives an error around 20mins into a test, but after 5hours of prime 95 it gave no errors. Would that be too low soc voltage? Or timings?
Looking forward to this video because I have a G.Skill Trident Z Royal 2x48GB CL32 6400MHz 1.35v (native) kit that I just got about 2 weeks ago, & I'd love to see if my timings can be further tightened, compared to what I'm running now, which is based off your last 2x48GB video where you ran CL28 6200MHz - I'm using your timings from that but at CL30 6200MHz 1.4V (Voltage could probably be lowered) - I can say the RGB on these suckers adds temperature more than my previous kit, it makes temps go up about 3-4C if you set them to solid white. I think I'm more attached to running 2200MHz FCLK than I am trying to hit 6400MHz on my RAM though lol Edit: Also, what's the possibility that you get ahold of a cheapy 2x48GB kit, there's this one G.Skill makes that's $199, no RGB, it's CL40 5200MHz 1.1V or something like that stock - I'm wondering how those would fair with your tuning, if they can be pushed at all or not.
I also have the same question regarding the 2x48GB kits running 5200MHz and 5600Mhz CL40 on a Ryzen 9950X, X870 system. Is it possible to push them to 6000HMz? Buying the 6000MHz+ kits are almost the double the price.
@@Supercalafradjal I think most likely they are. Because those 5200MHz kits are running at 1.1v, so if you push the voltage they can probably hit 6000MHz with decent timings... Unless they're Samsung DDR5 lol that stuff is trash. Wish I could spend the $200 to find out but I can't right now.
@@Supercalafradjal I'm trying to egg Buildzoid on in the comments of new videos to give that $189 2x48GB CL40 1.1v kit a go, hoping he'll ask G.skill or just buy it himself to try. Best of luck. I'm betting CL30 6000MHz at ~1.35v should be simple IF it's Hynix chips.
I’ve got 96gb (2x48gb) Corsair dominator titanium (dual rank Hynix m-die) with 7000 & 7200 XMP profiles, alongside an ASUS ROG x870e Hero and a 9950x. These settings work wonderfully for me at 6400 but I’d like to see if I can get 8000 out of them. So far no luck but I’m way out of date when it comes to RAM tuning and overclocking. What are the key settings and voltages I should be looking to set to get 8000 stable? Of course given that they’re dual rank it may not even be possible but given the XMP profiles it feels worth a shot at least.
If nobody is selling any exhorbitantly priced kits with pre-rated xmp/expo timings at any given clock speed, it's fairly safe to assume that it's unachievable under anything approaching normal circumstances...
@@GR_Krusly that should get you the best results. I believe you can run that on the x670e or the x870e. I think the cpu memory controller is the limiter at this time. Because I had a 7950x on an ASRock Taichi Carrera and the best I could do was 5600 1:1. When I first got the 9950x that's all it would do with stability. Then the updates both windows and amd. Now both my 7950x and 9950x (on MSI Meg x670) run 6400 stable. Same ram, same MBs, same cpu. The software made the difference where the controller is I believe. My opinion only.
I haven't watched the video yet and will delete this comment if my questions answered in it but how does the x870e hero compare to good 1dpc boards like the x670e gene?
Any thoughts on X870E vs X670E and R9xxx vs R7xxx? In terms of OC potential \ performance \ quirks and features of the newer chips (and boards) vs the older ones? Would love to watch a video about your experiences so far, even if you ended up with a mediocre bin and what not, I'm just having a bit of a hard time seeing the differences between this gen and the previous one.
For memory overclocking there are 2 things that could help. All boards *should* have better traces layout from CPU -> RAM that should help getting better timing, and there's also an Asus exclusive (for 1 year) feature called Nitropath, that according to Der8auer, it should help in 4 memory slot boards to get better RAM speed when only using 2 memory sticks, because it fixes some interference from unpopulated slots.
@@gusgyn I don't need generic tips.. my current setup (AM4) is highly optimized and I'm looking to upgrade soon. I'm talking specifically about this gen vs previous gen.
Hi, do you have a video, where you explain what exactly all these different timings like Tcl, Trcp and so on mean, and what they affect?. I don´t want fancy overclocking for my next system. Just being more sure about what i am doing, when it comes to set the timings right.
I just upgraded to a Proart X870E, Ryzen 9950X and 2x32GB Corsair 32-40-40-84 DDR6400. 1:1 3200 isn’t stable at DOCP settings. I lowered to 6200 and dialled it in to 32-39-34-64 and some of the sub timings BZ suggested for Hynix previously. FCLK at 2067 and UCLK 1:1 and it’s passed 3000% Karhu and y-cruncher VT3, max temp 47C and it was a 26C day here in Aus. AIDA64 says 84K MB/s read and 63ns latency. DDR6400 seems elusive though with this kit and chip. Might be my silicon lottery ticket wasn’t as good?
I have the i have the same kit, corsair vengeance 2×32GB 6400. My motherboard is Asus TUF B650 with Ryzen 8700G. Same situation, I can only reach 6200, 6400 wont boot. However, on the ASUS support site it says that the 2×32 6000 kit is supported, no mentions of the 6400. Maybe a future bios update but I dont think its because of bad silicon.
I bought some G.SKILL Trident Z Royal 2x32GB (64GB) 6400MHz CL32 what settings are the most important for Performance and Latency? I heard tRFC and tREFI are important. Also are my kits supposed to be A-die?
Would these timings and setting work for Kingston fury beast 2x48g 6400 and my 9800x3d? I already have the stock 6400 Xmp running at 1:1 with 2200 fclk. Running an x870e-e
Asus exclusive (for 1 year) feature called Nitropath, that according to Der8auer, it should help in 4 memory slot boards to get better RAM speed when only using 2 memory sticks, because it fixes some interference from unpopulated slots. It should have no benefit for running 4 sticks of memory.
@@FPSDuLe ASUS X870 ITX is better than ASUS Crosshair X870E Hero. Crosshair X870E Hero is better than ASUS B650E/X670E. So it helps but not as much as removing the extra mem slots.
@@ActuallyHardcoreOverclocking Yeah its better. I did upgrade from X670E Hero to this one and now i get easy 6000MT/s Cl30 runing. I know thats nothing but i wasnt able to run it on X670E Hero out of box. So for me thast big upgrade like to stay on AM5 platform... 👀
I just tried to run my 9950x with a Patriot 8200mhz 2x24GB kit and got no luck either. Could boot in to windows with up to 8400mhz cl40, but not stable even on 8000mhz. Just not happening. The same kit happily runs identical settings what you have here. For now at least it seems that 8000+mhz ddr5 on Zen 5 stable is not an option.
I ordered G.Skill 4x16GB TZ5 Royal Neo Silver RGB EXPO 6000mhz CL28 for 9950x3d. Do you think this is a mistake? Should I return these and use dual? There are many people who say that they cannot get 6000mhz on the quad channel.. btw my motherboard msi mpg x870e carbon wifi
Your kit isn't quad channel, it is still dual channel just has the ranks split across DIMMs. That said, x2 2Rank kits are easier to stabilise than x4 1Rank kits. It's a function of the signalling to the second DIMM slot in each channel. Definitely return that kit and pick up a kit of 2 dual rank modules.
So really, what makes expensive boards like this with all the smart tech tags for optimizing >8000 Mts memory speeds useable when maybe 1 in 10000 (or even higher) 9xxx chips have a memory controller that will actually play along with a 8000 Mts memory config ? I get it .. the Hero board looks gorgeous and it have a nice boost in connectivity with 2x USB4 and 44 available pcie lanes, even it still need to disable the 2'nd Pcie slot if you want to eg. run 3x gen5 SSD's (is this a feature or "design flaw"?😅) And why do they keep putting 4 memory slots of DDR5 on all none ITX boards if DDR5 is so horrible at running 4 sticks at a time. Just a reason that they can write it supports up to 192Gb of Ram with heavy degradation in memory speed. Do they actually think AMD is going to fix the horrible memory controller before AM5 goes EOL ?
@@Vlad-jy9ls Problem is it's a waste of money as it isn't faster than a good tight 6000 kit. But it's more that what do you really get for all the money these Mobos cost beside a bit better connectivity. You still have to choose how you use the available pcie-lanes as the 44 lanes is still not enough to drive 2 pcie 5 gen slots + 2x USB 4 + 3x m.2 gen 5 SSD's.
@@danielt91we dont even know what itll cost, mobo sure but some people dont mind spending a lil on say an asrock taichi which cost nothing compared to other flagships. At this point its for fun
You're really good at overclocking but very bad at explaining it in a succinct manner. You should put the settings into a document or screen overlay, put it up, and quickly move through all the timings you got for the fastest performance and the voltage settings. You can add any notes in that are pertinent. The issue is that even someone like me, who is a pretty geeky hardware person, has my brain fuzz out and lose focus at the rambling, droning stream of consciousness with a full screen of random live data. Summarize. Synthesize any revelatory findings. Plan ahead on a format of some sort. Hey, here are the fastest timings I got. Here is the voltage I had to change. NEXT SECTION: Here is the testing for stability confirmation I use. NEXT SECTION: Here are some hurdles I had to go through. NEXT SECTION: My thoughts, conclusion.
These videos are actually a waste of time. Unless you have the same memory and mother board. With so many gullible people in the world there will always be people selling snake oil.
@actually hardcore overclocking I have hynix a-die that can’t do lower than 37 trp at 6200mhz. I’m wondering can it be because of too low tras (30) or trc? Is it worth it to try raise tras for lower trp or is the performance difference so low that i should not worry about it? Thanks for the help! 😊
As someone who is currently struggling to run a stable 96GB DDR5-6400 CL32 (F5-6400J3239F48GX2-RS5K) on an AsRock X870E Nova with a 9800X3D, I want to thank you a lot for this video. As always you are saving my ass while I'm trying to do something with the RAM!
@@Chicken-o5e Yes, but IMHO go with a 6000MT kit as Builzoid said on this Patreon. There is almost no performance uplift vs 6400MT and it will be way easier to run and tune.
let me know if it succeeds. use @ or i wont know
@@zahirkhan778 I dropped the clock to 6200MT with CL30 and the rest as in the rest as in the video with.
should have got the x870e/x670e taichi lite they go up to 8000 on AMD
4:21 maybe AMD could expose how much Infinity Fabric error correction is happening (preferably eroor counters: on the fly correctable errors if those could happen and uncorrectable errors that require retransmission) so overclocking FCLK could be easier.
thats why i hate memory oc on ryzen, too many variables could go wrong
It would decrease performance. Thats why they fully disabled reporting it on am5 becouse error correction is happening all the time.
@@ezioauditore1588 no, it shouldn't. Error reporting could be implemented by couple (for IOD-> CCD transfers - physycally located at CCD, for CCD-> IOD transfers - physycally located at IOD) of registers (initialized with zeroes at boot) that gets incremented each time infinity fabric error is found. On each error - data is retransmitted and there is definitely time to increase a counter by one.
I just bought a similar setup, I was really struggling with the bios settings. Huge performance increase! Thank you very much!
self note of bios 19:28 voltages 22:17 timings 25:55
Thank you, my ram has now been stabilized. Earned a sub.
My guy living in 2045 to give us knowledge from the future
explains the wisdom
32:10 but that is a design flaw, not a feature 😂👍🏽 made my day
Yeah I'm surprised we did not get a few OC flagship 2-dimm boards this cycle, given the RAM density increases. Nitropath is a nice innovation, though maybe short lived between temp exclusivity and if things transition to CAMM2.
@@Nunkuruji Nitropath is a good thing, yeah, but I doubt it could ever beat a board with just 2 normal DIMM slots. Sure there is a benefit for Workstation Boards.
But I fail to undestand why they go with 4 slots for the majority of the desktop lineup, given that the people building systems with it generally (should) have the required knowledge to know what is the best for a given use-case.
@@Nunkuruji I'm looking forward to CAMM2 aswell 😁 Next step before most CPUs will have on-package RAM.
Testing and Optimizing the 0x12b Microcode on the ASUS motherboard please 🙏
what's holding up 64GB per DIMM? Also looking forward to testing the new MSI UNIFY-X Z890!
You need 32gbit memory chips for that, we only have 16gbit and 24gbit DDR5 at the moment.
They weren't able to squeeze that much silicon into the RAM chips yet, memory just isn't shrinking as much with new silicon nodes as it used to. They'll make it happen eventually but for now we have to be satisfied with the 50% increase instead of the typical 100% increase.
Crucial has DDR5 CUDIMM memory coming soon, promised 64GB capacity in 2025. Micron and its sub-brand Crucial are introducing new CU-DIMM memory.
This is sick. Thanks 🙏
good result!!
How do you actual check if error correction of FCLK kicks in?
Random perf drops during ram stress testing, games that load the system will show errors as well
you can listen to music while stress testing and if it's unstable the audio should be distorted
Random stutters in daily usage and random restarts
@@catonturtle3202 tyvm random internet stranger! 💖Mine was actually one step too high apparantly, (spotify was cracking a bit, never noticed) I reduced it by one and my audio seems fine now.
Perfect... man, this is exactly my setup!
Only difference is that I have Dominator Titanium @ 6400.
Same motherboard, 96GB, 9950x...
And yes, it's a workstation. I don't even have a graphics card.
I'm going to copy all your settings and see what happens. Am I expecting any better CPU performance with better memory timings?
Also, as you have a 6000 kit going to 6400, whereas I have the higher speed out the box, can you recommend any alternative changes?
I believe I've got really good quality silicon because in many other videos, people are overclocking and tuning in order to hit 44,000 in Cinebench R23.
Whereas, out the box, I set the maximum temperature to 88 C (I was really shocked when it ramped up to 95C in 5 seconds - and I have a really high performance AIO water cooler)
My scores are 45,000 - with no tuning whatsoever...
I'm really just seeing what else I can squeeze out of it and also because my work benefits from memory speeds (audio processing in real time)
edit:
Copied all settings and although I can boot into Windows, my wallpaper was flickering. I still completed an R23 multicore test just fine though with a new score of 46,000!
I've reset everything back though as I don't want to cause any damage and getting an extra 500 points isn't the gains I was hoping for. Otherwise, the system runs fantastically as-is.
I was worried I'd damaged something as I have read that flickering windows can be permanent if you've done something too extreme. But resetting everything has solved the issue.
can please, at least see some performance results on videos like this? it would be really helpful to know what type of performance could be expected from this setup. thanks buildzoiid!~
You are a god bro
Gigabyte uses different names for certain memory timings, so I'd love to see you do a GSkill 8400 2x24 kit on an AORUS Master running a 9900x. 🙂
Similarly - for core oc's, SFT v3 (v4 is still beta and doesnt seem to hit the cores hard enough as of this posting) will get you failures faster than splitting it up. Also, SFT >>> P95.
Y cruncher is giving me a coefficient is to late error 20 mins in to a test, Karhu also gives an error around 20mins into a test, but after 5hours of prime 95 it gave no errors. Would that be too low soc voltage? Or timings?
"I don't like having my RAM stability in a superposition." 😂
Looking forward to this video because I have a G.Skill Trident Z Royal 2x48GB CL32 6400MHz 1.35v (native) kit that I just got about 2 weeks ago, & I'd love to see if my timings can be further tightened, compared to what I'm running now, which is based off your last 2x48GB video where you ran CL28 6200MHz - I'm using your timings from that but at CL30 6200MHz 1.4V (Voltage could probably be lowered) - I can say the RGB on these suckers adds temperature more than my previous kit, it makes temps go up about 3-4C if you set them to solid white.
I think I'm more attached to running 2200MHz FCLK than I am trying to hit 6400MHz on my RAM though lol
Edit: Also, what's the possibility that you get ahold of a cheapy 2x48GB kit, there's this one G.Skill makes that's $199, no RGB, it's CL40 5200MHz 1.1V or something like that stock - I'm wondering how those would fair with your tuning, if they can be pushed at all or not.
I also have the same question regarding the 2x48GB kits running 5200MHz and 5600Mhz CL40 on a Ryzen 9950X, X870 system. Is it possible to push them to 6000HMz? Buying the 6000MHz+ kits are almost the double the price.
@@Supercalafradjal I think most likely they are. Because those 5200MHz kits are running at 1.1v, so if you push the voltage they can probably hit 6000MHz with decent timings... Unless they're Samsung DDR5 lol that stuff is trash. Wish I could spend the $200 to find out but I can't right now.
@@LOLHoneybadger thanks for the reply. I plan to buy a pair next month. I will give you a head-up on the die and speeds I get.
@@Supercalafradjal I'm trying to egg Buildzoid on in the comments of new videos to give that $189 2x48GB CL40 1.1v kit a go, hoping he'll ask G.skill or just buy it himself to try. Best of luck. I'm betting CL30 6000MHz at ~1.35v should be simple IF it's Hynix chips.
Can you please link to the tools you use?
I’ve got 96gb (2x48gb) Corsair dominator titanium (dual rank Hynix m-die) with 7000 & 7200 XMP profiles, alongside an ASUS ROG x870e Hero and a 9950x. These settings work wonderfully for me at 6400 but I’d like to see if I can get 8000 out of them. So far no luck but I’m way out of date when it comes to RAM tuning and overclocking. What are the key settings and voltages I should be looking to set to get 8000 stable? Of course given that they’re dual rank it may not even be possible but given the XMP profiles it feels worth a shot at least.
If nobody is selling any exhorbitantly priced kits with pre-rated xmp/expo timings at any given clock speed, it's fairly safe to assume that it's unachievable under anything approaching normal circumstances...
these settings are applicable to 2x24gb 48 gb kits too right?
BZ, what sort of max temperature during stress test you find "safe" for use higher refresh interval values? 40000, 50000, 65535?
When I run Prime95, instead of straight lines I sometimes see weird squigglies on task manager. Is it because of clock stretching?
Intel says 8000 is the sweet spot for arrow lake. Also there will be some 1DPC oc boards available. That would be interesting.
😂😂😂😂😂😂
12 days later someone has reported disproportionally high gains with memory overclocking on arrow lake
Best way to test infinity fabric stability? Should it just be set as high as possible stable if you're in 1:2 mode?
I think you should change youre channel name to buildzoid, its a little confusing being known as two different names.
they say a x870e motherboard + 9000 series cpu can handle up to DDR5-8000. Is this true? or will Ryzen CPU's still struggle with higher than 6400?
8000 certainly is possible on some CPUs with some boards and some RAM kits. I don't think it's most of them.
They Will run 8200 but not 1:1. Best I can get 1:1 is 6600 stable (tried both 2x16 and 2x24)
@walterrankin200 i see. Thanks, i'm thinking to go for 2 sticks @24gb 6400 30cl each. if available
@@GR_Krusly that should get you the best results. I believe you can run that on the x670e or the x870e. I think the cpu memory controller is the limiter at this time. Because I had a 7950x on an ASRock Taichi Carrera and the best I could do was 5600 1:1. When I first got the 9950x that's all it would do with stability. Then the updates both windows and amd. Now both my 7950x and 9950x (on MSI Meg x670) run 6400 stable. Same ram, same MBs, same cpu. The software made the difference where the controller is I believe. My opinion only.
I haven't watched the video yet and will delete this comment if my questions answered in it but how does the x870e hero compare to good 1dpc boards like the x670e gene?
I just bought 9950X with Corsair 96GB (2x48GB) DDR5 7000MHz CL40 Dominator. Hoping for it to work :D
Any thoughts?
Did it work?
Did it work
Any thoughts on X870E vs X670E and R9xxx vs R7xxx? In terms of OC potential \ performance \ quirks and features of the newer chips (and boards) vs the older ones? Would love to watch a video about your experiences so far, even if you ended up with a mediocre bin and what not, I'm just having a bit of a hard time seeing the differences between this gen and the previous one.
For memory overclocking there are 2 things that could help. All boards *should* have better traces layout from CPU -> RAM that should help getting better timing, and there's also an Asus exclusive (for 1 year) feature called Nitropath, that according to Der8auer, it should help in 4 memory slot boards to get better RAM speed when only using 2 memory sticks, because it fixes some interference from unpopulated slots.
@@gusgyn I don't need generic tips.. my current setup (AM4) is highly optimized and I'm looking to upgrade soon. I'm talking specifically about this gen vs previous gen.
Hi, do you have a video, where you explain what exactly all these different timings like Tcl, Trcp and so on mean, and what they affect?. I don´t want fancy overclocking for my next system. Just being more sure about what i am doing, when it comes to set the timings right.
He has a playlist where he explains all the timings
@@compumaniacos do you mean the ddr4 timings explained? i saw this but this is ddr5. doesnt this matter?
@@Mullekularall the language is the same between ddr4 and ddr5 so it’s basically the same.
I just upgraded to a Proart X870E, Ryzen 9950X and 2x32GB Corsair 32-40-40-84 DDR6400. 1:1 3200 isn’t stable at DOCP settings. I lowered to 6200 and dialled it in to 32-39-34-64 and some of the sub timings BZ suggested for Hynix previously. FCLK at 2067 and UCLK 1:1 and it’s passed 3000% Karhu and y-cruncher VT3, max temp 47C and it was a 26C day here in Aus. AIDA64 says 84K MB/s read and 63ns latency. DDR6400 seems elusive though with this kit and chip. Might be my silicon lottery ticket wasn’t as good?
I have the i have the same kit, corsair vengeance 2×32GB 6400. My motherboard is Asus TUF B650 with Ryzen 8700G. Same situation, I can only reach 6200, 6400 wont boot. However, on the ASUS support site it says that the 2×32 6000 kit is supported, no mentions of the 6400. Maybe a future bios update but I dont think its because of bad silicon.
same kit, 1:1 isn't stable at 6400. I've to push the soc all the way to 1.3 to make it a little bit stable but still not good enough.
@@MobGamingYT 1 4V is standard setting for 6400 mts
@@ÁguiaDoSenhor yeah but I'm talking about SoC voltage and u can't exceed 1.3 on that
The kit was super cheap so not complaining. I think Phoronix did a test with the same kit and had some success.
I bought some G.SKILL Trident Z Royal 2x32GB (64GB) 6400MHz CL32 what settings are the most important for Performance and Latency? I heard tRFC and tREFI are important. Also are my kits supposed to be A-die?
Can i follow the timings for hynix a die 32gb?
I also have had the new sensor glitch it’s rather interesting
Would these timings and setting work for Kingston fury beast 2x48g 6400 and my 9800x3d? I already have the stock 6400 Xmp running at 1:1 with 2200 fclk. Running an x870e-e
Very cool!
Have you ever tried 4x48gb DDR5? What kind of speeds are realistic to attain with a 9950 or 7950 in your opinion?
I need help with my x670e hero. trying to run 64gb ddr5 4x16 @ 5600-6000. Im kinda new to pcs
Is this ram Kit with SKU number: CMH96GX5M2B6000C30 ?
The answer is yes
how to check fclk errors?
Question: What cooler are you using that has the fan cooling the RAM like that.. ah you explained it at 9:50
4x48GB please.
I hope he does 4x32!
don't have enough dimms for that
It's not gonna run pass 5000hx
Yes, please! Whatever configuration is stable and even slightly faster than stock would be perfect! Thank you!
Any 4 dim configutation would be greatly appreciated!
Are these tuning done without CPU PBO? I want to do both MEM tuning as well as CPU PBO, which should I enable first? 😢
RAM first. Because the extra mem performance can de-stabilize tests that were previously mem bottlenecked.
How is the nitropath feature? Any benefits from that?
Asus exclusive (for 1 year) feature called Nitropath, that according to Der8auer, it should help in 4 memory slot boards to get better RAM speed when only using 2 memory sticks, because it fixes some interference from unpopulated slots. It should have no benefit for running 4 sticks of memory.
@@gusgyn Yeah i know, and i am using only 2 sticks. Its something more like APEX version but only with 4 slots board
@@FPSDuLe ASUS X870 ITX is better than ASUS Crosshair X870E Hero. Crosshair X870E Hero is better than ASUS B650E/X670E. So it helps but not as much as removing the extra mem slots.
@@ActuallyHardcoreOverclocking Yeah its better. I did upgrade from X670E Hero to this one and now i get easy 6000MT/s Cl30 runing. I know thats nothing but i wasnt able to run it on X670E Hero out of box. So for me thast big upgrade like to stay on AM5 platform... 👀
so... how much performance does this give?
more
I just tried to run my 9950x with a Patriot 8200mhz 2x24GB kit and got no luck either. Could boot in to windows with up to 8400mhz cl40, but not stable even on 8000mhz. Just not happening. The same kit happily runs identical settings what you have here. For now at least it seems that 8000+mhz ddr5 on Zen 5 stable is not an option.
I ordered G.Skill 4x16GB TZ5 Royal Neo Silver RGB EXPO 6000mhz CL28 for 9950x3d. Do you think this is a mistake? Should I return these and use dual? There are many people who say that they cannot get 6000mhz on the quad channel.. btw my motherboard msi mpg x870e carbon wifi
Your kit isn't quad channel, it is still dual channel just has the ranks split across DIMMs. That said, x2 2Rank kits are easier to stabilise than x4 1Rank kits. It's a function of the signalling to the second DIMM slot in each channel. Definitely return that kit and pick up a kit of 2 dual rank modules.
@@ryanreynolds2401 I returned it and purchased 6400mhz 32x2 thank you
And like all asus boards will not be able to boot as they keep booking all bioses.
Weird people are running the Gskill Z5 Royal series 8200hz.
this is crazy, and my 7700 machine still lags in windows randomly, god i hate ddr5 so much
👍👍👍
ima say my cpu runs SOC at 1.12 at 2133 and it cannot run ddr5 6600 at 1:1 😢 or even boot
Can confirm I have had crashes while fftv4 was running.
Restart to bios/uefi = .cmd file with this:
shutdown -r -fw -t 1
save it, run as admin ;)
overcocking
What cpu cooler is that and what kind of fan is that cooling the ram?
for the ram fan it's an amd stock cooler with the heatsink removed
So really, what makes expensive boards like this with all the smart tech tags for optimizing >8000 Mts memory speeds useable when maybe 1 in 10000 (or even higher) 9xxx chips have a memory controller that will actually play along with a 8000 Mts memory config ? I get it .. the Hero board looks gorgeous and it have a nice boost in connectivity with 2x USB4 and 44 available pcie lanes, even it still need to disable the 2'nd Pcie slot if you want to eg. run 3x gen5 SSD's (is this a feature or "design flaw"?😅) And why do they keep putting 4 memory slots of DDR5 on all none ITX boards if DDR5 is so horrible at running 4 sticks at a time. Just a reason that they can write it supports up to 192Gb of Ram with heavy degradation in memory speed. Do they actually think AMD is going to fix the horrible memory controller before AM5 goes EOL ?
Most of the Ryzen chips can do 8000 in 2/1 mode
@@Vlad-jy9ls Problem is it's a waste of money as it isn't faster than a good tight 6000 kit. But it's more that what do you really get for all the money these Mobos cost beside a bit better connectivity. You still have to choose how you use the available pcie-lanes as the 44 lanes is still not enough to drive 2 pcie 5 gen slots + 2x USB 4 + 3x m.2 gen 5 SSD's.
@@mrdali67 i agree with you
It doesn't go over 6400 stable bro!
I want to build it when a 9800X3D + X870E taichi is released! Pluss DDR5 6400 1.1.1 ratio!
overpriced cpu, overpriced motherboard, ram oc that doesn't matter that much on a x3d cpu.
@@danielt91we dont even know what itll cost, mobo sure but some people dont mind spending a lil on say an asrock taichi which cost nothing compared to other flagships. At this point its for fun
1.1.1 is impossible you can't clock infinity fabric to 3.2ghz
@@danielt91 You sound a bit jealous Daniel
@@danielt91it matters a lot for 1% and 0.1% lows, and overall latency/frametimes, you are just clueless
expo ram runs on intel.. just let it go
You're really good at overclocking but very bad at explaining it in a succinct manner. You should put the settings into a document or screen overlay, put it up, and quickly move through all the timings you got for the fastest performance and the voltage settings. You can add any notes in that are pertinent. The issue is that even someone like me, who is a pretty geeky hardware person, has my brain fuzz out and lose focus at the rambling, droning stream of consciousness with a full screen of random live data. Summarize. Synthesize any revelatory findings. Plan ahead on a format of some sort.
Hey, here are the fastest timings I got. Here is the voltage I had to change. NEXT SECTION: Here is the testing for stability confirmation I use. NEXT SECTION: Here are some hurdles I had to go through. NEXT SECTION: My thoughts, conclusion.
These videos are actually a waste of time. Unless you have the same memory and mother board. With so many gullible people in the world there will always be people selling snake oil.
@actually hardcore overclocking I have hynix a-die that can’t do lower than 37 trp at 6200mhz. I’m wondering can it be because of too low tras (30) or trc? Is it worth it to try raise tras for lower trp or is the performance difference so low that i should not worry about it? Thanks for the help! 😊