I do find it interesting that AMD engineered a solution to its weak memory controller (and infinity fabric latency penalties). It seems AMD relies much more heavily on larger L3 cache, and extremely optimised cache prediction / high cache hit rates. I can see Intel brute forces it with sheer memory speeds and a very strong memory controller, but it is clear that isn't the only road to better performance, it's just that Intel really needs it a lot more than AMD. So many routes to performance improvement.
Gains from memory speed tend to ramp up very fast up to 6000 and starts to plateau after 6400, especially on AMD AM5 CPUs. AMD's problem is an inability to run a 1:1 ratio much past 6400 MT/s. Arrow Lake apparently requires as fast of DDR5 as it can get. It starts competing well with AMD in gaming around and after 8000 MT/s. CUDIMM is godsend for Intel, but it's wild that they need such speeds t o keep up. I have a 7800X3D+7900GRE, 9800X3D+4090, and am building a 265K system with CUDIMM so this topic has been on my mind recently. Arrow Lake is a monster for productivity and virtual machines.
I know with my AMD R5 5600X on a MSI B550 TOMOHAWK the only difference I see is loading times when I moved to faster ram. Everything from level loading in games to everything on the computer loads a lot faster. That's moving to ram with faster speeds and tighter timings
Intel slowly killing overclocking for the average Joe. Locking DLVR in bios updates limits any overclocking. Instead, they thought let's give the buyers of K series sku's more knobs and buttons but keep them away from voltage increases. sad day :'(
He did not mention the importance of CAS Latency (CL). 40 is quite high these days. For gaming at least. 30 is best in price to performance for the time being. 6000CL30
Something wrong with ur 13600k, the 7400 can’t be stable as the 14600k is much faster in most games in ur test dispute them being basically the same cpu with 200mhz difference Also its a i5 not a i7 like ur graph says lol 😂
BUT you are keeping the same slow CL40. Above 200 ratio is good as per the original speeds 8200:cl40 and the 6000:cl40 is just 150 BAD. Try tuning the CL
WHY not comment on CAS Latency or CL? HOW do Intel CPUs perform from 6000CL30? Clearly CAS Latency is more important than Megatransfers or Megahertz as you still called them(the industry changed it to MegaTransfers or MT). Perhaps there is not there not significant difference in 30-34, but 30-40 , 40 seems quite proble problematic .
No it’s not. We did that on 7800x3d and showed it makes performance worse and that 6000mhz cl30 is the sweet spot so what I’m trying to say is your comment is useless 😂
I do find it interesting that AMD engineered a solution to its weak memory controller (and infinity fabric latency penalties). It seems AMD relies much more heavily on larger L3 cache, and extremely optimised cache prediction / high cache hit rates. I can see Intel brute forces it with sheer memory speeds and a very strong memory controller, but it is clear that isn't the only road to better performance, it's just that Intel really needs it a lot more than AMD. So many routes to performance improvement.
Thanks for keeping the CL constant and testing!
Any time!
might matter in productivity
Gains from memory speed tend to ramp up very fast up to 6000 and starts to plateau after 6400, especially on AMD AM5 CPUs. AMD's problem is an inability to run a 1:1 ratio much past 6400 MT/s. Arrow Lake apparently requires as fast of DDR5 as it can get. It starts competing well with AMD in gaming around and after 8000 MT/s. CUDIMM is godsend for Intel, but it's wild that they need such speeds t o keep up. I have a 7800X3D+7900GRE, 9800X3D+4090, and am building a 265K system with CUDIMM so this topic has been on my mind recently. Arrow Lake is a monster for productivity and virtual machines.
I know with my AMD R5 5600X on a MSI B550 TOMOHAWK the only difference I see is loading times when I moved to faster ram. Everything from level loading in games to everything on the computer loads a lot faster. That's moving to ram with faster speeds and tighter timings
Jump on the DDR5 ram kit train when DDR6 is near or released. Wonder how much faster/ stable DDR5 will get nearing EOL?
Intel slowly killing overclocking for the average Joe. Locking DLVR in bios updates limits any overclocking. Instead, they thought let's give the buyers of K series sku's more knobs and buttons but keep them away from voltage increases. sad day :'(
Do the Scottish say CUDIMMY?
you think he is Scottish?
@ no
Вы лучшие спасибо за информацию Интел может подымиться когда нибудь 🙅♂️❤
He did not mention the importance of CAS Latency (CL).
40 is quite high these days.
For gaming at least.
30 is best in price to performance for the time being.
6000CL30
Something wrong with ur 13600k, the 7400 can’t be stable as the 14600k is much faster in most games in ur test dispute them being basically the same cpu with 200mhz difference
Also its a i5 not a i7 like ur graph says lol 😂
BUT you are keeping the same slow CL40.
Above 200 ratio is good as per the original speeds 8200:cl40 and the 6000:cl40 is just 150 BAD.
Try tuning the CL
This is about speed, not cas latency. That’s for another video!
@@eTeknix
CAS Latency clearly influences performance.
Yes, but this is titled about speed, not timings.
WHY not comment on CAS Latency or CL?
HOW do Intel CPUs perform from 6000CL30?
Clearly CAS Latency is more important than Megatransfers or Megahertz as you still called them(the industry changed it to MegaTransfers or MT).
Perhaps there is not there not significant difference in 30-34, but 30-40 , 40 seems quite proble problematic .
Just take AMD!
no 9800x3d ? useless test
No it’s not. We did that on 7800x3d and showed it makes performance worse and that 6000mhz cl30 is the sweet spot so what I’m trying to say is your comment is useless 😂
@eTeknix link ?
He speaking the truth and he is also not the only channel on TH-cam
@ sooooooooooooooooooooo no link ok ......